31st annual symposium on space science and technology 2015.pdf
TRANSCRIPT
Abstracts
ISRO-IISc Space Technology CellIndian Institute Science, Bangalore
8-9 January 2015
31st Annual Symposium on Space Science and Technology
31st Annual Symposium on Space Science and Technology Abstracts
1. A setup for the study of thermal properties of nanoscale devices of ultrathin layered materials Suman Sarkar, Kazi Rafsanjani Amin and Aveek bid
2.Low temperature giant magnetocaloric effect in magnetic ferroelectric GdMnO3 single crystals Aditya A. Wagh, P. S. Anil Kumar and Suja Elizabeth
3. Anomalous piezoelectric response in BaTiO3 based lead-free piezoceramics Ajay Kumar Kalyani, Kumar Brajesh and Rajeev Ranjan
4. Geospatial scenario based modelling of urban revolution in five major cities in India T.V. Ramachandra, Bharath H. A, Vinay S., Venugopal Rao K and. N V Joshi
5. Quantum communication and security using entangled states Ankur Raina, Shayan G. Srinivasa, Chithrabhanu P., Aadhi A., Shashi P. and R. P. Singh
6. Investigation of vegetation changes in the arid Trans-Himalayan ecosystem of northern India Sumanta Bagchi, Ekta Gupta, and Karthik Murthy
7. Study on self-assembly of donor-acceptor-donor molecular materials and melamine Nileshi Saraf, Joydeep Dhar and Satish Patil
8. High-resolution image restoration and enhancement using GPU Nithish Divakar and R. Venkatesh Babu
9.Turbulence-transport-chemistry interaction in statistically planar premixed flames in near isotropic turbulence A. Harshavardhana U, Swetaprovo Chaudhuri and K. N. Lakshmisha
10.Thermally activated unimolecular decomposition mechanisms of dimethylnitramine-aluminum (DMNA-Al) and dimethylnitramine-Zinc (DMNA-Zn) complex Anupam Bera and Atanu Bhattacharya
31st Annual Symposium on Space Science and Technology Abstracts
11. Investigation of spatial and temporal dynamics of vegetation in semi-arid-ecosystems Sumithra Sankaran, Sabiha Sachdeva, Ashwin Viswanathan and Vishwesha Guttal
12. Development of semiconductor nanocrystals for photovoltaics Biswajit Bhattacharyya, Rekha Mahadevu and Anshu Pandey
13. Die level 3D packaging of hybrid systems N. P. Vamsi Krishna and Prosenjit Sen
14. A small volume droplet dispenser based on electrohydrodynamic pulling K. R. Sreejith, and G. R. Jayanth
15. Synthesis optical and electrical characterization of II-VI colloidal quantum dots for IR applications Atul Prakash Abhale , Abhijit Chatterjee, Naresh Babu, Arup Banerjee and K.S.R. Koteswara Rao
16.Movement strategies for commensalism: coexistence of meso-carnivores in human-dominated landscapes Maria Thaker and AbiTamimVanak
17. Development and performance evaluation of a PCM coupled heat pipe G. M. Karthik, Girish Kini, A. Ambirajan and P. Kumar
18.Theoretical and computational study of frictional effects in viscoelastic and generalized elastic contacts D. Satish Kumar and Narayan K. Sundaram
19. Combustion studies of composite propellants containing nano- burn rate catalysts Charlie Oommen and R Arun Chandru
20. Electromagnetic field enhancement in sub-nm gaps Debadrita Paria and Ambarish Ghosh
21. Modelling primary atomization of liquid jets Santosh Hemchandra
31st Annual Symposium on Space Science and Technology Abstracts
22.Development of an optimization based image processing software system for Indian forest resource assessment using Radar Imaging Satellite (RISAT-1) Images S N Omkar and G. Narayana Naik
23. Numerical simulations of Darcy-Brinkman-Forccheimer equations for flow in porous media Karthick M and Gaurav Tomar
24. Double diffusive convection in the earth’s core Venkatesh Gopinath and Binod Sreenivasan
25.Utilizing ionic liquid and mixed solvent electrolytes to synthesize polymer electrolytes for lithium-based batteries Sudeshna Sen, Sneha Malunavar, C. Gouri and Aninda J. Bhattacharyya
26.Sparsity-based cross-terms-suppressed time-frequency distribution of multi-component linear frequency modulated signals Shreyas Hampali, Subhankar Ghose, and Chandra Sekhar Seelamantula
27. Investigating the origin of the Indian ocean Geoid low Attreyee Ghosh
28.Evolution of mesoproterozoic suture zones western India: Implications on India-Madagascar corre-lations C. Ishwar-Kumar and K. Sajeev
29.Estimation of soil hydraulic properties in a catchment using agro-hydrological models and micro-wave remote sensing K. Sreelash, M. Sekhar, S. K. Tomer, S. Bandyopadhyay and M.S. Mohan Kumar
30. Tribological properties of polytetrafluorethylene at cryogenic temperatures D.S. Nadig, V.K. Pavan and Paul P. George
31st Annual Symposium on Space Science and Technology Abstracts
31. Portable imaging flow analyzer for biological research applications in space Veerendra Kalyan Jagannadh and Sai Siva Gorthi
32.Processing of metallic thermal interface materials using liquid phase sintering followed by accumulative roll-bonding Deepak Sharma, Rajesh Kumar Tiwari, P. Ramesh Narayan and Praveen Kumar
33. Development of nanoparticle based radiation detectors for space applications B.H.M. Darukesha, V. Radhakrishna, M. Ravindra and K. Rajanna
34. Spatial coherence of tropical rain R. Ratan, V. Venugopal, J. Sukhatme and R. Murtugudde
35.Amorphous silicon carbide thinfilms by pulsed DC magnetron sputtering for micro electro-mechanical systems (MEMS) applications Habibuddin Shaik, M.A. Sumesh and G. Mohan Rao
36.Low temperature electrical transport studies on carbon nitride films prepared by chemical vapour deposition K. Ramesh, M. Prashantha, R.Venkatesh, N. Naresh and M.V.N. Prasad
37. Generation of BPSK/QPSK modulated microwave signals using optical techniques K R Yogesh Prasad, T Srinivas, Gopal Hegde and Abdul Hameed
38. Development and characterisation of nano-porous aluminium alloy surfaces Arti Yadav, Subrata Chakarbarti, A. Manimaran, M.S Bobji
A setup for the study of thermal properties of nanoscale devices of
ultrathin layered materials
Suman Sarkar, Kazi Rafsanjani Amin and Aveek bid
Department of Physics
Indian Institute of Science
Abstract: The aim of the proposed project
is to develop a research facility for
exploring thermal properties (Seebeck
effect, thermal conductivity etc.) of
nanoscale devices based on different forms
of layered semiconductor materials like
graphene and Bi2Se3/Bi2Te3. This will be
integrated to an electrical transport
characterization system to allow a
comprehensive study of transport and
thermodynamic properties in
nanostructured materials. In addition we
would like to measure fluctuations in
thermal conductivity to probe the effect of
defects on thermal conductivity of
technologically important materials. The
set up developed by us over the last few
months is capable of measuring voltage
fluctuations as low as 10-20 V2/Hz over a
wide band-width. We are currently in the
process of integrating the system with a low
temperature cryostat that will enable us to
carry out the measurements of thermal
conductivity over a temperature range 1.5-
300K. We have also done preliminary
measurements of single layer graphene
devices using this set-up.
I. INTRODUCTION In recent years layered semiconductor materials like
graphene and MoS2 have emerged as possible
candidates for post silicon technologies. Although
most present studies focus on their electrical properties,
their unique thermal properties make them very
interesting candidate for potential applications. It has
been proposed that the thermal conductivity of
graphene is highest ever measured in any material. For
graphene on silicon-dioxide substrate the thermal
conductivity is about 500W/mK, which is much
higher than that of bulk gold. For suspended graphene
devices the value of thermal conductivity can be as
high as 5000W/mK1, almost ten timed higher than that
of any metal. In addition to an exceptionally large
thermal conductivity, graphene is highly conductive
even in very low carrier density regimes2 and hence has
extremely low levels of Johnson-Nyquist thermal noise
as compared to semiconductor based sensors. It also
has fewer kinds of defects and hence has intrinsically
low levels of 1/f noise3 arising out of thermal switching
of defects. It is relatively easy to make four terminal
measurements on graphene strips making contact
resistances much easier to deal with than for example
in carbon nanotubes which share almost all other
advantages of graphene. On the mechanical side,
suspended graphene flakes have been shown to have a
very high Young’s modulus (~ 1TPa) and to have a
much higher elasticity than membrane materials like
silicon nitrate commonly used in NEMS4.
Both graphene and single-walled carbon nanotubes
exhibit excellent electrical conductivities and large
specific surface areas. Calculations show that a
covalently bonded graphene/single-walled carbon
nanotube hybrid material would extend those
properties to three dimensions. There have been several
proposals for exploiting the unique properties of
graphene-CNT hybrid for making body of aircraft.
Such a hybrid would have a reasonably good electrical
conductivity – an important criterion for spacecraft
safety. The large thermal conductivity of the matrix
could provide a strong, light weight and thermally
conducting outer shell for reentry vehicles. A study of
the thermoelectric figure of merit of the material would
give an indication of the applicability of these materials
as waste energy harvesters.
The measured thermal conductivity of graphene on
SiO2 substrate, although very high, is still lower than
the theoretical predicted value. In order to understand
the factors that limit the thermal conductivity of
substrated graphene below the theoretical limit
extensive measurements of the fluctuations (or noise)
in thermal conductivity of graphene need to be carried
out. To the best of our knowledge no such studies have
been carried out till date – either in pristine graphene
or in graphene based matrices.
.
II. METHODS
We have taken two different approaches to measuring
the thermal conductivity of layered semiconducting
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
materials. In the first approach, we have optimized a
technique to fabricate microheaters that have extremely
large intrinsic thermal time constants. This technique is
essential for devices where the thermal conductivity is
less than that of the substrate (for example
Bi2Se3/Bi2Te3 on SiO2).This will enable us to detect
small changes in the thermal signature of single or few
layered materials. The thermal conductivity of layered
semiconductors vary over several orders of magnitude-
from about 1W/mK in the case of BiSe to about 5000
W/mK for suspended graphene devices. The
microheaters have to be optimized to handle this entire
dynamic range of thermal properties. Figure 1 shows
the SEM image of one such device.
Figure 1: Microheater for measuring thermal conductivity of
layered materials.
In the second approach we are measuring the thermal
conductivity of graphene and BiSe systems using the 3
Omega technique. This is useful for devices where the
conductivity of the sample is larger than that of the
substrate (for example - graphene on SiO2 substrate)
Data for a typical measurement on graphene single
layer is shown in Figure 2. The data is plotted as a
function of the number density of charge carriers
introduced in the system by a back gate. We note that
as the carrier density increases the electronic
contribution to thermal conductivity increases as is
expected from Wiedemann-Franz law.
1 A. Balandia et.al., Nano letts. 8, 902 (2008). 2 K. S. Novoselov et.al. Nature 438, 197 (2005). 3 P. Dutta et.al. Rev. Mod. Phys. 53, 497 (1981).
Figure 2: Electronic contribution to the thermal conductivity
of graphene
We are currently also studying the thermal
conductivity of MoS2 - a material which is interesting
for many practical applications. A typical image of the
device is shown in figure
Figure 3: Optical microscope image of multi-layer MoS2
device.
III. CONCLUSION
We are in the process of developing the necessary
techniques for measuring thermal properties of single
layered and few layered semiconducting materials. The
measurements are extremely challenging because in
most cases the thermal signal from the substrates can
be orders of magnitude larger than that from the sample
under test. We have done preliminary tests and are now
in the process of measuring thermal properties of
graphene, MoS2 and BiSe single and few layered
devices.
4 I. W. Frank et.al. J. Vac. Sci. Technol. B 25 2558
(2007
Project Number:ISTC/PPH/SE/325
Low Temperature Giant Magnetocaloric Effect in Magnetic Ferroelectric GdMnO3 Single Crystals
Aditya A. Wagh,P. S. Anil Kumar, and Suja Elizabeth
Department of Physics, Indian Institute of Science, Bangalore 560012, India
In a typical magnetic material,
magnetic sublattice is sensitive to external magnetic fields and, due to this coupling, the magnetic part of the entropy changes with variation of magnetic field. Magnetocaloric effect (MCE) is generally demonstrated by isothermal change in magnetic entropy or adiabatic change in temperature with change in magnetic field. At room temperature, GdMnO3 is paramagnetic and paraelectric.It exhibits complex magnetic phase diagram (H-T phase diagram).1, 2 At 42 K (TN(Mn)), paramagnetic phase transforms to sinusoidally-modulated collinear antiferromagnetic phase with modulation vector directed along the ‘b’ axis which is incommensurate in nature where no spontaneous electric polarization is observed.1,. The competing AFM and ferromagnetic exchanges in ab-plane cause magnetic frustration, which in turn is responsible for long wavelength modulated AFM spin structures at low temperature.3, 4 Near 23 K (Tlock), the system transforms to a CA-AFM phase with weak canting component of Mn3+ spins lying along ‘c’ axis.2 GdMnO3 is known to display AFM ordering of rare earth (Gd3+) sublattice when the temperature is lowered further.2
Single crystals of GdMnO3 were grown using optical float-zone method. A four-mirror image furnace (argon ambience; flow rate ~ 1 lit/min) was employed which yields crystals of typical length 36 mm at growth rate of ~ 5mm/h. A photograph of the grown crystal is shown in Fig.1.
Fig.1: Single crystal of GdMnO3 M-H isotherms were recorded along
the three crystallographic axes using zero-
field protocol as shown in Fig.2. Magnetization as a function of temperature data were extracted from measured M-H isotherms and used for the calculation of ΔSM.
Fig.2: (a) Typical set of zero field-cooled magnetization – field (M-H) isotherms recorded at short temperature intervals along (a) `c' axis, (b) `a' axis and (c) `b' axis. The temperature evolution of numerically calculated ΔSMfor the three axes is plotted in Fig.3. The magnitude of
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
ΔSMincreases prominently below Tlock transition and records the highest value in the proximity of 7 K. This is likely to occur due to enhanced polarized paramagnetic moment of Gd3+ from weak ferromagnetic component of the canted Mn3+ moment in this phase. Further decrease in the temperature leads to a sharp fall in ΔSMresulting in an asymmetric broad peak. An anomaly is observed in ΔSMin the form of a sharp peak at 5 K along `c' axis, but it is feeble in other axes. This feature becomes pronounced as the field is increased and is perceived to be due to spin reorientation of the rare earth sublattice. Above a critical field, along `b' axis, the ferroelectric phase (ab-cycloidal) stabilizes which then transforms to paraelectric phase with increase in temperature. This manifests in step-like decrease in the ΔSMvs T (shown by arrow in Fig.3.c. The field-induced
transformation (FOPT) to ferroelectric phase (i.e. from CA-AFM to ab-cycloidal phase) along 'b' axis is clearly discernible as a step-like feature in the plot of field variation of ΔSMFig3.d. The distinct step-like feature, observed along `b' axis, is absent along other axes.The calculated value ofΔSM is very large (~ 31.8 J/kg K) along 'c' axis for the field variation 0-80 kOe at 7 K. The anisotropy inΔSMVs T is not strong (At ~7 K, ΔSM = 29.1 and 25.7 J/kg K for 'a' and 'b' axes, respectively). From the phase diagram, 'c' and 'a' are 'easy' and 'hard' axes. However, at higher fields along 'b' axis the system exists in ferroelectric cycloid AFM phase. In the respective ab-cycloid spin structure, the 'b' axis could be considered the 'hard' axis. It is prudent to compare ΔSMvalues in various AFM phases.
Fig. 3:Temperature evolution of estimated ΔSM for field variation up to 80 kOe along different crystallographic axes: (a) `c'axis, (b) `a' axis and (c) `b' axis. (d) ) Field variation of ΔSM along `b' axis. The inset shows the magnified view of the low-field region. References [1] T. Kimura, et al.,Physical ReviewB 71, 224425 (2005). [2] J. Hemberger, et al., Physical Review B 70, 024414 (2004). [3] M. Mochizuki, N. Furukawa, Physical Review B 80, 134416 (2009). [4] J. Ribeiro, Ferroelectrics 368, 114 (2008).
1
Anomalous Piezoelectric Response in BaTiO3based Lead-free Piezoceramics
Ajay Kumar Kalyani, Kumar Brajesh, Rajeev Ranjan
Department of Materials Engineering, Indian Institute of Science, Bangalore-560012, Karnataka, India
Abstract
The effect of Zr, Hf and Sn substitutions in BaTiO3has been investigated at close composition intervals in the dilute concentration limit. Detailed structural analysis by x-ray and neutron powder diffraction revealed that merely 2 mol percent of Zr, Sn and Hf stabilizes a coexistence of orthorhombic (Amm2) and tetragonal (P4mm) phases at room temperature. As a consequence all the three systems show substantial enhancement in the longitudinal piezoelectric coefficient (d33), with Snmodification exhibiting the highest value ~425 pC/N.
I. INTRODUCTION
Research on lead free piezoelectric materials have received considerable momentum in the recent past because of serious environmental concerns with regard to the toxicity of the commercial piezoelectric, lead zirconate titanate (PZT), which contains ~60 weight percent of lead. Lead free compounds such as Na1/2Bi1/2TiO3 (NBT). (K,Na)NbO3 (KNN) and BaTiO3 (BT) have been extensively investigated in pure and modified form with renewed interest. Both BaTiO3 and KNbO3 exhibit the same sequence of phase transformations, namely cubic(C)
tetragonal (T) orthorhombic (O) rhombohedral (R) on cooling from high temperature [1]. Though the high Curie point of KN (Tc ~410 C) as compared to BT (Tc ~120 C) is attractive, the difficulties associated with the fabrication of dense ceramic bodies, among other issues, have not yielded encouraging piezoelectric properties in KN based systems. The best property has been reported for a textured Li and Sb modified (K,Na)NbO3 [2]. Very high electric field driven strain values have been reported for certain compositions in the quaternary (K,Na)NbO3-NBT-BT system [3]. BT based ceramics have received attention only in the recent few years after discovery of anomalous properties in the (Ba, Ca)(Ti, Zr)O3system [4]. The essential idea used by Ren et al is to bring about a triple point – morphotropic phase boundary (TMPB) state, i.e a coexistence of tetragonal, orthorhombic and cubic phasesclose to room temperature. Though this approach can drastically increase the piezoelectric property due to the system’s proximity to triple point criticality, the piezoelectric properties would drastically deteriorate by slight increase in temperature since one of the phase at the triple point is cubic phase and therefore non-ferroelectric in nature. From this standpoint, it is always desirable to have piezoelectrics with Curie point well above room temperature.
The dielectric and piezoelectric properties of modified and pure BaTiO3 ceramics have been documented by Jaffe and Jaffe [1]. Of particular interest are the compositional modifications which can induce inter-ferroelectric instability, and coexistence of ferroelectric phases at room temperature. The two inter-ferroelectric transitions, tetragonal-orthorhombic and orthorhombic-rhombohedral occur at ~ 0 C and -90 C in pure BaTiO3. A perusal of literature suggests thatZr [5], Sn [6]and Hf [7]substitution in BaTiO3increase both the inter-ferroelectric transition temperatures from below room temperature. This offers interesting opportunity to chemically tune the orthorhombic-tetragonal and rhombohedral-orthorhombic phase transition near to room temperature for enhanced piezoelectric properties mimicking a morphotropic phase boundary scenario.We have carried out detailed structural, ferroelectric and piezoelectric measurements on a series of compositionally modified (Zr, Sn and Hf) BaTiO3. It was found that 2 mole percent of the other two substituents,SnZr and Hf remarkably increase the piezoelectric property by stabilizing coexistence of ferroelectric tetragonal and orthorhombic phases at room temperature. A longitudinal direct piezoelectric coefficient of 425 pC/N was obtained for Sn modified BaTiO3.
II. EXPERIMENTAL
Solid solution of Ba(Ti1-xZrx)O3(BZT), Ba(Ti1-
xSnx)O3(BST), Ba(Ti1-xHfx)O3(BHT) were prepared by solid state route using high purity BaCO3, TiO2, ZrO2, SnO2, HfO2 compounds. Powders were milled in planetary ball mill (P5, Fritch) using acetone medium for 10-12 hrs. Milled powder were dried and calcined at temperature of 1100 C for 4hrs. Sintering was done by two step process first heated to 1300 for 4 hrs and then heated to 1500 C for 6hrs. X-ray powder diffraction was done using Bruker (D8, advance). Neutron powder-diffraction (NPD) experiment was carried out at the SPODI diffractometer at FRM II, Germany using a wavelength of 1.548 Å. Dielectric measurement were done on Novocontrol (Alpha AN) impedance analyzer. Measurement of longitudinal piezoelectric coefficient (d33) was carried using Piezotest PM 300 by poling the pellets at room temperature for ~ 1 hour at a field of ~ 2 kV/cm. Rietveld refinement was carried out using FULLPROF software[8].
Project No. ISTC/MET/RR/323
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
2
III. RESULTS AND DISCUSSION
Figure 1 shows thecomposition dependence of the longitudinal piezoelectric coefficient (d33) ofBZT, BST and BHTsystems. All the three systems exhibits substantial enhancement of the piezoelectric coefficient. In BST d33increases from 190 pC/N (for x=0) to ~270 pC/N at x = 0.01, reaching a maximum ~425 pC/N at x = 0.02. Then onwards the value exhibits saturation after a slight decrease.For BZT, fig 1(b), a similar enhancement in d33 is observed reaching a maximum of ~370 pC/N at x = 0.02. After a slight drop at x=0.03, d33is nearly constant in the range 0.03≤x≤0.06. For BHT (fig 1(c)) maximum d33 (~370 pC/N) was again obtained at x = 0.02.
0 2 4 6 8
100
200
300
400
0 2 4 6 8 10
100
200
300
400
0 2 4 6 8 10
100
200
300
400
Mole % Hf
(c)
(b)
Mole % Sn
(a)
d 33 (p
C/N
)
Mole % Zr
Figure 1. Composition dependence of longitudinal piezoelectric coefficient (d33) in BaTi1-y XyO3. (a) X = Sn, (b) X = Zr and (c) X =Hf.
Fig 2 shows the temperature variation of relative permittivity of BZT, BHT and BST. For 2 mol % substitution all the three systems exhibit sharp dielectric anomaly corresponding to the tetragonal-cubic transition at Tc ~120 C. Interestingly the tetragonal-orthorhombic dielectric anomaly TT-O is shifted from 0 C to just above room temperature ~35 C for BZT (x=0.02) and BHT (x=0.02). This anomaly was not captured above room temperature in BST (x=0.02), which exhibits the highest d33 among the three systems but was clearly revealed for higher Snconcentration (4 mol %).
40 80 120 160 2000
3000
6000
9000
12000
15000
0
3000
6000
9000
12000
15000
30 60
30 40 50
Rel
ativ
e pe
rmitt
ivity
Sn = 0.04 Sn = 0.02(b)
Temperature (°C)
Zr = 0.02 Hf = 0.02(a)
Figure 2 Temperature dependence of dielectric constant of BaTi1-yXyO3 (X = Zr, Hf, Sn). The insets show the magnified plot near room temperature
0 2 4 6 8 10 12
50
100
150
0 2 4 6 8 10 12
50
100
150
0 2 4 6 8 10 12 14 16
50
100
150 Mole % Sn
Mole % Zr
TO-R
TT-O
Tran
sitio
n te
mpe
ratu
res
(°C
)
Mole % Hf
Tc
TO-R
TT-O
Tc
TT-O
Tc
(c)
(b)
(a)
Figure 3: Phase diagram of BaTi1-yXyO3 (X = Sn, Zr, Hf) constructed from the anomaly temperatures in dielectric studies.
Based on the dielectric anomaly temperatures measured for different compositions of the three systems under investigation, a phase diagram has been plotted in Fig. 3. The three substitutions Zr, Hf and Sn raise the orthorhombic-tetragonal (TO-T) and the rhombohedral-orthorhombic (TR-O) transition temperatures and decrease the tetragonal-cubic transition (Tc) temperature. The phase
3
45 46 65.8 66.5 83.5 84.0
45 46 65.6 66.4 83.5 84.0
44.8 45.6 65.6 66.4 83.2 84.0
45 46 65.6 66.4 83 84
45 46 65.6 66.4 83 84
45 46 65.6 66.4 83.5 84.0
45 46 65.6 66.4 83.5 84.0
(c)
(b)
(a)
Sn = 0.04
Sn = 0.02
BT
Inte
nsity
(arb
. uni
ts)
2θ (degree)
Zr = 0.02
Zr = 0.03
Hf = 0.02
Hf = 0.03
O
OO
(g)
(f)
O
O
(e)
O(d)
O
222C220
C
200C
O
O
Figure 4. X-ray powder Bragg profiles of pseudo cubic 200c, 220c and 222c reflections of BaTi1-xXyO3 The arrows indicates the reflections from the respective phases ( O = orthorhombic reflection)
diagrams based on dielectric measurements suggest that the minimum concentrations of the substituents which can help stabilize the orthorhombic phase at room temperature is x=0.02 for BZT and BHT and x=0.04 for BST. Both BZT and BHT exhibits rhombohedral-orthorhombic dielectric anomaly above room temperature in the composition range 0.08<x≤0.011. This anomaly could not be captured in the Sn modified BaTiO3. From the trend of the three phase boundaries in the phase diagrams of BZT and BHTthe possibility of the merging of the boundaries can be anticipated somewhere in the composition range 0.11<x<0.15. Though this scenario may suggest the possibility of coexistence of four phases-cubic, tetragonal, orthorhombic and rhombohedral, as has been reported in a recent work by Yaoet al [9],this is not feasible from thermodynamic view point since for a (pseudo)binary systemGibb’s phase rule does not allow more than three phases to coexist in equilibrium at any given pressure.
20 40 60 80 100 120 140
0
4
8
12
100 101 102 118 119 120
20 40 60 80 100 120 140
0
2
4
6
118.5 120.0100 102Inte
nsity
(cou
nts
104 )
Yobs Ycal Y diff peak position
Zr = 2 mole %(a)
204C400C
2θ (degree)
Sn = 4 mole %(b)
Amm2P4mm
Amm2
P4mm
240C
400C
Fig 5 Rietveld fitted neutron powder diffraction patterns of (a) BaTi0.98Zr0.02O3 and (b) BaTi0.96Sn0.04O3 with tetragonal (P4mm) + orthorhombic (Amm2) phase coexistence model. The insets show magnified plot of two pseudocubic reflections 400c and 240c
Sinceas per the dielectric study the highest d33 in BST seems to occur in the tetragonal phase whereas in BZT and BHT in the orthorhombic phase, a unified structure-property correlation for all the three systems was not possible to establish. A detailed structural characterization was therefore carried out. Figure 4(a-g) shows the x-ray powder Bragg profiles of three pseudocubic reflections 200C, 220C and 222C for different compositions of the three systems.For sake of direct comparison, the corresponding Bragg profiles of unmodified BaTiO3 are also given (figure 4a).Interestingly, in contrast to the dielectric study which suggests a tetragonal phase for BST x=0.02, the XRD pattern of this composition revealed additional features (pointed with arrow) apart from the tetragonal peaks. Further, though the phase diagram suggests orthorhombic phase for BST with x=0.04 at room temperature, tetragonal peaks are also present along with the orthorhombic peaks for this composition. The same phase coexistence was found in both BZT and BHT with x=0.02. Thus irrespective of whether the tetragonal-orthorhombic criticality as revealed by the dielectric anomaly is above or below room temperature, the structural analysis clearly proves that all the three systems exhibit coexistence orthorhombic and tetragonal phases at room temperature for the composition (x=0.02). Fig. 5 shows Rietveld fitted neutron diffraction patterns of BZT (x=0.02) and BST (x=0.04) with tetragonal (P4mm) + orthorhombic (Amm2) phase coexistence model.The insetsshow magnified plot of fits of two high angle pseudocubic reflections 400c and 204c. For both the cases, if the structure were single
4
phase (either tetragonal or orthorhombic)the each of the pseudocubic 400Cand 204c profiles wouldbe expected to split into two peaks. The presence of the extra reflectionsand their satisfactory modeling confirms the chosen phase coexistence model. The refined structural parametersare given in tables 1 and 2.
The observation of enhanced piezoelectric response in a system comprising of two ferroelectric phases is analogous to PZT which exhibits anomalous piezoelectric response at the morphotropic phase boundary (MPB) compositions exhibiting coexistence tetragonal + rhombohedral/monoclinic ferroelectric phases. The anisotropic lowering of the energy barrier separating two different ferroelectric phases with different orientation of
the polarization vector near a temperature/composition/pressuredriven critical point provides low energy polarization rotation pathway [10]. Similar to BaTiO3, KNN also exhibits three polymorphic phase transitions (Tc ~410 oC, TT-O ~200 oC, and TO-R ~ -123 oC). In this system also, the enhanced electromechanical response is observed by tailoring the TT-
O, Li-substitution at the (K0.5Na0.5)site in KNN has been reported to lower the TT-Otransition temperature close to room temperature, and stabilizes the two ferroelectric (P4mm+Amm2) phases [11]. Similarly ‘Sb’ substitution at the Nb-site in KNN raises the TO-Rtransition temperature.
Table1: Refined structural parameters and agreement factors for Ba(Ti0.98Zr0.02)O3 using Tetragonal(P4mm) + orthorhombic(Amm2) phase coexistence models.
Space group: P4mm Space group: Amm2
Atoms x y z B(Å 2) x y z B(Å2)
Ba 0.000 0.000 0.000 0.46(5) 0.000 0.000 0.000 0.1(1)
Ti/Zr 0.500 0.5000 0.470(1) 0.3(1) 0.500 0.000 0.526(1) 0.31(9)
O1 0.500 0.5000 0.029(1) 0.31(6) 0.000 0.000 0.5114(9) 0.62(5)
O2 0.500 0.000 0.523(1) 0.37(1) 0.500 0.2491(4) 0.2464(5) 0.61(4)
a=3.99782(7) Å, c=4.03493(9) Å
v= 64.488(2) Å3, %Phase = 33(1)
a= 3.99597(5) Å, b=5.69071(8) Å,
c=5.67698(8) v=129.094(3) Å3, %Phase = 67(1)
Rp: 6.80, Rwp: 5.06, Rexp: 3.05, Chi2: 2.75
Table2: Refined structural parameters and agreement factors for Ba(Ti0.96Sn0.04)O3 using Tetragonal(P4mm) + orthorhombic(Amm2) phase coexistence models.
Space group: P4mm Space group: Amm2
Atoms x y z B(Å 2) x y z B(Å2)
Ba 0.000 0.000 0.000 0.5 (1) 0.000 0.000 0.000 0.07(9)
Ti/Sn 0.500 0.5000 0.471(2) 0.2(2) 0.500 0.000 0.495(2) 1.3(1)
O1 0.500 0.5000 0.024(1) 0.1(1) 0.000 0.000 0.498(2) 0.6(1)
O2 0.500 0.000 0.510(2) 0.5(1) 0.500 0.2577(8) 0.2371(5) 0.83(6)
a=4.00183(8) Å, c=4.02807(12) Å
v= 64.508(3) Å3, %Phase = 39(2)
a= 3.99980(8) Å, b=5.6857(1) Å,
c=5.6754(1) v=129.096(5) Å3, %Phase = 61(2)
Rp: 8.90, Rwp: 5.94, Rexp: 4.16, Chi2: 2.04
5
Thus the mechanism associated with the enhanced piezo-response in BaTiO3 and KNN based systems is primarily based on the similarity of the topology of their respective phase diagrams. While two different types of chemical substitutions are required to stabilize two different combinations of ferroelectric phasesat room temperature in KNN,in BaTiO3 the same substituent (either Zr, or Hf or Sn) does the job by varying the concentration. The first order nature of the P4mm-Amm2 and R3m-Amm2 transitions in these systems ensures their coexistence around the critical point. The temperature range of this coexistence islikely to be increased due to random elastic strain induced in the lattice by substitution of Zr/Hf/Sn at the Ti site. This explains why BST still exhibits the same phase coexistence as BZT and BHT for x=0.02 inspite of the difference in their respective critical temperatures.
IV. CONCLUSIONS
In conclusion, we have shown that merely 1-3 mole percent of Zr, Sn and Hf substitution in
BaTiO3considerably enhances the piezoelectric response of the system with Sn modification exhibiting the highest piezoelectric response of 425 pC/N. A detailed structural analysis by x-ray and neutron powder diffraction revealed that these dilute modifications stabilize a coexistence of ferroelectric phases, orthorhombic and tetragonal,at room temperature. It is this phase coexistence that is responsible for the anomalous piezoelectric response exhibited by these systems. Our results suggests the possibilities that there might be other elements as well which can enhance the piezoelectric properties of BaTiO3. More work needs to be carried out to explore this possibility.
Acknowledgement
RR thanks the ISRO-IISc Space Technology Cell for financial assistance.
References [1] B. Jaffe, Piezoelectric Ceramics (Elsevier Science,
(2012). [2] Y. Saito, H. Takao, T. Tani, T. Nonoyama, K. Takatori, T. Homma, T. Nagaya, M. Nakamura, Nature, 432, 84 (2004). [3] W. Jo, T. Granzow, E. Aulbach, J. Rödel, and D.
Damjanovic, J. Appl. Phys. 105, 094102 (2009). [4] W. Liu and X. Ren, Phys. Rev. Lett. 103, 257602(2009)
[5] G. H. Jonker, and W. Kwestroo, J. Am. Ceram. Soc. 41, 390 (1958). [6] G. A. Smolenskii, V. A. Isupov, Dokl. Akad. Nauk USSR, 96, 53 (1954).
[7] E. G. Fesenko, O. I. Prokopalo, Soviet Phys. Cryst. 6, 373 (1961)
[8] J. R. Carvajal, FULLPROF, A Rietveld Refinement and Pattern Matching Analysis Program, Laboratoire Leon Brillouin (CEA-CNRS), France.
[9] Y. Yao, C. Zhou, D. Lv, D. Wang, H. Wu, Y. Yang, and X. Ren, Europhys. Lett, 98, 27008 (2012).
[10] D. Damjanovic, Appl. Phys. Lett. 97, 062906 (2010). [11] Y. Dai, X. Zhang, G. Zhou, Appl. Phys. Lett, 90,
262903 (2007
[12] R. Zuo, J. Fu, D. Lv, and Y. Liu, J. Am. Ceram. Soc. 93, 2783 (2010).
1 ISTC/BES/TVR/313
Geospatial scenario based modelling of urban revolution in five major cities in India T.V. Ramachandra, Bharath H. A, Vinay S., Venugopal Rao K and. Joshi N V
Abstract - Urbanisation being irreversible and very rapid with fast growth of population and settlements during the last century, needs to be monitored and visualised for evolving strategies towards sustainable development approaches. This study visualises the growth of major Tier I cities of India such as Delhi, Mumbai, Pune, Chennai and Coimbatore, India through AHP Fuzzy based Cellular Automata Markov model. CA Markov model is considered to be one of most effective algorithm to visualise the growth of urban spatial structures. The analysis performed in previous studies such as spatial pattern of land use change in the area was considered and the future growth was modelled considering agents of growth applying fuzzy AHP CA-Markov model. The projection as predicted using the model was validated to obtain better validation and was then used to predict future projections. Modelling suggested a clear trend of various land use classes’ transformation in the area of urban built up expansions.
I. INTRODUCTION
Rapid urbanization is one of the most important factor affecting the local ecology and loss of biodiversity in India during the last two decades (Shivaramakrishnan et al., 2005; MOUD., India, 2011; Ramachandra et al., 2012; Bharath H. A., 2012; Ramachandra et al., 2014a). Urbanisation is a form of growth with implications of economic, social, and political forces and to the physical geography of an area (Sudhira et al., 2007; Ramachandra et al., 2014a). The sprawl takes place at the urban fringes resulted in radial development of the urban areas or development along the highways results in the elongated development of urban forms (Sudhira et al., 2003).
Dr. T V Ramachandra, Coordinator of Energy and Wetlands
Research Group (EWRG), Centre for Ecological Sciences (CES),
Indian Institute of Science (IISc), Bangalore.
Bharath H Aithal, is a Junior Research Fellow at Indian Institute
of Science (IISc). [email protected]
Vinay S, is a research scholar at Energy and Wetlands Research
Group (EWRG), Centre for Ecological Sciences (CES), Indian
Institute of Science (IISc). [email protected]
Joshi N V, Faculty at Centre for Ecological Sciences (CES), Indian
Institute of Science (IISc), Bangalore
Venugopal Rao K, is Group Head, Urban Studies &
Geoinformatics, National Remote Sensing Centre (NRSC), Indian
Space Research Organisation, Hyderabad, India.
It can also be defined as a finite cycle through which nations evolve to form industrially dominant regions, which further results in rural push and spreading of city towards outskirts (Ramachandra et al., 2013) also refers to urban sprawl.
This urban development in the fringes is called sprawl. The study on urban sprawl was attempted by various researchers across the globe (Batty et al., 1999; Torrens, 2000; Sudhira et al., 2004; Huang et.al 2007; Bhatta, 2009a, 2009b, 2010; Ramachandra et al., 2012). These sprawl areas do not have a fixed plan or process of development due to which the process of preparing visionary documents such as developmental plans, specific corridors of developments are being ineffective considering the fact that spatial patterns and dynamic behaviour of growth and also may be attributed to lack of skills and tools to help in informed, accurate decision making (Adhvaryu, 2011; Bharath H.A., et al., 2014). This can be improved and timely decisions may be enabled using technological improvements such as remote sensing and tools such as Geographic Information system (GIS).
Remote Sensing data acquired through space borne remote sensors enables a bird eye view of the landscape at low cost (Lillesand and Kiefer, 2005). The advantage of remote sensing data is to acquire repeated measurements of the same area on periodic basis which helps in detection and monitoring of LULCC and surveillance of problematic sites (Campbell, 2002). The analysis of changes at local, regional and global scales is possible through the collection of remote sensed data covering the larger spatial extent. Remote sensing aids in identification and assessment of land use patterns which is important for environmental management and decision making. Further it is essential to visualise and provide better planning strategies for future urban growth. This can be planned and visualised using various modelling techniques.
Traditional large-scale urban simulation approaches of early 90’s were based on theories, and suffered from significant weaknesses such as poor handling of space-time dynamics and too much generalisation of data. The integration of space, time, and attributes in modelling was further enhanced with the implementation of Cellular automata (CA) models (Allen 1997; Batty 1999; EPA 2000; Alberti and Waddell 2000). CA modelling is capable of addressing the spatial complexity with discrete time change. A
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
2 ISTC/BES/TVR/313
number of CA-based models of urban growth have produced satisfactory simulations of spatial urban expansion over time (Clarke et al., 1997; Leao et al. 2004; Bharath and Ramachandra, 2013; Ramachandra et al., 2013; Arsanjani et al., 2013).The main advantages of CA are simplicity, easy integration with raster GIS, and adaptability to various urban growth situations. CA models can realistically generate and represent complex patterns through the use of simple rules and considering its neighbouring properties since these models operate on basis of cell states, size, neighbourhood and transition rules (White and Engelen 2000). This communication presents the Land use change modeller (Bharath H.A. et al., 2013) and Fuzzy AHP based CA (Bharath H.A et al., 2014) models implemented to visualise the urban growth in five Tier I cities in India.
II. DATA AND METHOD
Temporal remote sensing data of Landsat TM and ETM+ downloaded from GLCF were preprocessed to correct geometrical and radiometrical accuracy USGS (http://www.usgs.gov). This was further used to analyse and model LULC changes. Remote sensing data were supplemented with the Survey of India topographic maps (of 1:50000 and 1:250000 scale), which were used to generate base layers of the administrative boundary, drainage network, Road network etc. Slope map was extracted using ASTER data (30 m) downloaded from USGS (www.usgs.gov). Ground control points (GCPs) and training data were collected using pre calibrated Global Positioning System (GPS) and virtual online spatial maps such as Bhuvan and Google Earth. GCPs were useful in geometric correction of remote sensing data. Census data (1991, 2001 and 2011) was used to capture population dynamics. Modelling of urbanization and sprawl involved: i) Remote Sensing data acquisition, geometric correction, field data collection, ii) Classification of remote sensing data and accuracy assessment using GRASS, iii) Land use analysis, iv) Identification of agents and development of attribute information, v) Prediction: Designing scenarios of urban growth and calibrating the model to find out the best weights based on the influence on the neighborhood pixels vi) Accuracy assessment and validation of the model, vii) Prediction of future growth based on validated data. Land use analysis was carried out using supervised pattern classifier - Gaussian maximum likelihood algorithm based on probability and cost functions (Duda et al., 2000). Land use with gradient analysis results were further used in Modelling. These results can be accessed in previous working literatures
(Ramachandra et al., 2014a, 2014b, 2014c, 2014d, Chandan et al., 2014). These data was used in Modelling and visualizing the growth of these cities.
III. MODELLING USING FUZZY AHP-CA
Using the combination of Fuzzy Logic, Analytical Hierarchical Process (AHP), Multi Criteria Evaluation (MCE), Markov chains and Cellular Automata (CA). Agents of urbanisation such as roads, industries, educational institutions, bus stands, railway stations, metro, population, etc. were normalized. Conservation regions as per city development plan (CDP) water bodies were considered as constraints. The fuzzy based analysis is used to normalize the contributing factors between 0 and 255, where 255 showing the maximum probability of change and 0 indicating no change, for different land uses. The normalized agents were taken as input to AHP to determine the weights of driving factors using pair wise comparisons ith weights as Eigen vectors. The weights analysed and calibrated through AHP is verified using measured consistency ratio (CR). CR below 0.1, the model is consistent and used for subsequent processes.
These weights along with the factors of growth are combined along with the constraints to obtain site suitability maps for different land uses using equation below
LC =
∗ ∑ ∗
……. (1)
Where LC is the linear combination of weights, n is the number of factors, D decision factor, W is the weight of the factor.
The Markov chains are used to determine the change probability between two historical datasets to derive the growth in the future scenarios based on different criteria’s. The Markovian transition matrix indicates the probability of the particular land use being converted to other land uses on single time step.
Criteria Factors and Constraints Without
CDP as a
constraint
Slope, Distance from roads, Distance to
industries, Distance to Bus stops and Railway
stations, Distance from metro, Distance from
educational institutions, Population Density
With CDP Slope, Distance from roads, Distance to
industries, Distance to Bus stops and Railway
stations, Distance from metro, Distance from
educational institutions, City Development
Plan, Population Density
Table 1: Criteria’s for simulating and predicting urban sprawl
The cellular automata based on the site suitability and the transition matrix is used to spatially predict the changes in land use based on current land use at every
3 ISTC/BES/TVR/313
single time step, based on the neighbouring pixels. Two scenarios were designed to predict the land use changes as shown in table 1.
Validation of the simulated datasets of were performed with classified datasets through kappa indices, as a measure of agreement. Once these data and agents are trained and validated, data is used to model and simulate for the year 2030 (ten years) with definite time steps.
IV. RESULTS
Geo-visualisation of urbanisation of five tier I cities are depicted in fig.1 to fig.5 and results are provided in tables 2 to 6 respectively. The cities on an average would grow by 1.5 to over 2 times the current state in next decade. By 2025, it is predicted that built up area in these cities and surroundings, grows over 57% (Delhi), 27% (Mumbai), 45.8% (Chennai), 50% (Pune) and 37% (Coimbatore) respectively. The various drivers of growth for different cities are as in annexure 1. In all these cases, spatially it could be understood that the CDP if implemented properly would play a major role in curtailing the unsustainable growth of the city in its limits, while some growth still takes place at the outskirts. Prime factors of growth include the transportation network, industrialisation, and educational sector.
Modelling growth of Delhi
Predicted 2017 Predicted 2024
Predicted 2031 Fig.1: Predicted landscape dynamics of Delhi
Year Built up Vegetation Water Others
2017 45.80 17.98 1.25 34.97
2024 57.37 8.77 1.18 32.68
2031 70.86 3.76 1.19 24.18 All units as percentage area
Table 2: Predicted landscape dynamics of Delhi
Modelling growth of Mumbai
Predicted 2020 Predicted 2031
Fig.2: Predicted landscape dynamics of Mumbai
Year Built up Vegetation Water Others
2020 25.83 9.09 44.52 20.56
2031 31.27 6.33 44.52 17.88 All units as percentage area
Table 3: Predicted landscape dynamics of Mumbai
Modelling growth of Chennai
Predicted 2026
Fig.3: Predicted landscape dynamics of Chennai
Year Built up Vegetation Water Others
2026 45.80 17.98 1.25 34.97 All units as percentage area
Table 4: Predicted landscape dynamics of Chennai
4 ISTC/BES/TVR/313
Modelling growth of Pune
Predicted 2016 Predicted 2019
Predicted 2022 Predicted 2025
Fig.4: Predicted landscape dynamics of Pune
Year Built up Water Vegetation Others
2016 37.78 1.75 16.37 44.11
2019 41.64 1.75 20.16 36.45
2022 47.89 1.75 20.16 30.20
2025 50.02 1.75 20.16 28.06 All units as percentage area
Table 5: Predicted landscape dynamics of Pune
Modelling growth of Coimbatore
Predicted 2023 Predicted 2033
Fig.5: Predicted landscape dynamics of Coimbatore
Year Built up Water Vegetation Others
2023 32.64 0.29 17.14 49.94
2033 42.92 0.29 17.58 39.21 All units as percentage area
Table 6: Predicted landscape dynamics of Coimbatore
V. CONCLUSION
This study demonstrates the application of temporal remote sensing data and Geoinformatics in mapping and understanding of urban dynamics. Advance geo-visualisation of urban growth would aid in decision making towards sustainable cities with basic infrastructure and amenities. Identification of regional factors that are most likely to influence a land-use changes has improved the accuracy of prediction. The predictions of land use/cover changes through CA-Markov model suggest a continual increase in urban settlements with a decline in local ecology and natural vegetation covers. The prediction using CA help to design sustainable urban transportation system.
VI. ACKNOWLEDGEMENT
We are grateful to ISRO-IISc Space Technology Cell, Indian Institute of Science and NRDMS division, Ministry of Science and Technology, DST, Government of India for the financial support. We thank USGS Earth Resources Observation and Science (EROS) Center for providing the environmental layers and Global Land Cover Facility (GLCF) for providing Landsat data.
VII. Publications
1. Ramachandra, T. V., Bharath, H. A. and Sowmyashree, M.V., 2014. Urban structure in Kolkata: metrics and modelling through geo-informatics, Appl Geomat 6(22): DOI 10.1007/s12518-014-0135-y
2. Ramachandra, T. V., Bharath, H. A. and Sowmyashree, M.V., 2014. Monitoring spatial dynamics of urbanisation in Ahmed city, Textile hub of India, SPATIUM International review, 31(2):85-91
3. Ramachandra, T. V., Bharath S., Bharath, H. A. 2014. Spatio-temporal dynamics along the terrain gradient of diverse landscape, Journal of Environmental Engineering and Landscape Management, 22(1): 50–63 doi:10.3846/16486897.2014.808639.
4. Ramachandra, T.V., Bharath H. Aithal and Sowmyashree M V., 2014. Monitoring urbanization and its implications in a mega city from space: Spatiotemporal patterns and its indicators, Journal of Environmental Management http://dx.doi.org/10.1016/j.jenvman.2014.02.015
5. Ramachandra, T. V., Bharath, H. A. and Barik, B., 2014. Urbanisation Pattern of Incipient Mega Region in India, TEMA - Journal of Land Use, Mobility and Environment, 7(1): 83-100.
6. Ramachandra, T. V., Bharath, H. A. and Sowmyashree, M.V., 2014. Urban footprint of Mumbai - the commercial capital of India, Journal of Urban and Regional Analysis, VI (1): p. 71 – 94
7. Ramachandra, T. V., Bharath, H. A., 2014. Geo-visualisation of Urbanisation in Greater Bangalore, India Geospatial Digest, September 2014, pp 12-18 http://www.geospatialworld.net/Paper/Application/ArticleView.aspx?aid=31130#sthash.NZScwjHH.dpuf
5 ISTC/BES/TVR/313
8. Ramachandra, T. V., Bharath, H. A. and Sowmyashree, M.V., 2013. Analysis of spatial patterns of urbanisation using geo-informatics and spatial metrics, Theoretical and Empirical Research in Urban Management, 8(4):5-24
9. Ramachandra, T.V., Bharath, H. A., and Vinay S, Land Use Land Cover Dynamics in a Rapidly Urbanising Landscape, SCIT Journal, Volume XIII, August 2013, XIII(1):1-13
10. Bharath, S., Rajan, K.S., Ramachandra, T.V., 2013. Land Surface Temperature Responses to Land Use Land Cover Dynamics. Journal of Geoinformatics and Geostatistics: An Overview 1(4):1-10. doi:10.4172/2327-4581.1000112.
11. Ramachandra, T.V., Meera, D.S., and Alakananda B., 2013. Influence of Catchment Land Cover Dynamics on the Physical, Chemical and Biological Integrity of Wetlands, Environment & We -International Journal of Science & Technology - (EWIJST), pp 8(1): 37-54
12. Ramachandra T. V., Sreejith K. and Bharath H. A., 2014. Sector-Wise Assessment of Carbon Footprint across Major Cities in India, Assessment of Carbon Footprint in Different Industrial Sectors, Volume 2, Muthu S. S. (ed.), Eco-Production, DOI: 10.1007/978-981-4585-75-0_8, Springer Science Business Media Singapore, 2014.
13. Ramachandra, T. V., Shwetmala, K., Dania, T.M., 2014. Carbon footprint of solid waste sector in Greater Bangalore, India, In: Assessment of Carbon Footprint in Different Industrial Sectors (ed. S.S. Muthu), Volume 1, Eco-Production, DOI: 10.1007/978-981-4560-41-2_11, Springer Science, Singapore Pp 265-292
VIII. References 1. Campbell, J. B., 2002. Introduction to Remote
Sensing (3rd Edn). Taylor & Francis, London, UK, 605.
2. Lillesand, T. M., Kiefer, R. W., 2002. Remote Sensing and Image Interpretation, 4th ed. John Wiley and Sons, 215–216.
3. Ramachandra, T.V., Bharath, H.A., Barik, B., 2014c. Urbanisation Pattern of Incipient Mega Region in India, Tema. Journal of Land Use, Mobility and Environment, 7(1), 83-100.
4. Ramachandra, T.V., Bharath, H. A., Sowmyashree, M. V., 2014d. Urban Footprint of Mumbai - The Commercial Capital of India, Journal of Urban and Regional Analysis, Vol. 6(1), 71-94.
5. Ramachandra, T.V., Bharath H. Aithal and Durgappa D. Sanna, 2012. Insights to Urban Dynamics through Landscape Spatial Pattern Analysis., International Journal of Applied Earth Observation and Geoinformation, Vol. 18, Pp. 329-343
6. Ramachandra, T.V., Bharath, H.A., Sowmyashree M. V., 2014a. Monitoring urbanization and its implications in a mega city from space: Spatiotemporal patterns and its indicators, Journal of Environmental Management, accepted, in press, doi:10.1016/j.jenvman.2014.02.015
7. Ramachandra, T.V., Bharath, H.A., Sowmyashree M. V., 2014b. Urban Structure in Kolkata: Metrics and
Modeling through Geo-informatics, Applied Geomatics, accepted, in press.
8. Chandan, M.C., Bharath, H.A., Ramachandra, T.V., 2014. Quantifying urbanisation using geospatial data and spatial metrics-a case study of madras, In proceedings of Lake 2014: Conference on Conservation and Sustainable Management of Wetland Ecosystems in Western Ghats, 13th -15th Nov., Uttara Kannada, India
9. Bharath, H. A., Vinay, S., Ramachandra, T.V., 2014. Landscape dynamics modelling through integrated Markov, Fuzzy-AHP and Cellular Automata, in the proceeding of International Geoscience and Remote Sensing Symposium (IEEE IGARSS 2014), July 13th – July 19th 2014, Quebec City convention centre, Quebec, Canada. Available at http://www.igarss2014.org/Papers/ViewPapers_MS.asp?PaperNum=1010
10. Bharath, H.A., Bharath, S., Sannadurgappa, D., Ramachandra, T. V., 2012, Effectiveness of landscape Spatial Metrics with reference to the Spatial Resolutions of Remote Sensing Data, Proceedings of India Conference on Geo-spatial Technologies & Applications 2012, IIT Bombay, Mumbai, India, 12-14 April, 2012.
11. Bharath, H.A., Vinay, S., Ramachandra, T.V., 2013. Prediction of Land use dynamics in the rapidly urbanising landscape using land change modeller In proceedings of Fourth International Joint Conference on Advances in Engineering and Technology, AET 2013, December 13-14, NCR Delhi, India.
12. MOUD. 2011.Urban development management for the formulation of the twelfth five year plan (2012–2017): Report of the working group on capacity building for the twelfth plan. New Delhi
13. Sivaramakrishnan, K.C., Kundu, A., Singh, B.N. 2005. Handbook of Urbanization in India, Oxford University Press, New Delhi, India.
14. Ramachandra, T.V., Bharath, H. A., Vinay, S., Joshi, N. V., Kumar, U., Venugopal, R. K., 2013. Modelling Urban Revolution in Greater Bangalore, India , 30th Annual In-House Symposium on Space Science and Technology, ISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 7-8 November 2013.
15. Sudhira H. S., Ramachandra T. V., Subrahmanya, M. H., Bala, 2007. Urban Sprawl Management: Need for an Integrated Spatial Planning Support System, Technical Report: 119, Centre for Ecological science, Indian Institute of science, Bangalore.
16. Sudhira, H.S., Ramachandra, T.V., Raj, K.S., Jagadish, K.S., 2003. Urban Growth Analysis using Spatial and Temporal Data, Photonirvachak, Journal of Indian Society of Remote Sensing, 31(4), 299-311.
17. Batty, M., Xie, Y., Sun, Z., 1999. Modeling urban dynamics through GIS-based cellular automata. Computers, Environment and Urban Systems, 23(3), 205–233
18. Torrens, P. M., 2000. How cellular models of urban systems work, Centre for Advanced Spatial Analysis, University College London, Working Paper Series, Paper 28.
6 ISTC/BES/TVR/313
19. Bhatta, B., 2009a. Analysis of urban growth pattern using remote sensing and GIS: a case study of Kolkata, India. International Journal of Remote Sensing, 30(18):4733–4746.
20. Bhatta, B., 2009b. Modeling of urban growth boundary using geoinformatics. International Journal of Digital Earth, 2(4):359–381.
21. Bhatta, B., 2010. Analysis of urban growth and sprawl from remote sensing data. Berlin, Heidelberg: Springer-Verlag.
22. Sudhira, H.S., Ramachandra, T.V. and Jagadish, K.S., 2004. Urban sprawl: metrics, dynamics and modelling using GIS”, International Journal of Applied Earth Observation and Geoinformation, 5(1):29-39.
23. Allen, J., Lu, K., 2003. Modeling and prediction of future urban growth in the Charleston region of South Carolina: A GIS-based integrated approach. Ecology and Society, 8(2), 2
24. EPA, 2000. Projecting land-use change: A summary of models for assessing the effects of community growth and change on land-use patterns. EPA/600/R-00/098, Office of Research and Development, Washington DC, USA.
25. Alberti, M., Waddell, P., 2000. An integrated urban development and ecological simulation model. Integr. Assess. 1(3), 213-227.
26. Arsanjani, J. J., Helbich, M., Kainz, W., Darvishi Boloorani, A., 2013. Integration of logistic regression, Markov chain and cellular automata models to simulate urban expansion. Int. J. of Appl Earth Observation and Geo., 21, 265–275.
27. Clarke, K. C., Hoppen, S., Gaydos, L., 1997. A self-modifying cellular automation model of historical urbanization in the San Francisco Bay area. Environment and Planning B: Planning and Design, 24(2), 247–261
28. Leao, S., Bishop, I., and Evans, D., 2004. Simulating urban growth in a developing nation’s region using a cellular automata-based model. J. Urban Plann. Dev., 3(145), 145–158.
29. White, R., Engelen, G., 2000. High-resolution integrated modelling of the spatial dynamics of urban and regional systems. Comput. Environ. Urban Syst., 24(5), 383–400.
30. Adhvaryu, B., 2011. The Ahmedabad urban development plan-making process: A critical review. Planning Practice and Research, 26(2), 229–250
Annexure
Road DEM CPD
Annexure 1: Factors and constraints of growth for Delhi
Road DEM CPD
Annexure 2: Factors and constraints of growth for Mumbai
7 ISTC/BES/TVR/313
Road and railway DEM CPD
Annexure 3: Factors and constraints of growth for Chennai
Road DEM CPD
Annexure 4: Factors and constraints of growth for Pune
Road DEM CPD Annexure 5: Factors and constraints of growth for Coimbatore
Quantum communication and security usingentangled states
Ankur Raina and Shayan G. SrinivasaDept. of Electronics Systems Engineering
Indian Institute of Science, Bengaluru, IndiaEmail correspondence: [email protected]
Chithrabhanu P., Aadhi A., Shashi P., R. P. SinghDivision of Quantum Optics
Physical Research Labs, Ahmedabad, IndiaEmail correspondence: [email protected]
Abstract—This research report has two main parts relatedto the theoretical and experimental investigations into quantumcommunications and security. In the first part, we consider superdense coding over the noisy bit flip channel with a shared Bell pairand show that the encoding scheme is perfectly secure using theinformation-theoretic criterion. We also extend the super densecoding idea to the tripartite system. In the second part, using athree particle hyper entangled state, we set up an experimentalplatform for illustrating the idea of multiparty quantum keydistribution protocol using photons. Our results are useful forrealizing the fundamental limits of transmission and informationsecurity over quantum channels with an eye towards experimentalrealization of the same via photonics.
I. INTRODUCTION
Super dense coding [1], a well known quantum protocolwas one of the first few pioneering discoveries in quantuminformation processing, that exploits the advantage of quantumentanglement. Ever since the discovery, quantum entanglementis considered as a resource in quantum information processing.It lets users pre-sharing an entangled pair of qubits, commu-nicate two classical bits. The protocol as given by Bennettand Wiesner is as follows - An entangled pair of qubits, alsocalled the Bell pair, is prepared and distributed to two nodes.One of the nodes performs unitary encoding and transfersits qubit to the other node via noiseless quantum channel.The receiving node performs measurement and decodes theinformation conveyed. However, there is one caveat. Thequantum channel transferring the qubit from the source to thedestination is assumed to be noiseless. In practice, no channel,classical or quantum is physically noiseless.
Quantum communication using entanglement is increas-ingly receiving attention and is an actively pursued topic ofcurrent research. In this work, we consider the super densecoding over the noisy bit flip channel. The quantum advantageof using entanglement is brought out. We also extend the ideaof super dense coding to a tripartite system initially in the GHZstate. We assume that the GHZ state is somehow prepared anddistributed to three parties namely A, B and C. We address thequestion whether 3 classical bits can be communicated usingsuch a tripartite state.
With entanglement between two quantum bits, protocolshave been demonstrated for teleporting an unknown quan-tum state [2], super dense coding of information [1] and
This is an ongoing collaboration between IISc and PRL funded byISTC/EED/SGS/311. The first part of the work is from IISc and the secondpart is from PRL.
secure communication [3]. We describe a hyper-entangled stateformed by orbital angular momentum (OAM) and polarizationdegree of freedom and use it for multiparty quantum key dis-tribution (QKD) protocol. The scheme is being experimentallyinvestigated via photonics.
The report is organized as follows. In sections 2 and 3,we describe super dense coding over the bit flip channel forbipartitite and tripartite states respectively. We also presentinformation security results over noisy quantum channels. Insection 4, we describe the construction of hyper entangledstates. We describe the experimental set up for state preparationin section 5 followed by the description of the QKD protocolin section 6 and conclusions.
II. SUPER DENSE CODING OF BELL STATES THROUGH ABIT FLIP CHANNEL
We present the protocol for communication between nodesA and B through a quantum bit flip channel [4]. The action ofthe quantum bit flip channel with parameter p on any state Φis given by [5]
N (Φ) = pσXΦσX + (1− p)Φ. (1)
We assume that a fully entangled bipartite state namely theBell pair φ+ is shared between nodes A and B.To encode classical information, A encodes its qubit usingPauli matrices σi, i ∈ 0, 1, 2, 3 [5]. A chooses I, σ1, σ2, σ3to respectively communicate two bits 00, 01, 10, and 11.Based on what A chooses, we get the following Bell pairsshared between A and B.
φ± =00± 11√
2, ψ± =
01± 10√2
.
The encoding operation by A is described using densitymatrices as follows:
ρi = (σi ⊗ I) ρ (σi ⊗ I)† , (2)
where ρ = φ+φ+, the initially shared Bell pair.
A. Calculation of Holevo capacity for a bit flip channel
We know from [6], that capacity of a classical channelis obtained by maximizing the mutual information betweencommunicating nodes, over all input distributions. When itcomes to the quantum scenario, the situation is very interesing.There exist notions of classical capacity and quantum capacity.Morever, there is an additional resource which can be used,namely entanglement. In this way four types of capacities have
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
been defined [7]. Holevo [8], Hausladen et al. [9], gave theexpression for classical capacity of a quantum channel usingquantum states. We are interested in the classical capacity of abit flip channel with entanglement shared between nodes. Wehave what is known as the HSW theorem, which states theclassical capacity, also called the Holevo capacity of a quantumchannel N , using the ensemble pi, ρi at the encoder, is givenby,
χ(pi, ρi) = S
((N ⊗ I)
(∑i
piρi))
−∑i
piS
((N ⊗ I)(ρi)
), (3)
where S denotes the Von Neumann entropy. It can be shownthat the uniform distribution also maximizes the Holevo capac-ity for the noisy quantum bit flip channel. The Holevo capacityevaluates to the following expression:
χ = 2− h(p). (4)
Observing (4), we find it is very akin to the capacity of a clas-sical binary symmetric channel with a cross over probabilityof p. When the channel is most random (p = 0.5), we get acapacity of zero for the classical channel but a capacity of 1 bitfor the quantum channel. There seems to be a clear advantagein using entangled quantum states for communication!
B. Connections to perfect security
Classically, Shannon [10] defined informationally secureencryption for message M and encrypted message E as
I(M ;E) = H(E)−H(E|M) = 0. (5)
That is, message should be statistically independent of theencrypted message for the encryption to be perfectly se-cure. The necessary condition for perfect secrecy implies thatH(K) ≥ H(M), where K represents the random variable forthe key. Boykin and Roychowdhury [11] provide conditionsfor perfect secrecy of qubits. In that work, condition for secureencryption of n qubits is formalized. The protocol is secureif for every message state ρ, the encrypted state, ρc is themaximally mixed state, i.e.,∑
k
pk Uk ρU†k =
I2n2n
. (6)
Note that the encoding by A gives
3∑i=0
pi(σi ⊗ I)ρ(σi ⊗ I)† =3∑i=0
piρi =I44.
Thus, this scheme leads to a perfectly secure encryption [11].It can be shown that the encrypted state does not lose itsmixedness even after it goes through a noisy bit flip channel.Infact, it is the unitality of the quantum channels that keepsthe perfect secrecy intact.
III. SUPER DENSE CODING USING TRIPARTITE STATES
We now explore tripartite system with prior entanglement.This is a natural extension from the bipartite case and helpsus take a peep into the world of higher dimensional Hilbertspaces. Considerable amount of research has been done onthe GHZ state [12], [13] and its preparation [14], [15]. Goingforward, one can think of having a network N nodes withprior entanglement [16] among more than two nodes. With thisinsight in mind, various communication scenarios like pointto point communication, multicast communication, broadcastcommunication gain prominence in the quantum world. Thelinks between various nodes in a physical setting would benoisy and information carrying capacities of such links wouldbe of immediate interest. The GHZ state is defined as
ψABC =000 + 111√
2.
A. Protocol for the noiseless case
N1
N2
A
B C
Fig. 1. Super dense coding over tripartite state.
Referring to Figure 1, A and B together encode theirqubits and send them over noiseless quantum channels N1 andN2 respectively. Upon receiving the two qubits, C performsmeasurement in an appropriate basis. We show that it ispossible to send 3 classical bits by joint encoding by A andB. A chooses one of the two unitary operations namely I, σ1and B chooses one of the four unitary operations namelyI, σ1, σ2, σ3 to encode their respective qubits. Without lossof generality, let A encode first followed by B. The encodingcan be represented mathematically by
ρ(A)i = (σi ⊗ I ⊗ I) ρABC (σi ⊗ I ⊗ I)†. (7)
ρi = ρ(B)i = (I ⊗ σi ⊗ I)ρ
(A)i (I ⊗ σi ⊗ I)†. (8)
Note that (7) represents the joint state after encoding byA and (8) represents the joint state after encoding by B.Depending upon the pair of encodings by A and B, thejoint state ends up being one of the eight orthonormal states.Hence, measurement in this basis gives one of the aboveeight possible outcomes. This brings us to summarizing ourprotocol as:
If A wants to communicate 1 bit and B wants to communicate2 bits then,1. A chooses either I or σ1 to respectively transmit 0 or 1.2. B chooses among I ,σ1, σ2, σ3 to respectively transmit00, 01, 10, 11.3. Both A and B send their qubits to C, which performsmeasurement in the appropriate basis.4. Based on the measurement outcome, C unambiguouslyrecovers three bits of classical data.5. The roles of A and B can be interchanged if A wants tocommunicate 2 bits and B wants to communicate 1 bit.
IV. HYPER ENTANGLED QUANTUM STATES
We consider a hyper-entangled state formed by OAMand polarization degree of freedom and use it for multipartyQKD protocol. In this protocol, one person can distribute twoindependent quantum keys with two different parties at thesame time.
Consider a system of particles in such a way that oneparticle is entangled to all other particles in different degreesof freedom (DOF). Let us consider a system consisting of threephotons as described in Fig. 2. Here, photon 2 is entangled withphotons 1 and 3 in different degrees of freedom namely OAMand polarization respectively. The polarization state of photon1 and the OAM state of photon 3 are arbitrary or unknown.
1 2 3
Entangled in OAM Entangled in Polarization
Unknown Polarization Unknown OAM
Fig. 2. The physical description of the defined quantum state.
Since OAM of a photon can have infinite values, one canhave higher dimensional entangled states. We take an arbitrarytwo dimensional subspace of the infinite dimensional OAMbasis as |l〉, |l′〉.
The described state can be prepared using a pair ofHadamard and CNOT gates in different DOFs which corre-spond to polarization and OAM. The initial states of three
1
2
3
123
Polarization
OAM
Fig. 3. Schematic diagram for preparation of the initial state.
particles can be written as
|1〉 = |ξp〉1|l〉1 ; |2〉 = |H〉2|l〉2 ; |3〉 = |H〉3|χo〉3 (9)
|ξp〉 is the unknown state of polarization of photon 1, while|χo〉 is the unknown OAM state of photon 3. |H〉 representshorizontal and |V 〉 represents vertical polarization states of aphoton. The schematic diagram is given in Fig. 3. To entanglephotons 1 and 2 in OAM, a Hadamard gate (Ho) on photon1 and subsequent CNOT gate on photon 1 and 2 are applied.Both gates act in OAM degree of freedom. Similarly, anotherHadamard (Hp) acting on photon 2 and a CNOT betweenphotons 2 and 3 entangle them in polarization. Here bothgates are acting on the polarization states of photons. Afterperforming the gate operations mentioned above, the final statebecomes,
|Ψ〉123 = |ξp〉1 ⊗ (|l, l′〉12 + |l′, l〉12)⊗ (|HV 〉23 + |V H〉23)|χo〉3.(10)
This state corresponds to a description given in Fig 2.
V. EXPERIMENTAL SCHEME FOR STATE PREPARATION
We are proposing a method in which initially two photonsare entangled in OAM and another photon is in an independentpure state of OAM and polarization. By polarization gateoperations in one of the entangled photons and the independentphoton, one can arrive at the described state. Experimentalscheme for generation of the state is given in Fig. 4.
BBO
SM
Pol
C NotHWP
1 V
! " Hoe # $%
123&
1
3
2
Fig. 4. Schematic experimental set up for the state preparation. SM- Simon-Mukunda gadget, HWP- half wave plate, BBO- second order nonlinear crystal(Beta Barium Borate) .
To generate the described state, we start with a Type Ispontaneous parametric down conversion of light in a secondorder nonlinear crystal that results a pair of photons entangledin OAM [17]. A vertically polarized optical vortex beam ofazimuthal index 1 has been considered as pump. The statecorresponding to this pair of photons is given by
|Ψ〉12 =+∞∑
m=−∞cm|m〉1|1−m〉2 ⊗ |H〉1|H〉2 (11)
For the ease of experimental realization, we reduce theinfinite dimensional entangled state to a simple two-qubitentangled state by grouping all even and odd OAM states andrewrite the expression for the OAM state in Eq. 11 as
+∞∑m=−∞
cm(|m〉1|1−m〉2) =+∞∑
k=−∞
c2k(|2k〉1|1− 2k〉2) +
+∞∑k=−∞
c2k+1(|2k + 1〉1| − 2k〉2).
(12)
We define a transformation from the general OAM spaceto the even-odd OAM space as
g(|u〉1|v〉2) = f(|u〉1)f(|v〉2), where
f(|x〉) =
|E〉, for even x;
|O〉, for odd x.
Note that g is nonlinear, but f is linear. Applying g on theleft-hand-side of Eq. 12 yields 1√
2(|E〉1|O〉2 + |O〉1|E〉2).
From the conservation of OAM, we have∑+∞k=−∞(c2k)2 =∑+∞
k=−∞(c2k+1)2 = 12
∑+∞m=−∞(cm)2 = 1
2 . Thus, we canrewrite Eq. 11 as
|Ψ〉12 =1√2
(|E〉1|O〉2 + |O〉1|E〉2)⊗ |H〉1|H〉2. (13)
Note that the photons are entangled in even/odd OAMstates. Let photon 1 pass through a Simon-Mukunda polarizinggadget which can convert its polarization to any arbitrary state[18], [19]. Photon 2 pass through a half wave plate at π8 whichperforms a Hadamard operation. Thus polarization state ofparticle 1 is encoded as the unknown state a|H〉1 + b|V 〉1and state of photon 2 is encoded as 1√
2(|H〉2 + |V 〉2). Action
of HWP on photon 2 is a Hadamard operation. Thus, the statebecomes
|Ψ〉12 =1√2
(|E〉1|O〉2 + |O〉1|E〉2) (a|H〉1 + b|V 〉1)⊗1√2
(|H〉2 + |V 〉2). (14)
Now, consider photon 3 with unknown superposition state inOAM and with definite polarization. Its state can be expressedas,
|Ψ〉3 = (α|q〉3 + β|1− q〉3)⊗ |V 〉3 (15)
where q is an integer. A polarization CNOT gate [20] is appliedon photon 2 (control) and 3 (target). This operation leads topolarization entanglement between photon 2 and 3. Thus, thethree particle state will become
|Ψ〉123 =1
2(a|H〉1 + b|V 〉1) (|E〉1|O〉2 + |O〉1|E〉2)
(|HV 〉23 + |V H〉23)(α|q〉3 + β|1− q〉3)(16)
This is the proposed three particle entangled state given in Eq.10
VI. QUANTUM KEY DISTRIBUTION
Quantum key distribution is essential for secure communi-cation to share a confidential message between two indentedparties, namely Alice and Bob. A QKD protocol can beimplemented by a single photon source or entangled photonsource [21], [22]. Entangled photons with one to one photoncorrelations are better source for QKD than a single photonsource. This strong correlation helps to share the key and getthe information about eavesdropping [23]. QKD is perfectlysecure, subject to all kinds of attacking strategies [24]. Thereare various protocols for QKD with the utilization of entangledphotons in different degrees of freedom such as polarization,OAM, and time-bin [25], [26], [27], [28]. Entanglement basedprotocols have been studied and implemented since long be-cause of its interesting properties such as strong correlation andnon-locality. Multipartite entanglement increases the codingcapacity and security of the key. It also provides multipartycommunication through the same channel [29].
Here we are proposing a multiparty QKD protocol usingthe entangled system described in section IV. We take a similarstate as in Eq. 10 given by
|Ψ〉123 = |ξp〉1 ⊗ (|00〉12 + |1− 1〉12 + | − 11〉12)⊗(|HH〉23 + |V V 〉23)|χo〉3, (17)
where |0〉, |1〉 and | − 1〉 are OAM states of photons.
Here we take a three dimensional subspace of infinitedimensional OAM entangled state produced by SPDC processwith Gaussian beam as pump instead of even/odd entangledstate. This state can be used for multiparty QKD. The advan-tage of this state is that it can generate two sets of independentkeys, one for Alice-Bob and another for Alice-Charlie. Inordinary three particle QKD, all the three parties will share thesame key and one can not get the key without the help of other[30]. So if Alice wants to pass some information to Bob, notto Charlie, this protocol is no longer useful. We have resolvedthis problem with the use of a quantum state mentioned in Eq.17.
Source+1
-1
0
D1
D2
D3D10
D11 BobCharly
Hologram
D9
D8
D4 D5 D6
+1
+1-1 0
0
-1 D7
Alice
Polarizing
Beam splitter
OAM sorter
Detector
13
2
HWP
Fig. 5. Schematic diagram for quantum key distribution between Alice-Bob and Alice-Charlie. Alice- Bob has Polarization entanglement and Alice-Charlie has OAM entanglement. Di - detectors, HWP - half wave plate, PBS- polarizing beam splitter.
The experimental implementation of the protocol is shownin Fig. 5. Bob, Alice and Charlie will receive the particles1, 2 and 3 respectively. Alice’s photon is entangled withBob’s photon in polarization and it is entangled with Charlie’sphoton in OAM. Bob’s and Charlie’s photons do not have anycorrelation between them. Alice will measure both OAM andPolarization states, Bob will measure polarization and Charliewill measure OAM of their received photons.
Alice and Bob share their keys by polarization measure-ments. They randomly measure polarization state of theirphotons in the following set of basis (αi = 0, 22.5, 45,67.5), (βi = 22.5, 45, 67.5, 180) respectively. Thesemeasurements can be done by a half wave plate and polarizingbeam splitter. The pairs of angles used by Alice and Bob forwhich the sum is 0 and 180 will give perfect correlationbetween them. The data corresponding to these correlatedphotons have same bits and can be used for the key. Alice
and Bob will compare their polarization measurement basisfor key distribution as well as for security of the key. Aftersufficient number of measurements 4/16 of the data are usefulfor key, two sets of 4/16 of the data are used to check CHSHinequalities (S and S′) and the remaining 4/16 are discarded.
Alice and Charlie share their keys by OAM correlationof entangled photons. Here we follow the key sharing schemeused in Ref [31]. In this scheme Alice and Charlie measure theOAM state of their photons in three different randomly chosenbases A1, A2, A3 and C1, C2, C3 respectively. This is doneby a pair of shifted holograms and an OAM sorter. The set ofbases (A3 and C3) has perfect correlation and the coincidencecorresponding to it can be used to generate the key. Alice andCharlie will compare their hologram settings for QKD andsecurity check. In total there are 9 possible measurements.After enough number of measurements, 1/9 of the produceddata can be used for the key. For Bell-type inequality check,4/9 of the data will be used which confirms the security of thekey and the remaining 4/9 of the data are useless.
If Ekert protocol [3] is used in a pairwise fashion, thenaround 8n pairs of photons (in practice, little more than 8n) arerequired to establish a pair of secret keys of length n betweenAlice and Bob and between Alice and Charlie. In our approach,around 6n pairs of photons would be required for the samepurpose. Hence, in terms of number of photons required, ourscheme is 33% more efficient than Ekert protocol.
On the other hand, if the same number of photons are usedin Ekert and our protocol and if the target key length is alsothe same, then because of more entanglement resource, ourprotocol would have more redundancy and hence can toleratemore noise. It is easy to see that the security in each degreeof freedom is equivalent to that of the Ekert protocol.
The security of the protocol is mainly checked by theviolation of Bell’s inequality test. All the three parties shouldcheck Bell like inequality with their data in order to checkeavesdropping. If there is violation of inequality the entan-glement is preserved and there is no eavesdropping in thechannel. The CHSH parameters S and S′ for photons entangledin polarization DOF are given by [32]
S = E(α1, β1)− E(α1, β3) + E(α3, β1) + E(α3, β3) (18a)
S′ = E(α2, β2) +E(α2, β4) +E(α4, β2)−E(α4, β4) (18b)
with
E(α, β) =R12(α, β) +R1′2′(α, β)−R12′(α, β)−R1′2(α, β)
R12(α, β) +R1′2′(α, β) +R12′(α, β) +R1′2(α, β)(19)
where R12, R1′2′ , R12′ and R1′2 are the coincidencesP (D11, D9 + D8 + D7), P (D10, D4 + D5 +D6), P (D11, D4 +D5 +D6) and P (D10, D9 +D8 +D7)respectively.
For any local realistic theory, the CHSH parameters S, S′ ≤2. Non-local nature of the entanglement will violate any ofthese inequalities and this violation can be used for checkingthe security of the key shared by Alice and Bob. The combi-nation of α and β for Bell’s inequality test and for the keydistribution is given in Table I.
Alice α1 α2 α3 α4
Bobβ1 S Key S ×β2 × S′ Key S′
β3 S × S Keyβ4 Key S′ × S′
TABLE I. DESCRIPTION OF DATA USAGE CORRESPONDING TOALICE’S AND BOB’S MEASUREMENT ANGLES. HERE S AND S′ ARE USED
FOR SECURITY CHECK THROUGH CHSH INEQUALITY AND × IS THEDISCARDED DATA.
Particles of Alice and Charlie have OAM correlation, so itwill violate the following Bell’s inequality for 3 dimensionalcase [31], [33]
S =P (A1 = C1) + P (A2 = C1 − 1) + P (A2 = C2) + P (A1 = C2)
− P (A1 = C1 − 1)− P (A2 = C1) + P (A2 = C2 − 1)+
P (A1 = C2 + 1) ≤ 2, where (20)
P (Aa = Cb+k) =2∑j=0
P (Aa = j, Cb = (j+k)Mod 3). (21)
The shifts of holograms are chosen in such a way thatthey maximally violate the Bell-type inequality. j=0,1,2 corre-sponds to the detection of OAM states 0, 1 and -1 respectively.The coincidence measurements with the combinations of P(D1,D6+D9), P(D1, D5+D8), P(D2, D4+D7), P(D2, D6+D9),P(D2, D5+D8), P(D2, D4+D7), P(D3, D6+D9), P(D3, D5+D8)and P(D3, D4+D7) are required for the key distribution and tocheck the security of the key using Eq. 20 and 21.
VII. CONCLUSION AND FUTURE WORK
We evaluated the Holevo capacity of the bit flip channelexisting between two entanglement sharing nodes. We alsopoint out that the encoding scheme ensures perfect secrecyunaffected by the noisy bit flip channel. We extended theideas of super dense coding to three nodes with a briefpractical motivation for the generalization. Theoretical under-standing into the limits of transmission capacity and securityare important for quantum communication channels. From anexperimental angle, we looked into the possibility of a newentangled state and its application in multiparty QKD. Withthe new QKD protocol, a common sender can share differentand independent keys with many parties. Our eventual goal isto unify the theoretical investigations within an experimentalframework via novel quantum coding schemes we are buildingto demonstrate the feasibility of a quantum communicationsystem.
REFERENCES
[1] C. Bennett and S. J. Wiesner, “Communication via one- and two-particleoperators on Einstein-Podolsky-Rosen states,” Phys. Rev. Lett. vol. 69,pp. 2881-2884, 1992.
[2] C. H. Bennett, G. Brassard, C. Crepeau, R. Jozsa, A. Peres, and W. K.Wootters. Teleporting an unknown quantum state via dual classical andeinstein-podolsky-rosen channels. Phys. Rev. Lett., 70:1895, 1993.
[3] A. K. Ekert. Quantum cryptography based on bell’s theorem. Phys. Rev.Lett., 67:661–663, 1991.
[4] A. Raina and S. G. Srinivasa, “Quantum communication over bit flipchannels using entangled bipartite and tripartite states,” IEEE. Allerton.Conf. Control. Computers and Comm. , Oct. 2014.
[5] M. A. Nielsen and I. L. Chuang, Quantum computation and QuantumInformation, Cambridge University Press, 2010.
[6] T. M. Cover and J. A. Thomas, Elements of Information Theory, Wiley,New York, 2006
[7] C. H. Bennett, P. W. Shor, J. A. Smolin and A. V. Thapliyal, “Entanglement-Assisted Capacity of a Quantum Channel and the ReverseShannon Theorem,” IEEE Trans. Inf. Theory, vol. 48, pp. 2637-2655,2002.
[8] A. S. Holevo, “The capacity of the quantum channel with general signalstates,” IEEE Trans. Info. Theory., vol. 44, pp. 269-273, 1998.
[9] P. Hausladen, R. Jozsa, B. Schumacher, M. D. Westmoreland and W. K.Wooters, “Classical information capacity of a quantum channel,” Phys.Rev. A, vol. 54, p. 1869, 1996.
[10] C. E. Shannon, “Communication theory of secrecy systems,” Bell Syst.Tech. J., vol. 28, pp. 656-715, 1949.
[11] P. O. Boykin and V. Roychowdhury, “Optimal encryption of quantumbits,” Physical Review A, vol. 67, p. 042317, 2003.
[12] J. Pan and A. Zeilinger, “Greenberger-Horne-Zeilinger-state analyzer,”Phys. Rev. A, vol. 57, pp. 2208-2211, 1998.
[13] C. Eltschka and J. Siewert, “Entanglement of Three-Qubit Greenberger-Horne-Zeilinger–Symmetric States,” Phys. Rev. Lett., vol. 108, p. 020502,2012.
[14] D. Bouwmeester, J. Pan, M. Daniell, H. Weinfurter and A. Zeilinger,“Observation of Three-Photon Greenberger-Horne-Zeilinger Entangle-ment,” Phys. Rev. Lett., vol. 82, pp. 1345-1349, 1999.
[15] L. Xin, L. Qing-Hong, F. Guang-Yu, W. Yue-Yuan and L. Shu-Tian,“Resonant interaction scheme for GHZ state preparation and quantumphase gate with superconducting qubits in a cavity,” Chinese Phys. B,vol. 23, p. 020311, 2014.
[16] Y. Xia, J. Song, H. Song, “Remote preparation of the N-particle GHZstate using quantum statistics,” Optics Communications, vol. 277, pp.219-222, 2007.
[17] S. Walborn, A. de Oliveira, R. Thebaldi, and C. Monken. Entanglementand conservation of orbital angular momentum in spontaneous parametricdown-conversion. Phys. Rev. A, 69:023811, 2004.
[18] R. Simon and N. Mukunda. Minimal three-component su (2) gadgetfor polarization optics. Phys. Lett. A, 143:165–169, 1990.
[19] Salla Gangi Reddy, Shashi Prabhakar, A. Aadhi, Ashok Kumar, MeghShah, R. P. Singh, and R. Simon. Measuring the mueller matrix of anarbitrary optical element with a universal su(2) polarization gadget. J.Opt. Soc. Am. A, 31(3):610–615, Mar 2014.
[20] X. H. Bao, T. Y. Chen, Q. Zhang, J. Yang, H. Zhang, T. Yang, and J. W.Pan. Optical nondestructive controlled-not gate without using entangledphotons. Phys. Rev. Lett., 98:170502, 2007.
[21] A. Beveratos, R. Brouri, T. Gacoin, A. Villing, J. P. Poizat, andP. Grangier. Single photon quantum cryptography. Phys. Rev. Lett.,89:187901, 2002.
[22] T. Jennewein, C. Simon, G. Weihs, H. Weinfurter, and A. Zeilinger.Quantum cryptography with entangled photons. Phys. Rev. Lett., 84:4729,2000.
[23] D. Bouwmeester, A. K. Ekert, and A. Zeilinger. The physics of quantuminformation, volume 38. Springer Berlin, 2000.
[24] E. Waks, A. Zeevi, and Y. Yamamoto. Security of quantum keydistribution with entangled photons against individual attacks. Phys. Rev.A, 65:052310, 2002.
[25] R. Ursin, F. Tiefenbacher, T. Schmitt-Manderbach, H. Weier, T. Scheidl,M. Lindenthal, B. Blauensteiner, T. Jennewein, J. Perdigues, P. Trojek,B. Omer, M. Furst, M. Meyenburg, J. Rarity, Z. Sodnik, C. Barbieri,H. Weinfurter, and A. Zeilinger. Entanglement-based quantum commu-nication over 144km. Nat. Phys., 3:481–486, 2007.
[26] D. S. Naik, C. G. Peterson, A. G. White, A. J. Berglund, and P. G.Kwiat. Entangled state quantum cryptography: Eavesdropping on theekert protocol. Phys. Rev. Lett., 84:4733, 2000.
[27] M. Mafu, A. Dudley, S. Goyal, D. Giovannini, M. McLaren, M. J. Pad-gett, T. Konrad, F. Petruccione, N. Lutkenhaus, and A. Forbes. Higher-dimensional orbital-angular-momentum-based quantum key distributionwith mutually unbiased bases. Phys. Rev. A, 88:032305, 2013.
[28] A. Martin, F. Kaiser, A. Vernier, A. Beveratos, V. Scarani, and
S. Tanzilli. Cross time-bin photonic entanglement for quantum keydistribution. Phys. Rev. A, 87:020301, 2013.
[29] J. Kempe. Multiparticle entanglement and its applications to cryptog-raphy. Phys. Rev. A, 60:910–916, 1999.
[30] W. Tittel, H. Zbinden, and N. Gisin. Experimental demonstration ofquantum secret sharing. Phys. Rev. A, 63:042301, 2001.
[31] S. Groblacher, T. Jennewein, A. Vaziri, G. Weihs, and A. Zeilinger.Experimental quantum cryptography with qutrits. New J. Phys., 8:75,2006.
[32] J. F. Clauser and M. A. Horne. Proposed experiment to test localhidden-variable theories. Phys. Rev. Lett., 23:880–884, 1969.
[33] D. Collins, N. Gisin, N. Linden, S. Massar, and S. Popescu. Bellinequalities for arbitrarily high-dimensional systems. Phys. Rev. Lett.,88:040404, 2002.
Dr. Sumanta Bagchi is Assistant Professor at Centre for Ecological Sciences, Indian Institute of Science (CES-IISc), Bangalore 560012, India. [email protected] Ekta Gupta is a research fellow at CES-IISc. Karthik Murthy is a Ph.D. student at CES-IISc
Project number: ISTC/BES/SB/332
Investigation of vegetation changes in the arid Trans-Himalayan ecosystem of northern India
Sumanta Bagchi, Ekta Gupta, Karthik Murthy
Abstract – Understanding ecosystem responses to current and projected climate change is important for sus-tainable land-use policies. The cold and arid Trans-Himalayan ecosystem is experiencing changes in regional climate (precipitation and tempera-ture) amidst concerns over land-degradation. We are assessing vegeta-tion growth and distribution using sat-ellite imagery to investigate the nature and severity of degradation. We ana-lyzed changes in Normalized Differ-ence Vegetation Index (NDVI), for phenological and biomass indicators in the watersheds of major wetlands in Ladakh region of northern India. The results will help resolve long-standing concerns over climate-induced deserti-fication in the Trans-Himalaya.
I. INTRODUCTION Central Asian highlands are a vast ex-panse of mountainous terrain that in-cludes of Trans-Himalaya in northern India. These represent arid ecosystems of high anthropological, biogeochemical, ecological, and hydrological significance (1-3). But they are also highly suscepti-ble to ongoing and projected changes in regional climate (4, 5). The Trans-Himalayan ecosystem covers large ex-panses of arid and semiarid rangelands in Ladakh (India), Spiti (India), Tibet (China) and Mustang (Nepal), and is ex-
periencing the effects of changes in both precipitation and temperature. There are widespread concerns over land-degradation and desertification that have adverse effects on biodiversity as well as on human livelihoods (5, 6).
The general climatic trend over the last three decades in this region, and more broadly across the western Hima-layan alpine shrub and meadow biome (5) is towards: (i) increasing mean an-nual temperature, (ii) decreasing winter precipitation (November-February), and (iii) increasing summer precipitation (August-September). Although there may be a net increase in total annual pre-cipitation, counter-intuitively, there is also evidence for desertification. It re-mains unclear whether and how climatic changes may play a major role in land degradation and desertification. The ex-tent, intensity, and magnitude of land degradation also remain poorly docu-mented.
In this project, we are analyzing temporal patterns of change in vegeta-tion phenology and biomass in the Trans-Himalayas using satellite-derived indices (Normalized Difference Vegeta-tion Index, NDVI). Initially we investi-gated changes in watersheds of major wetlands in Ladakh region, as these sup-port the highest densities of livestock grazers, and are likely highly susceptible to vegetation degradation.
II. METHODS
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
We used available NDVI data from MODIS for 500 m resolution. We ex-tracted NDVI data from 16-day interval MODIS images between 2004 and 2013. From this we extracted pixels from dif-ferent major watersheds in Ladakh, namely: Tso-Kar, Tso-Moriri, and Pan-gong-Tso. We screened these pixels based on additional information (pixel quality and data reliability).
Vegetation growth, as indicated by NDVI scores, follows a well-known double-sigmoidal pattern over an annual cycle (January to December, Fig. 1). We used nonlinear regression models in R 3.0.3 software to estimate the following parameters for each year: time of onset of growth, rate of increase after onset, peak vegetation biomass, rate of decline after peak, and duration of growth sea-son.
0 50 100 150 200 250 300 350
0.0
0.2
0.4
0.6
0.8
1.0
Fig. 1: Illustrative example for vegetation growth
Julian day (Jan-Dec)
ND
VI s
core
peak
rate of increaserate of decrease
onset end
III. RESULTS & DISCUSSION
Here we show results for the Tso-Kar watershed (Fig. 2). Key results of nonlinear regression models for 2003 are summarized in Table 1. In general, the statistical models perform well in esti-mating four of the parameters, but need further improvements in estimates of onset of growth season. In future work, we will compare estimates across time.
0 50 100 150 200 250 300 350
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Fig. 2: Tso-Kar watershed, 2003
Julian day (Jan-Dec)
ND
VI s
core
Table 1. Model parameters estimated by nonlinear (double-sigmoidal) regression for five parameters of vegetation growth.
Parameter Estimate 95% Confi-
dence limits
Peak growth 0.8 0.6-1.6
Rate of increase 0.01 day-1 0.009-0.013
Onset of growth 272nd day 215-355
Rate of decrease 0.03 day-1 0.02-0.04
End of growth 289th day 281-301
References
1. G. Miehe et al., Palaeogeography, Palaeoclimatology, Palaeoecology 276, 130 (2009).
2. S. Bagchi, M. E. Ritchie, Ecol. Lett. 13, 959 (2010).
3. W. R. Boos, Z. Kuang, Nature 463, 218 (2010).
4. X. Liu, B. Chen, Int J Climatol 20, 1729 (2000).
5. U. B. Shreshta, S. Gautam, K. S. Bawa, PLoS One 7(5), e36741 (2012).
6. W. Shaohong, Y. Yunhe, Z. Du, Y. Qinye, J Geogr Sci 17, 141 (2007).
ISTC/CSS/STP/0339 Page 1
Study on Self-assembly of Donor-Acceptor-Donor Molecular Materials and Melamine
Nileshi Saraf, Joydeep Dhar and Satish Patil* Solid State and Structural Chemistry Unit, Indian Institute of Science, Bangalore, India 560012
Email: [email protected]
Abstract
In the present work, the interaction between donor-acceptor-donor molecular oligomers (TTB) and melamine is studied with the help of various spectroscopic and electron microscopy techniques. TTB is a π -conjugated molecule which has donor-acceptor-donor (D-A-D) structure and can be used as an active chromophore in solar cell devices. However, we observed that the self-assembled ribbon like morphology of TTB results in low PCE due to phase separation within the active layer of bulk heterojunction solar cells. Melamine is introduced to disrupt the intermolecular hydrogen bonding within the TTB molecules which results in change in optical properties. The FTIR and NMR analysis suggests the presence of hydrogen bonding between the barbituric unit of TTB and amine group of melamine. AFM studies show the morphological variations in the films after adding melamine in different ratios.
Introduction
The donor-acceptor-donor (D-A-D) based small molecules and polymers have been extensively investigated for applications in organic FETs and solar cell devices.1-3 In this context, our group explored self-assembly approach of barbituric acid TTB (D-A-D) molecules in organic solar cells and field-effect transistors.4 Barbituric acid based TTB molecule is synthesized by Knoevenagel condensation of terthiophenecarbaldehyde and barbiturate appended pyran derivative as discussed in our earlier studies.5 The barbiturate and thiophene units act as donor and acceptor group, respectively. This molecule undergoes inter-molecular hydrogen bonding due to the presence of acidic proton (N−H) and electronegative oxygen (C=O) of barbiturate functional group. The H-bond mediated self-assembled morphology of TTB results in long ribbon like structures which might be disrupted by introducing melamine as shown in Fig. 1.
Figure 1: Plausible intermolecular H-bonding between TTB and Melamine
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
ISTC/CSS/STP/0339 Page 2
Results and Discussions
The interaction between TTB and melamine was examined with the help of various characterization techniques. Photophysical properties were investigated by UV-vis spectroscopy and fluorescence studies. Atomic force microscopy (AFM) was used to observe morphological changes in the film after adding melamine. FTIR and NMR results demonstrate the presence of hydrogen bonding between the molecules of melamine and TTB.
A) NMR spectroscopy The NMR study of TTB and TTB-
melamine at different molar ratio (1:0.25, 1:0.5, 1:1, 1:2, 1:4) is shown in the Fig. 2. The samples were prepared in CDCl3. It can be seen that there is an apparent shift in the peak (~7.54) of proton of the barbiturate unit present in TTB with the addition of melamine. The maximum downfield shift (~7.57) with the broadening of the peak-width is observed at 1:1 molar ratio of TTB/melamine indicating strongest H-bonding interaction. Further, the peak position slightly shifts upfield and broadens with further increase in melamine concentration. This suggests that with further increase in melamine concentration number of H-bond increases, but the strength of the H-bond decreases.
Figure 2: NMR analysis of TTB and TTB-Melamine with varying molar ratio of melamine.
B) FTIR analysis
Figure 3: FTIR spectra of TTB and TTB-Melamine
The FTIR analysis of TTB and TTB-melamine composite (Fig. 3) was performed in KBr pellet method. It was observed that the C-N stretching at 1357 cm-1 and 1295 cm-1 in TTB molecules diminished after adding melamine. The peak due to C=O stretching (1620 cm-1, 1640 cm-1) and –NH bending (amide II) (1496 cm-1) in TTB shifts at higher frequency after addition of melamine due to the formation of H-bond between TTB and melamine.
C) Optical Properties
The thin film normalized absorption spectra of TTB and TTB/melamine at different molar ratio are shown in Fig. 4(a). The band at 390 nm corresponds to the π-π* transition and the band at 517 nm is due to the intramolecular charge transfer from the donor to the acceptor moiety in TTB. It can be seen that there is no shift in the absorption maxima of TTB with gradual addition of melamine. The thin film emission spectra (Fig. 4(b)) shows that the fluorescence intensity varies with varying melamine concentration. The intensity is maximum at 1:1 molar ratio followed by 1:2 whereas 1:4, 1:2 and TTB shows similar emission intensity. Therefore, the results of
ISTC/CSS/STP/0339 Page 3
the fluorescence study indicate that at 1:1 molar ratio the favorable interaction between TTB and melamine efficiently restricts the self-quenching phenomenon of TTB.
(a)
(b)
Figure 4: (a) Normalized Absorption and (b) Emission Spectra of TTB and TTB-Melamine at different molar ratio of melamine in thin film
D) AFM analysis
The p-type Si (100) wafers were pre-cleaned by repeated sonication in acetone followed by piranha treatment. The wafers were then thoroughly rinsed with distilled water and dried with nitrogen flow. Oxygen plasma treatment was done for 15 min before depositing films in order to make the surface hydrophilic to ensure the uniformity of the deposited films. Thin films of TTB and TTB-melamine dissolved in chlorobenzene
were spin-coated onto pre-cleaned Si wafers at 1000rpm for 60 sec. The AFM image of TTB (1mg/ml) as shown in Fig. 5 shows nanoribbon-like structure with width around 60nm and height ranging from 5-15 nm (Fig. 5(b)).
Figure 5(a): AFM analysis of TTB dissolved in chlorobenzene (1mg/ml)
Figure 5(b): Depth profile of AFM image of TTB dissolved in chlorobenzene (1mg/ml)
Melamine in the molar ratio of 1:0.05,
1:0.1, 1:0.2, 1:0.4 was added in TTB and changes in the morphology is studied as shown in Fig. 6(a-d). It can be seen that as the concentration of melamine is increased, the TTB network is becoming denser. Melamine assists the TTB molecule to form a better interdigitated network enhancing the area of coverage.
ISTC/CSS/STP/0339 Page 4
(a) (b)
(c) (d)
Figure 6: AFM images of TTB-Melamine at (a) 1:0.05, (b) 1:0.1, (c) 1:0.2 and (d) 1:0.4 molar ratios.
E) DSC measurements
Figure 7: DSC thermogram of TTB and TTB-Melamine
with different molar ratio of melamine
The thermal properties of TTB and
TTB-Melamine were investigated by using differential scanning calorimetry (DSC). It was observed that TTB shows a higher melting point of around 237°C which is due to the intermolecular hydrogen bonding. DSC results as shown in Fig. 7 suggests that as melamine concentration is increased, the melting point shifts to lower value with minimum of 174°C in 1:1 molar ratio due to
the disruption of strong H-bonding between the TTB molecules. The crystallization behavior in cooling cycle was not observed.
Conclusion and Future Plan
In conclusion, we have investigated the morphological, thermal and optical properties of TTB and TTB-Melamine composite. The results reported here clearly states that melamine disrupts the self-assembly between the TTB molecules. The TTB-Melamine in a ratio of 1:1 shows promising behavior for further use in solar cells with strongest hydrogen bond interaction. Further studies on its electronic properties are under progress in order to investigate the utility of these materials for solar cells.
Acknowledgement
We sincerely acknowledge funding from ISRO-STC.
References
1. Garnier F.; Hajlaoui R.; Yassa A.; Srivastava P. Science 1994, 265, 1684–1688
2. Burroughes J.; Bradley C; Brown A.; Marks R.; Mackay K.; Friend R.; Burns, P.; Holmes A.; Nature 1990, 347, 539–540.
3. Liang, Y.; Feng, D.; Wu, Y.; Tsai, S. T.; Li, G.; Ray, C.; Yu, L.; J. Am. Chem. Soc. 2009, 131, 7792–7799.
4. Bhaskar, R.; Tandy, K.; Horecha, M.; Formanek, P.; Stamm, M.; Gevorgyan, S.; Krebs, F.; Kiriy, A.; Meredith, P.; Burn, P.; Namdas, E.; Patil, S.; J. Phys. Chem. C 2011, 115, 14369–14376.
5. Bhaskar, R.; Stephen, M.; Ali, F.; Patil, S.; J. Phys. Chem. C 2013, 117, 9129−9136.
High-resolution Image restorationand enhancement using GPU
Nithish Divakar and R. Venkatesh Babu
Abstract—The Weighted Nuclear Norm Minimisation fordenoising is an effective denoising algorithm which has a verygood denoising performance at high noise levels. But it requiresa huge amount of computational time. Here we propose amethodology by which the computationally intensive part ofWNNM is accelerated using GPUs. The acceleration requirescomputation of singular value decomposition of large number ofaverage sized matrices. We also discuss a better way to place datain GPU memory so that it results in high memory bandwidth.
I. INTRODUCTION
A digital image can be considered as a discrete represen-tation of continuous 2D signal. A generic noisy image modelcan be expressed as follows.
y(p) = x(p) + η(p)
where x is original image,y is the noisy image and η is thenoise that gets mixed. Image denoising algorithms attempt toobtain an approximation x of the the signal x, when only thesignal y and some knowledge of noise η is known.
Image denoising using weights nuclear norm minimisationis a denoising algorithm which shows a very good recoveryrate even at high noise levels [1], but WNNM denoising re-quires huge computational time. Here we propose an approachwhich can accelerate this algorithm using GPUs.
II. RELATED WORK
Many state of the art denoising algorithms exist for imagedenoising, most, differing only in their model of what noiseis. From the assumptions that natural images constitutes piece-wise smooth functions, image denoising algorithms like theones given by Lindenbaum et al. in [2] uses smoothening filterto remove noise. But this approaches fails to preserve imageedges as their noise model includes all sudden variations inthe signal.
Non Local means algorithm suggested by Buades et al.in [3] removes noise by taking contribution from only thosepixels which are similar to the key pixel.
BM3D algorithm proposed by Dabov et al. in [4] computesneighbourhood of blocks in transform domain and collabora-tive filtering is used to estimate pixel intensities.
Although the non local neighbourhood model appears tobe concrete in estimating the noise model, it is intrinsicallydepending on the fact that a major portion of original image
Nithish Divakar is masters student in the SERC De-partment, Indian Institute of Science, Bangalore, [email protected]
R. Venkatesh Babu is an assistant professor in the SERC Department, IndianInstitute of Science, Bangalore, India [email protected]
details can still be extracted despite the corruption from noise.This assumption fails at higher noise levels.
Low rank matrix approximation based image denoisingmethods uphold there denoising efficiency even at higher noiselevels. Weighted nuclear norm minimisation based denoisingproposed by Gu et al. in [1] is such a strategy which usesweighted reduction of data variability along different dimen-sions to reduce noise.
Although this method shows good denoising performance,the algorithm’s heavy resource consumption in terms of com-putation time makes it unusable for application that require onthe line processing. This research effort is to improve WNNMdenoising algorithm’s execution time.
III. WEIGHTED NUCLEAR NORM MINIMISATION FORIMAGE DENOISING
Weighted nuclear norm minimisation is a problem in lowrank matrix approximation theory [5], which uses a nuclearnorm as the objective function for low rank matrix completion.The problem can be stated as follows
X = arg minX‖Y −X‖2F + ‖X‖w,∗
w is a weight vector and ‖X‖w,∗, is defined to be dotproduct of the singular values of X and the weight vector.The algorithm is throughtly described in Gu et al. [1]
WNNM is used in image denoising by constructing a matrixwhose columns are vectorised patches which are obtained froma local neighbourhood of a key patch. Local neighbourhoodof each patch is defined as the set of patches which are closerto it in terms of some distance measure. Usually, `2 normdistance is used.
Denoising is essentially constructing the low rank approxi-mation of the original matrix. WNNM helps us in achievingthis.
IV. PROPOSED APPROACH
The most computationally intense part of the WNNMalgorithm is to compute the singular value decomposition ofthe neighbourhood matrix for evech key patch. But the com-putation corresponding to one key patch is totally independentof another.
The key part of the proposed approach is a technique tocompute all the required SVDs on a GPU device. Since theneighbourhood matrix is not very large, the usual algorithmsfor computing singular value decomposition does not providean optimum performance as they have been designed to handlelarge matrices.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
More over, the WNNM algorithm requires computation ofSVDs for a large number of matrices.
V. ACCELERATION USING GPUAn easy approach for computing all the required SVDs
would be to simply offload the computation to GPU suchthat a single thread compute SVD per matrix. The executiontimings of this approach can be seen in Table I. This approachhas many possible improvement because it uses only a smallfraction of the thread pool available.
A better idea would be to use multiple threads to computethe SVDs in parallel, leading to 2 levels of parallelism.
We have used the lowest number of threads which haveimplicit synchronisation between them. This atomic set ofthreads are called as warps. The hardware on which ourexperiments were conducted restricts this number to be 32.But the following methodology can be easily adapted to anyother hardware specification.
Our specific problem structure advocates us to use a dif-ferent algorithm for SVD computations. Biorthogonalisationalgorithm [6] for svd computation offers a faster and moreunifrm memory access pattern when used in parallel thanstandard bi-diagonalisation algorithm.
Biorthogonalisation procedure was first proposed byHestenes in [6]. A good discussion about better accuracy ofbiorthogonalisation procedure and its implementation can befound in [7]. The Complete algorithm is given in Figure 1.
1: procedure BIORTHOGONALISATION(A)2: [a1|a2| . . . |an] = A conv = false . Columns of A3: V = In di = aTi ai ∀i = 1, . . . , n . 4: while 30 or less iteration and conv = false do5: for Each column pair (ai, aj) : i < j do6: if |aTi aj | > mε
√didj then .
7: Compute the givens rotation angles8: Rotate column vector through the angle9: Update corresponding dot products
10: end if11: end for12: Set conv = true if no columns were updated13: end while14: if conv=true, then set Σii =
√di and ui = aiΣii .
15: U = [u1|u2| . . . |un]16: end procedure
Fig. 1. SVD decomposition by Biorthogonalisation
A. Data distribution in GPU memory for optimum perfor-mance
The statements marked by “” in Figure 1 can be executedin parallel by a a set of GPU threads. During the comutation,because of the structure of the algorithm, adjacent threads inthe warp access elements of a olumn i matrix together. Thissuggest us to use a stacked transpose storage of matrices inthe memory for faciliting non strided memory access. So wewill use a P ×N ×M order stacking where P is number ofmatrices and M ×N is matrix dimensions.
VI. EXPERIMENTAL RESULTS
All experiments were carried out in a system with IntelXeon CPU E5-1650 v2 processor and Tesla K20 GPU cardwith 5 GB of GDDR5 memory.
The patch size used by the algorithm is 9. So the dimensionof the vectorised patches is fixed at 81 components. Also, thenumber of SVDs to be performed throughout the algorithmalso stays the same at 7056 for an image size of 256× 256.
The running times are calculated for matrix size which thealgorithm encounters during a successful run.
TABLE IEXECUTION TIME OF BOTH SINGLE AND 32 THREAD PARALLEL SVD
COMPUTATION
Problem size (m× n× p) One Thread 32 Threads81× 130× 7056 10.4893s 1201ms81× 120× 7056 9.5463s 954ms81× 110× 7056 8.0076s 800ms81× 100× 7056 7.3253s 653ms81× 90× 7056 6.9263s 490ms81× 80× 7056 5.0035s 383ms81× 70× 7056 4.5902s 271ms
A. Final speedup
The original algorithm spent about 575.739 seconds for allrequired SVD computations. Our method is able to performsame amount of comutation in 9.504 seconds. Thus we haveobtained about 60x acceleration in required decompositionpart.
VII. CONCLUSION
The results from the experiments suggested that WNNMmethod has lots of potential to become a fast and efficientdenoising algorithm. We were able to reduce the time taken forthe most computationally intensive part of the algorithm by 60times. If other parts of the algorithm are equally accelerated,WNNM can become a fast algorithm indeed.
REFERENCES
[1] S. Gu, L. Zhang, W. Zuo, and X. Feng, “Weighted nuclear normminimization with application to image denoising,” in IEEE Conf. onComputer Vision and Pattern Recognition, 2014.
[2] M. Lindenbaum, M. Fischer, and A. Bruckstein, “On gabor’s contributionto image enhancement,” Pattern Recognition, vol. 27, no. 1, pp. 1–8, 1994.
[3] A. Buades, B. Coll, and J.-M. Morel, “A non-local algorithm for imagedenoising,” in Computer Vision and Pattern Recognition, 2005. CVPR2005. IEEE Computer Society Conference on, vol. 2. IEEE, 2005, pp.60–65.
[4] K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, “Image denoising bysparse 3-d transform-domain collaborative filtering,” Image Processing,IEEE Transactions on, vol. 16, no. 8, pp. 2080–2095, 2007.
[5] I. Markovsky and K. USEVICH, Low rank approximation. Springer,2012.
[6] M. R. Hestenes, “Inversion of matrices by biorthogonalization and relatedresults,” Journal of the Society for Industrial & Applied Mathematics,vol. 6, no. 1, pp. 51–90, 1958.
[7] Z. Drmac, “Implementation of jacobi rotations for accurate singular valuecomputation in floating point arithmetic,” SIAM Journal on ScientificComputing, vol. 18, no. 4, pp. 1200–1222, 1997.
*Corresponding Author, Email: [email protected]. IISC-ISRO STC Project Code Number: ISTC/MAE/SC/334 Part of this work appeared as Technical Publication Nos. ASME GTINDIA 2014-8340 in the 2014 ASME GT India Conference, New Delhi
Turbulence-transport-chemistry interaction in statistically planar premixed flames in near isotropic turbulence
Harshavardhana A. U, Swetaprovo Chaudhuri* and K. N. Lakshmisha
Department of Aerospace Engineering, Indian Institute of Science, Bangalore, INDIA
Abstract - Turbulence-transport-chemistry interaction plays a crucial role on the flame surface geometry, local and global reaction rates, and therefore, on the propagation and extinction characteristics of intensely turbulent flames encountered high speed combustors. The aim of the present work is to understand these interaction effects on the flame surface. This will be useful in developing an insight into the annihilation and extinction mechanism of lean premixed flames, interacting with near isotropic turbulence. As an example case, lean premixed H2-air mixture is considered with detailed chemistry effects in Direct Numerical Simulations (DNS). The work is carried out in two cases where a statistically planar flame interacts with near isotropic turbulence of different turbulence intensities (Case 1 ReT = 146 and Case 2 ReT = 351). A recently proposed Flame Particle Tracking (FPT) technique was used for this purpose. Flame particles are surface points residing and comoving with an iso-scalar surface within a premixed flame. Tracking flame particles allows us to study the evolution of propagating surface locations uniquely identified with time along with the evolution of various properties at those locations. In this work, using DNS and FPT we study the flame speed, reaction rate and transport histories of such flame particles residing on iso-scalar surfaces. An analytical expression for the local displacement flame speed (Sd) is derived, and the contribution of transport and chemistry on the displacement flame speed is identified. An examination of the results of both the cases leads to a conclusion that the cause of variation in Sd may be attributed to the effects of turbulent transport and curvature.
I. INTRODUCTION
Turbulent combustion can be understood to a greater extent in terms of propagating iso-scalar surfaces particularly for studying the phenomena such as turbulent mixing and reactions on these surfaces [1]. This approach is followed by many researchers working on the turbulent combustion by using detailed, reduced and/or single step chemistry to model the reactions.
The previous studies on the statistically planar premixed flames considered an iso-scalar surface of mass fraction [2-3] and reaction progress variable [4]. In the work of Echekki et al. [2-3] the effects of strain rate, and curvature are examined on the propagation of an iso-scalar surface of mass fraction of CH4. Further, in [4] the effects of non-
unity Lewis number were assessed on the behaviour of the displacement flame speed. These studies used either reduced or single step chemistry to model the reactions. In spite of important insights gained from these studies, the focus was on understanding the global effects of the turbulence flame interaction and the local effects are not studied.
The effects of turbulent mixing and dispersion are understood to a great extent with Lagrangian type of flow description. However, a semi-Lagrangian approach is an attractive way to study the interaction of premixed flame with turbulence. A brief of this approach is described below.
In the recent work of S. Chaudhuri [5], a new technique is proposed to study the local effects on the flame surface by
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
tracking the evolution of discrete locations on an iso-scalar surface within a premixed flame and is called as Flame Particle Tracking (FPT). Further, the FPT technique is used to study the kinematic effects such as strain rate and curvature on propagation characteristics of the premixed flame.
In the present work utilizing the FPT technique, the interaction between
turbulence-transport-chemistry is studied by considering statistically planar premixed flames interacting with near isotropic turbulence with two turbulence intensity levels namely,
1. Low turbulence intensity level (Case 1, ReT = 146)
2. High turbulence intensity level (Case 2, ReT = 351)
Figure 1 Computational domain of statistically planar flame in near isotropic turbulence. a) 2D schematic and b) 3D domain showing flame surface with black dots representing the flame particles
II. COMPUTATIONS
A. Direct Numerical Simulations
In the present work, three dimensional Direct Numerical Simulations of lean, premixed, H2-air mixture are performed using the open source PENCIL CODE [6]. The combustion chemistry part in the pencil code is implemented by Babkovskaia et al. [7], where the Navier-Stokes, energy and species equations are solved in the compressible form for a turbulent, reacting mixture. Further, for both the cases the combustion chemistry is modelled by using detailed H2-air mechanism developed by Li et al. [8], consisting of 19 elementary reactions and 9 species. The inlet conditions namely, temperature (T) = 310 K, pressure (P) = 1 atm, equivalence ratio () = 0.81 and Lewis number (Le) = 0.84 (diffusivity
calculated with that of the deficient species H2), are maintained same for both cases. Also, all the variables pertaining to both the cases are listed in Tab. 1.
B. Flame Particle Tracking Algorithm
The DNS data obtained from the above simulation serves as an input to the FPT algorithm. On a given iso-scalar surface various parameters are calculated for each flame particle and for the iso-surface. The FPT algorithm is described briefly in the following paragraph and readers can refer to [5] for further details. The position of the flame particle at any time instant (t) is governed by,
where, ([], ) is the flow velocity; (
[], ) is the local displacement flame speed of the iso-surface and ([], ) local unit normal. Each of these variables are dependent on () at time t.
In the FPT technique, a constant property surface which propagates into the flow is chosen and the discrete locations on the surface are marked and are called as “flame particles”. This flame particles are surface points [1] on a reacting propagating surface. The evolution of
flame particles with time obeys the Eq. (1). The particle locations are advanced with the time and the co-ordinates of the flame particles at time level t+t is obtained from the previous time level t using the triangle-ray intersection algorithm. The noticeable distinction between the present method and the previous works on particle tracking is that, the flame particles are reactive and they are confined to a specific iso-scalar surface.
III. RESULTS AND DISCUSSION
A. Statistically planar flames
In [5] two statistically planar flame cases (Case 1 and Case2) experiencing different turbulent intensity levels were investigated using FPT. It was found that the tangential strain rate and also, curvature increased to large values towards the end of the flame particle lifetime (Here, the negative is used to indicate that the flame surface is concave towards reactants. A positive value indicates that flame surface is convex towards reactants). These observations were attributed to the
reorientation of local surface normal w.r.t. the most extensive component of the principal strain rate. This leads to persistently increasing tangential strain rate, cusp formation and further leading to annihilation of the flame surface. It was also shown that the local displacement flame speed Sd sharply increased towards the end of flame particles. In recent computational work [9], similar phenomena was also observed, albeit with single step chemistry.
Figure 2 Displacement flame speed variation with normalized time for Tiso = 1493K (a) Case 1 (b) Case 2
(a) (b)
(1)
Figure 3 (a) Heat flux (b) Diffusion flux and (c) Heat release rate variation with normalized time for Tiso = 1493K and Case 1
Figure 4 (a) Heat flux, (b) Diffusion flux and (c) Heat release rate variation with normalized time for Tiso = 1493K and Case 2
Figure 5 Magnitude of gradient of T variation with normalized time (a) Case 1 and (b) Case 2 for Tiso = 1493K
For an iso-temperature surface the analytical expression for Sd is as follows,
In the Figs. 2 to 5, the red lines represent the variation in the property for individual particles and the black line with circles is the ensemble average of the property over all the particles at that instant.
Figures 2a and 2b shows the
variation of Sd with , for Case 1 and
Case 2, respectively for the iso-temperature surface of Tiso = 1493K. Here, , is the life time of the flame particle. It is total duration between the instant at
which the particle is placed on the iso-surface to the time instant at which it is lost. It is observed from Fig. 2a that the Sd
remains constant up to , ≈ 0.8 and
increases rapidly towards the end. To assess this behavior, individual terms of Sd equation (Ref. Eq.(2)) are analyzed.
The first term of the Sd equation is the heat flux term and it shows an increase with the normalized time, as shown in Fig. 3a. The Figs. 3b and 3c shows the behavior of
(a)
(a)
(c) (b)
(b)
(a)
(b) (c)
=1
||. () +
1
−ℎ
()
diffusion flux term and the heat release rate term, respectively. These terms show a decreasing trend with normalized time signifying negative contribution to the Sd. Further, the |∇| in the denominator of the expression, as plotted in Fig. 5a, shows a decreasing trend with normalized time. The negative contribution of diffusion flux and heat release rate to Sd is overcome by the drop in |∇|, leading to an overall increase in resultant. Since, the |∇| can be related to the curvature, it can be argued that for a constant temperature surface the increase in Sd is due to curvature and heat focusing effects.
Similar to the Case 1 of Tiso=1493K, Fig. 2b shows the variation of Sd for the Case 2 for the same iso-surface. Here though the Sd value increases continuously from the initial time itself, the variation
however, is more steep from , ≈ 0.8
onwards. To understand this behavior of Sd, again the individual terms of Eq. (2) are analyzed. The behavior of individual terms pertaining to Case 2 and Tiso = 1493K are shown in Fig. 4 and Fig. 5b. The nature of heat flux, diffusion flux and heat release rate is shown in Fig. 4a, 4b and 4c respectively. Further, Fig. 5b shows the variation in magnitude of gradient of T with normalized time. Similar to the Case 1, the analysis of the individual terms shows that the increase in Sd towards the end of particle life time can be attributed to the heat focusing and curvature effects.
Thus, from the study presented in this paper, it can be concluded that the increase in Sd is universal for the constant temperature iso-surfaces towards the end of the flame particle lifetime. Further, the increase in Sd is mainly attributed to pure curvature and heat focusing effects.
IV. CONCLUSIONS AND FUTURE WORK
In the present work, by using flame particle tracking algorithm, the flame surface annihilation and propagation of statistically planar premixed flames is studied. The increase in the displacement flame speed towards the end of the particle lifetime is universal irrespective of the turbulence intensity level and initial condition of the flame surface. This
increase in Sd is attributed to decrease in |∇| which influences the Sd through the curvature increase to large negative values.
The FPT technique is an important tool in understanding the local phenomena of the flame surface. Further, FPT technique will be employed to understand the flames in the shear layers in high speed flows. The stabilization of flames at high
Case 1 Case 2
Grid 256x128x128 256x128x128
Domain (cm) 1x0.5x0.5 1x0.5x0.5
∆ 9 9
⟨⟩ 300 500
(cm/s) 192 503
146 351
(cm) 0.119 0.109
() 627 217
() 28 13
() 52 12
(K) 310 310
0.81 0.81 Flame
Thickness (μm)
361 361
Δ() 4 2.5
() 680 292
Table 1 List of Parameters
speeds is a challenging task as the flow time scales are very small compared to chemical time scales. The phenomenon such as local to global extinction transition
is very crucial to understand the flame stabilization. The FPT technique offers us elegant way to study this extinction transition.
REFERNCES
1. Pope, S., 1988. “The evolution of surfaces in turbulence”. International journal of engineering science, 26(5),pp. 445–469.
2. Echekki, T., and Chen, J. H., 1996. “Unsteady strain rate and curvature effects in turbulent premixed methane-air flames”. Combustion and Flame, 106(1), pp. 184–202.
3. Echekki, T., and Chen, J. H., 1999. “Analysis of the contribution of curvature to premixed flame propagation”. Combustion and Flame, 118(1), pp. 308–311.
4. Chakraborty, N., and Cant, R. S., 2005. “Influence of lewis number on curvature effects in turbulent premixed flame propagation in the thin reaction zones regime”. Physics of Fluids (1994-present), 17(10), p. 105105.
5. Chaudhuri, S., 2014. “Life of flame particles embedded in premixed flames interacting with near isotropic turbulence”. Proc. Combust. Inst. (2014),http://dx.doi.org/10.1016/j.proci.2014.08.007.
6. http://www.nordita.org/software/pencil-code/.
7. Babkovskaia, N., Haugen, N., and Brandenburg, A., 2011. “A high-order public domain code for direct numerical simulations of turbulent combustion”. Journal of computational physics, 230(1), pp. 1–12.
8. Li, J., Zhao, Z., Kazakov, A., and Dryer, F. L., 2004. “An updated comprehensive kinetic model of hydrogen combustion”. International Journal of Chemical Kinetics, 36(10), pp. 566–575.
9. Andrei N. Lipatnikov, Vladimir A. Sabelnikov, S. N., and Hasegawa, T., 2014. “Unburned mixture fingers in premixed turbulent flames”. Proc. Combust. Inst. (2014),
http://dx.doi.org/10.1016/j.proci.2014.06.081.
Thermally Activated Unimolecular Decomposition Mechanisms of Dimethylnitramine-Aluminum (DMNA-Al) and Dimethylnitramine-Zinc
(DMNA-Zn) Complex
Anupam Bera and Atanu Bhattacharya*
Department of Inorganic and Physical Chemistry, Indian Institute of Science, Bangalore –560012, India *[email protected]
Abstract In this report we discuss the effects of metal atoms in the ground state decomposition mechanisms of dimethylnitramine (DMNA), which is the simplest nitramine molecule, by means of quantum chemical calculations. For the present study, we have considered two important metal atoms, aluminum (Al) and zinc (Zn). The decomposition pathways of isolated DMNA, DMNA-Al and DMNA-Zn have been explored at the MP2/6-31G(d) level of theory. The reaction pathways of the isolated DMNA, DMNA-Al, and DMNA-Zn are also compared and contrasted in this report. MP2 level of theory predicts that isolated DMNA can follow N-NO2 dissociation as well as nitro-nitrite isomerization pathways; however, N-NO2 bond dissociation pathway is associated with the lowest activation energy barrier. DMNA-Al complex, on the other hand, shows significantly different decomposition pathway: it involves several steps, such as Al-O bond dissociation, and then N-N bond dissociation followed by an isomerization. DMNA-Zn complex exhibits, first, N-N bond dissociation, and after transforming to a very stable dimethyl (µ2-nitro) amine zinc complex it undergoes a Zn-O bond dissociation which leads to the similar nitrite isomerization product and finally NO elimination. In contrary to the DMNA results, NO elimination is found to be the lowest energy pathway for both DMNA-Al and DMNA-Zn complexes.
Introduction
For a long time, molecular energetic materials (EMs), produced by mixing oxidizer and fuel constituents into one molecule, (e.g., nitroglycerine, RDX, HMX, etc.), have been the centre of attraction for high energy propellant applications. However, recently, novel composite metalized energetic materials (mEMs),[1] containing metal particles (e.g., Fe, Al, etc.) and traditional molecular EMs, have drawn significant attention due to their superior performance in advanced propellant applications.[2] They are found to release more than twice as much energy as the best molecular explosives do.[3] As compared to the “conventional” molecular EMs, mEMs offer the possibility of faster energy release, more complete combustion and greater control over performance. Burn rates of metalized energetic propellants can be accelerated by controlling the size of the constituent metal particles. Tuning the surface-to-volume ratio of the metal particles, the rate of chemical reactions of composite mEMs can also be controlled.1
Enhanced performance of mEM is often accounted for the exothermicity of oxide formation reaction for constituent metal during combustion. For an example, heat of formation of alumina from aluminum (4Al+3O2→2Al2O3) is -1676kJ/mol.[4] During burning of molecular EMs in mEM, the Al particles are oxidized and
Project # ISTC/CIP/AB/333
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
the exothermic heat is added to the total heat of reaction. This thermodynamic explanation of superior performance of mEM over molecular EM is well accepted in energetic material community. However, another part of this problem has never been seriously considered: the presence of metal (particle) surface can also alter the decomposition mechanisms and dynamics of molecular EMs. How does metal surface alter decomposition mechanisms and dynamics of molecular EMs? How does metal surface transfer energy to the molecular EMs during burning? These questions have mostly remained unanswered, thus far. In the quest of the answers to these questions, this article presents theoretically predicted decomposition pathways of a simple analogue nitraminmolecule and its complex with aluminum atom.
RDX(hexahydro-1,3,5-trinitrotriazine), HMX (Octahydro-1,3,5,71,3,5,7-tetrazocine, C4H8N8O8) and CL(2,4,6,8,10,12-hexanitro-2,4,6,8,10,12hexaazaisowurtzitane) are well known nitramine (N-NO2)-based energetic materials. Their chemical structures are depicted in Figure 1. These molecules possess a number of N(nitramine energetic moieties. Due to the complex structures of nitramine energetic molecules, often a structurally simple analmolecule dimethylnitramine (DMNA, see Figure 1 for structure), containing only one Nenergetic moiety, is subjected to the laboratory experiment to explore intrinsic decomposition mechanisms and dynamics of a single nitramine moiety. [5]. To date, thermally activated decomposition (which occurs on ground electronic state surface) of isolated DMNA has been investigated experimentally [6]theoretically [7]; however, thermal decomposition of DMNA-Al and DMNAcomplexes, which can be considered to be a simple analogue molecule of metalized nitramine energetic material, is hithertounexplored. Furthermore, a comparative study of the decomposition mechanisms oDMNA, DMNA-Al, and DMNA-Zn complexes at the ground electronic state will enable us to predict how presence of metal atom can change/alter the decomposition behavior of single nitramine moiety.
the exothermic heat is added to the total heat of reaction. This thermodynamic explanation of superior performance of mEM over molecular EM is well accepted in energetic material community. However, another part of this
er been seriously considered: the presence of metal (particle) surface can also alter the decomposition mechanisms and dynamics of molecular EMs. How does metal surface alter decomposition mechanisms and dynamics of molecular EMs? How does metal
ansfer energy to the molecular EMs during burning? These questions have mostly remained unanswered, thus far. In the quest of the answers to these questions, this article presents theoretically predicted decomposition pathways of a simple analogue nitramine molecule and its complex with aluminum atom.
trinitro-1,3,5-1,3,5,7-tetranitro-) and CL-20
2,4,6,8,10,12-hexaazaisowurtzitane) are well known nitramine
energetic materials. Their chemical structures are depicted in Figure 1. These molecules possess a number of N-NO2 (nitramine energetic moieties. Due to the complex structures of nitramine energetic molecules, often a structurally simple analogue molecule dimethylnitramine (DMNA, see Figure 1 for structure), containing only one N-NO2 energetic moiety, is subjected to the laboratory experiment to explore intrinsic decomposition mechanisms and dynamics of a single nitramine
thermally activated decomposition (which occurs on ground electronic state surface) of isolated DMNA has been investigated experimentally [6] and
however, thermal Al and DMNA-Zn
complexes, which can be considered to be a simple analogue molecule of metalized nitramine energetic material, is hitherto-unexplored. Furthermore, a comparative study of the decomposition mechanisms of isolated
Zn complexes at the ground electronic state will enable us to predict how presence of metal atom can change/alter the decomposition behavior of
Figure 1: Structure of RDX, HMX, DMNA
In the present work, ground electronic
state decomposition pathways of isolated DMNA and DMNA-Al complex have been explored utilizing second order MØller(MP2) perturbation theory with 6set. The decomposition pathways osystems are compared and contrasted. Isolated DMNA exhibits N-NO2 bond dissociation (rendering NO2 product) as the lowest energy pathway; however, both DMNADMNA-Al complexes exhibit a novel lowest energy decomposition pathway, which is to nitro-nitrite isomerization; however, the pathway finally yields NO as a final decomposition product.
Theoretical Procedure All geometry optimizations at the
ground state of DMNA, DMNADMNA-Zn are performed at the MP2/6level of theory using Gaussian 09.[8] MP2/631G(d) was previously found to be optimum level of theory to estimate activation barriers for DMNA system. [9] The transition state is confirmed by performing frequency calculation. The unstable normal mode of vibration atransition state shows an imaginary frequency and an intrinsic reaction coordinate (IRC) calculation starting from the relevant transition state is indicative of the related pathway. The basis set superposition error (BSSE) associated with the NO and NO2 elimination pathways are corrected by counterpoise method, as
Figure 1: Structure of RDX, HMX, CL20 and
In the present work, ground electronic state decomposition pathways of isolated
Al complex have been explored utilizing second order MØller-Plesset (MP2) perturbation theory with 6-31(d) basis set. The decomposition pathways of these two systems are compared and contrasted. Isolated
bond dissociation product) as the lowest energy
pathway; however, both DMNA-Zn and Al complexes exhibit a novel lowest
energy decomposition pathway, which is similar nitrite isomerization; however, the
pathway finally yields NO as a final
All geometry optimizations at the ground state of DMNA, DMNA-Al, and
Zn are performed at the MP2/6-31(d) theory using Gaussian 09.[8] MP2/6-
31G(d) was previously found to be optimum level of theory to estimate activation barriers for DMNA system. [9] The transition state is confirmed by performing frequency calculation. The unstable normal mode of vibration at the transition state shows an imaginary frequency and an intrinsic reaction coordinate (IRC) calculation starting from the relevant transition state is indicative of the related pathway. The basis set superposition error (BSSE) associated
elimination pathways are corrected by counterpoise method, as
implemented in Gasssian09. Transition states for both the NO and NO2 eliminations are also performed through potential energy scanning. A relaxed potential energy scan is performed using Z-matrix by changing the scan variable (respective bond distance).
Results and Discussion
As previous experimental investigations on the thermal decomposition of gas phase isolated DMNA [6] shows that DMNA can dissociate through two plausible mechanisms: NO2 elimination and nitro-nitrite isomerization followed by NO elimination, in the present work we have explored these two reaction channels for isolated DMNA, and DMNA-Al and DMNA-Zn complexes. The results are presented and contrasted below. DMNA: The reaction pathways with relative energies (in Kcal/mole) and structures of the optimized geometries at different critical points for DMNA are depicted in Figure 2. Relative energies are given in Table 1. Previously, decomposition pathways of DMNA in the ground electronic state was explored by Thomson and coworkers [7(a)] at MP2/6-311G(d,p) level of they and NO2 elimination was predicted to be the lowest energy dissociation channel as compared to nitro-nitrite isomerization, which is in concordance with the present results. The nature of the transition state (TS) associated with the nitro-nitrite isomerization dissociation channel of DMNA has been a subject of intense discussion for a long time. [9] Both tight and loose TSs on the ground electronic state surface of DMNA are localized at MP2 level. The activation energy barrier to the loose TS is calculated to be 69 kcal/mol, significantly lower in energy than that of tight TS, as predicted at the MP2/6-31G(d) level. NO2 elimination barrier, however, is calculated to be 30 kcal/mol, which is significantly lower in energy than that of both isomerization TSs.
Figure 2: Decomposition pathway of DMNA in
the ground electronic state, as revealed at the
MP2/6-31G(d) level of theory. The relevant
energies are in Kcal/mole.
DMNA-Al: Figure 3 shows decomposition pathway of DMAN-Al complex and the relative energies are given in Table 1, as predicted at MP2/6-31G(d) level of theory. Optimized geometry of DMNA-Al complex at different critical points are also depicted the figure. Figure 3 shows that the Al atom is located below the N-NO2 plane: this complexation finally results in a non planar geometry of the N-NO2 moiety in DMNA-Al complex (Opt1), which is in contrast to that of DMNA. The Al atom shows bonding with both oxygen atoms (of NO2 moiety) and perhaps with the (N1) nitrogen atom (of N-NO2 moiety). Furthermore, figure 3 clearly shows that decomposition of DMNA-Al follows a number of steps, which can finally render NO and NO2 eliminations. However, NO elimination channel shows the lowest energy pathway.
The first transition state (Ts1), in figure 3, suggests Al-O bond dissociation with an activation energy of 23 kcal/mole. This transition state is connected to an intermediate (Opt2) in the forward direction, which is again connected to a transition state (Ts2) for N-N bond dissociation. Following this transition state DMAN-Al complex can be converted to an isomeric form (Opt3). This two-step
0
20
40
60
80
100
En
erg
y(K
ca
l/m
ole
)
24
0
69
94
30
NO elimination
39
NO2 elimination
Opt1
Ts1
Ts2
Opt3
isomerization pathway of DMNA-Al complex is similar to a nitro-nitrite isomerization channel of isolated DMNA. Interestingly, the overall isomerization process is an exothermic process in the case of DMNA-Al complex and is endothermic process for isolated DMNA molecule. The energy barriers associated with NO elimination and the NO2 elimination with respect to Opt3 are estimated to be 13 and 43 kcal/mol, respectively, which undoubtedly indicates that the NO elimination following isomerization is the lowest energy pathway. Here note that NO2 elimination is predicted to be the lowest energy dissociation channel for isolated DMNA.
Figure 3: Decomposition pathway of DMNA
complex in ground electronic state, as revealed
at the MP2/6-31G(d) level of theory. The
relative energies are given in Kcal/mole.
Potential energy surface scan is
performed along the O1-Al1 bond to explore NO2 dissociation reaction coordinate starting from the Opt3 (see Figure 3) of DMNAcomplex. Scan results are depicted in Figure 4. This figure shows that the energy increases with the increase of O1-Al1 bond length; however, after certain O1-Al1 bond distance (~2.8 Å), energy starts decreasing due to shortening of the other O2-Al2 bond distance. Thus, in attempt to remove NO2 moiety, we have found that, due to strong interaction of Al and O atoms, NO
Al complex is nitrite isomerization channel of
isolated DMNA. Interestingly, the overall isomerization process is an exothermic process
Al complex and is endothermic process for isolated DMNA
barriers associated with elimination with
respect to Opt3 are estimated to be 13 and 43 kcal/mol, respectively, which undoubtedly indicates that the NO elimination following isomerization is the lowest energy pathway.
elimination is predicted to be the lowest energy dissociation channel for
Figure 3: Decomposition pathway of DMNA-Al
complex in ground electronic state, as revealed
31G(d) level of theory. The
relative energies are given in Kcal/mole.
Potential energy surface scan is bond to explore
dissociation reaction coordinate starting from the Opt3 (see Figure 3) of DMNA-Al complex. Scan results are depicted in Figure 4.
his figure shows that the energy increases with Al1 bond length; however,
Al1 bond distance (~2.8 Å), energy starts decreasing due to shortening of the
Al2 bond distance. Thus, in attempt to e have found that, due to
strong interaction of Al and O atoms, NO2
elimination is unlikely to occur on the ground electronic state surface: this conclusion is, however, reflected clearly in Figure 3, in which NO elimination pathway is shown to be energetically the most favorable process as compared to the NO2 elimination.
Figure 4: Potential energy surface scan from
Opt3 along O1-Al1 bond elongation reaction
coordinate for DMNA-Al complex
DMNA-Zn: The DMNA-Zn complex shows geometry similar to that of DMNAwhere the metal Zn is located below the Nplane, rendering non-planar NFigure 5 shows decomposition pathway of DMAN-Zn complex and relevant relative energies are given in Table I, as predicted at MP2/6-31G(d) level of theory. As illustrated in Figure 5, first, DMNA-Zn complex can follow a transition state associated with Ndissociation from Opt1, which results in a very stable geometry dimethyl (µ2
complex. This geometry is stabilized by 44 kcal/mole because of formation of two stable Zn-O bonds. Thereafter, the complex undergoes Zn-O bond dissociation which leads to the nitronitrite isomerization followed by NO elimination. The isomerized produexothermic in nature compared to starting geometry both for DMNA-Al and DMNAcomplex. From the energy profile in Figure 5 it is evident that the NO elimination is endothermic for DMNA-Zn system.
elimination is unlikely to occur on the ground electronic state surface: this conclusion is, however, reflected clearly in Figure 3, in which NO elimination pathway is shown to be
cally the most favorable process as elimination.
Figure 4: Potential energy surface scan from
Al1 bond elongation reaction
Al complex.
Zn complex shows geometry similar to that of DMNA-Al complex, where the metal Zn is located below the N-NO2
planar N-NO2 moiety. Figure 5 shows decomposition pathway of
Zn complex and relevant relative in Table I, as predicted at
31G(d) level of theory. As illustrated in Zn complex can follow a
transition state associated with N-N bond dissociation from Opt1, which results in a very
2-nitro) amine zinc complex. This geometry is stabilized by 44 kcal/mole because of formation of two stable
O bonds. Thereafter, the complex undergoes O bond dissociation which leads to the nitro-
nitrite isomerization followed by NO elimination. The isomerized product is exothermic in nature compared to starting
Al and DMNA-Zn complex. From the energy profile in Figure 5 it is evident that the NO elimination is
Zn system.
Figure 5: Decomposition pathway of DMNA
complex in ground electronic state, as revealed
at the MP2/6-31G(d) level of theory. The
relative energies are given in Kcal/mole.
Comparison: Energy profile diagrams for both DMNA-Al and DMNA-Zn are illustrated together in Figure 6 for comparison. Both complexes exhibit non-planar N-NOThe starting geometries with different bond angles and bond distances are given in the figure 6. The initial steps of decomposition of these two complexes, however, are different. The first transition state for DMNA-Al is associated with Al-O bond dissociation, whereas, that same for DMNA-Zn is N-N bond dissociation. In spite of an exhaustive search of Zn-O bond dissociation transition state, we have not found any Zntransition state for DMNA-Zn, similar to DMNA-Al. This difference leads to different geometries at Opt2 state for these two complexes. However, both complexes, in the end, are showing nitro-nitrite isomerization and NO elimination as the ultimate decomposition product but their pathways to reach the products are different. DMNA-Al exhibits Albond dissociation followed by Ndissociation which results in nitroisomerization product. On the other hand, presence of transition metal like Zn, Ndissociation becomes the first step, yielding dimethyl (µ2-nitro) amine zinc complex. Then it undergoes Zn-O bond dissociation which leads to the nitro- nitrite isomerized product and
Figure 5: Decomposition pathway of DMNA-Zn
mplex in ground electronic state, as revealed
31G(d) level of theory. The
relative energies are given in Kcal/mole.
Energy profile diagrams for both Zn are illustrated
together in Figure 6 for comparison. Both NO2 moieties.
The starting geometries with different bond angles and bond distances are given in the figure 6. The initial steps of decomposition of these two complexes, however, are different. The first
Al is associated with O bond dissociation, whereas, that same for
N bond dissociation. In spite of O bond dissociation
transition state, we have not found any Zn-O Zn, similar to
Al. This difference leads to different geometries at Opt2 state for these two complexes. However, both complexes, in the
nitrite isomerization and NO elimination as the ultimate decomposition product but their pathways to reach the ultimate
Al exhibits Al-O bond dissociation followed by N-N bond dissociation which results in nitro-nitrite isomerization product. On the other hand, presence of transition metal like Zn, N-N bond
st step, yielding nitro) amine zinc complex. Then it
O bond dissociation which leads nitrite isomerized product and
finally NO elimination occurs. The isomerization is exothermic in nature for DMNA-Al and DMNA-Zn comthe subsequent NO elimination is exothermic for DMNA-Al but endothermic for DMNAsystem. Table1: All the energies (Kcal/mole) are
calculated with respect to the energy of Opt1, at
the MP2/6-31G (d) level of theory. Ts1 for
DMNA is N-N bond dissociation and for DMNA
Zn but for DMNA-Al is Al-O bond dissociation
Whereas Ts2 is N-N bond dissociation for
DMNA-Al and Zn-O for DMNA
Critical
Points
DMNA
DMNA
Al
Opt1 0
Ts1 94
Opt2 24
Ts2 _
Opt3 _
NO
elimination
39
NO2
elimination
30
finally NO elimination occurs. The isomerization is exothermic in nature for
Zn complexes; however, the subsequent NO elimination is exothermic for
Al but endothermic for DMNA-Zn
All the energies (Kcal/mole) are
calculated with respect to the energy of Opt1, at
31G (d) level of theory. Ts1 for
N bond dissociation and for DMNA-
O bond dissociation
N bond dissociation for
O for DMNA-Zn complex.
DMNA-
Al
DMNA-
Zn
0 0
23 12
21 -44
33 -26
-20 -32
-7 28
23 40
Figure 6: Comparative study of the
decomposition mechanisms of DMNA
DMNA-Zn complexes in ground electronic state,
as revealed at the MP2/6-31G(d) level of theory.
The relative energies are given in Kcal/mole.
Conclusions
In the work presented in this article, the ground state decomposition reaction pathways are calculated for isolated DMNA, DMNAand DMNA-Zn complexes at the MP2/6level of theory. Theoretical results suggest that DMNA can follow N-NO2 bond dissociation as well as nitro-nitrite isomerization pathways; however, N-NO2 bond dissociation pathway is associated with the lowest activation energy barrier. DMNA-Al complex, on the other hand, shows significantly different multidecomposition pathway: first, Aldissociation, then N-N bond dissociation followed by an isomerization and finally NO elimination. Our calculations also show that NO elimination reaction pathway (followed by nitronitrite isomerization) of isolated DMNA is energetic unfavorable at the ground electronic state; but similar isomerization is favorable for DMNA-Al complex. Therefore, the decomposition behavior of nitramine en
Figure 6: Comparative study of the
decomposition mechanisms of DMNA-Al and
Zn complexes in ground electronic state,
31G(d) level of theory.
The relative energies are given in Kcal/mole.
In the work presented in this article, the ground state decomposition reaction pathways are calculated for isolated DMNA, DMNA-Al,
es at the MP2/6-31G(d) level of theory. Theoretical results suggest that
bond dissociation as nitrite isomerization pathways;
bond dissociation pathway is associated with the lowest activation energy
Al complex, on the other hand, shows significantly different multi-step decomposition pathway: first, Al-O bond
N bond dissociation followed by an isomerization and finally NO elimination. Our calculations also show that NO
imination reaction pathway (followed by nitro-nitrite isomerization) of isolated DMNA is energetic unfavorable at the ground electronic state; but similar isomerization is favorable for
Al complex. Therefore, the decomposition behavior of nitramine energetic
moiety is predicted to be altered completely in presence of Al atom. Furthermore, overall reaction of DMNA decomposition is found to be endothermic; however, that of DMNAcomplex shows overall exothermic reaction. Similar behavior can be anticipcomplex metalized nitramine energetic materials.
Now by changing the metal atom from a non-transition metal aluminum to a transition metal zinc what change can we anticipate in the decomposition of DMNA? DMNAalso shows isomerization and finally NO elimination in concordance with DMNAhowever, on the contrary with DMNADMNA-Zn complex shows first Ndissociation and then metal-O bond dissociation. Furthermore, the NO elimination process for DMNA-Zn system is found to bnature as compared to DMNApresent results show that only metal atoms can alter the decomposition of nitramine moieties, but also nature of metal atoms can have significant effect on the decomposition of nitramines. Our next goal is to experimentally investigate these changes using ultrafast spectroscopy.
References
[1] (a) Fried, L.E.; Manaa, M.R.; Pagoria,
P.F.; Simpson, R. L. Annu. Re
Res., 31 (2001) 291-321. (b) Tilton, T.M.; Gash, A.E.; Simpson, R.L.; Hrubesh, L.W.; Satcher, J. H.; Poco, J.FJ. Non Cryst. solids, 285 (2001) 338
[2] Ramaswamy, A. L. Expansion and Shock waves
131-137. [3] (a) Politzer, P.; Lane, P.; Grice, M. E.
Phys. Chem. A., 105(2001)(b) Swihart, M. T.; Catoire, L. Flame, 121 (2000), 210-222.
[4] Snyder, E.P; Seltz, H. J. Am. Chem.
Soc., 67 (1945) 683-685. [5] Guo, Y. Q ; Greenfield, M
Bhattacharya, A ; Bernstein, E. R. Chem. Phys. 127 (2007) 154301
moiety is predicted to be altered completely in presence of Al atom. Furthermore, overall reaction of DMNA decomposition is found to be endothermic; however, that of DMNA-Al complex shows overall exothermic reaction. Similar behavior can be anticipated for more complex metalized nitramine energetic
Now by changing the metal atom from a transition metal aluminum to a transition
metal zinc what change can we anticipate in the decomposition of DMNA? DMNA-Zn complex
on and finally NO elimination in concordance with DMNA-Al; however, on the contrary with DMNA-Al,
Zn complex shows first N-N bond O bond dissociation.
Furthermore, the NO elimination process for Zn system is found to be endothermic in
nature as compared to DMNA-Al system. Thus present results show that only metal atoms can alter the decomposition of nitramine moieties, but also nature of metal atoms can have significant effect on the decomposition of
t goal is to experimentally investigate these changes using ultrafast
(a) Fried, L.E.; Manaa, M.R.; Pagoria, Annu. ReV. Mater.
321. (b) Tilton, T.M.; Gash, A.E.; Simpson, R.L.;
W.; Satcher, J. H.; Poco, J.F.
285 (2001) 338-345 Combustion,
Expansion and Shock waves., 36(2000)
(a) Politzer, P.; Lane, P.; Grice, M. E. J.
105(2001) 7473-7480. (b) Swihart, M. T.; Catoire, L. Combust.
222. J. Am. Chem.
; Greenfield, M ; ; Bernstein, E. R. J.
127 (2007) 154301-10.
[6] (a) Stewart, P. H.; Jeffries, J. B.; Zellweger, J. M.; McMillen, D. F.; Golden, D. V. J. Phys. Chem., 93 (1989) 3557. (b) Nigenda, S. E.; McMillen, D. F.; Golden, D. M. J. Phys. Chem. , 93 (1989) 1124. (c) Lloyd, S. A.; Umstead, M. E.; Lin, M. C. J. Energ. Mater., 3 (1985) 187. (d) Flournoy, J. M. J.
Chem. Phys., 36 (1962) 1106. (e) Lazarou, Y. G.; Papagiannakopoulos, P. J. Phys. Chem. 94 (1990) 7114. (f) Lazarou, Y. G.; Papagiannakopoulos, P. Laser Chem. 13 (1993) 101.
[7] (a) Velardez, G. F.; Alavi, S.; Thompson, D. L. J. Chem. Phys. 123 (2005) 074313. (b) Sumpter, B. G.; Thompson, D. L. J. Chem. Phys. 86 (1987) 3301. (c) Sumpter, B. G.; Thompson, D. L. J. Chem. Phys. 88(1988) 6889.
[8] Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. A., Jr.; Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin, K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; e Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, N. J.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, Ö.; Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J.A. Gaussian 09,
Revision D.01,Inc,Wallingford CT,(2009)
[9] (a) Bhattacharya, A.; Guo, Y. Q.; Bernstein, E. R. J. Phys. Chem. A 113 (2009) 811. (b) Bhattacharya, A.; Guo, Y. Q.; Bernstein, E. R. Acc. Chem. Res. 43 (2010) 1476-85.
Investigation of spatial and temporal dynamics of vegetation
in semi-arid-ecosystems11
Sumithra Sankaran, Sabiha Sachdeva, Ashwin Viswanathan and Vishwesha
Guttal (PI)
Centre for Ecological Sciences, Indian Institute of Science, Bangalore, 560012
Abstract:Semi-arid ecosystems can exhibit striking vegetation patterns, which may have no characteristic size of patchiness. Elucidating local scale processes that generate these macroscopic patterns is of fundamental ecological importance as well as in being able to forecast the future dynamics of these highly vulnerable ecosystems. Recent studies have shown that, these patterns can be explained by local facilitative interactions with global competition for resources and have suggested that these patterns can be considered signatures of a system’s resilience. However, no alternative models that could potentially explain similar patterns have been investigated. In this work, we compare and contrast the predictions of a spatially explicit model of vegetation in which both, facilitation and competition, occur at local scales with that of the previous models. We characterize spatial patterns by using the distribution of patch sizes and power spectrum analyses. We show that power-law distribution of patch sizes can result from local competitive constraints combined with local facilitative interactions, that a global scale of competition is not necessary. Further we also find that power law patch size distributions are not necessarily indicative of high resilience but can be even observed in systems approaching regime shifts, depending on the scale of underlying interactions. We find no means of discerning
1Project number: STC/P‐315.
the scale of processes in a system based on results from power spectrum analyses. I. INTRODUCTION Semi-arid ecosystems support critical flora, fauna together with large livestock and human populations. Under threat due to increasing anthropogenic activities and climate change, they are prone to regime shifts, i.e., abrupt changes in their states and thus, potentially leading to desertification [1]. Vegetation of these ecosystems may exhibit striking self-organized patterns ranging from regular periodic patterns such as spots, labyrinths to irregular “scale-free patchiness”. Recent studies based on mathematical models and computer simulation studies argue that spatial patterns of vegetation in semi-arid ecosystems may offer signatures of resilience and its vulnerability to perturbations [2, 3. 4, 5]. In this context, understanding the local processes that generate large scale self-organized spatial patterns in semi-arid ecosystems and how these patterns are affected by stressors are questions of both basic and applied ecological significance. Many ecological interactions between species of the same trophic level (such as between different plants) can be broadly classified into two fundamental types‚ competition and facilitation. The relative scales over which interactions of each type are operational could potentially determine the nature of the spatial patterns observed in the landscape (Levin 1992).Some of the early work in this area, in late 1990’s and early 2000’s, developed various mathematicalmodels to explain these vegetation patterns. A key ecological assumption of these models was that of local positive feedback that helps
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
increased germination of seeds in the vicinity of other plants because of reduced evaporation and/or increased infiltration of water. However, competition sets in at local yet relatively longer spatial scales since plants draw water towards them from nearby regions. Two key studies[3, 4] published in 2007 then showed, using computer simulation models, that scale-free distribution of patchiness in vegetation can be explained by assuming local or near-neighbor facilitative interactions amongst plants in combination with a global scale of competition (motivated by studies showing that rainfall constrains global control of tree cover). Although the competition in resource limited regions such as semi-arid ecosystems can indeed extend to long distances, the assumption of a global scale constraint seems unrealistic. Moreover, analyzing a single metric of spatial pattern suchas patch-size distribution, could potentially be misleading; for instance, we know from the literature on percolation transitions in physics that patch sizes can exhibit scale-free features even in an entirely non-interacting system.Given the implications of such spatial patterns on the stability of these systems, we ask: (1) Can local processes (such as
competition for resources and facilitation among nearby plants) alone, in the absence of global constraints, explain scale-free patchiness of vegetation patterns found in semi-arid ecosystems?
(2) Can we discern scale of interactions by analyses of spatial patterns, for example those available from aerial imagery?
II. METHODS We analyse two cellular automata based spatially explicit stochastic models, differing in their interaction rules; specifically, the length scales over which competitive interactions operate. We compare various characterizations of spatial patterns resulting from these two models, henceforth called Kéfi’s and Lubeck’s models, along their respective phase diagrams.Both models assume a discrete space where each lattice cell represents a small spatial scale that could be occupied by an individual (plant). Each cell has an associated probability of switching from one state to another, such as plant or no-plant state. Competition and facilitation are captured by making these transition rates/probabilities conditional on the states of other cells in the lattice. Note that the two models are qualitatively similar in that the average density/vegetation cover as a function of birth ratecan show discontinuous or continuous transitions (for high and low death rates respectively) from vegetated to barren state. Kéfi’s model: In this model, meant to capture the biology of semi-arid ecosystems, there are three possible states for each cell: vegetated, empty and degraded. Local facilitation interaction is incorporated via a transition rate of degraded to empty-inhabitable cells which positively depends on the proportion of vegetated cells in the neighbourhood of the focal degraded cell. Competition is assumed to occur on a global scale and is incorporated by making the transition rate (denoted by ‘b’) of empty to vegetated cells depend inversely on the mean global vegetation density in the lattice. ‘b’ is also
constrained by other biological factors such as aridity of the landscape. Mortality rate (denoted by ‘m’) is the rate at which vegetated cells transition to occupied cells. It is the density independent parameter and captures biological factors such as extent of grazing in the landscape. (See [3] for detailed description of the model and its parameters). Lubeck’s model: This model, also known as tri-critical directed percolation model and based on stochastic process popularly known as contact process in the Mathematics literature, was developed in the Physics literature to investigate nonequilibrium phase transitions. Here, we employ the model with ecological interpretation of its rules. Each lattice cell can occur in two possible states corresponding to occupied and unoccupied by a plant. At each discrete time step, a randomly chosen vegetation site becomes empty with a probability (denoted by p), or an empty cell in the vicinity of that vegetated site undergoes a transition to a vegetated state with another probability. Positive local feedbacks are incorporated in this model through the increased probability (denoted by q) of transitioning from unoccupied to occupied cell as a consequence of the neighbourhood of the parent cell. Competition plays out through the reduction in available (empty) cells nearby the focal plant for propagation of its seeds as the overall density increases. In simple terms, this model assumes local facilitation and local competition to determine vegetation growth. (See [7] for detailed description of the model and its parameters).
We simulate these models in two dimensional discrete space of 1024 x 1024 grids. We use power spectrum analyses and calculate patch size distributions (using Maximum Likelihood Estimate methods) to characterize the resulting spatial patterns. III. RESULTS The vegetation patterns realised from Kefi’s model, across the parameter space (all values of ‘m’), show a power law distribution of patch sizes away from the transition (high values of ‘b’), which becomes increasingly truncated as the system approaches the critical point (or as ‘b’ is decreased) [as already shown by 3,4]. Vegetation patterns realized from Lubeck’s model also show power law patch size distributions, however, only when parameter values correspond to a discontinuous regime shift (q=0.91) and only when the system is very near the critical point (b=0.2951) (Fig. 1; Table 1). This shows that only local processes can also explain scale-free patch size distributions. Additionally it also challenges the reliability of patch size distributions as an indicator of approaching regime shifts.
Fig.of pmodusinmet
Tabthe real
ThespatquanspatrandcontwheshowpatcpatcSpefromfollocasethe shif
. 1: Inversepatch sizesdels. All pang Maximuthod.
ble 1: (MLEbest fit di
lized from th
e power sptial patternsntifies contial perioddom pattetributions ereas a periow a wavelech-size anches. ectral Analm both modow an expe of paramecontinuous
fts).
e Cumulativs realized arameters w
um Likeliho
E) Parameteistribution he two mod
pectrum des into periodntributions
dicities (waern would
from alodic patchy
ength corresnd/or dista
yses of spdels (Fig. 2)ponential deters corresp
and discon
ve Distribufrom the
were estimaood Estima
er estimatesof patch s
dels
ecomposes dic patterns
of diffeavelengths)
have eql frequen
y system wosponding toance betw
patial patte) were foundistribution ponding to bntinuous reg
ution two ated
ation
s for sizes
the and
erent A qual
ncies ould
o the ween
erns nd to
(in both gime
Fig. 2: Potwo modcorrespondshifts. Thediscontinuoto be simila
IV. DISCU Spatial chdepend on close to or shown by system is discontinuoof theseinteraction in patch sizmoves towon the scsystem.Kefdistributionwhile Lubedistributionubiquitous distributionThis challesize distriapproachin It has beenfree patccharacteristlandscapescombinatiointeractionsconstraintsthat this p
wer spectrudels for ding to ce (qualitativous regimear.
USSION
haracteristicboth, whefar from th
[3,4]), as wundergoin
ous regime two modescales, sho
ze distributwards a trancale of infi’s model n away fromek’s model n near the cr
pattern ns away frenges the rbutions as
ng regime sh
n widely acch size tic of s, result spe
on of s with gl. Our workattern of sp
um results parameter ontinuous ve) results shifts wer
cs of a ether the syhe critical pwell as wheng continushift. Our a
els with dows that thetions as thesition is de
nteractions shows a scm the criticshows a scritical point
of patchrom the trareliability os an indichifts.
ccepted thadistribution
several seecifically fr
local lobal comk, howeverpatial organ
for the values regime of the
e found
system ystem is point (as ether the uous or analyses different e change e system pendent in the
cale free cal point cale free t and no h size ansition. of patch cator of
at “scale ns”, a emi-arid rom the positive
mpetition , shows nization
can also from only local scales of competition coupled with local positive feedbacks. Further, the scale of interactions in the system can also not discerned from power spectrum analyses of the spatial organization of the vegetation. We therefore conclude that it may not be easy to distinguish the origin or details of processes by analyzing spatial vegetation patterns with these metrics alone. We speculate that temporal dynamics of spatialpatterns may yet offer signatures of interaction scales in the system as the response to perturbations of a system with global competitive constraints will, in all likelihood, differ from that with solely local constraints. It is also highly probable that natural ecosystems, due to continuous disturbances, are seldom found in their steady states. We now plan to study the non-steady state behaviours of Kefi and Lubeck’s models to investigate the effect of noise on the spatial patterning of vegetation in ecosystems with varying scales of processes.
References
[1]Scheffer et al, 2001, Catastrophic regime shifts in ecosystems, Nature, Vol 413, 591-596.
[2] Rietkerk, et al, 2004, Self-organized patchiness and Catastrophic Shifts in Ecosystems, Science, Vol 305, 1926-1929.
[3] Kefi et al, 2007, Spatial vegetation patterns and imminent desertification in Mediterranean aris sytems, Nature, Vol 449, 213-218.
[4] Scanlon et al, 2007, Positive feedbacks promote power-law clustering of Kalahari vegetation, Nature, Vol 449, 209-213.
[5] Guttal & Jayaprakash, 2009, Spatial variance and spatial skewness: leading indicators of regime shifts in spatial ecological systems, Theoretical Ecology, 2: 1-11.
[6] Kefi, Guttal, et al, 2014, Early warning signals of ecological transitions: Methods for spatial patterns, Plos ONE., Vol 9: e92097.
[7] Lübeck 2006, Tricritical directed percolation, Journal of Statistical Physics, Vol 123,193-224.
ISTC/CSS/AP/312 Project Status Report
Development of Semiconductor Nanocrystals for Photovoltaics
Biswajit Bhattacharyya, Rekha Mahadevu and Anshu Pandey
Abstract: Photovoltaics require the
development of Semiconductors with
precisely tuned optical properties. Here
we report the development of materials
with very precise tuning of radiative
lifetimes and anomalous band gaps. The
synthesis of this novel class of materials
will motivate precise engineering of
material properties in order to optimize
device performance.
INTRODUCTION
Control over emission rates is of
fundamental importance to a number of
areas of research and technology. Long
spontaneous decay rates are extremely
desirable in photovoltaics. A number of
strategies to regulating the rates of
spontaneous decay already exist. Most of
these techniques –plasmonics, photonic
crystals and cavities, high dielectric media,
etc. – exclusively focus on tuning photon
density of states. Advances in colloidal
chemistry have enabled the synthesis of
semiconductor quantum dots with a wide
range of formulations, compositions and
morphologies. This flexibility naturally
provides several potential routes for
manipulation and control over the physical
properties of these materials. While
considerable attention has been devoted
towards utilizing QD structure to control
properties such as blinking, multiexciton
decay, luminescence efficiency, etc., the
deliberate use of structure to manipulate
radiative decay dynamics of single
excitons remains an untapped possibility.
In this article we demonstrate that dopants
may be used to regulate the density of
states available for excitonic
recombination. This provides a very
straightforward route to regulating the
radiative rates of carrier recombination in
QDs.
Copper is a known substitutional
impurity in II-VI QDs. When introduced
into II-VI materials such as ZnSe and
CdSe, Copper is known to exclusively
exist in a +2 oxidation state. The 3d9
configuration of Cu+2 substitutional
impurities has been demonstrated to
exhibit exchange with semiconductor host,
causing the emergence of diluted magnetic
semiconducting character in doped QDs.
Being Jahn-Teller ions, substitutional Cu+2
species are EPR (electron paramagnetic
resonance)–silent due to a combination of
spin-orbit and electron-lattice coupling. In
a II-VI lattice in particular, Cu+2 ions give
rise to an emission band that arises from
the radiative transition of a conduction
band electron into the copper d9 level.
From this level, the electron is believed to
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
ISTC/CSS/AP/312 Project Status Report
relax nonradiatively into the valence band
hole. This process is highlighted in Figure
1b. The CB-Cu radiative transition (thin
arrow) presents a significantly weaker
transition dipole than the band edge
excitonic decay (thick arrow). Both
processes are however emissive and lead
to the emission of a photon along with the
de-excitation of the quantum dot.
Despite its inherent weakness, the CB-Cu
transition is behaviorally very different
from other radiative decay channels
available within a quantum dot (figure 1c).
Regulation of the amount of copper doped
into a semiconductor lattice provides a
direct handle on the available density of
states for the radiative relaxation of the
conduction band electron. A typical copper
doped QD shows both band edge as well
as dopant related emission(Fig 2a). While
band edge emission typically occurs with a
lifetime of the order of 10 ns, the copper
emission band is typically seen to decay in
about 500 ns(Fig 2b). The doping levels
for the Cu+2 impurity in II-VI QDs
typically correspond to the presence of
one-two dopant ions in QDs.
QDs were characterized through electron
microscopy, X-ray diffraction (XRD) and
Inductively Coupled Plasma-Optical
Emission spectroscopy (ICP-OES).
Studies of QD composition by ICP-OES,
Figure 1 a. Spontaneous emission involves the relaxation of an excited electron to its ground state. Material properties like density of states do not play a role in the process. b. A copper containing QD has two alternate pathways for radiative exciton recombination; either direct recombination with a band edge hole or else through the copper center. c. Enhancing the number of copper atoms in the QD increases the number of available channels for radiative recombination, speeding up radiative decay.
revealed that these QDs were composed of
a cadmium-copper-zinc selenide alloy. By
varying the molar ratio of copper acetate to
zinc and cadmium acetates during
synthesis, the copper inclusion may be
tuned from 0% to 35%. Figure 3a shows
the XRD patterns of a sample containing
copper, Zinc and Cadmium (ICP-OES
analysis on cations basis). While the XRD
patterns are indeed consistent with a Zinc
Blende material, the positions of the peaks
of signal do not correspond to either pure
ZnSe or to pure CdSe.
Rather, the peak positions are intermediate
to ZnSe and CdSe, and correspond to a
semiconductor alloy. A quantitative
statement may be made using Vegard’s
law.
ISTC/CSS/AP/312 Project Status Report
Figure 2 a. Absorption and emission spectrum of a 1% doped Cu doped CdZnSe QD. b. Emission decay kinetics of the host and the copper band.
Copper and Zinc have similar ionic radii
and structure factors. This allows us to
write the lattice parameter of the alloy as a
weighted mean of Zinc Selenide and
Cadmium Selenide lattice parameters. This
leads us to expect a lattice parameter of
>0.5 nm for the alloy. Figure 3a also
presents a simulated pattern based on this
expectation. The excellent agreement
between the simulated pattern of the alloy
and the pattern that is observed in practice
as well as the absence of features
corresponding to potential impurity phases
such as CdSe, ZnSe or CuSe in our
patterns prove that these QDs are indeed
pure CdCuZnSe alloys. Figures 3b and 3c
show TEM images of QDs containing 1%
and 10% Cu (measured relative to the total
cation content). Consistent with the XRD
patterns, TEM images reveal a
homogenous distribution of particles, with
no evidence of extraneous phases. These
analyses thus suggest successful
substitution of a large fraction of the Cd
and Zn ions by copper.
We studied the electronic structure
of these materials under a probabilistic
model. In particular, the probability of
cluster formation is expected to determine
the electronic and optical properties of the
materials. The probability of formation of
such clusters may be evaluated in the case
of a completely random placement of
copper ions. In this situation, the
probability of occurrence of two nearest
neighbor copper ions is proportional to x2,
where x is the mole fraction of copper. In a
similar manner, the probability of
occurrence of a cluster of size s is
proportional to xs.
Of significant importance for our
discussion is the cluster connectivity or
degeneracy. Consider for instance the
simple case of a quasi-spherical
arrangement of copper ions. In such a
situation, if V is the volume of such a
cluster, then the radius of such a cluster is
related to V1/3. Of course, the number of
atoms in the cluster is related to V. i.e.
∝ . If we consider the addition of one
more atom into the cluster, the ways of
doing so are linked to the surface area of
the cluster, viz. V2/3 or /. When
dealing with mole fractions, one can
similarly relate N to x. This simple
argument illustrates two important points:
ISTC/CSS/AP/312 Project Status Report
firstly, the cluster formation probability is
dependent on the connectivity of a cluster
with one lesser atom. Secondly, this also
illustrates that for a spherical arrangement,
a cluster is expected to have a sublinear
dependence of available cluster
morphologies on the cluster size.
In any practical situation the
occurrence of a spheroidal cluster is
extremely unlikely. Instead, a cluster with
much higher surface area is expected
simply because of a high degeneracy and
more available conformations. In this
situation, one can write ∝
, where s is the size of the cluster. The
probability expression for a cluster of size
s is given by . We are of course
interested in the total probability of
formation of any sized cluster. This is
given by = ∑ ~ ∫
.
This expression leads to a series of the
form. =
−
() + ⋯
. For
sufficiently small x, only the first term
contributes. It is further evaluated to be
~[() − 2]
~
.
Therefore, for small x, the total probability
of the occurrence of a cluster of any size is
related to
. Any property that is affected
by the formation of clusters of copper may
therefore be expected to have the
dependence +
on the mole
fraction. Here A and B are constants that
relate the variation of a particular property
to the probability of a cluster. This
expression is used to generate the fit to
experimental fluorescence data.
In most samples the decays are
adequately described by biexponentials. A
mean photon lifetime may be extracted by
considering the areas under various parts
of the PL decay curve. We therefore define
the photon lifetime by ∑ ∗
∑ =
∑
∑ . Here n
are the number of photons under a
particular component of the fit, while t is
the associated time constant. An increase
in the number of copper centers causes a
direct shortening of sample lifetimes. By
changing the copper inclusion levels from
1% to 31% the lifetimes from 600 ns to 1
ns.
In most systems the exciton
lifetime arises from the interplay of
radiative and nonradiative rates. Simply
increasing the nonradiative decay channels
in any system therefore also speeds up
exciton decay, although of course, this is
of little practical or scientific value. We
therefore turn to investigate the actual
causes of speeding up of the excitonic
decay rates in our systems.
ISTC/CSS/AP/312 Project Status Report
Figure 3. a. Observed and Simulated XRD
patterns of CdCuZnSe QDs along with
standard patterns of CdSe (top) and ZnSe
(bottom). b. TEM micrograph of QDs
with 1% and c. 10 % Cu. Scale bars are 20
nm.
For example, consider the PL
kinetics of a sample with a >30% inclusion
of copper. The sample exhibits a quantum
yield of 17 %, and an exciton lifetime of
1.4 ns. The sub-unity quantum yield could
possibly originate from the presence of
non-emissive QDs in the ensemble or else
from the existence of a non-radiative decay
mechanism in all the QDs. In the former
event, the spontaneous lifetimes are simply
equal to the observed excitonic lifetimes.
Even in the latter case, with a 17%
quantum yield, the spontaneous emission
lifetime is estimated to be 8 ns. While the
actual spontaneous lifetime is expected to
lie somewhere between these two
estimates, even the upper bound is two
orders of magnitude lower than the
typically observed lifetimes in copper
containing QDs. We thus confirm that our
procedure represents the development of a
route to tune spontaneous emission
processes in QDs.
The procedure presented thus far
provides a straightforward recipe for
shortening spontaneous emission lifetimes
in QDs. We now turn to a method to
lengthen the same. While shortening
lifetimes relied on the propensity of
individual copper ions to offer alternate
decay channels, longer lifetimes may be
attained by regulating electron-copper
overlap. We start with ZnSe QDs where
copper ions are doped into the QD core.
An electron extracting type-II CdS shell is
now grown over the core. This leads to
reduced overlap between the conduction
band electron and the copper impurity,
causing increase in the copper lifetimes.
Figure 4a provides a schematic to describe
this effect. Aliquots were obtained during
the process of shell growth.
We analyzed sample composition
in each case and observed no variation in
Copper:Zinc ratio, indicating that no
copper escaped from the QDs during the
growth process. Figure 5b shows the
variation of the emission kinetics of these
materials as a function of energy of the
emission maximum. As anticipated from
ISTC/CSS/AP/312 Project Status Report
the straightforward arguments relating to
electron-impurity overlap, we observe that
the emission slows down substantially.
Figure 4a. Schematic of the change in
impurity-electron ovelap in a core/shell
structure. b. Variation of exciton lifetimes
as a function of emission maximum. c. A
combination of these approaches allow us
to tune the emission maxima of these
materials from 1 ns to over 1 s.
Growth of CdS over the Zinc Selenide
core allows for tuning of the emission
lifetime from 650 ns to 1600 ns. We also
note that a shifting of the emission
maximum can by itself cause the slowing
down of emission. The estimated
magnitude of this effect is represented by a
dashed line in Figure 4b. The variation in
excitonic lifetimes clearly occurs both due
to a red shifting of the emission band as
well as the reduced carrier-impurity
overlap. The methods suggested here
nevertheless allow us to develop materials
that show a remarkable tuning of excitonic
lifetimes, from 1 ns to over 1600 ns
(Figure 4c).
CONCLUSIONS
The development of custom
chromophores and sensitizers is of
paramount importance for the next
generation of devices. Copper based
materials are of particular significance due
to the recent demonstration of feasibility of
these substances in charge transfer
processes. By employing dopants, we have
been able to significantly enhance the
concepts of idealized photovoltaic
chromophores that we have developed
during the course of this project. Future
studies will aim to apply these concepts in
practical devices containing ZnTe/CdS and
doped materials.
REFERENCES
(1) Bhattacharyya B. et. al. (Submitted)
(2) Mahadevu, R.; Yelameli, A. R.; Panigrahy, B.; Pandey, A., Controlling Light Absorption in Charge-Separating Core/Shell Semiconductor Nanocrystals. ACS Nano 2013, 7, 11055-11063.
Die Level 3D Packaging of Hybrid Systems N. P. Vamsi Krishna and Prosenjit Sen
Centre for Nano Science and Engineering, Indian Institute of Science, Bangalore
Abstract – The aim of this work1 is to develop and optimize design rules and processing technologies required for 3-D die level packaging of hybrid systems including MEMS components. In this paper we report the development of the dielectric filler process which works as the intermediate layer and isolates the various chips. We also report the initial results in using conductive printing to obtain electrical interconnects between layers.
I. INTRODUCTION
Transistor scaling has led to ever smaller chips with larger device density in integrated circuits. The reduction in size is reaching a saturation where further reduction of device dimensions is limited by fundamental problems. Further reduction in system weight and volume is possible through realizing that the existing technologies arrange dies in a horizontal fashion on either a PCB or a multichip module [1-2]. The semiconductor industry is seeking solutions in 2.5D-3D systems using vertical integration of devices using thru-silicon via (TSV) technology. The technology being developed is however expensive and available only to the big fabrication facilities. The technology that has been developed also does not accommodate MEMS devices with moving parts, which often have more severe restrictions in packaging specifications. In order to address the above issues we propose to develop processing technologies which will allow system developers to access 3D packaging through post processing on individual dies. The technology developed will also seek to address issues related to integration of MEMS devices
1 Project#: ISTC0336 (2014-2017)
with moving parts. An example of such a 3D packaged hybrid system is shown below.
Figure 1: A 3D integrated hybrid system
II. APPROACH AND RESULTS
The proposed technology for optimal vertical stacking will require the thinning of dies from standard 500-600 um thickness to less than 150 um thickness. This thinning process is expected to have no effect on the performance of electronic chips. For MEMS devices with structural layer less than 100 μm this thinning process will not change its performance. The die thinning can be carried out using back-grind and chemical mechanical polishing process. Wafer and die thinning process is a well-established process in industry and we will bypass this step by fabricating our dummy dies on thinned wafers.
Figure 2: Dielectric filler processing
The process starts with performing a die attach to a handle substrate. The handle substrate can be pre-patterned to provide interconnects as shown in Figure 3. This layer also contains the alignment marks for subsequent die attach steps.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Figure 3: Layer 1 mask for handle substrate
A 150 μm thick cover slip is used as a dummy die in the current process. The dummy die is also patterned using lift-off or etching to form the metal contact pads, devices and interconnects. The die is then aligned and bonded to the handle layer using UV curable epoxy EPO-TEK- UJ 1190. The aligned wafers are exposed to an UV lamp to harden the epoxy.
Figure 4: Wafer temperature under the UV lamp
Figure 4 shows the rise in temperature of the wafer during the epoxy exposure step. The temperature remains low at less than 70 °C.
Figure 5: Handle with bonded die.
After the bonding a spin coated dielectric is used to fill and planarize the substrates. For this project we are currently using SU-8 2150 [3]. The SU-8 is spin coated at 3000 rpm to coat and planarize the 150 μm. The SU-8 is soft baked at 65 °C for 7 mins and at 95 °C for 15 mins. The temperature was ramped to reduce stress issues. After the soft bake the wafers were allowed relax at room temperature for 15 mins.
Figure 6: Spin coated dielectric filler SU-8
The wafers were exposed using a mask aligner to 450 mJ/cm2. This was followed by post exposure bake at 65 °C and 95 °C for 7 and 15 mins respectively. The exposed wafers were developed for 16 mins. The initial runs faced issues with poor adhesion of SU-8 on the substrate. This lead to SU-8 crack and peel of due to stress in the SU-8 layer as seen in Figure 7. This issue was solved by improving the adhesion of SU-8 to the glass surface using a 2 min RIE O2 plasma treatment.
Figure 7: Crack and peel off in the SU-8 layer.
Interconnections between the handle wafer and the dummy die can be made in several ways. Currently we are considering jet printing of a silver paint and screen printing of conductive paste. Figure 8 shows a SU-8 planarized sample in which interconnects have been printed using
the jetting of silver ink using a Diamatix [4] printer. As is seen in the image the lines are rough and at some places discontinuous. We are in the process of improving the printing process parameters such as number of drops, number of runs and the surface condition of the SU-8.
Figure 8: Interconnects fabricated using jest printing of silver ink.
III. CONCLUSIONS AND FUTURE WORK
We have been able to optimize the dielectric filler process that is essential in die level 3D assembly. We are currently optimizing interconnect printing technology. Following optimization of the interconnect technology we will work on demonstrating a 3 layer stack of interconnected devices. We also intend to explore several other dielectric filler materials. Inspite of several attempts we were unable to obtain photo-pattern-able polyimide due to export control. We intend to try PDMS as another option.
REFERENCES
[1].http://en.wikipedia.org/wiki/System_in_pack
age
[2]. http://www.amkor.com/go/SiP
[3].http://www.microchem.com/pdf/SU-
82000DataSheet2100and2150Ver5.pdf
[4].http://www.fujifilmusa.com/products/industrial_inkjet_printheads/
Project #ISTC 0317
1
Abstract- This paper describes the design and development of a
small volume liquid dispenser that is capable of real-time
measurement of the mass of liquid dispensed. These features of
the dispenser are enabled by appropriate design of the three
subsystems of the dispenser, namely, the droplet generation
sub-system, the droplet mass measurement sub-system and the
droplet deposition sub-system. The droplet generation system
comprises a micro-pipette with a silvered surface along with a
syringe. The droplet mass measurement system enables
measurement of the volume of liquid dispensed by measuring
the resulting shift in resonant frequency of the micro-pipette.
The droplet deposition system employs electro-hydrodynamic
pulling and comprises a 330V voltage supply. All the three
systems are designed and fabricated. The developed overall
system is demonstrated to generate, measure and dispense
liquid droplets of volume down to about 260pL.
I INTRODUCTION
Microdispensing is the technique of producing liquid
dosages in volumes less than one microlitre.
Microdispensing finds applications in drug delivery, inkjet
printing, cell biology research, microelectromechanical
systems (MEMS), lubrication in automobiles, precision
deposition of soldering paste in electronic circuits industry,
among others. The continuing miniaturization in almost all
technical areas drives a constant need to improve the
microdispensing technologies. Such requirements include
the need to dispense smaller amounts of adhesive, liquid, oil,
grease and a multitude of other media reliably, accurately,
both in dosage and placement, and within small time.
The precise positioning and quantity of the dispensed fluids
such as glue and reagents have a great influence on the
overall quality of the final product.
Dispensing can be broadly considered to be of two types,
namely, contact type dispensing and the non-contact type
dispensing. In contact dispensing, the drop forms at the exit
of a nozzle, and is deposited by contact while the drop is still
on the nozzle. In non-contact dispensing, the drop formed at
K. R. Sreejith and G. R Jayanth are with the Department of
Instrumentation and Applied Physics. e-mail:[email protected],
jayanth@ isu.iisc.ernet.in, Phone: 22933197.
the end of a nozzle is deposited on a target area that is
situated relatively far away from the nozzle. Thus, the drop
separates from the nozzle before it hits the target area.
Compared to non-contact type dispensing, contact type
dispensing has a few important disadvantages. Since the
nozzle which forms the droplet has to touch the target area,
the target could be damaged. This is especially true with
micro-fabricated parts, such as MEMS structures. Further,
since viscous liquid forms threads, reproducibility in the
amount of liquid dispensed liquid amounts is poor. Thus,
owing to increasing requirements in regards to cycle time
and accuracy in almost all areas of production, non-contact
dispensing has steadily gained in importance. The different
technologies available for dispensing liquids in a
non-contact manner include piezoelectric [1], direct liquid
displacement [2], and pyroelectrodynamic shooting [3].
Here too, an important requirement to deposit droplets in a
controlled manner is to measure the volume of the liquid
precisely. The technologies that have been employed to
measure mass or volume of liquid dispensed include optical
techniques [4] capacitive type measurement [5][6], and flow
sensor based measurement[7].
This paper reports the design and development of a
non-contact type droplet dispenser based on
electrohydrodynamic pulling. The dispenser is capable of
real time measurement of the mass of liquid dispensed and
thus, facilitates precise control of the mass of the liquid
deposited.
The rest of the paper is divided as follows: Section II
describes the principle and provides an overview of the
construction of the droplet dispenser. Section III describes
the development of each of the subsystems that constitute the
droplet dispenser. Section IV describes the results of liquid
dispensing.
II PRINCIPLE OF OPERATION
Fig. 1 is a schematic showing the construction of the
droplet dispenser. The experimental setup consists of a
syringe attached with a micropipette. The syringe can be
actuated to form a liquid drop at the tip of the micropipette.
An optical beam deflection system is employed to measure
the mass of the liquid droplet formed at the end of the syringe
A Small Volume Droplet Dispenser based on
Electrohydrodynamic Pulling
K. R. Sreejith, and G. R. Jayanth
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Project #ISTC 0317
2
in real-time. Finally, a high voltage power supply is used to
deposit liquid droplet on to the target.
Fig. 1: Schematics showing the construction of the droplet dispenser.
To generate a droplet, the piston of the syringe is actuated
using a suitable actuator, such as a piezo- or a thermal-
actuator, to produce a liquid drop of volume in the range
nanoliters-picoliters. In order to measure the mass of the
droplet formed, the micro-pipette is excited and the resulting
motion is measured by means of optical beam deflection.
The change in resonant frequency of the micropipette upon
formation of a droplet is employed to determine the mass of
the droplet. After the formation of a droplet of the desired
volume, a potential difference is applied between the droplet
and the target. The liquid droplet is attracted towards the
conductive target due to coulomb force of attraction and gets
deposited on the target. By controlling the duration of
application of potential difference, the quantity of liquid
dispensed can be controlled. While the target is expected to
be electrically conductive, droplets can also be deposited on
insulating targets by mounting them on a conductive
substrate. The design and development of the entire system
is described in Section III.
III DESIGN AND DEVELOPMENT OF THE DROPLET DISPENSER
Section II reveals that the overall droplet dispenser can be
divided into three sub systems, namely, the droplet
generation system, droplet mass measurement system and
the droplet deposition system. This section describes the
design and development of each of these subsystems.
A. Design and development of the droplet generation system
The main objective of the droplet generation system is to
generate liquid droplet of volumes in the range of nanoliters
to picolitres and subsequently facilitate their deposition. In
order to generate liquid droplets of such low volume, a
conventional syringe attached with micropipette is
employed. Since the volume of the liquid and the dynamic
characteristics of the micro-pipette are decided by its
geometry, the performance of the system depends on the
ability to adequately control the taper and tip diameter of the
micro-pipette during its fabrication process. Accordingly, a
micro-pipette puller was designed and developed in order to
achieve control over these parameters. In order to facilitate
deposition of the droplet by means of electrohydrodynamic
pulling, the surface of the micropipette was rendered
conductive by coating it with silver using Brashear’s
process. The optimization and subsequent evaluation of the
pipette puller is described in Section III A(a). The process
employed to coat the pipette is described in Section III A(b).
(a) Development of a pipette puller for fabrication of glass
micro-pipettes
(a)
Ball screw
Pulley
Timer belt
HolderChannel
(b) (c)
(d)
Fig. 2: (a) Schematic showing the construction of the micro-pipette puller.
(b)-(d) Photographs showing the construction of the pipette puller.
Fig. 2(a) is a schematic showing the principle behind the
operation of the pipette puller [8]. The pipette puller heats a
glass tube to a temperature hT over a length h such that
hT exceeds the softening temperature sT of glass. It is
subsequently pulled in opposite directions along the
longitudinal axis of the tube, viz., the X-axis, by means of
two motion stages. One stage is capable only of linear
motion along the X-axis and assists in pulling the pipette,
while the other is capable of three-axis in-plane motion, viz.,
linear motion along X- and Y-axes and angular motion about
Z-axis, and thereby assists also in bending the pipette. The
one-axis motion stage employs a guiding channel to restrict
the motion of the pipette holder along the axis of the channel,
viz., the X-axis. The three-axis stage and the one-axis stage
are connected by an inextensible wire such that, upon
actuation, the two stages can be retracted in opposite
directions at identical velocities. When the two motion
stages are pulled in opposite directions with velocity v, the
length p of the region that gets pulled is that whose
temperature exceeds sT . If the time constant of cooling of
the micropipette after emerging from the heated region is τc,
p is given by
Project #ISTC 0317
3
2 ln hp h c
s
Tv
T (1)
To control the taper of the pipette, the pulling velocity v is
controlled to be sufficiently small so that p h , or
equivalently that the pulling velocity satisfies
2 ln hh c
s
Tv
T . The resulting dependence of taper on
h is obtained by assuming that the pipette possesses nearly
uniform cross-sectional area A within the region h . Assuming that the cross section preserves its shape during
pulling, i.e., the ratio of the inner diameter to the outer
diameter din/d= η remains constant, the profile d(x) of the
outer diameter of the pulled micropipette is given by
d(x) = d0 exp(− x/ ℓh) (2)
If 'v' is the velocity with which the pipette is being pulled
and’t’ is the time of pulling, x = vt. Thus, the taper of the
micropipette is controlled by controlling the length h
through which the glass capillary tube is heated. To achieve
active control over the length of capillary tube that is being
heated, the capillary tube is passed through a ceramic tube.
By adjusting the position of the ceramic tube relative to the
coils, the number of heater coil turns that are exposed to the
glass capillary tube can be controlled. To control the tip diameter, the glass tube is pulled for a
duration t1 after which the heater is switched off. Continued
pulling fractures the glass tube in the middle, thereby
creating two micropipettes of identical diameters d1 given by
d1 ≈ d0 exp(−vt1/ ℓh). Thus, the time t1 at which the heater is to
be switched off in order to achieve a tip diameter d1 is given
by
1)0 1( / ) ln( /ht v d d (3)
In the experiments, the glass capillary was heated for 135
seconds and then pulled. Experiments were carried out with
various values of h ranging from 4mm to 16 mm. The
profile of the pipettes obtained in each case is shown in Fig
3(a). Figure 3(b) plots ln [d(x)/d0] as a function of x, where x
is the distance from the base of the micro-pipette and d0 is the
original diameter of the glass capillary, and shows that the
dependence is linear in each case. Further, the slopes of the
four lines decrease monotonically with increased h , in
accordance with the theoretical result in Eq. (2). Figure 4(a)
is a micrograph showing tips of diameter d1 ranging from
500 μm down to 20 μm, obtained for values of t1 ranging
from 10s to 25s. The figure demonstrates that the tip
diameter reduces monotonically with increased t1. Figure
4(b) plots ln [d1/d0] as function of t1 and demonstrates a
linear relationship between the two in accordance with the
theoretical result obtained in Eq. (3).
Fig 3. (a) Micrograph showing the profiles of pipettes pulled with different
values of h and (b) plot of ln [d(x)/d0] as function of the distance x.
Fig 4. (a)Micrograph showing the tip diameters of pipettes obtained for
different values of t1. In all cases h was set to be 16 mm. (b) Plot of ln [d1/
d0] as function of the time t1.
(b) Silver coating of the fabricated micro-pipettes
After pulling a pipette of the required taper and tip
diameter, the pipette is coated with silver in order to render
its surface conductive. Silver coating is done by means of
Brashear’s process. Brashear’s process facilitates
deposition of silver almost uniformly on any glass substrate.
It is a chemical process in which large surface area can be
coated in less than 15 minutes. The steps involved in
Brashear's process involve first dissolving equal amounts of
silver nitrate and sodium hydroxide in water. Next, the two
solutions are mixed so that a black precipitate of silver oxide
forms. Third, ammonia is added until the precipitate
re-dissolves. Fourth, sugar is added and stirred until it
dissolves in this solution. Finally, the article to be coated is
placed in this solution while the solution is gently heated.
This results in deposition of silver on the article. Fig. 5 is a
micrograph showing a micropipette coated with silver.
500μm Fig 5. Micrograph showing a micropipette coated with silver
Project #ISTC 0317
4
B. Droplet mass measurement system
In order to measure mass of the liquid droplet, the shift in
resonance frequency caused by the droplet is measured. The
undamped resonant frequency of the micropipette is a
function of its stiffness k and effective mass m and is given
by 0 k m . Thus, for a small change in mass m , the
corresponding change in resonant frequency is given by
0 m1
2 m
(4)
Thus, as m increases, the resonant frequency of the
micropipette gets reduced. If the density of the liquid is
known, it is possible to determine the volume of the droplet
formed at the tip of the micropipette. In the proposed system,
the resonant frequency is obtained by exciting the
micro-pipette and subsequently measuring its response by
means of optical beam deflection. The optical beam
deflection-based experimental setup for measurement of
droplet volume measurement consists of three components,
namely, a source of laser light, external actuator to vibrate
micropipette and an optical detector (Fig. 6).
A B
C D
I/V Converter/
Amplifier
Data Acquisition
and Processing
Focusing Optics
Laser Source
Syringe
Piezo Actuator
Silver Coated
Micropipette
Droplet
Quadrant
Photo
Detector
Fig 6. Schematic diagram of droplet volume measurement subsystem.
Light from a laser diode (wavelength 633 nm, power 3
mW) was focused using a converging lens on to the surface
of the micropipette. The micropipette attached with syringe
was arranged in such a way that the reflecting surface formed
an angle with the optical axis of the laser beam. The reflected
light from the surface of micropipette was collected on a
quadrant photo detector (QP100-6, First Sensor GmbH,
active diameter DD = 11 mm, responsivity 0.4 /A W )
kept at a distance of around 0.5mm from the reflecting
surface. It was ensured that the reflected spot was within the
active area of the detector. The photo-currents from the
quadrants of the detector were fed to the signal conditioning
circuitry, composed of I-V converters of gain 106 .The
voltages from each quadrant VA,VB, VC and VD were fed to
real-time controller (DS1103, dSPACE GmbH) operated at
20 kHz update rate. Within the real-time controller, the value
(VA+VB)-(VC+VD) was computed. The resulting signal is
proportional to the amplitude of angular oscillation of the
micropipette at the site of measurement. In order to excite the
micropipette, a piezo-buzzer was employed. In order to
obtain the resonant frequency of the micropipette, the
piezobuzzer was vibrated by providing a sinusoidal input
and the frequency of excitation was automatically swept
between two specified limits. Fig. 7(a) shows the photograph
of the measurement system. The normalized frequency
responses are plotted in Fig. 7(b) and clearly depict the
reduction in resonant frequency caused by generation of a
droplet.
Laser Source
Microscope
Piezo Actuator
Quadrant
Photo
Detector
Syringe With
Silver Coated
Micropipette
I/V
Converter/
Amplifier
1500 1750 2000 2250 2500 2750 3000-0.5
0
0.5
1
Am
pli
tud
e
Frequency (Hz)
Without Droplet
With 260pL Droplet
With 1696pL Droplet
(a)
(b)
Fig 7. (a) Photograph showing the measurement system (b) normalized
frequency responses of the micro-pipette carrying droplets of two different
volumes.
B. Droplet deposition system
After the formation of liquid droplet at the tip of the
syringe needle, the droplet is dispensed by applying a
voltage between the syringe needle and a target aluminium
plate. Upon application of the potential difference, the drop
becomes polarised and gets attracted to the plate owing to
coulomb force. If the force of attraction is larger than the
force due to surface tension of the liquid, it separates from
the pipette and gets deposited on the plate. This phenomenon
is known as electrohydrodynamic pulling. The shooting of
charged droplets depends on viscosity, the surface tension,
the density, the electrical permittivity and the electrical
conductivity of the fluid. In the experiments a 330 V, 100
mA dc power supply was constructed so that the distance
Project #ISTC 0317
5
between the micro-pipette and the target could be maintained
around 100-150μm. The circuit diagram of the power supply
is shown in Fig 8(a). It employed a 230/230V transformer
followed by a bridge rectifier and a filter. The power supply
also comprised a short circuit protection circuit designed
using MOSFET (BUZ326). Fig 8(b) shows the photograph
of the power supply.
(a)
(b)
Fig. 8: (a) Circuit diagram of the power supply (b) photograph of the
fabricated circuit.
IV EXPERIMENTAL RESULTS
Syringe
Power Supply
Target
(Aluminium Plate)
Droplet
Silver Coated
Micropipette
Actuator
(Screw Driven
Stage)
(a)
(b)
Fig 14. Photograph of droplet generation system
The silver-coated micro-pipette was attached at the end of
a 1ml insulin syringe. In order to produce a droplet, a
screw-driven stage was attached with the piston of the
syringe. The schematic diagram of the setup is shown in Fig
9(a) and the photograph of the system is shown in Fig 9(b). The pitch of the screw used was 0.5 mm. The radius of the
syringe cavity was 0.36 cm. One full rotation of the bolt
displaced a liquid volume of around 20.3μL. In this setup,
with careful manual actuation, it was possible to produce
liquid droplets of volume around 260 picolitres consistently.
The corresponding radius of the droplet was around 40 um.
The micrographs showing the formation and shooting of a
liquid droplet of volume about 260 pL on application of a
high voltage is shown in Fig. 15.
500um
Fig. 15. Electro hydrodynamic pulling of glue of volume about 260pL
IV CONCLUSION
This paper reported the design and development of a small
volume droplet dispenser based on Electro Hydrodynamic
pulling. The dispenser is capable of real time measurement
of the mass of liquid dispensed. The developed liquid
dispenser consists of thee subsystems, namely, the droplet
generation subsystem, droplet mass measurement subsystem
and the non-contact type liquid droplet deposition
subsystem. The droplet deposition subsystem comprised a
silver coated micropipette attached to a syringe along with an
actuation system. The micropipette of desired characteristics
and dimensions was made using a pipette puller developed in
house. The silver coating on the micropipette was done by
Brashear’s process. The droplet mass measurement
subsystem works on the principle of shift in resonant
frequency of a micropipette due to formation of a liquid
droplet. The shift in resonant frequency was obtained by
employing an optical beam deflection measurement system.
The principle of the third subsystem, namely the non-contact
type liquid droplet deposition system is electro
hydrodynamic shooting, wherein an electric field is applied
between the droplet and the target to result in deposition of
the droplet. A 330 V power supply was designed and
developed to realize this subsystem. All of these developed
subsystems were experimentally evaluated. The droplet
generation subsystem was able to generate liquid droplets of
260pL consistently. The shift in resonant frequency of the
micropipette due to the formation of 260 pL and 1696 pL
liquid droplets were measured in real time with the liquid
droplet mass measurement subsystem. Finally, the droplets
Project #ISTC 0317
6
of different volumes were successfully deposited on a target
by means of electrohydrodynamic pulling.
REFERENCES
1. A. S.Yang, C. H. Cheng and F. S. Hsu, “PZT actuator applied to a
femto-liter droplet ejector”, Journal of Mechanical Science and
Technology 21, pp. 621, (2007) .
2. B.Heij, C. Steinert, H. Sandmaier, R.Zengerle, “A tuneable and
highly-parallel picolitre-dispenser based on direct liquid displacement”,
Sensors and Actuators A 103 (2003).
3. P. Ferraro, S. Coppola, S. Grilli, M. Paturzo and V. Vespini,
“Dispensing nano–pico droplets and liquid patterning by
pyroelectrodynamic shooting”, Nature Nanotechnology 5(2010).
4. K.Thurow, T.Kruger , N. Stoll, “An optical approach for determination
of droplet volumes in nano dispensing”, Journal of automated methods
and management in Chemistry 2009(2009).
5. Ren, R.B Fair, M.G. Pollack, “Automated onchip droplet dispensing
with volume control by electro wetting actuation and capacitance
metering”, Sensors and Actuators B 98(2004).
6. Ernst, W. Streule, R. Zengerle, P Koltay, “Quantitative Volume
Determination of Dispensed Nanoliter droplets on the fly”, IEEE-
Transducers 2009.
7. C. Haber, M. Boillat, B.V. D Schoot, “Flow sensing driven nano
dispensing: The path to more reliable liquid handling operations”,
Application Note: American Laboratory October 2004.
8. R. Tamizhanban, K. R. Sreejith, and G. R. Jayanth, “An automated
pipette puller for fabrication of glass micropipettes”, 85, 055105 (5pp),
2014.
Synthesis Optical and Electrical Characterization of II-VI colloidal quantum dots for IR applications
Atul Prakash Abhalea , Abhijit Chatterjeeb, Naresh Babub, Arup Banerjeeb and K.S.R. Koteswara Raoa,*
a Department of Physics, Indian Institute of Science, Bangalore- 560012 b SAC (ISRO), Sensors Development Area, Ahmedabad -15
Abstract: colloidal quantum dots (CQDs) can provide tuneable bandgap, cost effective and relatively low processing temperature materials for photo-detection. In this paper we have discussed the synthesis of HgCdTe and PbS colloidal quantum dots and their application as IR photodetectors. I. Introduction
Colloidal quantum dots are potential
candidates for the light detection applications,1 due to its exceptional characteristics such as tuneable bandgap, comparatively easy and cost effective synthesis and tuneable properties by post synthesis processes.1,2 In the CQD materials bandgaps are tuneable with the values greater than that of the bulk ones and hence only low band gap compound semiconductors can match the spectral range suitable for Infrared photodetection applications. Among several low band gap compound semi-conductors HgCdTe (MCT), PbS are suitable materials with higher exciton Bohr radius (~18 nm), which allows bandgap tuneablity and low binding energy due to its high dielectric constant (~17.2), which helps relatively easy exciton dissociation and charge extraction. Recent work on the PbS-CQD based schottky barrier photodiodes highlights the importance of the CQD based photodetection.3,4
II. Experimental
Monodispersed colloidal quantum dots
of MCT and lead sulphide are synthesized with the method reported by the MC Weidman et. al.,5 with excitonic absorption peaks from ~1300-1600 nm. This method provides narrow size distribution of the quantum dots, which are essential for the better carrier transport in the
*Email Address: [email protected]
solid CQD films, which increases the hopping probability 6.
Fig. 1 The absorption spectrum of the PbS CQD solid films.
Fig. 1 shows a typical absorption
spectrum of solid CQD films, where quantum dots are grown at 120 ºC for different time periods varying from 5 – 60 minutes. The presence of clear first and second excitonic peaks in all the spectra confirms the narrow size distribution, which is also visible from the TEM image in Fig. 2.
Fig. 2. Transmission electron micrograph of the PbS quantum dots.
In the device point of view, schottky barrier devices are fabricated with Aluminium and Indium Tin Oxide (ITO) contacts, where ITO works as Ohmic contact. Similarly PbS-
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
CQD/Si <100> othe schois utilisearray odependawhich geometrare depo~50 nm patterneWhen pcontactsthe croshave efcompletnumber
Fig. 3 C
in the dshown iunder thin the degradatschottkyof the Ato avoidCQD/p-observedcharacteshown inSi and Pfrom thewith the~1100nm
heterojuncoriented singlottky devicesed (Fig. 3) tf size 300
ant photoresis analogou
ry, patternedosited on ththick CQD d Al is deppotential is s only one Css junction offective cone 2N pixelsof contacts.
Crossbar cont
I-V characte
dark and undin Fig. 4. Ihe light photreverse biastion is obse
y diodes, proAl, indicates a
d degradatioSi heterojud under theristics of n Fig. 5, whePbS-CQD. Te CQD film
e Si filter whm were passe
tions are fle crystal silis particular dthis can be × 300 μm
sponse can us to the d transparenhe glass subsfilm is spin
posited as shapplied to
CQD area, whof the two linntribution. I can be ha
tact CQD pix
eristics of theder 40 W tuIt is clear frtocurrent dens. Over a perved in thobably due ta need for a on.7 In the unctions, phe light illu
these heterere a contribTo observe t
m, devices where only waved through.
fabricated onicon wafers. device geomused as a p
m and positbe collec
ROIC. In nt ITO contstrate on whcoated and t
hown in Figany two cr
hich appearsnear arrays In this wayandled with
xel array.
e schottky diungsten lampfrom the fignsity dominperiod of t
he Al/PbS-Cto the oxidaprotective lacase of P
photocurrentumination. rojunctions bution from bthe contribu
were illuminavelengths ab
n a For etry
pixel tion
cted, this acts hich then g. 3. ross s on will y a 2N
iode p is ure, ates time
CQD tion ayer PbS-
is I-V are
both tion ated
bove
Figsch
FigPb
III
wiCuwhthiarefileRCshaspadevsigsigdet
g. 4 I-V chahottky diode
g. 5 Dark abS-CQD/Si h
I. LBIC PbS-C
ith the remourrent (RC-Lhich are presis technique ea on the deed is presentC-LBIC imaaped deviceatial distancvice and co
gnal. In Figgnal is plottetails will be
aracteristics .
and light I-Vheterojunction
C Imaging
QD / p-Si ote contact LLBIC) technsented else-wgives informevice, wheret. Fig. 6(a) shage of the Pe. Here, theces parallel lour represe
g. 6(b) the ed, which is bdiscussed du
of the PbS-
V Charactern under reve
devices areLight Beamnique, the dwhere. Imag
mation about e the built inhows a 3-dim
PbS-CQD/p-Se axes repreto the plannts the magcross-sectio
bipolar in nauring the pres
-CQD/Al
ristics of erse bias.
e imaged m Induced details of ging with an active
n electric mensional Si square esent the ne of the gnitude of on of the ature. The sentation.
Fig. 6 a). 3D plot of RC-LBIC image. b) a cross-section of 3D signal for PbS/Si heterostructure.
IV. Conclusion We have synthesized monodispersed colloidal MCT and PbS quantum dots with chemical bath deposition technique with variable excitonic absorption peaks. Schottky, as well as heterojunction devices are fabricate with solid film of CQD on Al and Silicon by spin coating technique. In order to study the photodetection capabilities of the devices I-V characteristics under the dark and light conditions are studied, where a good photoresponse is observed for both schottky as well as heterojunction devices.
References
1 G. Konstantatos and E.H. Sargent, Infrared Phys. Technol. 54, 278 (2011). 2 E. Sargent, Nat. Nanotechnol. 7, 349 (2012). 3 J.P. Clifford, K.W. Johnston, L. Levina, and E.H. Sargent, Appl. Phys. Lett. 91, 253117 (2007). 4 E. Heves, C. Ozturk, V. Ozguz, and Y. Gurbuz, Electron Device Lett. 34, 662 (2013).
5 M. Weidman, M. Beck, and R. Hoffman, ACS Nano (2014). 6 P. Guyot-Sionnest, J. Phys. Chem. Lett. 3, 1169 (2012). 7 H. Fu and S.-W. Tsang, Nanoscale 4, 2187 (2012).
Movement strategies for commensalism: coexistence of meso-carnivores in human-dominated landscapes
Maria Thaker1 and AbiTamimVanak2
1Centre for Ecological Sciences, Indian Institute of Science
1Ashoka Trust for Research in Ecology and the Environment Abstract - In the increasingly human-dominated landscapes of India, the survival and long-term persistence of mammalian carnivores is a key conservation challenge. This is especially the case for species that survive in the semi-arid savanna biomes of central India. The overall aim of this study is to determine the ecological parameters and behavioural strategies that enable existence of the golden jackal,Indian fox, and jungle cat in human-dominated systems. Here, we present the first part of the study, which was to generate detailed maps of the study areasin Maharastrausing supervised and rule-based classification ofLISS IV imagery. Comparison of these methods indicates that object-based methods have higher accuracy and is more applicable th 1 an supervised classification methods for delineating habitat patches in grassland areas. On-ground surveys of these areas confirm the presence of focal species (Golden Jackal, Indian fox and Jungle cat). During the next phase of the project, we will obtain fine-scaled movement data of the focal species using advanced GPS/UHF telemetry. These data will be superimposed onto the high-resolution maps generated thus far to determine patterns of resource use and movement ofmeso-carnivores across the landscape mosaic. I. INTRODUCTION
Tropical landscapes worldwide are 1Project code: IST/BES/MT/329
being rapidly transformed into a complex matrix of human-dominated multiple use land cover types with differing intensities of use (Gibbs et al. 2010). With the second largest human population in only 2.4% of the world’s land area, natural habitats in India have undergone some of the greatest rates of transformation (FSI 2011).The consequence for organisms that once inhabited large, unbroken tracts of native habitats has been either population decline to local extinction on the one hand, or tenuous survival on the other (Karanth et al. 2010). The majority of research has focused on species in decline, with particular attention given to the effects of habitat fragmentation and general loss of biodiversity, yet there are a range of species that seem to be effective commensals in human-dominated landscapes.
An understanding of how willing species are to traverse or even utilize potentially inhospitable and risky habitats is critical to understanding coexistence (Russell et al. 2003). Adaptations of animals to such risk in the landscape can be measured in terms of timing (whether at particular times of day or night), residence (how long they stay in landscape elements) and speed of movement (Graham et al. 2009). Thus, understanding how animals make decisions regarding movement and incorporating behavioral decisions into population models is an important step for predicting the consequences of fragmentation for population
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
persistence (Russell et al. 2003; Fahrig 2007; Vanak et al. 2010). Resource selection by animals is a hierarchical process of behavioral responses to particular environmental characteristics (Horne et al. 2008). Different environmental factors and scales may influence the movement of individuals within a landscape, and therefore animal-landscape relationships should be examined across a range of scales (Anderson et al. 2005; Boyce 2006). This is particularly important when examining resource selection dynamics of species inhabiting fragmented or human-dominated landscapes (Anderson et al. 2005).
In India much wildlife exists in human-dominated landscapes across a variety of biomes. Yet, there are very few detailed studies on the status, ecology and behavioural strategies of species living in these landscapes (Akhtar et al. 2004; Athreya et al. 2007) or even of the semi-arid tropics in particular (Vanak and Gompper 2010). The first step and a key challenge in understanding how animals utilize human-dominated areas isin distinguishing human-modified from natural habitats using remotely sensed data. Here, we report the results of our analyses of high-resolution imageries using supervised and rule-based classification to effectively delineate habitat patches at a fine resolution.We focus on the dry savanna grassland habitat, because the spectral heterogeneity of these landscapes are particularly difficult to classify and because small carnivores that utilize these areas are poorly studied. These detailed maps of native and human-modified habitat patches have been surveyed on ground, and will later be superimposed with the movement information of Indian foxes (Vulpesbengalensis), golden jackals (Canisaureus) and jungle cats
(Felischaus). In doing so, we will have fine-scaled data on the habitat requirements and levels of tolerances for human modification for each of these species. II- METHODS A. Study area
The dry savannah grasslands of peninsular India are unique habitats that support a vast proportion of India’s agro-pastoralist community. These ecosystems have also been subject to land-use change (e.g. irrigation or conversion to agriculture or tree-plantations) resulting in major decrease in extent. We focus our study on the areas near the villages of Takali-Dhokeshwar and Kedagaon-Kurkumbhof Maharashtra. The landscape there is typical of much of the central Indian plateau, with a matrix of savannah grasslands, agriculture land, and settlements. B. Classification of remote sensing data
Landcover data were obtained from high-resolution satellite imageries (IRS LISS IV, dated 18 January 2013) were acquired for the two study areas. After atmospheric correction, we conducted a supervised classification using the Maximum Likelihood Classifier in Erdas Imagine, using ground truthed data from field survey as training data for the classifier. We mapped the following and cover classes: grasslands, agriculture, fallow, water and builtup/bare (Fig. 1a).
As a comparison, we also conducted an object-based classification of the same LISS IV images in eCognition developer 8.7, based on a rule-set (Fig. 1b). All images were subjected to multiresolution segmentation in order to cluster spectrally homogenous entities into one object (Hofmann and
Puzicha, 1998). We classified built-up areas and bare rocky areas using brightness and green band values to avoid spectral mixing with barren grasslands. To identify agricultural fields with and without standing crops, we incorporated two vegetation indices, Normalized Differential Vegetation Index (NDVI) and Normalized Differential Vegetation Structural Index (NDVSI) within eCognition as object features. NDVI relates directly to vegetation health, while NDVSI shows vegetation structure (Yang, Willis and Mueller, 2008)
One of the biggest hindrances in grassland mapping is the difficulty in distinguishing between grasslands and fallow agricultural fields. If a field is left uncultivated for period of time, we find little to no spectral difference between these fallow fields and natural grasslands. We thus utilized a combination of rules to separate the fallow land from grasslands; for example, the mean value of the IR band in summer images for grassland “objects” was higher compared to fallow fields “objects”. We set the threshold for object classification accordingly.
We further classified grasslands
into different classes using the GLCM (Grey Level Co-occurrence Matrix) texture feature within eCognition(similar to Guo et al. 2004). The GLCM is produced by putting each pixel and its neighbour within an object into a 256x 256 matrix, and the number of pair occurrences is final value of the GLCM feature. Since the study area included a wide range of spectral classes, GLCM dissimilarity and homogeneity were used to extract grasslands. Finally, patches that were left unclassified because of the heterogeneity of the landscape were classified using relations to
neighbouring objects properties. For example, smaller patches of land between agricultural fields were assigned as agriculture and not grassland. And grasslands with undulating topography were included in the grassland class if surrounded by other grassland 'objects'.
Figure 1.(a) Pixel- and (b) Object-based classification of the Takali-Dhokeshwar (Maharastra) area.
C. Ground-truthing and animal presence surveys The land cover maps that we generated were exported as shapefiles and largest contiguous 'core' areas within our study area were extracted using Patch Analyst extension in ArcGIS. The core areas were divided into 13km2 grids and the grids above 60% grassland cover were selected for survey (Fig. 2). Extensive field-work, which included transects and random sampling, was conducted to survey presence of grassland specific species, especially golden jackals, Indian foxes, and
Legend
Grassland
Agriculture
Fallow land
Water
Vegetation
Builtup area/Bare(b) Object based classification
(a) Pixel based classification
Jungle cats. We searched grassland and the surrounding areas along existing trails and tracks for species-specific signs, such as carnivore scat, track impressions, herbivore pellets, and opportunistic sightings. Local information was used to confirm presence of endangered and grassland specialist species.
Figure 2. Map of the TakaliDhokeshwarstudy area, depicting the grids (white) and survey tracksfor animal presence (green).Image courtesy of Google earth.
III- CONCLUSIONS AND FUTURE WORK Thus far, we have generated detailed maps of our study area using both maximum likelihood supervised classification and object based classification methods. As expected, supervised classification of complex heterogeneous landscapes, such as grassland habitats, is limited. We find that this classification method produces a ‘salt and pepper effect’ which greatly contributes to the inaccuracy of the classification (see also de Jong et al. 2001; Gao and Mas 2008; Van de Voorde et al. 2004). This problem arises because supervised
classification assigns classes to pixels ignoring their related spatial and contextual information (Weih, Jr and Riggan, Jr), which results in incorrectly classifying a map as overly fragmented. Instead, the object based classification created a grassland map, which incorporated the heterogeneity of the class. For example, hilly grasslands, ridges, scrub forest, and grass covered plains were all classified as a single class: grassland. This method therefore the land cover class as a habitat rather than a heterogeneous milieu. Not surprisingly, comparison between these methods shows that object based classification had a higher overall accuracy (81% accuracy) than supervised classification (64%) for detailed mapping of semi-arid grassland areas.
On ground surveys of both study areas confirm the presence of the target study species. During the next phase of the project, we will capture and collar the target animals (n = 10 per species) using GPS/UHF telemetry devices. The detailed movement information we then obtain from these tracked animals will inform us of the behavioural strategies that permit use of the habitat matrix. ACKNOWLEDGEMENTS We thank AmeyaGode for help in the analysis of remote sensing data, and AbhijeetKulkarniand HarshalBhosalefor help in the field surveys. REFERENCES [1] Akhtar, N., H. Singh Bargali, & N.
Chauhan. (2004). Sloth bear habitat use in disturbed and unprotected areas of Madhya Pradesh, India. Ursus, 15, 203-211.
[2] Anderson, P., M. G. Turner, J. D. Forester, J. Zhu, M. S. Boyce, H. Beyer, & L. Stowell. (2005). Scale-
dependent summer resource selection by reintroduced elk in Wisconsin, USA. Journal of Wildlife Management, 69, 298-310.
[3] Athreya, V. R., S. S. Thakur, S. Chaudhuri, & A. V. Belsare. (2007). Leopards in human-dominated areas: a spillover from sustained translocations into nearby forests? Journal of the Bombay Natural History Society, 104, 45-50.
[4] Boyce, M. S. (2006). Scale for resource selection functions. Diversity and Distributions, 12, 269-276.
[5] de Jong, S. M., Hornstra, T., & Maas, H.-G. (2001). An integrated spatial and spectral approach to the classification of Mediterranean land cover types: the SSC method. International Journal of Applied Earth Observation and Geoinformation, 3(2), 176-183.
[6] Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity. Annual review of ecology, evolution, and systematics, 487-515.
[7] Gao, Y., & Mas, J. F. (2008). A Comparison of the Performance of Pixel Based and Object Based Classifications over Images with Various Spatial Resolutions. Online journal of earth sciences, 2(1), 27-35.
[8] Gibbs, H., A. Ruesch, F. Achard, M. Clayton, P. Holmgren, N. Ramankutty, & J. Foley. (2010). Tropical forests were the primary sources of new agricultural land in the 1980s and 1990s. Proceedings of the National Academy of Sciences, 107, 16732-16737.
[9] Graham, M. D., I. DouglasHamilton, W. M. Adams, & P. C. Lee. (2009). The movement of African elephants in a humandominated landuse mosaic. Animal Conservation, 12, 445-455.
[10] Guo, X., et al. (2004). Measuring spatial and vertical heterogeneity of grasslands using remote sensing techniques. Journal of Environmental Informatics, 3(1), 24-32.
[11] Hofmann, T., & Puzicha, J. (1998). Statistical models for co-occurrence data.Technical report, Artifical Intelligence Laboratory Memo 1625, M.I.T., 1998.
[12] Horne, J. S., E. O. Garton, & J. L. Rachlow. (2008). A synoptic model of animal space use: Simultaneous estimation of home range, habitat selection, and inter/intra-specific relationships. Ecological Modelling, 214, 338-348.
[13] Karanth, K. K., J. D. Nichols, K. U. Karanth, J. E. Hines, & N. L. Christensen. (2010). The shrinking ark: patterns of large mammal extinctions in India.Proceedings of the Royal Society London-B. 277: 1971-1979.
[14] Russell, R. E., R. K. Swihart, & Z. Feng. (2003). Population consequences of movement decisions in a patchy landscape. Oikos, 103, 142-152.
[15] Van de Voorde, T., De Genst, W., Canters, F., Stephenne, N., Wolff, E., & Binard, M. (2004). Extraction of land use/land cover related information from very high resolution data in urban and suburban areas. Remote Sensing in Transition, 237-244.
[16] Vanak, A. T., & M. E. Gompper. (2010). Multi-scale resource selection and spatial ecology of the Indian fox in a human-dominated dry grassland ecosystem. Journal of Zoology, 281, 140-148.
[17] Weih Jr, R. C., & Riggan Jr, N. D. Object-Based Classification vs. Pixel-Based Classification: Comparative Importance of Multi-Resolution Imagery. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 38, 4.
[18] Yang, Z., et al. (2008). Impact of band-ratio enhanced AWIFS image to crop classification accuracy. Proc. Pecora 17 http://www. asprs.org/a/publications/proceedings/pecora17/0041. pdf.
ISTC0321
Development and Performance Evaluation of a PCM coupled Heat PipeG. M. Karthik, Girish Kini, A. Ambirajan and P. Kumar
Abstract - Designing a suitable thermal control system forpresent day electronics is becoming more and morecomplex with the invention of low mass(compact) highdissipation electronic components. Heat pipe is aninnovative device used for removal of heat from a remotesource to far-away heat sink. Thermal control design of aspacecraft requires transportation of heat from theelectronic packages which are housed inside, to a thermalradiator which is located outside. Heat from thermalradiator is eventually dissipated into the space by radiation.Even if the spacecraft heat loads are cyclic in nature,thermal radiators are being designed to meet peak loadrequirements. To avoid oversizing of a thermal radiator, aheat storage unit may be employed to store the peak loadand release it over a longer time. The present projectexplores the possibility of integrating a Phase ChangeMaterial (PCM) module to a conventional heat pipe to takecare of periodic heat load transience. The PCM acts as athermal capacitor storing heat during transient peaks in heatloads and releasing it during the off peak load conditions tothe radiator.
NOMENCALTURE
A area (m2)Dh hydraulic diameter (m)hfg latent heat of vaporization (J/kg)L length (m)ṁ mass flow rate (kg/s)N number of groovesp pressure (Pa)Q heat rate (W)rc capillary radius (m)u velocity (m/s)w groove width (m)x coordinate along the heat pipe (m)
Greek Symbolssurface tension coefficient (N/m)density (kg/m3)
μ viscosity (kg/m/s)
Subscriptsa adiabatic sectionc condensere evaporatorin inputl liquid
v vapor
I. INTRODUCTION
Axially grooved Aluminium Ammonia Heat pipes are usedin spacecraft applications to transport heat from electronicpackages (heat source) to heat radiator (heat sink). Heatpipes are generally embedded in the honeycomb panels onwhich electronic packages are mounted on the inside andheat is radiated to space from the outside. A typical heatpipe consists of a hermetically sealed envelope containingwick structure filled with saturated working fluid in liquidphase and remaining space (core) filled with vapour. Heattransferred to the heat pipe (at evaporator) is removed byevaporation of the working fluid in the wick, therebyincreasing the vapor pressure in the core causing flow ofvapor towards the cooler end, where it condenses (atcondenser) and heat is released. The condensate returns tothe evaporator due to pressure gradient along the length ofthe wick caused by surface tension forces (capillarypumping).
In thermal management of electronics, Phase ChangeMaterials (PCMs) are extensively used to maintaintemperatures during transient heating. PCMs are highlyeffective heat storage materials that undergo a phasechange at a particular temperature or a range oftemperatures. These materials generally have a high latentheat of fusion allowing a small amount of PCM to meet theheat storage needs of a majority of applications.
A key challenge in the design of a thermal control systemof a spacecraft, while ensuring its reliability andeffectiveness, lies in accommodating high heat dissipationsand varying orbital heat loads. When the thermal load is notsteady, as in the case of remote sensing satellites, thethermal system adopted demands similar heat removalcapacity as that at constant thermal load. The reason forthis is that the number of heat pipes and radiator arearequirement are based on the peak thermal load and not theorbital average thermal load. One method of optimallymanaging the issue of varying heat load withoutoverdesigning the thermal system is by the use of a suitablePhase Change Material in conjunction with heat pipe forefficient and stable heat transfer. This concept would notonly optimize the thermal design, it would also result inlesser number of heat pipes and reduced radiator area. This
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
in turn would result in a significant reduction of overallmass of the spacecraft thermal control system.
II. MATHEMATICAL MODEL OF HEAT PIPE
The heat transport capability (measured in Watt-meter) isone of the key factors in the thermal performance of heatpipes. The mathematical model is developed to estimateheat transport capability of a heat pipe based on capillarylimitation. Heat pipe with Aluminium alloy as envelopematerial, extruded axial grooves as wick and ultrapureanhydrous ammonia as working fluid is considered for thisstudy. The variation of the radius of curvature of themeniscus at the liquid-vapor interface is used to calculatethe maximum heat transport capability of the heat pipe.
The radius of the meniscus which is formed at the liquid-vapor interface is related to the pressure difference betweenthe liquid and vapor by the Young-Laplace equation [4],which in differential form can be written as
− = − ( ) ( ) (1)Both liquid and vapor flow are modeled as laminar pipeflow. The liquid pressure gradient in the grooves and vaporpressure gradient in the core can be expressed as, [5],
= −32, , = −32 , (2)The axial conduction of heat in the envelope is neglected.Hence by conservation of energy, the mass flow rate can beexpressed as
= − = − ( )ℎ (3)The liquid and vapor velocities are expressed as follows
= ( )ℎ , = ( )ℎ (4)The heat load in the axial direction, [3], is
Combining equations (1) and (2) we obtain−32 , + −32, = − ( ) ( ) (6)
Further substituting for the liquid and vapor velocities fromequations (5) and (6)32 ( ), ℎ + 32 ( ), ℎ = − ( ) ( ) (7)The boundary condition having radius of meniscus at thecondenser end equal to the radius of the vapor groove [1] isconsidered. The equation (9) is solved by the fourth-orderRunge-Kutta method and variation of curvature over thelength of the groove is obtained. The maximum heat thatcan be transported by the heat pipe is obtained by equatingthe radius of meniscus at the evaporator end to half ofgroove width [2].
= 2 (8)Maximum heat transport capability is obtained bymultiplying the maximum heat transported with effectivelength of the heat pipe. For a given heat pipe geometry,heat transport capability is a function of temperature. Thisis due to the fact that working fluid properties like density,viscosity, surface tension, latent heat, etc. varies withtemperature. Variation of computed maximum heattransport capability over different operating temperatures isplotted in the following figure.
Fig. 1. Variation of maximum Heat Transport Capability withoperating temperature
It can be seen from the plot that, maximum heat transportcapability reaches to its peak at a given temperature andfalls to minimum on either side.
III. HEAT PIPE EXPERIMENTAL SETUP
An experimental setup is built to measure maximum heattransport capability of the given heat pipe. Theexperimental setup consists of a horizontal workingplatform, height adjustable fixture, instrumented heat pipe,DC power supply, constant temperature bath and a dataacquisition unit, as shown in the following figure.
100
150
200
250
300
350
-40 -30 -20 -10 0 10 20 30 40 50 60
Hea
t Tra
nspo
rt C
apab
ility
(Wm
)
Temperature(°C)
( ) = ⎩⎪⎨⎪⎧ , 0 < <, < < ++ + − , + < < + + (5)
Fig. 2. Schematic showing experimental setup
The heat pipe is divided into three zones, namely,evaporator (where heat is added), adiabatic (where there isno heat transfer across the heat pipe, to externalenvironment) and condenser (where heat is removed). Thegiven heat pipe is instrumented with resistance foil heaterson the evaporator end and micro channel heat exchangers atthe condenser end. The zone between evaporator andcondenser (adiabatic) is covered with insulation material.T-type thermocouples are mounted on external surface atvarious locations along the length of the heat pipe. Theentire setup is connected to a data acquisition and controlunit (cRIO-9082 system). A Virtual Interface (VI) iscreated using Lab-View program through which theexperiments are controlled.
Fig. 3. Complete Experimental Setup
The maximum heat transport capability values arecalculated for highest value of heat input before dry outlimit is reached, for different operating temperatures.Figure 4 gives experimentally obtained values at -20oC,0oC, 20oC and 40oC along with theoretical estimate.
It shall be noted that, at -20 oC experimental value is lowerthan theoretical value and at higher temperatures,experimental results are higher than theoretical estimates.This is due to the fact that, mass of the working fluid isestimated considering liquid volume equal to groovesvolume and vapour volume equal to core volume at 0oC. Attemperatures below 0oC, specific volume of liquidammonia decreases causing shortage of liquid in thegrooves. While at higher temperatures, specific volume ofliquid ammonia increases causing availability of moreliquid in the grooves to absorb more heat in the evaporator.This liquid puddle effect at higher temperatures will beconsidered in the theoretical computation as a part of futurework.
Fig. 4. Experimental data of maximum Heat Transport Capabilitycompared with theoretical estimation
IV. DESIGN OF PCM MODULE
The Phase Change Material (PCM) module is designed soas to store approximately 30 kJ of heat. Consideringparaffin wax or Eicosane as the candidate phase changematerial, the size of the module was worked out to be100mm x 100mm x 20mm based on the latent heat of thechosen PCM material. Flanges are provided on all foursides at the top to have Copper is chosen as the material ofthe module. Two pressure relief valves are provided on thetop plate of the module to provide pressure relief in casesof pressure build up inside the PCM module during itsoperation. A silicone gasket of 1mm thickness is usedbetween the two parts of the PCM module to ensureadequate sealing of the phase change material duringmelting. The module also has a slot, at its base, throughwhich the heat pipe can be coupled thermally to the PCMwith minimum resistance.
Fig. 5. PCM module assembled with a top plate
Fig. 2. Schematic showing experimental setup
The heat pipe is divided into three zones, namely,evaporator (where heat is added), adiabatic (where there isno heat transfer across the heat pipe, to externalenvironment) and condenser (where heat is removed). Thegiven heat pipe is instrumented with resistance foil heaterson the evaporator end and micro channel heat exchangers atthe condenser end. The zone between evaporator andcondenser (adiabatic) is covered with insulation material.T-type thermocouples are mounted on external surface atvarious locations along the length of the heat pipe. Theentire setup is connected to a data acquisition and controlunit (cRIO-9082 system). A Virtual Interface (VI) iscreated using Lab-View program through which theexperiments are controlled.
Fig. 3. Complete Experimental Setup
The maximum heat transport capability values arecalculated for highest value of heat input before dry outlimit is reached, for different operating temperatures.Figure 4 gives experimentally obtained values at -20oC,0oC, 20oC and 40oC along with theoretical estimate.
It shall be noted that, at -20 oC experimental value is lowerthan theoretical value and at higher temperatures,experimental results are higher than theoretical estimates.This is due to the fact that, mass of the working fluid isestimated considering liquid volume equal to groovesvolume and vapour volume equal to core volume at 0oC. Attemperatures below 0oC, specific volume of liquidammonia decreases causing shortage of liquid in thegrooves. While at higher temperatures, specific volume ofliquid ammonia increases causing availability of moreliquid in the grooves to absorb more heat in the evaporator.This liquid puddle effect at higher temperatures will beconsidered in the theoretical computation as a part of futurework.
Fig. 4. Experimental data of maximum Heat Transport Capabilitycompared with theoretical estimation
IV. DESIGN OF PCM MODULE
The Phase Change Material (PCM) module is designed soas to store approximately 30 kJ of heat. Consideringparaffin wax or Eicosane as the candidate phase changematerial, the size of the module was worked out to be100mm x 100mm x 20mm based on the latent heat of thechosen PCM material. Flanges are provided on all foursides at the top to have Copper is chosen as the material ofthe module. Two pressure relief valves are provided on thetop plate of the module to provide pressure relief in casesof pressure build up inside the PCM module during itsoperation. A silicone gasket of 1mm thickness is usedbetween the two parts of the PCM module to ensureadequate sealing of the phase change material duringmelting. The module also has a slot, at its base, throughwhich the heat pipe can be coupled thermally to the PCMwith minimum resistance.
Fig. 5. PCM module assembled with a top plate
100
150
200
250
300
350
-40 -30 -20 -10 0 10 20
Hea
t Tra
nspo
rt C
apab
ility
(Wm
)
Temperature( C)
Fig. 2. Schematic showing experimental setup
The heat pipe is divided into three zones, namely,evaporator (where heat is added), adiabatic (where there isno heat transfer across the heat pipe, to externalenvironment) and condenser (where heat is removed). Thegiven heat pipe is instrumented with resistance foil heaterson the evaporator end and micro channel heat exchangers atthe condenser end. The zone between evaporator andcondenser (adiabatic) is covered with insulation material.T-type thermocouples are mounted on external surface atvarious locations along the length of the heat pipe. Theentire setup is connected to a data acquisition and controlunit (cRIO-9082 system). A Virtual Interface (VI) iscreated using Lab-View program through which theexperiments are controlled.
Fig. 3. Complete Experimental Setup
The maximum heat transport capability values arecalculated for highest value of heat input before dry outlimit is reached, for different operating temperatures.Figure 4 gives experimentally obtained values at -20oC,0oC, 20oC and 40oC along with theoretical estimate.
It shall be noted that, at -20 oC experimental value is lowerthan theoretical value and at higher temperatures,experimental results are higher than theoretical estimates.This is due to the fact that, mass of the working fluid isestimated considering liquid volume equal to groovesvolume and vapour volume equal to core volume at 0oC. Attemperatures below 0oC, specific volume of liquidammonia decreases causing shortage of liquid in thegrooves. While at higher temperatures, specific volume ofliquid ammonia increases causing availability of moreliquid in the grooves to absorb more heat in the evaporator.This liquid puddle effect at higher temperatures will beconsidered in the theoretical computation as a part of futurework.
Fig. 4. Experimental data of maximum Heat Transport Capabilitycompared with theoretical estimation
IV. DESIGN OF PCM MODULE
The Phase Change Material (PCM) module is designed soas to store approximately 30 kJ of heat. Consideringparaffin wax or Eicosane as the candidate phase changematerial, the size of the module was worked out to be100mm x 100mm x 20mm based on the latent heat of thechosen PCM material. Flanges are provided on all foursides at the top to have Copper is chosen as the material ofthe module. Two pressure relief valves are provided on thetop plate of the module to provide pressure relief in casesof pressure build up inside the PCM module during itsoperation. A silicone gasket of 1mm thickness is usedbetween the two parts of the PCM module to ensureadequate sealing of the phase change material duringmelting. The module also has a slot, at its base, throughwhich the heat pipe can be coupled thermally to the PCMwith minimum resistance.
Fig. 5. PCM module assembled with a top plate
20 30 40 50 60
Temperature( C)
Theoretical
Experimental
The thermo-physical properties of Paraffin Wax andEicosane are given in the following tables 1 and 2.
Material Paraffin Wax(MERCK make)Melting Point (°C) 58-60
Density (kg/m3) 790 (liquid, 65 °C)916 (solid, 24 °C)
Heat of Fusion (kJ/kg) 173.6Thermal Conductivity(W/m-K)
0.167 (liquid, 63.5 °C)0.346 (solid, 33.6 °C)
Table 1: Properties of Paraffin Wax
Material Eicosane (Sigma-Aldrich)Melting Point(°C) 35-37
Density(kg/m3) 770 (liquid)810 (solid)
Heat of Fusion(kJ/kg) 241.0Thermal Conductivity(W/m-K)
0.157 (liquid)0.390 (solid)
Table 2: Properties of Eicosane
V. PCM MODULE CHARACTERIZATION STUDIES
PCM module is filled with 150g of Paraffin wax and isinstrumented with a resistance foil heater at its base.Several thermocouples are mounted at various locations onthe external surfaces of PCM module. In addition, onethermocouple is left floating inside the PCM module tomeasure the temperature of the candidate PCM. Thethermocouples are connected to a NI 9213 DAQ cardwhich is connected to NI cRIO system. The heater ispowered through a programmable DC power supply unit.The measured data are continuously stored on a dedicatedcomputer through Lab-View interface.
Few experiments are conducted to study phase changebehaviour of PCM inside the module, to verify heat storagecapacity and to measure temperature distribution within themodule. The experiments were repeated for variousconfigurations of the module, different heat inputs andoperating conditions.
It is found that for an input of 20 W of heater power for4630 sec, 1.0 W was consumed by PCM module as sensibleheat, 3.5 W was expended as sensible heat within the PCM,9.5 W of heat is lost through convection and remaining 6.0W is utilised in melting the PCM. Approximately 28 kJ ofheat can be stored as latent heat using Paraffin wax as PCMin this module. Since Copper is chosen as material forbuilding PCM module, temperature difference under stablecondition was found to be less than 5oC within the module.Module temperatures and PCM melting characteristics areshown in the following plot.
Fig. 6. Experimental data showing temperature distribution on thePCM module
VI. HEAT PIPE TRANSIENT EXPERIMENTS
Experiments were carried out to characterize the transientbehavior of the heat pipe. These experimental results serveas a baseline reference to understand the characteristics of aheat pipe when coupled with PCM module. To simulate 30kJ of heat input, the heat pipe evaporator is powered to 50W for 10 minutes in the two experiments described below.
1. In the first experiment, 50 W of heat applied for 10minutes at the evaporator is removed from thecondenser, exposed to the surroundings, through naturalconvection.
2. In the second experiment, heat exchanger is attached tothe heat pipe condenser which in turn is connected toconstant temperature bath. 50 W of heat is removed at aconstant temperature of 30oC through the continuousflow of fluid inside the heat exchanger.
Heat pipe transient temperatures in the above experimentsare plotted in Figure 7.
Fig. 7. Experimental data showing evaporator temperature duringheat pipe transient experiments
0
10
20
30
40
50
60
70
80
90
100
0 1200 2400 3600 4800 6000 7200 8400
Tem
pera
ture
(°C
)
Time(s)
T_wallT_heaterT_pcm
25
30
35
40
45
50
55
60
0 150 300 450 600 750 900
Tem
pera
ture
(°C
)
Time(s)
Bath onNo Condenser
It may be noticed from Figure 7 that, in the absence of heatstorage unit, 25°C temperature raise is observed in the heatpipe within 10 minutes of heat input. To maintain the heatpipe at constant temperature during heat addition,continuous removal of heat is required. This can beachieved either through the use of complex active devicelike constant temperature bath / heat exchanger or a simplePCM module coupling.
VII. HEAT PIPE COUPLED WITH PCM MODULE
The PCM module is coupled to the heat pipe at theevaporator with the help of thermal grease to ensure goodthermal contact. The other surface of the heat pipe isinstrumented with foil heater, as shown in the followingfigure.
Fig. 8. Schematic showing Heat Pipe Coupled with PCM Module
The PCM module is filled with Paraffin wax (% fill). Twoexperiments are carried out again. In the first experiment,50 W of heat for 10 minutes is removed from the heat pipecondenser through natural convection. Following plotshows heat pipe transient temperatures without PCMmodule and by coupling the PCM module.
Fig. 9. Heat Pipe evaporator temperature when coupled with PCMModule (Paraffin Wax) having condenser section exposed toambient.
It may be noticed that, by introducing PCM module on theheat pipe evaporator, temperature raise falls by 7 to 8oC.However, since heat source is coupled to the heat pipe, onlya small amount of heat flows towards PCM module.Experiments by coupling the heat source directly to thePCM module will be carried out as a part of future work.
In the second experiment, condenser end of the heat pipe isconnected to the heat exchanger, which in turn to theconstant temperature bath. 50 W of heat for 10 minutes is
removed at a constant temperature of 60oC through thefluid circulation in the heat exchanger. Transienttemperatures of the heat pipe without PCM module and bycoupling PCM module are shown in the following plot.
Fig. 10. Heat Pipe evaporator temperature for Heat pipe coupledwith PCM Module having condenser section connected to a heatexchanger
An interesting phenomenon was observed during theexperiment on PCM coupled heat pipe. Though thecondenser was maintained at 60oC, the evaporatortemperature was less than 60oC initially due to the presenceof PCM module. Heat pipe was operating with flow ofammonia vapour from condenser end to evaporator end andflow of liquid ammonia in the grooves from evaporator endto condenser end. When a heat load of 50 W was appliedon the evaporator, its temperature increased to condensertemperature and heat pipe operation ceased. This resultedin sudden rise of heat pipe temperature and PCM modulefor next 90 seconds. When evaporator temperaturesufficiently rose above condenser temperature, heat pipestarted operating by reversing the direction of working fluidin the vapour core and grooves. This flow reversalphenomenon may be observed in the following plot whichshows evaporator and condenser temperatures of the heatpipe.
Fig. 11. Temperature vs Time plot showing the flow reversalprocess
25
30
35
40
45
50
55
60
0 150 300 450 600 750 900
Tem
pera
ture
(°C
)
Time(s)
HP+PCMHP
58
59
60
61
62
63
64
65
66
0 150 300 450 600 750 900
Tem
pera
ture
(°C
)
Time(s)
PCM+HPHP
58
59
60
61
62
63
64
65
66
0 150 300 450 600 750 900
Tem
pera
ture
(°C
)
Time(s)
EvaporatorCondenserPCM Module
Heater Heat PipeEVA
PCM module
Heat of superheat equivalent to about 5oC temperature risefor 90 seconds was required to cause the reversal of fluidflow in the heat pipe. Such phenomenon is not reported inany known literature of open source.
One more experiment is conducted by filling the PCMmodule with n-Eicosane and coupling to heat pipeevaporator. 50 W of heat for 10 minutes is removed fromthe condenser through natural convection to thesurroundings. Following plot gives heat pipe transienttemperatures without PCM module and by coupling PCMmodule.
Fig. 12. Heat Pipe evaporator temperature when coupled withPCM Module (Paraffin Wax) having condenser section exposed toambient.
Since n-Eicosane is a single compound pure substancewhich melts at a temperature below 40oC and has samemelting point and solidification point temperature, it is apotential candidate for use in space applications.
VIII. CONCLUSIONS AND FUTURE WORK
Efforts have been put to develop 1-D mathematical modelto simulate steady state operation of an axially groovedheat pipe and estimate maximum heat transport capability.This model will be improved further to include puddleeffect of the working fluid at higher temperatures.Operation of heat pipe under transient heat loads will bestudied by developing a transient mathematical model as apart of future work.
Several experiments were conducted to characterise a givenheat pipe and calculate its maximum heat transportcapability at different temperatures. A PCM module isdesigned and thermal characterisation tests were conducted.This PCM module was thermally coupled to the evaporatorof the given heat pipe. Series of experiments wereconducted to study thermal behaviour of heat pipe in thepresence of PCM module. An interesting phenomenon offlow reversal was noticed inside the heat pipe.
Following activities are planned as a part of future work:
Coupling of PCM module in adiabatic zone andcondenser zone separately and analyse thermalcharacteristics.
Shift the heat source from heat pipe side to PCMmodule side and repeat all the above experiments.
Use thermal conductivity enhancers like finned topplate, metal foam or honeycomb structure inside thePCM module and repeat the experiments.
All efforts will be made to develop a state-of-the-art PCMcoupled heat pipe for spacecraft thermal controlapplications.
IX. REFERENCES
[1]Y. Chen, W. Zhu, C. Zhang and M. Shi, 'ThermalCharacteristics of Heat Pipe with Axially Swallow-tailedMicrogrooves', Chinese Journal of Chemical Engineering,vol. 18, no. 2, pp. 185-193, 2010.
[2]K. Do, S. Kim and S. Garimella, 'A mathematical modelfor analyzing the thermal characteristics of a flat micro heatpipe with a grooved wick', International Journal of Heatand Mass Transfer, vol. 51, no. 19-20, pp. 4637-4650,2008.
[3] Jeong-Se Suh and Young Sik Park, ‘Analysis ofThermal Performance in a Micro Flat Heat Pipe withAxially Trapezoidal Groove’, Tamkang Journal ofScience and Engineering, Vol. 6, No. 4, pp. 201-206 (2003)
[4]A. Faghri, Heat pipe science and technology.Washington, DC: Taylor & Francis, 1995.
[5]D. Reay, P. Kew and P. Dunn, Heat pipes. Oxford:Butterworth-Heinemann, 2000
25
30
35
40
45
50
55
0 150 300 450 600 750 900
Tem
pera
ture
(°C
)
Time(s)
HPHP+PCM
ISTC/CCE/NKS/335
Theoretical and Computational Study of Frictional Effects in Viscoelastic and
Generalized Elastic Contacts
D. Satish Kumar, Narayan K. Sundaram
I. INTRODUCTION
The problem of contacts between nominally clamped
engineering components under the influence of
external cyclic loads is important because of their as
initiators of failure. Micro-slip, high edge-of-contact
stresses and stress gradients combine to produce the
phenomenon of fretting failure. The highly local
nature of fretting contact failure necessitates high-
fidelity contact stress modeling in such applications
as bearings, riveted joints, etc.
In this context, metal-metals contacts have
been studied extensively using elastic contact
mechanics. However, metal-polymer (elastic-visco
elastic) contacts have been studied only sparingly,
with far fewer solutions to problems of fundamental
interest. The list of such solutions dwindles to none
when frictional effects and geometrically conforming
bodies are considered. The present work is a progress
report on a computational and theoretical study to
remedy these deficiencies and better understand the
role of friction in viscoelastic contacts better. A
secondary objective is to explore more general
geometries for elastic frictional contacts.
II. PRELIMINARY COMPUTATIONAL STUDIES
A geometrically conforming pin-plate configuration
was chosen for the problem. As a first approximation,
the elastic pin was modeled as rigid, and the material
of the plate modeled as a delayed viscoelastic
element with two springs and one time-constant. A
material time constant τ=5 s was assumed for
PMMA. The ABAQUS STANDARD implicit Finite
Element analysis (FEA) solver was used to model the
contact. A custom structured contact mesh-maker,
FEMESH2D, developed by one of the authors1 was
Narayan K Sundaram is an Assistant Professor in the
Department of Civil Engineering, IISc
used to generate the mesh for the plate , as shown in
Fig. (1). A total of 324 contact elements were used to
mesh the hole, of which roughly 100 elements were
in contact at peak load. The indenter was modeled
using discrete rigid R2D2 elements. The ABAQUS
four-noded, plane strain CPE4R element with
reduced integration was chosen. At least 100 time-
step increments were used per load step in the
analyses in order to capture the slip history and slip
reversals accurately.
The contact parameters are as follows: Hole
radius R=1 inch, pin-radius Rd=0.995 inch, E
(PMMA) =4.35x105 psi, Poisson’s ratio ν = 0.40. The
viscoelastic network parameters are τ=5 s, G1 = G2
= 2.14x105 psi. In the interest of simplicity, the
viscoelastic material’s Poisson’s ratio was treated as
a constant. The coefficient of friction µ=0.35.
Multiple loading scenarios were tested: 1)
Monotonically increasing, uni-directional pin-load 2)
Monotonically increasing, oblique pin-load 3) Cyclic
pin-load. A peak load of P=10,000 lb/in was chosen,
giving an effective conforming contact parameter
Lp = 2G(R-Rd)/(P(κ+1)) of 0.21. The lower the value
of this parameter, the more geometrically conforming
the contact, and the less valid the half-plane approx-
imation used in Hertz-type contact analyses.
In typical elastic contact analyses, a repeated
(fretting-type) load consists of the following: A
vertical load P applied in the first step with no
horizontal loading. In subsequent steps, P is held
constant, while horizontal load Q is cycled between
+Q0 and –Q0. In an elastic contact, it is well known
that the pressure and shear tractions reach a state
called shakedown, in which the tractions achieved at
D. Satish Kumar is an MSc student in the Department
of Civil Engineering, IISc
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
ISTC/CCE/NKS/335
he extremities of the load cycle do not change from
cycle to cycle.
In viscoelastic contacts, the expected lag in response
of the material makes the ratio of the time-period of
cycling to the time constant τ a critical parameter,
which is absent in purely elastic fretting contacts. We
selected a Q amplitude of 0.5 µ P, and the cycling
time period Tp of the horizontal load as 0.5 τ. The
analysis was run for total time of T=10 τ to capture a
sufficient number of cycles of response.
The pressure and shear tractions in the two cases are
shown in Figures (2) and (3) respectively. The purely
elastic tractions are also shown for comparison. It is
seen that the viscoelastic tractions also shakedown,
just as in the elastic case. The steady-state pressure
tractions differ by very little between the elastic and
viscoelastic cases; however, the peak shear tractions
in the two cases differ by as much as 9 percent. In
future studies, it is planned to do a complete
convergence study of the tractions. Subsequently,
subsurface stresses will be examined. The latter
analyses requires ultra-fine meshes, and replacement
of the contact boundary conditions by traction
boundary conditions using T, N obtained from the
contact analysis, with quadratic elements used
subsurface.
Figure 2: Normalized pressure tractions N(θ) R / P during horizontal load cycling. The elastic tractions (cyan and magenta lines) are shown
for reference. The time steps for these plots are t=2 τ (only P, black
lines). The blue and red lines are pressure tractions after 7 and
7.5 cycles of horizontal loading
Figure 3: Normalized shear tractions T(θ) R / P during
horizontal load cycling. The elastic tractions (cyan and
magenta lines) are shown for reference. The time steps
for these plots are t=2 τ (only P, black lines). The blue and
red lines are shear tractions after 7 and 7.5 cycles of
horizontal loading.
Figure 1 Close up view of the finite element mesh
REFERENCES
Golden and Graham 1988; Boundary Value
Problems in Linear Viscoelasticity, Springer
Sundaram N., Farris T.N., 2010; J. Mech. Phys.
Solids, 58(11) 1819-1833
ISTC/MAE/CO/322
Combustion studies of composite propellants
containing nano- burn rate catalysts
Charlie Oommen, R Arun Chandru
Department of Aerospace Engineering, Indian Institute of Science, Bangalore
and
R Rajeev
Vikram Sarabhai Space Center, Thiruvananthapuram
This paper presents our research initiative to
comprehensively study the combustion
properties of composite propellants in the
presence of nano-burn rate catalysts, and their
relation to the thermal decomposition properties
of AP, and the properties of the nano-catalysts.
Our preliminary results indicate that a simple
and direct co-relation between the thermal
decomposition properties of AP and the burning
rate of composite propellant, as suggested in
recent literature, does not necessarily hold true
in the case of nano-catalysts. On-going studies
will provide further insights into the factors and
mechanisms affecting the ballistic properties of
composite propellants in the presence of nano-
burning rate catalysts.
1. Introduction:
Ammonium Perchlorate (AP)-based solid rocket
propellants are extensively used in strategic
military vehicles and in launch vehicles for
space applications. The internal ballistic
properties of these composite propellants are
mainly governed by the thermal decomposition
characteristics of their major ingredient, AP (60–
90% w/w). The decomposition properties of AP
are known to be particularly sensitive to the
presence of certain additives, even in small
amounts. These additives are used as ballistic
modifiers to tailor the propellant’s ballistic
properties. Transition metal oxides are effective
catalysts for the thermal sensitization of AP, and
are consequently incorporated as burning rate
enhancers in AP-based propellant formulations
(0.1–3% w/w). Some proven transition metal
oxide catalysts used for burning rate
enhancement include Fe2O3, CuO, Cu–Cr–O and
MnO2 [1,2].
Of late, numerous nano-catalytic systems are
being evaluated for the thermal sensitization of
AP. It is reported that nano-metal oxides exhibit
superior catalytic activity than their micron-
sized counterparts, by way of lowering the
decomposition temperature, increasing the
reaction rate and enhancing the specific heat
release of AP decomposition. The enhanced
activity of these nano-metal oxide systems
towards AP decomposition has generally been
attributed to the higher surface area in the case
of nanoparticles, the enhanced defects in nano-
structures, and the synergistic effects of the
catalyst and support in nano-composites [3-5].
Although there are several reports on the
enhanced effect of nano-catalytic systems on AP
decomposition, reports regarding the ballistic
property modification of the composite
propellants with nano-catalysts are sketchy and
mostly contradictory. There are hardly any
studies that try to relate the properties of nano-
catalysts to the decomposition properties of AP,
and subsequently to the combustion properties
(burning rate, pressure exponent etc.) of the
processed AP-based composite propellant.
Our research group has previously worked on
the syntheses, characterization and evaluation of
various nano-catalytic systems for AP
decomposition. Some of the nano-catalysts that
we have synthesized and evaluated include
nano-structured Cu-Cr-O, Fe2O3, mesoporous
MnO2, ZnO etc. (see Fig. 1). Particularly, the
mesoporous MnO2 catalyst prepared by our
group shows the best catalytic activity towards
AP decomposition reported till date in open
literature [5].
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
This paper presents our research initiative to
comprehensively study the combustion
properties of AP-based composite propellants
containing various nano-burn rate catalysts, and
some preliminary results regarding the same.
Efforts to relate the combustion properties of the
propellant to the decomposition properties of
nano-catalyzed AP, and in turn to the nano-
structural properties of the catalysts are being
pursued. Various propellant processing aspects
involving the nano-catalysts like dispersion,
stability etc. are also being studied. During the
course of this research program, our objectives
mainly revolve around (a) creating a
comprehensive database of decomposition and
combustion characteristics of AP-based
composite propellants containing nano- burn rate
catalysts, (b) establishing co-relations between
the nano-catalyst properties, AP thermal
decomposition characteristics, and AP-based
composite propellant combustion characteristics,
and (c) investigating into the mechanism of
activity of nano- burn rate modifiers in AP-
based composite propellants.
Figure 1: Representative SEM images of some nano-
catalysts prepared by our research group:
(a) nano-structured Cu-Cr-O, (b) nano-structured
Fe2O3, and (c) mesoporous MnO2
2. Experimental
2.1 Materials: Various transition metal salts and
other chemicals used in the preparation of
(nano)-catalysts were all of analytical grade and
used as received without further purification. AP
used in our experiments was kindly provided by
IIT, Madras. Hydroxyl-terminated polybutadiene
(HTPB), Dioctyl adipate (DOA), and Isophorone
diisocyanate (IPDI) used for making the
composite propellants were provided by VSSC,
Thiruvananthapuram.
2.2 Syntheses and characterization of nano-
catalytic systems: Methods previously reported
in literature like homogeneous co-precipitation,
mesoporous templating, electrospinning etc.
were used to synthesize transition metal nano-
catalysts. Also, synthesis methods developed in-
house by our research group like impinging
streams and submerged spray precipitation were
used. Some of the nano-catalysts that we have
synthesized and evaluated include nano-
structured Cu-Cr-O, Fe2O3, mesoporous MnO2,
ZnO etc. The nano-catalysts were characterized
via techniques such as UV-Vis-IR Spectroscopy,
XRD, SEM, TEM, EDS and ICP-AES.
2.3 Evaluation of nano-catalytic systems: AP-
HTPB (80:20% w/w) composite propellant
formulations containing nano-burn rate catalysts
were prepared by standard methodology.
Custom-designed micro-V-blender and planetary
micro-mixer with hot-water circulation jacket
were used to mix the propellant compositions.
Multiple blade configurations were designed and
tested experimentally. The mixing homogeneity
achieved using our custom-designed micro-
mixers were comparable to that reported in
literature, as demonstrated using optical and
SEM imaging. Both micron-sized and nano-
structured catalysts could be processed
efficiently, under slightly different conditions.
Cylindrical strands of 50mm length were cast
under vacuum using custom-made casting
apparatus, and cured in a PID controlled hot-air
oven. The mixing homogeneity in the presence
of nano-catalysts was verified by optical and
SEM images. The propellant strands were
ignited using a heated nichrome wire, and the
burning rate was acquired via fuse wires
connected to a data acquisition system. Burning
consistency was monitored via a long-range
high-speed optical camera. Experiments at high
pressures upto 120 bars are planned in our high
pressure propellant strand burner being
fabricated. Thermal analyses techniques like
DTA-TG, DSC and EGA with FTIR & GC-MS
are also being planned to investigate into the
mechanism of activity of nano- burn rate
modifiers in AP-based composite propellants.
Table 1: Summary of catalyst surface area, decomposition temperature of catalyzed AP and burning rate
of AP-based composite propellant
3. Results and Discussion:
Fig. 1 shows representative SEM images of (a)
nano-structured Cu-Cr-O, (b) nano-structured
Fe2O3, and (c) mesoporous MnO2, prepared by
our research group. As shown in Fig. 2, the
addition of 2% w/w mesoporous MnO2 to AP
exhibits exceptional catalytic activity, as
compared to its micron-sized counterpart;
manifested as unprecedentedly low
decomposition temperatures, fast reaction rates
and enhanced heat releases. Structural features
of the nano-catalyst like its high surface area and
the presence of a 3D mesoporous network,
which facilitate the adsorption and desorption of
reactive molecules during the secondary gas-
phase processes, account for its observed
excellent catalytic activity.
Figure 2: Non-Isothermal TG and DTA curves of (a)
pure AP, (b) AP + 2% w/w micron-sized β-MnO2 and
(c) AP + 2% w/w mesoporous β-MnO2
One of the main arguments reported in literature
is that nano-catalysts, which are effective for
thermal sensitization of AP, when incorporated
in AP-based composite propellants, lead to
significantly higher burning rates [3-5].
However, as seen from Fig.2 and 3, and Table 1,
this is not necessarily true. Though mesoporous
MnO2 exhibits the best catalytic activity towards
AP decomposition reported in literature till date,
its effect on the burning rate of AP-HTPB
composite propellants is modest compared to
Fe2O3. These preliminary results with
atmospheric pressure burning rates reveal that
there is no direct co-relation between the AP
thermal decomposition characteristics, and AP-
based composite propellant combustion
characteristics in the presence of nano-catalysts.
Properties of the nano-catalysts such as its
phase, crystallanity, surface area and porosity
may be playing underlying roles in such
observation. Such a mechanism would be clear
after the on-going studies involving Fe2O3 nano-
catalysts with varied morphologies and surface
areas. These Fe2O3 catalysts are being prepared
by our in-house developed continuous
submerged spray precipitation technique,
wherein the nano-catalyst properties can be
varied in a facile manner by varying the reactant
conc. and/ or the inlet air pressure. Further, the
combustion studies in a high-pressure strand
burner are being intended, and are expected to
provide insights on the burning rates of nano-
catalyzed composite propellants at various
pressures and their corresponding pressure
exponents.
Figure 3: Representative images of combustion of
AP-HTPB composite propellant strands (a) without
catalyst, (b) with 2% w/w mesoporous MnO2 and (c)
with 2% w/w nano-structured Fe2O3,
SampleSurface area of
catalyst
Decomposition
temperature
Burning rate
(atm)
m 2 /g o C mm/sec
Pure AP - 426 2.1
AP+2% w/w micron-sized Fe2O3 14.5 361 3.0
AP+2% w/w nano-structured Fe2O3 66.9 310 3.8
AP+2% w/w mesoporous MnO2 86.7 273 3.2
Conclusion
This paper briefs about our research program
and progress regarding the combustion studies
of AP-based composite solid rocket propellants
containing nano- burn rate catalysts. Various
experimental set-ups have been/ are being
designed, fabricated and validated. Ongoing
studies indicate that the effectiveness of a nano-
catalyst in the thermal decomposition of AP
does not always translate into higher burning
rates of the composite propellant.
References
1. P. W. M. Jacobs and H. M. Whitehead,
Decomposition and combustion of
ammonium perchlorate; Chemical Reviews,
1969, 69 (4), 551–590.
2. David L. Reid, Antonio E. Russo, Rodolphe
V. Carro, Matthew A. Stephens, Alexander
R. LePage, Thomas C. Spalding, Eric L.
Petersen, and Sudipta Seal, Nanoscale
Additives Tailor Energetic Materials; Nano
Letters, 2007, 7 (7), 2157–2161.
3. Wei Li and Hua Cheng, Cu–Cr–O
nanocomposites: Synthesis and
characterization as catalysts for solid state
propellants; Solid State Sciences, 2007,
9(8), 750–755.
4. Satyawati S. Joshi, , Prajakta R. Patil and V.
N. Krishnamurthy, Thermal Decomposition
of ammmonium perchlorate in the presence
of nanosized ferric oxide; Defence Science
Journal, 2008, 58(6), 721-727.
5. R. Arun Chandru, Snehangshu Patra, Charlie
Oommen, N. Munichandraiah and B. N.
Raghunandan, Exceptional activity of
mesoporous β-MnO2 in the catalytic thermal
sensitization of ammonium perchlorate;
Journal of Materials Chemistry, 2012, 22,
6536-6538.
Debadrita Paria is a PhD student in the Centre for Nano Science and Engineering at the Indian Institute of Science, Bangalore, 560012, INDIA. [email protected] Ambarish Ghosh is an Assistant Professor at the Centre for Nano Science and Engineering, and an associate faculty in the Departments of Physics and Electrical Communication Engineering, Indian Institute of Science, Bangalore, 560012, INDIA. [email protected]
Project Ref No.: STC/P-337 dtd 25.3.2014
Electromagnetic field enhancement in sub-nm gaps
Debadrita Paria and Ambarish Ghosh
Abstract- We describe computational results on the
calculation of electromagnetic field enhancement at the
junction of plasmonic nanoparticles. In particular, we
assume a pair of Silver nanoparticles at variable
distances from each other and calculate the wavelength
and gap size dependence of the electromagnetic field
enhancement.
I. INTRODUCTION
The large concentration of electromagnetic fields near
metal nanoparticles upon illumination of light is one of the
defining concepts of light matter interactions at the
nanoscale [1,2]. The strength of this so called “near field”
can be very large at the junction of two metal nanoparticles,
which has facilitated promising technologies, such as the
detection of single molecules [3,4] through a large
enhancement of the Raman scattering signal. The degree of
enhancement of the near field increases rapidly as the
particles are brought closer, although at separations below a
few Angstorms, quantum tunneling of plasmons can limit
the enhancement. The purpose of this paper is to calculate
the degree of enhancement as a function of gap distance
down to few Angstorms, and investigate the wavelength
dependence of the field enhancement.
II. PLASMONIC COUPLING
In electrostatics, the electric field between two oppositely
charged plates varies inversely as the distance between
them, which arises due to Coulomb interactions between
the charges residing on the two plates. Similarly, the
dependence of the EM field enhancement9 on the size of the
gap (d) originates in the electromagnetic interactions
between the localized plasmon modes of two neighboring
nanoparticles (radius a). In the simplest approximation,
each particle can be approximated as an electric dipole,
whose strength is proportional to the net electric field
( = ) at its location, where is the polarizabiltiy of
the particle at the wavelength of illumination. The
contributions to the net electric field for particle 1 come
from the incident field ( ), as well as the electric field due
to the dipole () of particle2 at the location of particle 1,
where the dominant term of ∝ , thus implying a
strong distance dependence of the dipolar coupling. This
approximation breaks down for < , beyond which
higher order multipolar contributions needs to be taken into
account, where numerical calculations can provide an
accurate picture. Also important is the direction of
propagation and state of polarization of the incident beam,
and their relation to the interparticle axis.
III. RESULTS AND DISCUSSIONS
Figure 1: Light incident along the axis of a nanodimer. See text for details.
As an example, depicted schematically in Fig. 1, we have
considered linearly polarized monochromatic light incident
along the axis of the interparticle separation for two silver
nanospheres of radius a = 70 nm separated by a distance d.
The calculated value (using Comsol) of the squared
enhancement (|| ||⁄ )of the electric field, shows
concentration of the electric field at the junction of the
particles, which increases manifold as the distance d is
reduced. The maximum value of the enhancement within
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
the junction, given by (|| ||⁄ ) ,depended strongly on
the illumination wavelength and the interparticle separation,
as can be seen in Fig. 2a. The smallest d considered in
these simulations corresponded to the thickness of an atom,
which resulted in maximum value of(|| ||⁄ ) ≥ 10 in
the wavelength regime 405 nm to 530 nm (approx),
implying a system of plasmonic dimers with sub-nm
separation can achieve the strongest EM field
enhancements. The distance dependence of
(|| ||⁄ ) for three illumination wavelengths is shown
in Fig. 2b.
Figure 2:Light of certain wavelength, linearly polarized along the x-axis, incident along the interparticle axis (z-axis) of two silver nanospheres separated by a distance d. a, Spectral dependence of the maximum value of the enhanced field(|| ||⁄ )in the dimer junction for various d (labeled in nm). b, Dependence of the maximum enhancement (|| ||⁄ )
at the dimer junction on the size of the gap d = 0.34 nm, for three illumination wavelengths. The particular choices of the wavelengths (red – 620 nm, green – 530 nm, blue – 470 nm) are arbitrary, covering the visible range of the electromagnetic spectrum.
Interestingly, at small gaps (d< 0.68 nm) the dipolar
plasmonic modes, which are typically in the UV, were seen
to shift in the visible regime. This is interesting, as one of
our major aims in this project is to develop radiation
detectors in the NIR and IR region of the electromagnetic
spectrum.
IV. SUMMARY AND FUTURE DIRECTIONS
Our efforts in last few months have been to develop a
computational model to calculate the electromagnetic field
enhancement at the junction of a nanosized plasmonic
dimer. These calculations although trivial at large values of
d, tend to become time consuming and difficult when d
becomes less than a nm. As expected, over six orders of
intensity enhancement could be observed. It is important to
note that the present calculation does not allow for quantum
tunneling of the plasmons [5], which can become a limiting
mechanism below a dimer gap of 0.3 nm. Accordingly, the
enhancements obtained here could be the largest possible,
as reducing the gap further will only result in a reduced
field magnitude.
Going forward, we wish to try experimental validation of
these calculations. Of particular interest would be to
investigate if the plasmonic enhancement can be moved
further to NIR and IR regimes of the EM spectrum, such as
to couple the system with a photodetector material.
REFERENCES
[1] Maier, S. A. Plasmonics: Fundamentals and Applications: Fundamentals and Applications. (Springer, 2007). [2] Novotny, L. & Hecht, B. Principles of nano-optics. (Cambridge university press, 2012). [3] Kneipp, K. et al. Single molecule detection using surface-enhanced Raman scattering (SERS). Physical review letters 78, 1667 (1997). [4] Nie, S. & Emory, S. R. Probing single molecules and single nanoparticles by surface-enhanced Raman scattering. science 275, 1102-1106 (1997). [5] Savage, K. J. et al. Revealing the quantum regime in tunnelling plasmonics. Nature491, 574-577 (2012).
Modelling primary atomization of liquid jets
Santosh Hemchandra∗†
1. Introduction
The objective of this project is to develop a modelling technique using a fully compressibleapproach for modelling the atomization of liquid jets. The approach comprises of three mainelements -
1. Level-set method solver to track the location of the gas-liquid interface
2. Compressible flow solvers for the liquid and gaseous phases
3. A Ghost-fluid method module that captures jump conditions arising from mass, momen-tum and energy balance across the gas-liquid interface.
During the period 2013-2014, we analyzed several options for reinitialization scheme for item1 above. These efforts were partially successful at the time of the previous interim reportsubmitted in April, 2014 due to the presence of bugs in the reinitialization scheme. We havesince fixed these bugs and are currently working on phase 2. All of these solvers are beingimplemented within the custom MultiSolv framework developed within our group to realizecomplex multi-physics flow simulations.The activities over the past six months have been dividedinto phases as follows
1. Debug and fix the CR2 [1] level-set reinitialization scheme (current status: done).
2. Incorporate the level-set solver within the MultiSolv framework - (current status: done).
3. Implement a coupling framework within the MultiSolv kernel in and demonstrate couplingbetween flow and level-set solver in order to learn how this can be achieved - (currentstatus: done).
4. Convert existing 2D flow solver to 3D - (current status: done)
5. Parallelize the 3D solver, using MPI, to use structured multi block meshes and supportinput from commercial grid generation software - (current status: done).
The rest of this report is organized as follows. Section 2 provides details of the solver couplingframework implemented within the kernel and presents results from test simulations with a 2Dcompressible flow solver and the level-set method solver. Section 3 provides details about thestatus of the 3D compressible solver development effort and presents some initial results fromongoing simulations. Finally, this report closes with an outlook on next steps for the third year.
∗Assistant Professor, IISc, Bangalore†E-mail: [email protected], STC project # ISTC/MAE/SH/319
1
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Figure 1: Schematic of level-set and flow solver overset grids
2. Solver coupling framework
The solver coupling framework and the mechanism of communication between the flow andlevel-set equation solvers is described in this section. The level-set solver advects gas-liquidinterface as per the velocity imposed by the flow field on either side. The objective of couplingthe level-set solver and flow solvers through an interface within the MultiSolv Kernel is toensure that the former can get the velocity information from flow solver and move the level-set/interface accordingly and the latter can get level-set field values from the former and applymatching conditions via the Ghost Fluid Method (GFM).
The level-set solver uses uniformly spaced structured grids and uses the local level-set methodto solve the level-set equation in order to reduce computational cost. This allows for refine-ment of the the level-set grid in order to resolve small interface features without increasingcomputational cost. The flow solvers on the other hand, use multi block structured grids withnon-uniform spacing that is necessary to correctly resolve flow features. Thus, coupling the twosolvers together requires mapping of the level-set solver grid on to flow solver grid and vice versaso that level-set function values and flow velocities can be interpolated between the two grids asneeded. The level set grid is overlaid on the flow grid as shown in fig. 1. It is created such thatit encloses whole flow domain. The grid size chosen for the level-set grid is ∆xLS = α ∆xF,min,where, ∆xF,min is the smallest distance between any two given points on the flow grid and α isa parameter. Bilinear interpolation is used to interpolate velocities from neighboring flow gridpoints to level-set grid points eg. see fig. 1 - the velocities at points marked ‘a’ and ‘b’ on thelevel-set grid are determined from the flow grid points ‘m’-‘p’ using bilinear interpolation.
Figure 1 shows that the grid points on the two grids do not exactly overlap. Thereforeto interpolate the data (velocity) from flow solver, each level-set grid point needs to haveinformation about its neighboring flow grid points. This is achieved by mapping level-set gridon the flow grid. Within the MultiSolv framework, each solver requests the other to map itspoints onto the latter’s grid and return an index handle per point during computation setup.Subsequent requests for function values at the the mapped points are obtained using thesehandles to address points. Thus, neither solver needs to be aware of the other’s internal datastructures allowing for great flexibility in choosing the numerical method etc. for either solver.We next show results from tracking the evolution of a closed material line in a two-dimensionaljet.
2
(a) (b)
Figure 2: (a) Schematic of flow configuration and boundary conditions (b) Initial level-setlocation
2.1 Test flow problem
Figure 2a schematically shows a two dimensional planar jet flow that is chosen as the baselineflow test case. Inflow forcing is applied such that the preferred varicose mode of the inflowprofile is excited in order to reduce the potential core breakdown length. The inflow velocityprofile is specified using the tanh function such that it varies smoothly from a value of 174.8m/s along the jet centerline to a value of 88.0 m/s in the co-flow region (see fig. xx). Theslot width is chosen as 1.25 mm for this test case, yielding Re = 6795. The solution domainis 50h long and is discretized with a uniform grid with ∆x = 0.09h and ∆y = 0.063h. Note,this flow configuration was chosen in order to leverage a pre-existing computation available.Both the shear layers are forced with and amplitude of 1% of the centerline velocity, in order toexcite the varicose mode. A time step of 0.001∆x is used and simulations are run for 0.5 flowthrough times (approximately 2000 time steps). All boundary conditions are applied using theNavier-Stokes characteristic boundary conditions (NSCBC) technique [2, 3].
Figure 2b shows the initial condition for the closed material curve as a red circle of diameter1h located along the centerline at a distance of 15h from the jet inflow plane. This position cor-responds to the start of the jet breakup location and would hence result in maximum distortionof the material surface within the given simulation time. Figure 3a shows several snapshots ofthe evolution of the the material curve at various times. The evolution of the curve is trackedusing the level-set solver which in turn, gets the velocity field from the flow solver. Note alsothat the curve is deformed into very thin filaments over time. These filaments then break upinto smaller pockets as the curve evolves. This can be explained as follows. The thickness ofthe filament reduces until such a point where it can no longer be locally resolved on the level-set mesh and hence breakup occurs. Physically, this would occur when the filament thicknesswas of the order of molecular length scales - a length scale that cannot be resolved within theframework of a continuum assumption based CFD solution. However, since the level-set meshcan be refined in our method independently of the flow mesh, filament breakup can be caused tooccur arbitrarily close to the physical value without significant increase in computational cost.This fact can be seen qualitatively in fig. 3 which shows the shape of the curve at t∗ = 0.172for various values of mesh refinement ratio, α. Notice that finer mesh, i.e. progressively smallervalues of α results in retention of longer unbroken filaments.
3
Figure 3: Evolution of the material curve shown at various instants in time
Figure 4: Influence of level-set mesh refinement ratio, α = ∆xLS/∆xFlow, on the materialline geometry (t∗ = 0.172). The background field shows vorticity magnitude
This concludes the summary of the level-set - flow solver coupling framework implementedwithin the overall framework of MultiSolv. We next summarize progress made on the devel-opment of a, 3D fully compressible, massively parallel, structured multi-block DNS/LES solversince April 2014.
3. Three dimensional flow solver
The 2D flow solver used for the above simulations has been extended to be able to computethree dimensional flows. Together with the explicit filtering technique, this solver can in generalbe used for LES of complex turbulent flows. The solver has also been parallelized using MPIand has been implemented within the framework of MultiSolv. As such, the solver couplingmethodology described in the previous section extends to the present three dimensional caseas well. Our implementation solves the Navier-Stokes equations in strong conservation form ingeneralized co-ordinates. As such, the present code can be used to compute flows in complex,industrially relevant geometries as well. As such, an input interface for grids specified in the
4
(a)
0 20 40 60 80 100 120 1400
20
40
60
80
100
120
140
Number of processes
Sp
ee
du
p [
t(1
)/t(
k)]
Speedup (1213)
Speedup(Linear)
(b)
Figure 5: (a) instantaneous slice plot of pressure (p-po) showing evolution of gaussianpressure pulse in a rectangular domain (b) a typical speedup plot up to 128 processors for1213 grid points
CGNS format has been implemented in the solver. CGNS is a fairly widely used industrystandard format that is supported by many standard mesh generation packages including inour case, ANSYS ICEM-CFD. Thus, the present 3D code has a significantly greater range ofproblems, i.e. from academic geometries to industrially relevant geometries, to which it can bepotentially applied.
The two dimensional NSCBC procedure [2, 3] has been reformulated for NS equations in gen-eralized co-ordinates. Currently, subsonic inlet, constant temperature no-slip wall and subsonicnon-reflecting outlet boundary conditions have been implemented as these will be necessary torealize the spray simulations that are the final goal of this project. Currently, validation anddebugging of each of these boundary condition implementation is ongoing and is expected to befinished by January 2014. We next present two test cases computed with this code and resultsfrom scaling tests on our in-house compute cluster.
3.1 Spherical pressure pulse
This case computes the evolution of a spherical pressure pulse in a cube of dimension 0.1 m. Thedomain is discretized using a uniform mesh with 121× 121× 121 points. No slip wall boundaryconditions are applied on all faces of the cube. The solution is initialized with a stationary flowat a temperature of 300 K. The wall temperature is assumed to be the same at all times. Theinitial pressure distribution is given by
p
po= 1 + 0.0001 × exp
(− (r−ro)2
2×c2)
(1)
where, po = 1.01325 × 105 and r is the radius vector referenced to the center of the cube.Figure 5a shows a typical snapshot of the the solution at a time t = 3× 10−5 s after the start ofthe simulation. This test case was used to test scaling results. A result from preliminary scalingtests with the inviscid version of the code, i.e. before transport terms were implemented, areshown in fig. 5b. This shows a reasonable level of scaling upto 128 processors. We have notbeen able to perform scaling tests with the current code at the moment as we are focusing ondebugging and implementing boundary conditions. We will perform scaling tests on the fullcode during the next year.
5
(a)
x/D
y/D
0 0.5 1 1.5 2 2.5−0.5
0
0.5
2.3
2.4
2.5
2.6
2.7
x 10−6
p
(b)
Figure 6: (a) schematic for the flow in a square duct showing boundary conditions andinlet velocity profile. (b) pressure and velocity fields for the square duct case
3.2 Flow in a square duct
Figure 5a schematically shows the flow setup for a flow in a square duct. The profile of thex-component of the velocity at the inflow is specified a top hat profile going from a centerlinevelocity, UCL, of 4 m/s to 0 m/s using smooth profiles generated using hyperbolic tangentfunctions. Likewize, the velocity profile at the inlet is extended into the domain along thestreamwise direction . The other velocity components are set to zero. The temperature at theinlet is specified as 300K. The reference pressure at the outlet for the NSCBC procedure isspecified as 1.01325 × 105 Nm−2. Figure 6b shows the solution at time t = .04 s correspondingto 0.32 flow through times (L/UCL). The field shows the instantaneous pressure distribution.Also shown are typical velocity vectors at selected locations close to the inlet. Note this solutionhad not reached steady state as of the time of writing this report.
4. Conclusion and next steps
The simple flows described in the previous section are being used at this time to debug andvalidate the 3D solver. We will next generalize the thermodynamics model in the solver to beable to use arbitrary equations of state, thereby, allowing the same code to be used for theliquid phase as well. Finally, we envisage coupling the this solver with the level-set solver andperforming first multiphase flow simulations begining April 2015.
References
[1] D. Hartmann, M. Meinke, and W. Schroder. Differential equation based constrained reini-tialization for level set methods. Journal of Computational Physics, 227(14):6821–6845,2008.
[2] T. J. Poinsot and S. K. Lele. Boundary conditions for direct simulations of compressibleviscous flows. Journal of computational physics, 101(1):104–129, 1992.
[3] K. W. Thompson. Time-dependent boundary conditions for hyperbolic systems, ii. Journalof Computational Physics, 89(2):439–461, 1990.
6
Development of an optimization based image processing software system for Indian forest resource assessment using Radar Imaging Satellite (RISAT-1)
Images
S N Omkar and G. Narayana Naik Principal Investigator, Dept of Aerospace Engineering, Indian Institute of Science, Bangalore
Ashoka Vanjare andSourabh Rao A Project Assistant, Dept of Aerospace Engineering, Indian Institute of Science, Bangalore
P G Diwakar Co- Investigator, NRSC, Hyderabad, India
Prashant BK Project Intern, Dept of Aerospace Engineering, Indian Institute of Science, Bangalore
Abstract- This paper examines the results of optimization based Image processing software system developed at Research and Development Center, Computational Intelligence Lab, Indian Institute of Science, Bangalore. Currently we are developing data processing system for processing forest land cover classification problem by using RISAT-1 satellite images. In this paper, we are exploring use of optimization techniques for improving the results and advantage in image analysis.Developed imaging system provides image enhancements, speckle reduction, and image classification; substantially increase the functionality of the software along with application of remote sensing data in different fields. This leads to optimal usage of remote sensing data.
I. INTRODUCTION Basic essential of life is land, water and air. Land cover refers to the physical and biological cover at the surface of the earth, including water, vegetation, bare ground, man-made structures, etc. Land cover information [1] acts as important piece of information for assessment of land utilization and resources optimally. Due to its advancement in digital technology, robustness and cost-effectiveness, remote sensing has been increasingly used to extract land cover information through either manual interpretation or automated classification. Further researchers are using optimization techniques [2] in order to improve usage of available resource. Pattern recognition techniques [3]are commonly conceived to have the capability of improving automated classification accuracy due to their distributed structure and strong capability of handling complex data. Many researchers are using neural networks techniques for land cover classification and comparing with the traditional statistical methods. This paper is divided into six sections, radar dataset, study area, data processing methods, results and conclusion.
II. RADAR IMAGE DATASET
In this section, different space Agencies and their SAR imaging products are given. 1) Indian Space Research Organization (ISRO) and National Remote Sensing Centre (NRSC) Example:RISAT-1 and 2 2)National Aeronautics and Space Administration (NASA) Example: radar Sensor SIR-A, SIR-B and SIR-C. 3) Japan Aerospace Exploration Agency (JAXA) Example: Advanced Land Observing Satellite ALOS 1 and 2 4) European Space Agency (ESA) Example: ENVISAT ASAR Different space agencies have launched different earth observation missions in order to collect earth land surface features. Initially data is acquired, calibration and validation, which involve sensor characterizations, image quality measurements, and accuracy improvement. Processed data is dissimilated to public use as standard products like L0, L1 and L2 products for remote sensing applications. ISRO recently launched a satellite mission with SAR sensor as payload called RISAT-1 satellite. RISAT-1 radar satellite images have information on land surface features. RISAT-1 dataset works in ScanSAR, strip and spot modes to provide images with coarse, fine and high spatial resolutions respectively. This is very useful in different remote sensing applications and one such application is forest resource assessment.
III. STUDY AREA In this section, study is discussed. We have chosen Bangalore region for studying urban vegetation features and Mandya and Mysore region for studying forest vegetation. Forest region consists of agricultural region, water region and hilly regions. We have initially used unsupervised classified techniques for forest region classification before collecting ground truth data.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
IV. DATA PROCESSING METHODS In this section, we are discussing different data processing methods. They are: 1) Preprocessing (Geometric correction and Radiometric calibration), 2) Speckle reduction and3) Image Classification.
1) Preprocessing: (Geometric correction and Radiometric calibration)
Initially Polarimetric SAR satellite image are pre-processed. Radio metrically [4] calibrate – Calculate Sigma nought and beta nought of the given image and after this step geometric correction is performed. Antenna pattern correction is done to calculate the variation and remove the variations caused in transmission and receiving of signals.
Fig: 1 a) RISAT-1 Polarimetric SAR (Combination hh-hv-hh) is overlaid on the Landsat 8 multispectral optical satellite image and b) RISAT-1 image is compared with Landsat 8 image for urban vegetation classification and c) MysoreForest region is considered as region of interest for forest region classification. Given polarimetric SAR satellite image (HH and HV) is geometrically registered [4-6] to optical satellite image in order to geocoding the data. The geocoded image is shown in figure -1.
2) Speckle reduction techniques: As polarimetric SAR image is affected by noise [4] so filtering techniques are applied in order to remove the noise. a) Lee Filter: Lee filter is used to smoothen the image and remove noise from data. Lee filtering is a standard deviation based filter that filters data based on statistics computed. Lee filters preserve image edges, sharpness and spatial information compared to low-pass smoothing filter. The image pixel value is replaced by a value calculated using the surrounding connected pixels. b) Enhanced Lee: Enhanced Lee filter is used to reduce speckle in radar imagery which preserving texture and edge information. The Enhanced Lee filter is an advanced version of the Lee filter. Enhanced Lee filter uses local statistics-coefficient of variation for window information. Neighboring image pixel value is put into one of three classes namely a) Homogeneous: the neighboring image pixel value is replaced by the average of the filter window, b) Heterogeneous: The neighboring image pixel value is replaced by a weighted average and c) Point target: The neighboring image pixel value is not changed. c) Frost filter: Frost filter is used to reduce speckle while preserving edges and retaining texture information in radar images. It is an exponentially damped circularly symmetric filter that uses local statistics. The pixel being filtered is replaced with a value calculated based on the distance from the filter center, the damping factor, and the local variance. d) Gamma filter:Gamma filter is used to reduce speckle while preserving edges in radar images. Gamma filter assumes that the data is gamma distributed and filtered image pixels are replaced with a neighbor image pixel value calculated based on the local statistics. e) Local Sigma filter:Local sigma filter is used to preserve fine detail in low, medium and fine resolution image and low contrast areas.It is to reduce speckle significantly by using the local standard deviation computed for the filter box to determine valid image pixels within the filter window. f) Bit Error filter:Bit error filter is used to remove bit-error noise, which is usually the result of speckles in the data caused by isolated pixels that have extreme values unrelated to the image scene. Bit-error removal method uses an adaptive based technique to replace speckle or spike pixels with the average of neighboring pixels. Local statistics like- mean and standard deviation are computed within the filter box and it is used to set a threshold for filtering the desired image quality. g) Gray Level Co-occurrence Matrix filter: In statistical texture analysis, texture features are computed from the
statistical distribution of observed combinations of intensities at specified positions relative to each other in the image. According to the number of intensity points (pixels) in each combination, statistics are classified into first-order, second-order and higher-order statistics. Gray Level Co-occurrence Matrix (GLCM) method [4-7] is a way of extracting second order statistical texture features. GLCM is a matrix where the number of rows and columns is equal to the number of gray levels, in the image. Suppose an image to be analyzed is rectangular and has Nx resolution cells in the horizontal direction and Ny resolution cells in the vertical direction.Let the gray tone appearing in each resolution cellsis Ng quantized levels. Let Lx = 1, 2… Nx be the horizontal spatial domain, Ly = 1, 2... Ny be the vertical spatial domain and G = 1, 2...Ng be the set of Ng quantized gray tones. The set LyxLx is the set of resolution cells of the image ordered by their row- column designations. The image I can be represented as a function which assigns some gray tone in G to each resolution cell or pair of coordinates in LyXLx, such that I: LyX Lx --> G. GLCM is a matrix of relative frequencies Pij with which two neighboring resolution cells separated by distance d occur on the image, one with gray tone i and the other with gray tone j. This matrix is a function of the angular relationship between the neighboring resolution cells as well as a function of the distance between them. Texture measurement requires the choice of window size, quantization levels, displacement value, and orientation factors for each texture measure. In this work, each texture measure/feature have been calculated in all 4 orientation angles (0°, 45°, 90°, 135°) and feature measures of the matrices of all the four orientations are then averaged, to get a directionally invariant matrix.
3) Image Classification methods: In this section, we are discussing different image classification methods [3-4] used in this paper. Unsupervised method: In this section, two unsupervised classification methods are discussed. K-means and ISO clustering method are applied to SAR image in order to identify the vegetation regions. a)K-means clustering: K-Means is an unsupervised classification technique which calculates initial class means in evenly distributed in the data space then iteratively clusters the pixels into the nearest class using a minimum distance technique.Each iteration recalculates class means and reclassifies pixels with respect to the new means. All pixels are classified to the nearest class. By setting a standard deviation value or distance threshold value unclassified image pixels are grouped till they meet threshold criteria. This process is continued till thegive number of pixels in each class are assigned till selected pixel change threshold value or the maximum number of iterations is reached.K means parameters
Mean 1752.132369, Standard deviation 1360.323747, Classes-8 and threshold-5 are used for all the filtered images. b)ISO-Data Clustering: ISO-DATA is an unsupervised data clustering technique which calculates class means considering as evenly distributed in the data space then iteratively clusters the remaining pixels using minimum distance techniques.In each iteration,mean is recalculated and the same image reclassified with respect to the new computed means. This process is repeated until the number of pixels in each class changes till selected threshold value is reached else the maximum number of iterations is attained. Supervised classification method: In this section, the three ArtificialNeural Network methods like Multi-Layer Perceptron (NN-MLP),Radial Basis Function - Neural Network(NN-RBF) andCircular Complex valued Extreme Learning Machine(CC-ELM) are discussed. These methods are applied for classification of Multispectral and Polarimetric SAR (Synthetic Aperture Radar) data. Artificial neural network methods c) Neural Networks-Multi Layer Perceptron: Neural Network is a layered feed-forward neural network classification technique. NN-MLP technique uses standard back propagation as supervised learning technique. It can select the number of hidden layers.It uses a logistic or hyperbolic activation function for classification process. NNMLP learns by adjusting the weights in the node to minimize the difference between the output node activation and the output. InNNMLP method, error is back propagated through the network and weight adjustment is made using a recursive process. NNMLP classification technique is used to perform non-linear classification and this method is used as optimized for pattern classification. d) Neural networks-Radial Basis Function:Neural Network - radial basis function (NN-RBF) also uses a feed forward network. It is structurally similar to MLP-NN, but the activation function in the hidden layer nodes is called radial basis activation function and it also uses back propagation as supervised technique. The output of the activation function depends on the location of the center of the function and the spread of the function. The output of a radial basis function can be defined as:
ɸ(x) = exp(-|| x – c ||2 / 2*σ2) Eq 1 Where, c is the center of the RBF unit,x is the input and σ is the spread of the RBF unit. The inputs are first normalized using suitable normalization. This activation function in the hidden layer produces a non-zero response when the input falls within kernel function. Each hidden unit has its own receptive fields in input space. The weights connecting the inputs to the hidden layer decide the spread of the activation function and the weights
connecting the hidden layer to the outputs is used as a scalar multiplier to the hidden layer outputs. The network output is the sum of weighted hidden layer outputs. The training of the network involves the adjustment of the two sets of weights and updating of centers of the hidden nodes. The mean squared error is determined between the network output and the target output values. The centers and weights are initialized randomly. The weights and centers are optimized using Gradient Descent Back Propagation algorithm, by minimizing the instantaneous mean squared error. The training algorithm aims at minimizing the error and the optimization of the weights and the center location. The training is carried out till the target performance is reached. The network is then tested for performance and generalization using the testing dataset. e) Circular Complex Extreme Learning Machine (CC-ELM): Circular Complex-valued Extreme Learning Machine (CC-ELM) [23] is designed for handling real-valued classification problems. CC-ELM is a single hidden layer network with non-linear input and hidden layers and a linear output layer. A circular transformation with a translational or rotational bias term that performs a one-to-one transformation of real-valued features to the complex plane is used as an activation function for the input neurons. The neurons in the hidden layer employ a fully complex-valued Gaussian based activation function. The input parameters of CC-ELM are chosen randomly and the output weights are computed analytically. The input and output configurations of CC-ELM are similar to that of MLP and RBF-NN. In CC-ELM the input layer activation function is ‘Circular Transformation’, which maps the real valued data to the complex domain. The transformation function is give in equation 2.
zt = sin ( axt + ibxt + α ) Eq 2 Where:xt is the input vector for the tth sample and observation,such that each of xti is normalized in [0,1]; 0 < a, b ≤ 1, and 0 <α< 2π. All the three NN-MLP, NN-RBF and NN-ELM methods are used for urban vegetation because of similar techniques for SAR image classification.
V. RESULTS In this section, two applications urban vegetation classification and forest vegetation classification is discussed.For urban vegetation classification, we have used Landsat 8 multispectral image and RISAT-1 SAR image. Initially the SAR image are preprocessed using frost and GLCM filters and enhanced images are classified using three different neural networks-NNMLP, NNRBF and CC-ELM methods. Here we have classified RISAT image and overlaid on multispectral image. Class Number of Pixels Urban 12405 Vegetation 3942 Water 5957
Training Set Size 300 Pixels per class Testing Set Size 750 Pixels per class Table 1: Comparison of different classification techniques In table-1, three different image pixels-urban, vegetation and water is given along with training and testing set size.
Fig 2: a) original Landsat image, b) MLP-NN classified SAR image, c) NN-RBF classified SAR image and d) CC-ELM classified SAR image. Figure 2 gives details of classified images by using three different methods-NN-MLP, NN-RBF and CC-ELM.Results of classified image can be compared with original image. Properties NN-MLP NN-RBF CC-ELM
Hidden neurons
10-30 (best 28) 100 max (accuracy
goal 0.007)
50
Train accuracy
97.30% 98.85% 98.45%
Testing accuracy
97.25% 98.63% 98.40%
Table 2: Comparison of different classification techniques From Table-2 we can see the classification accuracy. These three algorithms are optimized so we are getting good results. By quantitive analysis and value, we can understand by using optimization techniques we can improve classification results. For forest vegetation classification we have used RISAT-1 (HH and HV) images. For forest region classification application we applied six filters lee, enhanced lee, frost, gamma, local sigma, bit error.
Figure 3: Results of lee filtered SAR classified image
Figure 4: Results of enhanced lee filtered SAR classified image
Figure 5: Results of frost filtered SAR classified image
Figure 6: Results of gamma filtered SAR classified image
Figure 7: Results of local sigma filtered SAR classified image
Figure 8: Results of Bit error filtered SAR classified image Figure 3-8 gives the different filtering techniques applied and image quality is enhanced. Six filtered SAR images are classified using K Means and ISO Data clustering techniques into 8 different classes. As we don’t have ground truth data,hence we have usedK-means unsupervisedtechnique for forest region classification.Enhanced lee and frost filtering technique has improved image quality in comparison with other filtering techniques. By visual inspection, we have to determine the classified quality with the given image.
VI. CONCLUSION Speckle reduction technique plays very important role in radar image processing due to which pattern classification results
improves.Initially, we used filtering techniques for preprocessing and enhanced images were classified. For urban vegetation region as we use optimized neural network techniques, results are good and in case of NN-RBF it has classified the given image very well in comparisons to other techniques. To classify forest regions, we initially used unsupervised method to determine different unknown classes. K-means clustering method performs well in comparisons with ISO Data method. In urban vegetation applications, we observed that optimization techniques are very useful in order to improve classification results while in forest region classification application K-means gives initial class information.
FUTURE WORKS Develop different optimization techniques and compare for different forest land classification problem and these techniques are designed into asatellite image processing application. Further these techniques have to be improved by keeping time complexity.
ACKNOWLEDGMENT
The authors thank the reviewers for their critical comments which helped to improve the quality of the paper significantly. This work is supported by IISc-STC cell and carried out by research grant (Project no ISTC0340). Thanks to Aditi Kanjolia, Student, Indian Institute of Technology, Indore, Email: [email protected] and Sandesh C, Student, Indian Institute of Technology, Kharagpur, Email: [email protected]. We are grateful to Forest department of India, Forest survey of India, survey of India, NASA, JAXA and ESA for providing information and support.
REFERENCES
[1] Ulaby, F.T.; Kouyate, F.; Brisco, B.; Williams, T.H.L., "Textural Information in SAR Images," Geoscience and Remote Sensing, IEEE Transactions on , vol.GE-24, no.2, pp.235,245, March 1986 doi: 10.1109/TGRS.1986.289643
[2] Barber, David G., et al. "A comparison of second-order classifiers for SAR sea ice discrimination." Photogrammetric engineering and remote sensing 59.9 (1993): 1397-1408.
[3] Richard O. Duda, Peter E. Hart, David G. Stork, Pattern Classification, 2E" wiley
[4] Lee, J.S., Pottier, E., “Polarimetric Radar Imaging - From basics to applications” , CRC Press (2009))
[5] D. G. Barber, M. E. Shokr, R. A. Fernandes, E. D. Soulis, D. G. Flett, and E. F. LeDrew. A comparison of second-order classifiers for SAR seat ice discrimination. Photogrammetric Engineering &Remote Sensing, 59:1397-1408, 1993.
[6] R. T. Frankot and R. Chellappa. Lognormal randomfield models and their applications to radar image synthesis. IEEE Trans. Geosc. Remote Sensing, 25:195-206, 1987.
[7] Murni, N. Darwis, M. Mastur, and D. Hardianto. A texture classification experiment for SAR radar images. In Proceedings of Pattern Recognition in Practice IV, pages 213-224, Vlieland, Netherlands, June 1994.
[8] A. H. Schistad and A. K. Jain. Texture analysis in the presence of speckle noise. In Proceedings of the International Geoscience and Remote
Sensing Symposium (IGARSS), pages 884-886, Houston, Texas, May 1992.
[9] Berberoglu, Suha, et al. "Texture classification of Mediterranean land cover." International Journal of Applied Earth Observation and Geoinformation 9.3 (2007): 322-334.
[10] M. Tuceryan and A. K. Jain. Texture analysis. In C. H. Chen, L. F. Pau, and P. S. P. Wang, editors, Handbook of Pattern Recognition and Computer Vision, pages 235-276. World Scientific Publishing Company, 1993
[11] Holmes, Quentin A; Nuesch, Daniel R.; Shuchman, R.A, "Textural Analysis And Real-Time Classification of Sea-Ice Types Using Digital SAR Data," Geoscience and Remote Sensing, IEEE Transactions on , vol.GE-22, no.2, pp.113,120, March 1984 doi: 10.1109/TGRS.1984.350602
[12] Solberg, Anne H. Schistad, and Anil K. Jain. "Texture fusion and feature selection applied to SAR imagery." IEEE Transactions on Geoscience and Remote Sensing 35.2 (1997): 475-479.
[13] Schistad, AH.; Jain, AK., "Texture Analysis in the Presence of Speckle Noise," Geoscience and Remote Sensing Symposium, 1992. IGARSS '92. International, vol.2, no., pp.884,886, 26-29 May 1992 doi: 10.1109/IGARSS.1992.578286
[14] Jong-Sen Lee, "Digital Image Enhancement and Noise Filtering by Use of Local Statistics," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.PAMI-2, no.2, pp.165,168, March 1980 doi: 10.1109/TPAMI.1980.4766994
[15] Kuan, Darwin T.; Sawchuk, AA; Strand, Timothy C.; Chavel, P., "Adaptive Noise Smoothing Filter for Images with Signal Dependent Noise," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.PAMI-7, no.2, pp.165,177, March 1985doi: 10.1109/TPAMI.1985.4767641
[16] Frost, Victor S.; Stiles, Josephine Abbott; Shanmugan, K.S.; Holtzman, J., "A Model for Radar Images and Its Application to Adaptive Digital Filtering of Multiplicative Noise," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.PAMI-4, no.2, pp.157,166, March 1982 doi: 10.1109/TPAMI.1982.4767223
[17] Lopes, A; Touzi, R.; Nezry, E., "Adaptive speckle filters and scene heterogeneity," Geoscience and Remote Sensing, IEEE Transactions on , vol.28, no.6, pp.992,1000, Nov 1990 doi: 10.1109/36.62623
[18] Zhenghao Shi; Fung, K. B., "A comparison of digital speckle filters," Geoscience and Remote Sensing Symposium, 1994. IGARSS '94. Surface and Atmospheric Remote Sensing: Technologies, Data Analysis and Interpretation., International , vol.4, no., pp.2129,2133 vol.4, 8-12 Aug 1994 doi: 10.1109/IGARSS.1994.399671
[19]Haralick, R.M.; Shanmugam, K.; Dinstein, Its'Hak, "Textural Features for Image Classification," Systems, Man and Cybernetics, IEEE Transactions on , vol.SMC-3, no.6, pp.610,621, Nov. 1973 doi: 10.1109/TSMC.1973.4309314
[23] Fast learning Circular Complex-valued Extreme Learning Machine (CC-ELM) for real-valued classification problems R. Savitha , S. Suresh , N. Sundararajan
[24] Kim, T., Adali, T.: Fully complex multi-layer perceptron network for non-linear signal processing. Journal of VLSI signal processing 32(1/2), 29-43 (2002)
[25] Tohru Nitta: Orthogonality of decision boundaries in complex valued neural networks. Neural Computing. 2004 Jan; 16 (1): 73-97.
[26] Ortega, J.M.: Matrix Theory. Pienum Press, New York (1986) [27] A. H. Schistad Solberg, A. K. Jain, and T. Taxt, “Multisource
classification of remotely sensed data: Fusion of Landsat TM and SAR images,” IEEETrans. Geosci. Remote Sens., vol. 32, no. 4, pp. 768–778, Jul. 1994.
[28]Pacifici, Fabio, et al. "Urban mapping using coarse SAR and optical data: Outcome of the 2007 GRSS data fusion contest." Geoscience and Remote Sensing Letters, IEEE 5.3 (2008): 331-335.
[29] Ashoka Vanjare, S.N. Omkar, J.Senthilnath,"Satellite Image Processing for Land Use and Land Cover Mapping", IJIGSP, vol.6, no.10, pp.18-28, 2014.DOI: 10.5815/ijigsp.2014.10.03
[30]Ashoka Vanjare,S.N. Omkar,Akhilesh Koul,Devesh,"Aerial Video Processing for Land Use and Land Cover Mapping", IJIGSP, vol.5, no.8, pp.45-54, 2013.DOI: 10.5815/ijigsp.2013.08.06
[31] C.S. Arvind, Ashoka Vanjare, S.N. Omkar, J. Senthilnath, V. Mani and P.G. Diwakar, “Multi-temporal Satellite Image Analysis Using Unsupervised Techniques,” Advances in Computing and Information
Technology (Eds. Natarajan Meghanathan, DhinaharanNagamalai, NabenduChaki), Advances in Intelligent Systems and Computing, 2013, vol. 177, pp. 757-765, Springer Verlag, Berlin, Germany, ISBN 978-3-642-31551-0.
[32] J. Senthilnath, S. N. Omkar, V. Mani, Ashoka Vanjare, P. G. Diwakar"Multi-Temporal Satellite Image Analysis Using Gene Expression Programming"Proceedings of the Second International Conference on Soft Computing for Problem Solving (SocProS 2012), December 28-30, 2012Advances in Intelligent Systems and Computing Volume 236, 2014, pp 1039-1045.
[33] M. E. Shokr. Texture measures for sea-ice classification from radar images. In Proceedings of the International Geoscience and Remote Sensing Symposium (IGARSS), pages 763-768, Vancouver, Canada, 1989
[34] RISAT-1 Handbook, Indian Space Research Organization, India [35]ALOS Data Users Handbook, JAXA, Japan [36] Sentinel-1 User Handbook, European space agency [37] SAR user marine manual [38] CEO Users Guide book [39]J.N Bruning, Radar Imagery: Radarsat-1 [40] PolSAR Tutorials for researchers, European space agency
Numerical Simulations of Darcy-Brinkman-Forccheimer Equations
for Flow in Porous Media∗
Karthick M.†and Gaurav Tomar‡
Abstract – Implementation of Darcy-Brinkman-Forccheimer equations in the open source codeGerris. The adaptive mesh refinement andvariable time stepping in Gerris can be em-ployed to perform high resolution simulationsof two phase flow in porous media.
I Introduction
Multiphase flow in porous media has gained impor-tance in the last decades due to demand for increasedefficiency in extracting oil and other fuel sources suchas shale gas. Porous wicks are employed in the evapo-rators of modern heat pipes such as capillary pumpedloop and loop heat pipes. Liquid refrigerant floodsthe porous wick due to capillary suction and evap-orates due to the heat flux from a heat load. Thevapour so formed is pushed out of the vents into thevapour line and condenses away from the heat gener-ation point. This provides great flexibility in design-ing heat exchangers for electronic cooling in satelliteswhere equipment located in the core of the satelliteneed to dissipate heat in the outer space through ra-diation. Since the experimental investigation are lim-ited by the lack of access to the flow regions in suchsystems, numerical simulations can provide insightsinto the complex phase change heat transfer and fluidflow physics.
∗Project: ISTC/0327†Project Assistant in the Department of Mechanical Engi-
neering, IISc Bangalore‡Assistant Professor in the Department of Mechanical En-
gineerin at Indian Institute of Science Bangalore - 560012.Email: [email protected]
The flow through a porous medium is inherentlymultiscale involving small length scale local resis-tance to flow due to tortuous paths and flow overmultiple order larger domains. Therefore mathemat-ical models for flow in a porous media are derivedby generating volume averaged representations of theflow at smaller micro length scales [1, 2, 3]. Flow inporous media is generally governed by Darcy’s lawwhere the velocity, u, is directly proportional to thepressure gradient:
u = −Kµ
(∇p− ρg) (1)
The permeability, K, is a function of the porousstructure essentially pore diameter, porosity and tor-tuosity. However, the above equation is valid only formoderate flow rates and at higher flow rates convec-tive terms become important and should be included.
In the present study, we develop algorithms forsolving Darcy-Brinkman-Forccheimer equations forflow in porous channels. An operator split algo-rithm has been employed in the open source codeGerris which allows use of adaptive mesh refinementand time stepping. We validate the above numeri-cal implementation using an asymptotic Darcy flowin a channel and Poisseuille flow through a high per-meability channel. The developed algorithm can bereadily extended to simulate three dimensional flowswith complex geometries.
II Governing Equations
The dimensionless form of the governing equations fora non-homogeneous porous medium are given below.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Figure 1: Linear variation in Darcy flow velocity withpermeability. The theoretical stream wise velocity isUt and the numerical results are denoted by Un.
Continuity equation is given by:
∇ · u = 0 (2)
The x−direction momentum equation is given by:
1
ε
∂u
∂t+
1
ε2(u · ∇)u =
1
ρ
[−1
ε∇ (εp)
+µ
ε∇2u− µ
Ku− CF ρ√
k|u|u
](3)
and the y−direction momentum equation is given by,
1
ε
∂v
∂t+
1
ε2(u · ∇) v =
1
ρ
[−1
ε∇ (εp)
+µ
ε∇2v − µ
kv − CF ρ√
k|u| v
](4)
The second term in the above equations are due toBrinkman correction to porous media flow to includethe effect of viscous forces, third term is the lineardrag due to solid matrix resistance to the flow, andthe last term is the Forccheimer correction to includethe non-linear drag when inertial forces become im-portant.
III Numerical Formulation
An operator split algorithm has been employed forthe implicit implementation of the Darcy and For-cchiemer terms [4]. For constant permittivity flows,
Figure 2: Parabolic velocity profile for permeabil-ity K = 0.004 and µ = 0.05 in a channel withno-slip boundary conditions at the top and bottomwalls. The corresponding pressure decreases acrossthe channel linearly.
we solve for u/ε instead of u. This scaling with poros-ity, results in modified momentum equations that areidentical to Navier-Stokes, thus allowing us to employthe standard Godunov schemes for advection termsand Crank -Nicolson algorithm for viscous Brinkmanterms. The Darcy terms are modelled using an oper-ator split that results in:
∂u∗
∂t= − µ
Ku∗ (5)
which essentially results in point updates for individ-ual grid cells. A similar scheme has been used forForccheimer terms.
IV Results and Discussions
We validate the above discussed operator split al-gorithm by solving the Darcy velocity by choosingCF = 0 and not solving for the advection and theBrinkman terms. Figure 1 shows the linear varia-tion in the velocity for a specified pressure with per-meability. The theoretical and numerical curves su-perimpose thus validating the implementation of theDarcy flow. Figure 2 shows the parabolic profile withno slip boundary conditions imposed at the top andbottom walls. The initial uniform velocity profile de-velops into the theoretical fully developed parabolicprofile.
V Conclusions
The basic test cases validate the implementationof the Darcy-Brinkman-Forcchiemer equations. Thepresent code needs to be validated against strictertest cases such as natural convection in a porous me-dia [5]. Subsequently, a two phase flow would be im-plemented by modelling the saturation equation asan advection-diffusion equation [6, 7]. Further, heattransfer and phase change models would enable usto simulate evaporators of capillary pumped loops tounderstand the conditions under which depriming ofthe wick may occur.
References
[1] Bastian P., Numerical computation of multiphaseflows in porous media. Habilitation thesis, 1999.
[2] Bear J., Dynamics of Fluids in Porous Media.Dover Publications, 1972.
[3] Bear A. and D. Nield, Convection in Porous Me-dia. Springer, 2006.
[4] Versteeg H.K. and Malasekera W., An Introduc-tion to Computational Fluid Dynamics. LongmanScientific & Technical, 1995.
[5] Nithiarasu P. and Seetharamu K.N. and Sundara-jan T. Natural convective heat transfer in a fluidsaturated variable porosity medium Int. J. HeatMass Transfer, 40, 1997.
[6] Chavent G and J. Jaffre., Mathematical Mod-els and Finite Elements for Reservoir Simulations.North Holland, 1978.
[7] Tomar G., Numerical Simulations of Two PhaseFlow in a Porous Medium using Volume of FluidMethod STC Symposium, 2011.
Double diffusive convection in the Earth’s core
Venkatesh Gopinath and Binod Sreenivasan
Abstract—The present-day Earth’s dynamo is powered by therelease of light elements due to inner core freezing. Compositionalbuoyancy dominates thermal buoyancy in the Earth’s core, butbefore inner core nucleation, the geodynamo may have beendriven by purely thermal convection. A study is in progress tounderstand the combined effects of thermal and compositionalconvection in a rapidly rotating spherical system permeated bya magnetic field. The onset of convection will be investigated asan eigenvalue problem so that parameter regimes inaccessibleto direct numerical simulations can be reached. We shall theninvestigate the full nonlinear problem with Earth-like boundaryconditions for heat and composition. This study constitutes animportant step in building Earth-like dynamo models.
I. INTRODUCTION
Evidence for the existence of a geomagnetic field goes backto at least 3.5 Gyr. Estimates of the age of the inner core areon the order of 1 Gyr. The geodynamo has therefore operatedin its early stages on purely thermal convection, and later on amixture of thermal and compositional convection. Recent largeestimates of the thermal conductivity of the Earth’s core havereinforced the belief that present-day Earth’s core is largelycompositionally driven.
Double-diffusive phenomena have been studied in vari-ous contexts. However, in the context of the Earth’s core,numerical studies of the dynamo process have used equalthermal and compositional diffusivities. In this so-called ‘co-density’ formulation [1], only one scalar transport equationis solved for. A few recent studies [2], [3] deal with con-vective instabilities in thermal convection within a rapidlyrotating spherical shell. Where double diffusion is present, thebuoyancy profiles for heat and composition convection usedin these studies are not Earth-like. Moreover, no magneticfield effects have been considered. It is accepted that theEarth’s outer core convects more strongly at the bottom thanat the top. Whereas early studies have proposed that the top10 − 20% of the core may be thermally stratified, recentlarge estimates of thermal conductivity have pushed the stablestratification to about 40%. In simple terms, the ratio of thermalto compositional diffusivities in the core may vary significantlyacross different regions of the Earth’s core. Although directnumerical simulations of double diffusive convection appliedto planet Mercury are available [4], Earth-like regimes have notbeen investigated. Three-dimensional dynamo simulations withdifferent thermal and compositional diffusivities and shownthat finger-like narrow plumes penetrate into the thermallystable layer, causing an increase in magnetic field strength inthe stable layer [5] compared to dynamo models that used
Venkatesh Gopinath is a Senior Research Fellow at Centre of EarthSciences, IISc Bangalore, India. [email protected]
Dr. Binod Sreenivasan is an Associate Professor at Centre for EarthSciences, IISc, Bangalore, India. [email protected]
Project number: ISTC/EES/BS/331
the codensity formulation. These studies provide the impetusfor a systematic investigation of dynamo action with double-diffusive convection.
Linear stability analysis offer significant advantages overDirect Numerical Simulation (DNS) in Earth-like magneto-hydrodynamic (MHD) systems. One can obtain meaningfulresults at very low Ekman number E (ratio of viscous toCoriolis forces, ∼ 10−15 for the Earth’s core), which ispresently very difficult to obtain in fully non-linear simulationsbecause of computing constraints. In this study we perform alinear stability analysis of combined thermal and compositionalconvection in an applied magnetic field to obtain the criticalparameters at onset and the structure of convection.
II. GOVERNING EQUATIONS
The governing equations for the velocity U , the magneticfield B, the temperature T and the concentration C are solvedin the Boussinesq approximation. These equations are madedimensionless [5] and written as follows:
E
(∂U
∂t+ U · ∇U
)+ 2z × u +∇Π = E∇2U
+RaTPrTr
roT +RaCPrC
r
roC +
1
Pm(∇×B)×B (1)
∂B
∂t= ∇× (U ×B) +
1
Pm∇2B (2)
∂T
∂t+ (U · ∇)T =
1
PrT∇2T (3)
∂C
∂t+ (U · ∇)C =
1
PrC∇2C (4)
∇ ·U = 0 (5)∇ ·B = 0, (6)
where E is the Ekman number, RaT is the thermal Rayleighnumber, RaC is the compositional Rayleigh number, PrT isthe thermal Prandtl number, PrC is the compositional Prandtlnumber and Pm is the magnetic Prandtl number. These areour dimensionless control parameters. In order to study thegrowth of instability in the system, the initial state of thesystem is perturbed, and the perturbation variables u, b, Θ,T ′ and C ′ are solved for at onset of convection. The velocityand magnetic fields can be expressed as a sum of toroidal andpoloidal vectors:
u = ∇×Ψr + ∇×∇× Φr (7)b = ∇× Γr + ∇×∇× Ξr (8)
and with the application of the operators (∇×) and (∇×∇×)to the Navier–Stokes equation and (∇×) to the inductionequation and taking r-components of the equations, the lin-earized equations governing the perturbations are obtained.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
The resulting perturbations quantities are expanded in normalmodes:
X(r, θ, φ, t) = X(r)Pml (cos θ) exp(imφ+ iωt), (9)
where X = (Ψ,Φ,Γ,Ξ, T ′, C ′) and Pml (cos θ) is the associ-
ated Legendre function of degree l and order m.
III. METHOD OF SOLUTION
In our eigenvalue problem, the eigenvalue is the frequencyω. With temporal derivatives, onset of stationary convectionarises for Re(ω) = 0 and oscillatory convection arises forRe(ω) > 0. For a range of given Rayleigh numbers Ra, a bi-section algorithm is employed to find the eigenvalue ω. The Racorresponding to the eigenvalue obtained is the critical thermalRayleigh number RacT keeping the compositional Rayleighnumber fixed (or vice versa). The pseudospectral collocationmethod is being employed in the radial (r) direction. Theresulting generalized complex eigenvalue problem is given by
AX = λBX (10)
where λ = ω and X = (Ψ,Φ,Γ,Ξ, T ′, C ′).
IV. WORK IN PROGRESS
The eigensolver is presently being benchmarked againstknown thermal convection results. A systematic study will beundertaken at low Ekman numbers where all dimensionlesscontrol parameters are kept constant except for the ratio ofthermal to compositional diffusivities. The effect of this ratioon the structure of the flow at onset is not well understood.The role of a spatially varying azimuthal magnetic field withequatorial antisymmetry (that mimics an axial dipole) will alsobe examined. Finally, the onset will be studied at small radiusratios that represent early-Earth regimes.
Apart from the linear stability calculations, a three-dimensional nonlinear simulation study is also in progress, forwhich a dynamo code is already available. Even as the linearstudy gives insights into the low Ekman number regimes whichare otherwise difficult to simulate, the problem is not steppedforward in time and the back reaction of the magnetic fieldon the flow (via the magnetic Lorentz force) is not treatedadequately. Furthermore, the magnetic field is an applied onerather than a self-generated one, which makes DNS morerealistic than linear onset analyses. The Lewis number (ratioof thermal to concentration diffusivity) used in previous DNS[4] was 10, which is much lower than the value expected inplanetary cores. In our study, we use Lewis numbers ∼ 100-1000, and thermal buoyancy profiles that mimic stratificationin the upper regions of the core. As the dynamics of the innercore is thought to affect outer core mixing [6], determiningthe thermal and solutal conditions at the inner core boundaryposes an additional challenge in our study. It is hoped that ouranalysis would eventually provide valuable data on the pastand present-day evolution of the Earth’s magnetic field.
REFERENCES
[1] Braginsky, S.I., Roberts, P.H., Equations governing convection in Earth’score and the geodynamo, Geophys. Astrophys. Fluid Dyn. 79, 197(1995).
[2] Net, M., Garcia, F., Sanchez, J., On the onset of low-Prandtl-numberconvection in rotating spherical shells: no-slip boundary conditions, J.Fluid. Mech. 601, 317-337 (2008).
[3] Net, M., Garcia, F., Sanchez, J., Numerical study of the onset ofthermosolutal convection, Phys. Fluids. 24, 064101 (2012).
[4] Manglik, A., Wicht, J., Christensen, U.R., A dynamo model with doublediffusive convection for Mercurys core, Earth Planet. Sci. Lett. 289,619628 (2010).
[5] Christensen, U.R., Wicht, J., Models of magnetic field generation inpartly stable planetary cores: applications to Mercury and Saturn, Icarus196, 1634 (2008).
[6] Alboussiere, T., Deugen, R., Melzani, M., Melting-induced stratificationabove the Earths inner core due to convective translation Nature 466,744-747 (2010).
1
Utilizing Ionic Liquid and Mixed Solvent Electrolytes to Synthesize Polymer
Electrolytes for Lithium-based Batteries
Sudeshna Sen,1 Sneha Malunavar,2 C. Gouri3 and Aninda J. Bhattacharyya4
Abstract: We have developed a polymer electrolyte prepared from room temperature ionic liquid electrolytes. The polymer electrolyte possesses ambient temperature ionic conductivity ≥ 10-3 Ω-1cm-
1, mechanical strength with elastic modulus 1 ≈ MPa and electrochemical window ≈ 5 V. Electrolyte films of (20-30)µm thickness were assembled in laboratory cells with various nanostructured cathode and anode materials for application in lithium-ion and lithium-air batteries.
I INTRODUCTION
Over the last few decades, polymer electrolytes have been demonstrated as an important alternative to molecular liquid electrolyte. Polymer electrolytes exhibit compliable mechanical property compared to crystalline compounds and may also exhibit high ambient temperature ionic conductivity. All these combined together make polymer electrolytes very promising for prospective electrochemical devices. In polymer electrolyte, the polymer plays the role of a solid solvent. Polymers with different chemical compositions, architectures are used to completely or partially dissolve salts containing ions such as Li+ /Na+ / H+ salts. However, optimization and theoretical comprehension of material's ion transport mechanism is a non trivial issue. This is primarily related to the extensive spatial and temporal heterogeneity of the polymer host which significantly influences ion transport and in general the electrolyte’s mechanical and electrochemical properties. The focus of the present work is to synthesize cross-linked
polymer electrolytes from room temperature ionic liquids (Scheme 1). The motivation here is to employ them in lithium-ion batteries as an electrolyte and separator. We also wish to theoretically investigate the extent of influence of the host complexities on the ion transport in cross-linked polymer electrolytes. This will be achieved via modeling of experimental data using concepts based on statistical mechanics. The motivation here is to obtain various relaxation processes were analyzed using time-temperature scaling principles. The experimental investigations will include infrared spectroscopy, Brillouin scattering, and impedance spectroscopy. We summarize here the work done under this project.
Scheme 1: Schematic representation of the proposed gel polymer electrolyte
1Sudeshna Sen is a JRF at SSCU, IISc, Bangalore, India, [email protected] 2Sneha Malunavar, PA at SSCU, IISc, India, [email protected] 3C. Gouri is Scientist SG & Head, Lithium-ion Electrodes Section, Lithium-ion and Fuel Cell Division (LFCD) VSSC, Thiruvanthapuram 695022, India, [email protected] 4Prof. Aninda J. Bhattacharyya is an Associate Professor at SSCU, IISc. Bangalore 560012, India, [email protected]
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
2
II PROPOSED WORK AND GOALS A. Cross linked gel polymer electrolyte: Gel polymer electrolytes are attractive alternatives to liquid electrolytes in several electrochemical devices including lithium-based batteries. Gels display high ambient temperature ionic conductivity and highly compliable mechanical property. The main motivation for the usage of gels is mainly targeted towards the enhancement in energy density of batteries. Conventional gel polymers (polymer electrolyte + liquid solvent) do not perform well as their performance deteorates rapidly due to ageing factors leading to very poor battery performance. Thus, there is considerable need to design gel polymers with superior electrochemical and mechanical properties using newer methods and materials. We demonstrate here newer forms of multifunctional gel polymer electrolytes using low melting point solids such as room temperature ionic liquids. Room temperature ionic liquids (RTILs) have been extensively demonstrated as a viable alternative electrolyte in lithium-ion battery applications due to their high thermal stability, ionic conductivity and wide electrochemical voltage window. RTILs have also been used as a medium for synthesis of diverse compounds. In our studies, RTIL was used as a medium to synthesize gel polymer electrolytes. Earlier acrylonitrile was polymerized inside ionic liquid medium to obtain crosslinked gel system [ref 3]. Although, the resulting polymer electrolyte displayed high ionic conductivity, it’s the mechanical properties were poor. In this work we explore the possibility of improving the mechanical property of the gel polymer electrolyte by inclusion of more than one monomer to synthesize cross linked blend polymer-like electrolyte. Free radical polymerization of acrylonitrile and methacrylate (MPTMS) monomers was carried out in pyrolidinium ionic liquid to synthesize
novel crosslinked gel polymer electrolytes. Thermal stability, mechanical property and ionic conductivity of the synthesized cross linked blend-like polymer gel system were studied. From the thermogravimmetry analysis (TGA) data it was observed that this new system was thermally stabile up to 350 °C [figure 3]. At temperatures higher than 350 °C, significant weight loss ≈ 80 % (350-475 oC) was observed. However, the weight retention (> 475 oC) of the blend-like polymer was higher than the pure ionic liquid suggesting higher thermal stability.
Fig 1: Thermogravimetric analysis (above) and differential scanning caloriemetry (bottom; only relevant temperature regime shown) of cross-linked gel samples vis a vis the pure ionic liquid The ionic conductivity is measured by ac-impedance spectroscopy in the frequency range from 1 Hz to 1 MHz (figure 2). The
3
conductivity of this system was found to be of the order of 10-3 Ω-1cm-1 in the temperature range -20 to 60 °C, which is highly beneficial for operation in battery applications. The acrylate monomer was incorporated to improve mechanical property of PAN/IL system. The mechanical property of the blend-like gel polymer electrolytes were studied using static and dynamic rheology measurements (figure 3).
Fig 2: Temperature dependent ionic conductivity of crosslinked polymer gel electrolyte The polymer network structure in the gel polymer electrolyte was characterized by static rheology measurement. In static rheology measurement, the viscosity of the system was monitored as a function of shear rate. The decrease in viscosity with increasing shear rate indicated presence of a connected network in the gel. Along with the network formation, the viscoelastic behavior was also studied by dynamic rheology measurement. Cyclic voltammetry were also carried out to estimate the electrochemical voltage window of the synthesized materials. It was observed that the material is electrochemically stable in the potential range from -0.5 to 5 V. This makes it highly suitable for battery applications with high voltage electrode materials such as polyanions phosphates and silicates.
Fig 3: Photographs of the polymer elecrolytes at various monomer ratios (above). Dynamic rheology measurement of cross-linked blend-like gel polymer electrolyte (bottom) Additionally, the new cross-linked systems were galvanostatically cycled with LiFePO4 as a cathode material. Though the as-synthesized polymer electrolyte showed electrochemical performances, the applicability of the cross-linked systems in battery application is presently restricted by the inability to cast thin films (≈ 30-100 µm). It is felt that further improvement in the mechanical property of the gel polymer is expected to improve overall physical properties and device performance.
Fig 4: Electrochemical stability window of the gel polymer electrolytes
4
B: Mixed solvent based gel polymer electrolyte: Plastic crystals possess rotational disorder in their crystalline lattice structure. The molecule transformed to this rotator phase via solid-solid phase transition below their melting point. Plastic phase is achieved via solid-solid phase transition(s) below their melting point. Depending on the plastic crystalline compound, the role of rotators vary in their influence on the ionic conductivity. In insulating polar organic plastic crystalline compounds such as succinonitrile, the presence ofrotors significantly influence the effective conductivity of the electrolyte (SN + ionic salt, NaX, LiX). Succinonitrile consists of three geometrical isomers, two gauche and one trans above the normal to plastic transition temperature (Tnp). These two isomers exist in equilibrium in plastic phase and transform from one to another form via 120o rotation around C-C bond. The trans phase has been proposed to act as impurity phase which reduces activation energy barrier for ion transport. Such fenomenon of solvent dynamics, playing a important role in ion transport has also been obsereved in various ionic liquid, an important subclass of ionic plastic crystalline materials. Among several ionic liquid systems based on imidazolium, piperidinium, alkyl ammonium based ionic liquids, N-methyl-N-butyl pyrrolidinium bis(trifluoromethane sulfonyl) imide (PY14TFSI) has been well studied ionic liquid for battery applications it’s owing to several attractive physical properties. However the intrinsic ionic conductivity is lower compared to other ionic liquid.
We discuss here an electrolytic system where a mixture of solvents comprising of a plastic crystalline compound and RTIL has been employed to synthesize the electrolyte. Incorporation of small amount of ionic liquid into succinonitrile matrix resulted in a highly conducting electrolyte compared to both IL-
LiX or Succinonitrile-LiX systems. Similar system has been earlier reported in dye sensitized solar cell lead to an increase in efficiency of dye DSSC. The ionic liquid decreases melting point of the SN while keeping normal-plastic transition temperature unchanged. The room temperature ionic conductivity and cyclic voltammetry have been performed in order to investigate applicability of the mixed solvent in lithium battery application and a new type of gel polymer electrolyte has also been synthesized using this solvent matrix in this study.
Scheme 2: Schematic representation of the synthesis of gel polymer electrolyte using SN and IL as solvents Prior to prepare the above mentioned mixed plastic crystalline solvent, succinonitrile was sublimed twice to reduce impurity. In order to prepare mixed solvent based electrolytic system, initially SN-LiTFSI system was synthesized by dissolving required amount of lithium salt, lithium bis(trifluoromethane sulfonyl) imide (LiTFSI) in succinonitrile and stirred at 60 0C. The salt concentration was kept constant (0.5 M) for all the studies. The ionic liquid, 1-Butyl-1-Methyl bis(trifluoromethane sulfonyl) imide (PY14TFSI) was incorporated in the SN-LiTFSI system in three different molar ratios (with respect to SN) and stirred at room temperature to get homogeneous liquid electrolytic system. Three different molar ratios of SN to IL studied are 20:1, 10:1 and
5
5:1 keeping LiTFSI concentration same (0.5M). For the lower ionic liquid
Fig 5: Ionic conductivity of the gel polymer electrolytes using SN+IL as solvents concentration i.e. for the composition [SN:IL 20:1]-0.5M LiTFSI, an waxy appearance, characteristic to SN was observed at room temperature. But as we increase ionic liquid concentration, i.e, for the compositions [SN:IL 10:1]-0.5M LiTFSI and [SN:IL 5:1]-0.5M LiTFSI, a liquid appearance was observed at room temperature. This shows a clear indication of reduction of melting point of succinonitrile upon addition of small amount of ionic liquid.
Fig 6: Electrochemical stability window of the gel polymer electrolytes using SN+IL as solvents
The room temperature conductivities of mixed solvents with various IL concentrations, keeping salt concentration same, is presented in the table 1. Table 1: Room temperature conductivities of SN-IL systems of different solvent ratio
Solvent composition Conductivity (Ω-1cm-1)
SN-0.5M LiTFSI 2.4 x10-4
[SN:IL 20:1]-0.5M LiTFSI
2.3x10-3
[SN:IL 10:1]-0.5M LiTFSI
5.9 x10-3
[SN:IL 5:1]-0.5M LiTFSI 4.5 x10-3
IL 0.5M LiTFSI 2.5 x10-3 The above room temperature ionic conductivity data clearly indicates that room temeprature conductivity of the liquid electrolyte strongly depends on molar ratio of succinonitrile and ionic liquid. With increasing ionic liquid concentration, conductivity increases up to a maximum value (6.9 x10-3 Ω-
1cm-1 for the composition SN:IL 10:1). On further increasing ionic liquid concentration, i.e for the molar ratio 5:1, conductivity decreases. As the lithium salt concentration was kept constant, the effect of salt on the ionic conductivity has not been taken into account. Instead of the large variation in conductivities, all of the compositions of the plastic crystal mixed solvent (including the composition [SN:IL 5:1]-0.5M LiTFSI) have higher ionic conductivities compared to the pristine SN-LiTFSI and IL-LiTFSI systems. The composition, [SN:IL 10:1]-0.5M LiTFSI exhibits highest conductivity (5.9 x10-3 Ω-1cm-
1), almost comparable to widely used liquid EC/DMC based electrolyte. In order to investigate the reason behind this significant improvement in ionic conductivity using the mixed solvent compared to both pristine IL an succinonitrile, several factors like viscosity, solvent dynamics and ion association effect
6
should be taken into account carefully. Reduction of viscosity and ion association in SN-IL-LiTFSI systems compared to pristine IL-LiTFSI and SN-LiTFSI systems may be reason behind the improvement in ionic conductivity in the proposed mixed solvent. Besides, the decrease in ionic conductivity at higher ionic liquid concentration (for the composition [SN:IL 5:1]-0.5M LiTFSI) compared to other compositions, indicates possibility of increasing ion association in that concentration compared to lower concentration of ionic liquid. To explain the variation in ionic conductivities with ionic liquid concentration, the effect of ionic liquid incorporation on succinonitrile dynamics and nature of ion pairing in this solvent should be studied in more details The cyclic voltammetry was carried out for the highest conducting composition to monitor electrochemical window of the mixed plastic crystal solvent and represented in figure 6. The plot clearly indicates a wide electrochemical voltage window for the solvent from 0.5V to 4.7V. In this plot a strong cathodic peak at around 0.2 V and anodic peak at -0.5 V were observed for characteristic lithium striping and lithium plating processes. No further electrochemical degradation of electrolyte is observed up to the voltage 4.5V. The stability of the solvent at upper voltage region is close to ionic liquid PY14TFSI-0.5M LiTFSI system (5V) and thus can be applied for high voltage cathode materials.
III CONCLUSIONS We have demonstrated here a novel class of gel polymer electrolytes for possible applications in lithium-based batteries. The ion transport mechanism (not discussed here, c/f proceeding paper 2013 and progress report 2014) is affected by both the dynamics of the polymer and ion solvation. As other types of salts viz. magnesium, sodium can be easily incorporated in the synthesis, the gel polymer
electrolytes discussed here will also be promising for other battery chemistries.
REFERENCES [1] Das, S; Bhattacharyya, A. J. Phys. Chem. Lett. 2012, 3, 3550−3554 [2] Patel, M; Bhattacharyya, A. J. Phys. Chem. B 2010, 114, 5233–5240 [3] Patel, M; Gnanavel, M; Bhattacharyya, A. J. J. Mater. Chem. 2011, 43, 17419-17424 [4] S. Sen, Bhattacharyya, A. J. 2014, unpublished
SPARSITY-BASED CROSS-TERMS-SUPPRESSED TIME-FREQUENCY DISTRIBUTION OFMULTI-COMPONENT LINEAR FREQUENCY MODULATED SIGNALS
Shreyas Hampali, Subhankar Ghose, and Chandra Sekhar Seelamantula
ABSTRACTWe address the problem of estimating the parameters of multiplelinear frequency-modulated (LFM) chirp signals. We consider sig-nals corrupted by additive white noise and develop a new techniquebased on the properties of the instantaneous autocorrelation (IA)sequence of multiple LFM chirp signals. The IA matrix, whoserows are the IA sequences obtained at different time instances ofthe signal is introduced. We show that, by obtaining a sparse repre-sentation of the rows of IA matrix with a dictionary of sinusoids, theinherent cross terms are eliminated. Further, by obtaining a sparserepresentation of the columns of the IA matrix, the instantaneousfrequencies of the component chirps are estimated to obtain a highresolution cross-term suppressed sparse WVD (CTSS-WVD).Thecomponent chirps that manifest as straight lines in CTSS-WVDare estimated using an iterative line parameter estimation method.We show simulation results for four-component chirp signal withvarying SNR levels. We also validate the technique by suppressingcross-terms in a bat echolation signal.
Index Terms—Instantaneous autocorrelation sequence, Am-biguity function, sparsity, LASSO, Linear frequency modulatedchirps
I. INTRODUCTION
L INEAR frequency modulated (LFM) signals are a class ofnon-stationary signals widely used in applications such as
radar, sonar and wireless communication [1]. In all these casesthe transmitted signal undergoes a time-varying phase shift dueto relative motion between transmitter and the receiver. Further,due to presence of multiple targets in the radar case and multiplepaths in the communication case, the received signal is a sum ofLFM signals and analysis of such multicomponent LFM signalsreveal properties of targets or communication paths. The multi-component LFM signals also arise in applications such as electroniccounter measure systems for pulse-Doppler radars, where LFMsignals from various transmitters need to be analyzed. Due to itswide applications the study of multiple LFM signals in noise hasreceived considerable interest during the past two to three decades[2]–[5].
I-A. Contributions of this paperIn this paper, we propose a technique for obtaining cross-
term suppressed WVD (CTS-WVD) by sparsifying the discreteambiguity function for multi-component LFM chirp signals in whitenoise. Using CTS-WVD, we introduce high-resolution, noise andcross-term suppressed sparse WVD (CTSS-WVD) that is suitablefor estimating multi-component chirp parameters in low SNRs (0
This work was supported by the Indian Space Research Organization- Indian Institute of Science Space Technology Cell (Project:ISTC/EEE/CSS/0293).Authors email : [email protected], [email protected],[email protected] : +91 80 2293 2695
dB). The estimation of the chirp parameters in the CTSS-WVD iscomputationally far less complex than Hough transform. The CTS-WVD is shown to have lower variance, better resolution in the time-frequency (TF) distribution and comparable cross-term suppressionability than other cross-term suppression methods.
II. AMBIGUITY FUNCTION OF MULTICOMPONENTLFM CHIRP SIGNAL
The instantaneous autocorrelation (IA) sequence of a signal s[n]is defined as,
IA(n, k) = s∗[n− k]s[n+ k]. (1)
The ambiguity function (AF) of a signal is given in terms of IAsequence as
M(τ, θ) =∑n∈Z
IA(n, τ/2)ejθn (2)
For a mono-component LFM chirp signal, the AF is given by,
M(τ, θ) = b20eja0τ
∑n
eja1τnejθn (3)
Thus, for a given τ , the AF of an LFM chirp is the Fouriertransform of a sinusoid of frequency a1τ . In case of multiplechirps, equation (2) can be divided into two terms as cross-termcomponent, Mc(τ, θ) and auto-term component, Ma(τ, θ). For atwo component LFM chirp signal the AF is given by,
M(τ, θ) =
2∑i=1
b2i eja0iτ
∑n
eja1iτnejθn
︸ ︷︷ ︸Ma(τ,θ)
+∑n
ej(C01(τ)+C11(τ)n+C21n2)ejθn
+∑n
ej(C02(τ)+C12(τ)n+C22n2)ejθn
︸ ︷︷ ︸Mc(τ,θ)
, (4)
where,
C0i(τ) = a01τ
2+ a02
τ
2+ (−1)i
1
8(a11 − a12)τ2,
C1i(τ) = a11τ
2+ a12
τ
2+ (−1)i(a01 − a02),
C2i = (−1)i1
2(a11 − a12)
It can be verified that in the scenario a11 6= a12, the cross-terms are Fourier transforms of LFM chirps. The auto-terms arealways sinusoids with frequency a1iτ . When the chirp rates a1i
of the component chirps are equal, the cross terms are sinusoidsof frequency a1iτ ± (a01 − a02). In the presence of white noise,the signal to noise cross-terms manifest as random signal. Thus,the problem of cross-term suppression is that of extraction of lowfrequency sinusoids in the ambiguity function domain.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
III. CROSS-TERM SUPPRESSION USING A SPARSEREPRESENTATION FOR AMBIGUITY FUNCTION
We introduce the instantaneous autocorrelation (IA) matrix ofthe signal s[n] and denote it by A. The size of A is N
2×N and
the (i, jth) element is given by,
Ai,j = s∗[j − i] · s[j + i]. (5)
The AF can be expressed in terms of IA matrix as,
M(2τ, θ) = F B; B = AT τ, θ ≥ 0 (6)
where F is the discrete Fourier transform matrix.The rows of the IA matrix contain signal auto-terms, signal-to-
signal cross-terms and signal-to-noise cross-terms. The signal-to-signal cross-terms are LFM chirps when chirp rates are unequaland the signal-to-noise cross-terms are random signals. The signalauto-terms being the only sinusoids in this scenario, the cross-termsare suppressed by identifying the sinusoids in the IA sequence.The sinusoids can be identified by estimating its frequencies usingvarious parameter estimation techniques such as estimation ofsignal parameters via rotational invariance techniques (ESPRIT)[11] and multiple signal classification (MUSIC). However, suchtechniques require the model parameter which is the number ofsinusoids present in the signal and is unknown a priori. Further, thepresence of signal dependent cross-terms deviates the IA sequencefrom the signal model defined by sinusoidal parameter estimationmethods.
In our approach, we use sinusoidal dictionary based least abso-lute shrinkage and selection operator (LASSO) formulation definedin [12].
III-A. Cross-term suppression problem formulation usingLASSO in ambiguity function domain
The cross-terms in the rows of IA matrix in (5) for unequal ratechirp signals are separated by identifying the sinusoids, which arethe auto-terms in the IA sequence. A set of sinusoids in an IAsequence along with their magnitudes are estimated using LASSOformulation with dictionaries of sinusoids each of size P denotedby X(i) and X(r), respectively. The (i, j)th element of the Mn ×P, P ≥Mn dictionary matrices is given by,
X(i)i,j = sin
(2πi(j − P/2)
P
),
X(r)i,j = cos
(2πi(j − P/2)
P
),
j = 0, 1, 2, · · · , P − 1;
i = 0, 1, 2, · · · ,Mn − 1. (7)
The nth column vector of matrix B in (6) has Mn = N − 2nnon-zero elements in it as can be verified from the construction ofIA matrix in (5). This Mn element vector is denoted by Bn. AsBn is complex the auto-term identification problem is formulatedin the LASSO framework in two stages as,
minwn
‖ X(i)wn − imag(BnT) ‖22 +λ ‖ wn ‖1,
minwn
‖ X(r)wn − real(BnT) ‖22 +λ ‖ wn ‖1 . (8)
The two stages correspond to the imaginary and real parts of Bn.The optimization problem in (8) identifies the auto-terms in the nth
column of the matrix B. The performance of this two stage processis improved by retaining only those frequencies for the dictionaryin stage 2 whose amplitudes are non-zero in stage one. As the
coefficient vector w∗n denotes the auto-terms in the IA matrix, thecross-term suppressed sparse AF (CTSS-AF) is obtained as,
CTSS-AF = [w∗1,w∗2, · · · ,w∗N
2]
When the formulation in (8) is used for the case of a signalconstituted of equal rate LFM chirps, the cross-terms fail to getsuppressed and appear along with the auto-terms. This is because,the cross-terms too appear as sinusoids when rates of componentchirps are equal.
The cross-terms in case of equal rate LFM chirps have frequen-cies higher than that of auto-terms. As explained in the SectionII when a11 ≈ a12 in (4), cross-terms appear along coordinates(τ, a11τ ± (a01 − a02)
)in the discrete AF domain, whereas the
auto-terms are present at coordinates (τ, a11τ). Thus, for small τ ,the cross-terms appear on either side of the frequency origin asseen from Figure 2.
The cost function (8) imposes equal penalty (λ) on all thecoefficients in vector wi. Thus, we reformulate the auto-termidentification problem in (8) to encompass both the scenarios ofequal and unequal chirp rates and is given by,
minwn ‖ X(r)wn − real(ATn ) ‖22 +λ ‖ Dwn ‖1, (9)
where D ∈ Rp×p,
Di,j =
0 if i 6= j,
e(i−p/2)2
γ2 if i = j.
where γ is the penalty controlling factor. The diagonal penaltymatrix D satisfies the condition of imposing higher penalty forfrequency components that are away from zero. The penalty matrixD is motivated by the Choi-Williams kernel, Φ(τ, θ) used for cross-term suppression and is given by,
Φ(τ, θ) = e−α(τθ)2 .
It should be noted that, in case of wide-band (large a11) equalrate chirp signals, for large τ , (a11τ ± (a01 − a02))) is positiveif a11 > 0 and negative if a11 < 0. Hence, in this scenario thecross-terms that appear on the same side of the frequency originare not suppressed by the cost function (9).
Figure 1 shows a normalized slice of the CTSS-AF obtained us-ing the general formulation in (9). We observe that the cross-termswhich appear as sinusoids are suppressed by the new formulation.Figure 3 shows the AF of a three chirp signal obtained using (6) andFigure 4 shows the corresponding cross-term suppressed discretesparse AF.
−3 −2 −1 0 1 2 30
0.2
0.4
0.6
0.8
1
FREQUENCY
Fig. 1: A slice of the sparse-AF for two equal rate chirpsignals each at SNR = 0 dB.The sparse-AF is obtained us-ing cost function (9). Thecross-terms are suppressed.
τ
θ
50 100 150 200 250−2
−1
0
1
2
0
0.2
0.4
0.6
0.8
1
Cross−terms
Auto−term
Fig. 2: Ambiguity function of twoequal rate chirp signals with param-eters a01 = 0.2513, a02 = 0.6283,a11 = a12 = 7.33× 10−4.
τ
θ
20 40 60 80
−50
−25
0
25
50
0.2
0.4
0.6
0.8
1
Fig. 3: Ambiguity function ofa three component LFM chirpsignal each at high SNR
τ
θ
20 40 60
−50
−25
0
25
50 0
0.2
0.4
0.6
0.8
1
Fig. 4: CTSS-AF of a three com-ponent LFM chirp signal each athigh SNR
The length of the AF sequence in nth row of the IA matrix isN −2n. The ability of the cost function (9) to provide an accuratesparse representation of the signal is a function of the signal lengthas shown in [12]. Thus, for large τ , when the IA sequence lengthis small, the signal auto-terms are not consistently identified asverified in Figure 4.
Denoting the minimizer of (9) by w∗n, the sparse approximationof Bi denoted by Bi is given by,
Bn = (X(r) + jX(i))w∗n (10)
The cross-term suppressed IA (CTS-IA) matrix denoted by A isobtained by arranging the vectors Bi row-wise as,
A =[B1, B2, · · · , BN
2
]T(11)
Similar to AF, the Wigner-Ville distribution (WVD) of a signalis directly obtained from its IA matrix A as,
WVD(n,ω
2
)=∑k
IA(n, k)ejωk = 2 · real(F A
)(12)
By utilizing the cross-term suppressed IA matrix A, the cross-termsuppressed WVD is obtained as 2 · real
(F A
).
Figure 5 shows cross-term suppressed WVD obtained fromdifferent techniques in no noise scenario. The Choi-Williamsmethod which uses a low pass filter in the AF domain is knownto compromise on the TF resolution of the signal. The signaldecomposition based technique [7] which uses Gabor atoms isfree of cross-terms, however the TF resolution of the signal iscompromised as verified from Figure 5b. The proposed methodon the other hand has better TF resolution and also achieves crossterm suppression as observed in Figure 13, which shows a sliceof the TF distributions in Figure 5 along the dotted line. Theparameters considered in the simulation are a01 = 0.2513, a02 =0.7504, a11 = a12 = 7.36× 10−4, p = 1024, N = 512.
Figure 6 compares the performance of the proposed methodwhen a signal is constituted of four LFM chirp signal in a noisyscenario. The proposed method is seen to have better TF resolutionthan Gabor atoms based signal decomposition method [7] and a bet-ter cross-term suppression ability than Choi-Williams method. Theparameters considered are a01 = 0.2513, a02 = 0.5027, a03 =0.9425, a04 = 1.1310, a11 = 7.36 × 10−4, a12 = 18× 10−4,a13 = −6.14 × 10−4, a14 = 7.36 × 10−4, p = 1024, N = 512and SNR1 = SNR2 = SNR3 = SNR4 = 0 dB
IV. SPARSE WVD AND MULTI-COMPONENT CHIRPSIGNAL PARAMETERS ESTIMATION
In this section, we introduce cross-term suppressed sparse WVD(CTSS-WVD) using A. The CTSS-WVD is then used to esti-mate chirp parameters using an iterative line parameter estimation
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150
0.2
0.4
0.6
0.8
1
(a) Wigner-Ville distribution
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150 0
0.2
0.4
0.6
0.8
1
(b) TF distribution using match-ing pursuit method [7]
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150
0.2
0.4
0.6
0.8
1
(c) TF distribution using Choi-Williams kernel method
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150
0.2
0.4
0.6
0.8
1
(d) TF distribution using pro-posed method
Fig. 5: Comparison of the cross-term suppression methods for twoequal rate chirp signal in no noise scenario.
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150
0.2
0.4
0.6
0.8
1
(a) Wigner-Ville distribution
TIMEF
RE
QU
EN
CY
100 200 300 400 500
50
100
150 0
0.2
0.4
0.6
0.8
1
(b) TF distribution using match-ing pursuit method [7]
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
20
40
60
80
100
120
140
0.2
0.4
0.6
0.8
1
(c) TF distribution using Choi-Williams kernel method
TIME
FR
EQ
UE
NC
Y
100 200 300 400 500
50
100
150
(d) TF distribution using pro-posed method
Fig. 6: Comparison of the proposed cross-term suppression methodfor a four component chirp signal each with SNR = 0 dB
method. The CTSS-WVD is obtained by sparsifying CTS-WVDintroduced in Section III and has improved frequency resolutionaiding in accurate estimation of parameters of component chirpsignals.
The nth column of the CTS-IA matrix which we call as the IAvector is denoted by zn. The auto-term component in the IA vector
is a sum of complex sinusoids given by,
zn[k] =∑i
b2i ej2(a0,i+a1,in)k. (13)
We observe that the frequency of each sinusoid is twice theinstantaneous frequency of the corresponding chirp at time instantn. Thus, a sparse representation of the IA vector zn similar to(8), with a dictionary of sinusoids as defined in (7), providesinformation of the sinusoidal frequency content in the vector whichin turn is an estimate of the IFs of the component chirps. Thetwo cost functions for obtaining the sparse representation of zn isprovided below for clarity,
minwn
‖ X(i)wn − imag(zn) ‖22 +
p∑i=1
‖wn‖1, (14a)
minwn
‖ X(r)wn − real(zn) ‖22 +
p∑i=1
‖wn‖1, (14b)
where X(i) and X(r) are the sine and cosine dictionary matrix.As discussed in Section III, (14b) is used for suppressing spuriousfrequencies detected by optimizing (14a).
The sparse representation is obtained for each column of CTS-IA matrix and the minimizer of (14) is denoted by w∗n for columnn. The CTSS-WVD matrix is obtained from w∗n as,
CTSS-WVD = [w∗1,w∗2, · · · ,w∗N
2] (15)
Figure 7 shows the WVD of a four chirp signal. The SNR of eachof the component chirps is 4 dB. The CTSS-WVD obtained usingthe above described procedure is shown in Figure 8. A slice of theTF distributions along the dotted line in Figure 7 and 8 is shown inFigure 9. The resolution of CTSS-WVD is better than WVD due tothe the sparsity constraint used in the formulation. The cross-termsare also suppressed thus providing better estimate of the IF of thecomponent chirps.
IV-A. Detection and estimation of line parameters in sparseWVD
‘ We develop a new technique termed as iterative line parameterestimation method (ILPEM) that iteratively estimates the slope andintercept parameters of intersecting lines in a sparse distribution.ILPEM is an alternative to the conventional Hough transform andhas lower computational complexity.
Let the set of frequencies present at any time instant n, obtainedfrom CTSS-WVD be denoted by Γn. A single chirp scenario is firstconsidered and is later extended to multiple chirps at the end ofthis section. The problem of line parameter estimation in sparse TFplane with least squares cost function is written as,
mina,b,
η1∈Γ1,η2∈Γ2,··· ,ηN∈ΓN
N∑n=1
(ηn − (a · n+ b))2 , (16)
where, at each time instant n, an optimal frequency is chosen formthe set Γn such that the frequencies at all time instants are co-linearand is given by the line parameters (a, b). We propose a two stepsuboptimal iterative solution to the (N + 2)-dimensional searchproblem by assuming that the initial value of the line parametersis close to the minimizer.
η(i)n = arg min
ηn∈Γn
(ηn − (a(i−1) · n+ b(i−1))
)2
; (17a)
n = 0, 1, 2, · · · , N − 1
a(i), b(i) = arg mina,b
N∑n=1
(η(i)n − (a · n+ b)
)2
, (17b)
where a(i), b(i), η(i)1 , η
(i)2 · · · , η
(i)N represent the optimization pa-
rameters at ith iteration.The initial value of the line parameters in (17a) denoted by
(a(0), b(0)) is obtained using the frequency marginal which isshown in Figure 8. The frequency corresponding to the maximumvalue of the marginal is used as b(0). In our simulation, the valueof a(0) is set to 0 as no information about the chirp rate is availableinitially. For the case of multiple chirp lines, the frequencies η(i)
n
of the set Γi corresponding to a line as obtained from (17a) areremoved from their corresponding sets and the two-step iterativeprocedure is repeated.
We now evaluate the performance of the sparse WVD alongwith ILPEM (sparse-WVD-ILPEM) by estimating the parametersof a mono-component chirp signal.To facilitate comparison, thechirp parameters are also estimated by applying Hough transformto sparse-WVD (sparse-WVD-Hough). The case of a single chirpsignal in noise enables performance evaluation of sparse-WVD andILPEM in separation as the cross-term suppression described insection III is not required.
For simulations we consider signal length N = 300 and thenumber of sinusoids in the dictionary P = 750. Figure 10a and10b shows the variance of the chirp parameter estimates usingvarious techniques. The resolution of the transformed variablesin the Hough transform (θ, ρ) is set to 0.1. The performanceof the ILPEM is comparable with the Hough transform basedtechnique (sparse-WVD-Hough) for SNR > −2 dB and is betterfor SNR < −2 dB. A comparison with de-chirping based methoddiscussed by Peleg and Porat in [10] with 4096-point discreteFourier transform (DFT) shows that the threshold SNR of thesparse-WVD-ILPEM technique is about 4 dB lesser than the de-
TIME
FR
EQ
UE
NC
Y
100 200 300
50
100
150
200
250
300
350
0.2
0.4
0.6
0.8
1
Fig. 7: WVD of a four compo-nent LFM chirp signal each atSNR = 4 dB.
TIME
FR
EQ
UE
NC
Y
100 200 300
50
100
150
200
250
300
3500
0.2
0.4
0.6
0.8
1
Fig. 8: CTSS-WVD for a fourcomponent LFM chirp signaleach at SNR = 4dB along withfrequency marginal used for ini-tializing the line parameters. Thepeak of the frequency marginal isused to initialize the iterative lineparameter estimation method.
0 100 200 3000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Frequency
(a) WVD
0 100 200 3000
0.2
0.4
0.6
0.8
Frequency
Auto−terms
(b) CTSS-WVD
Fig. 9: A slice of the WVD and CTSS-WVD along the dotted lineshown in Figure 7 and 8
chirping method. Further, the proposed method is consistent withCRLB for SNR > −2 dB.
The advantages of the ILPEM over Hough transform are threefold. 1)The estimates obtained using the ILPEM have lower vari-ance than Hough transform based estimates. 2)For an N×N sparsematrix, the computational complexity of ILPEM is O(N2) (linearsearch in (17a) has a complexity of O(N)), whereas the complexityof Hough transform is O(N3), if the transformed space also has thedimension N ×N . 3) ILPEM has the line-fitting property and candetect lines in noisy scenarios. The Hough transform formulationyields poor results when the observations are not co-linear.
The iterative line parameter estimation method converges to localminimas of (16) when the distribution is noisy and less sparse. Atlow SNRs, the sparse-WVD is noisy and the estimates of chirpparameters are inaccurate as observed in Figures 10a and 10b.Simulation results show that the iterative line parameter estimationmethod converges to minima in less than 15 iterations for signallength N = 300, dictionary size P = 375 and SNR >= −2 dB.
V. PERFORMANCE EVALUATIONWe evaluate the robustness of the proposed algorithm by consid-
ering a four component chirp signal with varying SNR levels. Theperformance is compared with two other methods namely Wigner-Hough transform discussed by Barbarossa in [9] and producthigher order ambiguity function (PHAF) based method proposed byBarbarossa et al. in [13]. We obtain CTSS-WVD of a bat echolationsignal and prove that the technique efficiently suppresses cross-terms in real scenarios.
The multi-component signal considered is same as in Figure6 whose parameters are given in Section III. The size P of thesinusoidal dictionary defined in (7) is 750. For the PHAF method,four different lags are considered as discussed in [13] with 2048-point DFT. For Wigner-Hough based method, the size of thetransformed domain is 256 × 256. At low SNR, the probabilityof detecting a chirp line in a sparse-TF distribution or in theWigner-Hough domain decreases. However, such false detectionscan be identified by comparing the sum of squared errors asobtained in cost function (17b) with a threshold. Deriving anoptimal detector for detecting such false alarms is beyond the scopeof this work and hence, in this paper we compare the estimatedparameters with the ground truth and declare as detected if thedeviation is within 10% of its original value (oracle). We consider200 Monte-Carlo simulations to obtain variance of the estimatedchirp parameters. Figures 11 and 12 show the performance of theproposed method for each of the detected component chirp signalsand compare it with other techniques for different SNR levels. Thechirp parameter estimates obtained using CTSS-WVD is observedto have lower mean-squared-error (MSE) than Wigner-Hough orPHAF estimates for three component chirps (chirps 1,2 and 4). Theerror in estimation of chirp rate propagates in the PHAF methodand affects the carrier frequency estimates as discussed in [13]. Thecarrier frequency estimates using CTSS-WVD does not suffer fromerror propagation and hence has lower MSE than PHAF techniqueas seen from Figure 12. The average probability of detection (PD)of all the component chirps is shown in Figure 14. The Wigner-Hough method is benefited from the coherent integration and hasPD = 1 for large SNRs. The PD of the proposed method is morethan 0.95 for all SNRs and decreases with SNR as the iterative lineparameter estimation method converges to local minimas.
Figure 15 shows the TF distributions of a bat echolation signal1.The bat echolation signal used here has deviations from the linear
1The authors wish to thank Curtis Condon, Ken White, and Al Feng ofthe Beckman Institute of the University of Illinois for the bat data and forpermission to use it in this paper.
0 5 10
−60
−40
−20
0
20
SNR (dB)
VA
RIA
NC
E (
dB)
CRLBSPARSE−WVD−ILPEMDECHIRPINGMETHOD [8]SPARSE−WVD−HOUGH
(a) Comparison of carrier fre-quency estimates.
−4 −2 0 2 4 6 8 10
−100
−80
−60
−40
−20
SNR (dB)
VA
RIA
NC
E (
dB)
CRLBDECHIRPINGMETHOD [8]SPARSE−WVD−ILPEMSPARSE−WVD−HOUGH
(b) Comparison of chirp rate es-timates.
Fig. 10: Chirp parameter estimates of a mono-component LFMchirp in noise. The sparse WVD-Hough estimates are obtained byapplying Hough transform to the sparse WVD obtained using (14).
0 2 4 6 8−110
−105
−100
−95
−90
−85
−80
SNR (dB)M
SE
(dB
)
CTSS−WVDPHAFWIGNER−HOUGH
(a) Chirp 1
0 2 4 6 8−110
−105
−100
−95
−90
−85
−80
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(b) Chirp 2
0 2 4 6 8−100
−95
−90
−85
−80
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(c) Chirp 3
0 2 4 6 8−110
−105
−100
−95
−90
−85
−80
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(d) Chirp 4
Fig. 11: Performance comparison of chirp rate estimator for a fourcomponent LFM chirp signal with varying SNRs
chirp and the chirps do not occupy the full signal length. However,we observe from Figure 15c that the cross-terms are suppressedin CTSS-WVD without loss of TF frequency resolution. The lowamplitude second harmonic chirp at higher frequencies is notcompletely captured by CTSS-WVD.
VI. CONCLUSIONWe have proposed a novel technique for estimating the param-
eters of the multiple LFM chirp signals in noisy scenario. Thealgorithm is based on the sparse representation of the instantaneousautocorrelation sequence of the signal and is a three step process.The cross-terms present in the IA sequence are suppressed in thefirst step, followed by instantaneous frequency estimation in thesecond step and an line parameter estimation method in the laststep. The sparse representation is obtained using LASSO with adictionary of sinusoids. A high-resolution cross-term suppressedtime frequency representation of the multiple chirp signals is alsoobtained. Simulation results show the estimator performs efficientlyup to an SNR of -2dB for single chirp signals. We also evaluate theperformance using a four-component chirp signal in noise and showthat the resulting estimates are reliable. Finally, the effectivenessof the technique is proved by applying it to a bat echolation signalthat has minor deviations from the signal model considered.
0 2 4 6 8−70
−65
−60
−55
−50
−45
−40
−35
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(a) Chirp 1
0 2 4 6 8−70
−65
−60
−55
−50
−45
−40
−35
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(b) Chirp 2
0 2 4 6 8−70
−65
−60
−55
−50
−45
−40
−35
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(c) Chirp 3
0 2 4 6 8−70
−65
−60
−55
−50
−45
−40
−35
SNR (dB)
MS
E (
dB)
CTSS−WVDPHAFWIGNER−HOUGH
(d) Chirp 4
Fig. 12: Performance comparison of carrier frequency estimator fora four component LFM chirp signal with varying SNRs
VII. REFERENCES
[1] M. S. Wang, A. K. Chan, and C. K. Chui, “Linearfrequency-modulated signal detection using Radon-ambiguitytansform,” IEEE Trans. Signal Process., vol. 46(3), pp. 571–586, Mar 1998.
[2] Y. Sun and P. Willett, “Hough transform for long chirpdetection” IEEE Trans. Aerosp. Electron. Syst., vol. 38(2),pp. 553–569, Apr. 2002
[3] X-G Xia, “Discrete chirp-Fourier transform and its applicationto chirp rate estimation,” IEEE Trans. Signal Process., vol48(11), pp. 3122–3133, Nov. 2000
[4] L. Qi, R. Tao, S. Y. Zhou and Y. Wang, “Detection andparameter estimation of multicomponent LFM signal based onfractional Fourier transform,” Sci. China (Ser F.),, vol 47(2),pp. 184-198, 2004
[5] J. C. Wood and D. T. Barry, “Radon transformation oftime-frequency distributions for analysis of multicomponentsignals,” IEEE Trans. Signal Process., vol. 43, pp. 3166–3177,Nov. 1994.
[6] H. I. Choi and W. J. Williams “Improved time-frequencyrepresentation of multicomponent signals using exponential
0 50 100 1500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
FREQUENCY
NO
RM
ALI
ZE
D A
MP
LIT
UD
E
WVDMATCHING PURSUIT [15]CHOI−WILLIAMSSPARSE−AF
Fig. 13: A slice of differentTF distributions along the dot-ted line shown in Figure 5. Theproposed method has the bestresolution.
0 2 4 6 80.7
0.75
0.8
0.85
0.9
0.95
1
SNR (dB)
PD
CTSS−WVDPHAFWIGNER−HOUGH
Fig. 14: Comparison of aver-age probability of detection of afour component LFM chirp sig-nal with varying SNRs
TIME (SECS)
FR
EQ
UE
NC
Y (
Hz)
0 0.5 1 1.5 2 2.5
x 10−3
0
2
4
6
x 104
0.2
0.4
0.6
0.8
1
(a) Wigner-Ville distribution
TIME (SECS)
FR
EQ
UE
NC
Y (
Hz)
0 0.5 1 1.5 2 2.5
x 10−3
0
2
4
6
x 104
0.2
0.4
0.6
0.8
1
(b) Spectrogram
TIME (SECS)
FR
EQ
UE
NC
Y (
Hz)
0 0.5 1 1.5 2 2.5
x 10−3
0
2
4
6
x 104
0
0.2
0.4
0.6
0.8
1
(c) CTSS-WVD
Fig. 15: TF distribution of bat echolation signal using differenttechniques. The fundamental and the first harmonic are sharplyresolved in CTSS-WVD along with suppressing the cross-terms.
kernels,” Proc. IEEE Int. Conf. Acoust., Speech, Signal Pro-cess., vol. 37, pp. 862–871, Jun 1989.
[7] S. Mallat and Z. Zhang,“Matching pursuit with time-frequency dictionaries,” IEEE Trans. on Signal Process.,vol. 41, pp. 3397–3415, Dec. 1993.
[8] R. A. Altes, “Detection, estimation and classification withspectrograms,” J. Acoust. Soc. Am., vol. 67, pp. 1232–1246,Apr. 1980
[9] S. Barbarossa, “Analysis of multicomponent LFM signals bya combined Wigner-Hough transform,” IEEE Trans. SignalProcess., vol. 43. pp. 1511–1515, Jun. 1995.
[10] S. Peleg, B. Porat, “Linear FM signal parameter estimationfrom discrete-time observations,” IEEE Trans. Aerosp. Elec-tron. Syst., vol. 27, pp. 607-615, Jul. 1991.
[11] R. Roy and T. Kailath, “ESPRIT – Estimation of signalparameters via rotational invariance techniques,” IEEE Trans.Acoust., Speech, Signal Process.,vol. 37, pp. 984 – 995, Jul.1989
[12] Tibshirani, “Regression shrinkage and selection via the lasso,”J. Roy. Statist. Soc., Series Bvol. 58, pp. 267–288, 1996.
[13] S. Barbarossa and A. Scaglione, “Product high-order ambi-guity function for multicomponent polynomial-phase signalmodeling,” IEEE Trans. on Signal Process., vol. 46, pp. 691–708, Mar. 1998.
Investigating the Origin of the Indian Ocean Geoid Low
Attreyee Ghosh
December 24, 2014
Abstract
This is an ongoing effort to interpret the geoid low in the Indian Ocean just south of the Indian penin-
sula. Several theories exist regarding the origin of the geoid low, which happens to be the lowest gravity
anomaly on Earth. However, none of them are wholly successful in explaining it. We wish to test differ-
ent tomography models, both global and regional, as well as different viscosity models to investigate the
source of this gravity low.
Introduction
Very few methods exist that can provide us with information about the deep Earth. One way to learn about
the deep earth is through seismic tomography, which uses seismic waves to image the Earth’s interior.
Mapping gravity anomalies provides another way of understanding the Earth’s internal structure. It has
been shown by Hager [1984], Hager and Richards [1989] that the Earth’s long wavelength geoid is highly
correlated with density models of subducted slabs. Other studies have also shown how the observed geoid
anomalies can be explained by density anomalies in the Earth’s mantle. Recent satellite images from
GRACE and GOCE missions are yielding a more and more detailed map of the gravity anomalies and
hence a clearer image of the Earth’s mantle. Two of the most important parameters that control the dy-
namics of the deep Earth are density and rheology or viscosity of the rocks that make up the Earth’s mantle.
While we have some knowledge about the density structure from seismic tomography, the viscosity struc-
ture is largely unknown and the observed geoid anomalies can be crucial in constraining the viscosity
structure of the Earth.
The lowest geoid anomaly on Earth lies just south of the tip of the Indian peninsula in the Indian
Ocean. Origin of this geoid low is controversial. Several theories have been proposed to explain this geoid
low, most of which invoke past subduction [Chase and Sprowl, 1983; Hager and Richards, 1989; Richards
and Engebretson, 1992; Steinberger, 2000]. But, there is no general consensus regarding the source of this
geoid low. This project aims to investigate the source of this anomaly by using forward models of mantle
convection. The models aim to test various tomography models with different radial and lateral viscosity
variations. A proper reproduction of this gravity anomaly will not only provide a greater understanding
into the density and viscosity structure of the deep mantle, but may also provide insight regarding the role
of chemical composition in the mantle, which at present is poorly understood.
Numerical models of mantle convection
In the past few months, we have explored some preliminary models of mantle convection. These models
use different tomography (density) structures and different structures of radial viscosity variations.
1
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
0˚
0˚
30˚
30˚
60˚
60˚
90˚
90˚
120˚
120˚
150˚
150˚
180˚
180˚
−150˚
−150˚
−120˚
−120˚
−90˚
−90˚
−60˚
−60˚
−30˚
−30˚
0˚
0˚
−90˚ −90˚
−60˚ −60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚ 60˚
90˚ 90˚
0˚
0˚
30˚
30˚
60˚
60˚
90˚
90˚
120˚
120˚
150˚
150˚
180˚
180˚
−150˚
−150˚
−120˚
−120˚
−90˚
−90˚
−60˚
−60˚
−30˚
−30˚
0˚
0˚
−90˚ −90˚
−60˚ −60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚ 60˚
90˚ 90˚
−80 −40 0 40 80 120
GEOID[m]
Tomography:s20rts,Viscosity: Visc_D
Figure 1: Predicted global geoid height from S20RTS tomography model
with radial viscosity structure.
30˚
30˚
60˚
60˚
90˚
90˚
120˚
120˚
150˚
150˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚ 60˚
30˚
30˚
60˚
60˚
90˚
90˚
120˚
120˚
150˚
150˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚ 60˚
−90 −60 −30 0 30 60 90 120
GEOID[m]
Tomography: s20rts,Viscosity:Visc_D
Figure 2: Geoid anomaly zoomed over the Indian
Ocean region from the same model as in Figure 1.
All these models were run using the HC code
[Hager and O’Connell, 1981; Milner et al., 2009],
which is a semi-analytical code that solves the
equations of conservation of mass and momentum
as well as the constitutive equation between stress
and strain rate. Of the models run so far, a few of
them yield a better fit to the global geoid (Figure
1) as well as to the Indian Ocean geoid low (Fig
2). The next step would be to run models of higher
complexity, such as those with lateral strength vari-
ations, using the parallel finite element mantle con-
vection code CitcomS, which we have just success-
fully installed on one of the SERC supercomputers.
Acknowledgement
The above work has been accomplished with the help of project assistant Ms. Shree Sumanas Badrinath.
2
References
Chase, C. G., and D. R. Sprowl, The modern geoid and ancient plate boundaries, Earth Planet. Sci. Lett.,
62, 314–320, 1983.
Hager, B. H., Subducted slabs and the geoid: Constraints on mantle rheology and flow, J. Geophys. Res.,
89, 6003–6015, 1984.
Hager, B. H., and R. J. O’Connell, A simple global model of plate dynamics and mantle convection, J.
Geophys. Res., 86, 4843–4867, 1981.
Hager, B. H., and M. A. Richards, Long-wavelength variations in Earth’s geoid—physical models and
dynamic implications, Phil. Trans. R. Soc. Lond., 328, 309–327, 1989.
Milner, K., T. W. Becker, L. Boschi, J. Sain, D. Schorlemmer, and H. Waterhouse, The solid earth research
and teaching environment: a new software framework to share research tools in the classroom and across
disciplines, EOS Trans. AGU, 90, 12, 2009.
Richards, M., and D. Engebretson, Large-scale mantle convection and the history of subduction, Nature,
355, 437–440, 1992.
Steinberger, B., Slabs in the lower mantle - results of dynamic modelling compared with tomographic
images and the geoid, Phys. Earth Planet. Int., 118, 241–257, 2000.
3
Evolution of Mesoproterozoic suture zones western India: Implications on India-Madagascar correlations
C. Ishwar-Kumar and K. Sajeev
Abstract-The Kumta suture zone in western India is in the point of paleogeographic study of India and Madagascar. The ca. 1300 Ma Kumta suture separates the Karwar and Dharwar blocks within the western Dharwar craton of India (Ishwar-Kumar et al., 2013a). Towards the east the Sirsi shelf is unconformable on gneisses of the Dharwar craton and towards the west the Karwar block is mainly composed of undeformed tonalite-trondhjemite-granodiorite (TTG) with enclaves of amphibolites. The TTGs from Karwar block have U-Pb zircon magmatic ages of ca. 3200 Ma (Ishwar-Kumar et al., 2013a). The Coorg block mainly contains ca. 3200 Ma granulite grade rocks and separated from Dharwar craton by ca. 1200 Ma Coorg (Mercara) suture (Chetty et al., 2012; Ishwar-Kumar et al., 2013b; Santosh et al., 2014). Integration of the structural, geological and geochronological results suggests the presence of a 1300-1200 Ma suture in western India, which is possibly an eastern extension of the Betsimisaraka suture zone in Madagascar.
I. INTRODUCTION
The amalgamation of Gondwana supercontinent in the period 590-550 Ma and its subsequent fragmentation into the continental blocks of India, Madagascar, Australia, Africa, Sri Lanka, South America and Antarctica were key tectonic events in the earth history (Fig. 1). Much multi-disciplinary research in the last few decades has been focused on finding tectonic criteria to satisfactorily link, match and thus re-assemble the separated blocks into their former parent
(1e.g., Yoshida et al., 1992; Mishra et al., 2006; Grantham et al., 2008), and in particular Eastern Gondwana (Yoshida, 1995; Meert, 2003a; Collins and Pisarevsky, 2005). For example, much effort has been applied to finding ways to convincingly re-unite Madagascar and India (e.g., Katz and Premoli, 1979; Agarwal et al., 1992; Windley et al., 1994; Yoshida et al., 1999; Torsvik et al., 2000; Collins et al., 2007a).
Fig. 1. Map showing the assembly of Eastern Gondwana (after, Meert, 2003) consists of India, Madagascar, Sri Lanka, Antarctica, Australia, Congo craton and Kalahari craton. The study area India and Madagascar lies in the central part of Eastern Gondwana.
The position and morphological relationship of India relative to Madagascar in
ISTC/CEAS/SJK/291
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
the past is one of the most debated and current problems in understanding the tectonics of eastern Gondwana (e.g., Katz and Premoli, 1979; Collins and Windley, 2002; Braun and
Fig. 2. The regional geology map of southern India (modified after, Geological Survey of India, 2005), overlaid by reclassified shear zones. The rectangles show the detailed study area Karwar region and Coorg region. Acronyms: KSZ- Kumta Shear Zone, ChSZ- Chitradurga Shear Zone, MeSZ- Mettur Shear Zone, KolSZ- Kolar Shear Zone; NSZ- Nallamalai Shear Zone, MSZ- Moyar Shear Zone, McSZ- Mercara Shear Zone, BSZ- Bhavani Shear Zone, SASZ- Salem Attur Shear Zone, CaSZ- Cauvery Shear Zone, PCSZ-Palghat Cauvery Shear Zone, ASZ- Achankovil Shear Zone.
Kriegsman, 2003; Ghosh et al., 2004; Collins and Pisarevsky, 2005; Collins, 2006; Ashwal et al., 2013; Gibbons et al., 2013; Ishwar-Kumar et al., 2013; Rekha et al., 2013). The breakup of Gondwana started with the roughly simultaneous rifting of Madagascar, Seychelles, India, Antarctica and Australia from Africa at around ca. 150 Ma. Then, Madagascar, Seychelles and India separated together from Antarctica and Australia at about 128-130 Ma (Biswas, 1999).
II. GEOLOGICAL BACKGROUND
The period between the assembly of Rodinia and its break-up to form the Gondwana supercontinent in the Neoproterozoic marks an important stage in the tectonic history of southern Peninsular India. The position and morphological relationship of India relative to Madagascar in the past is one of the most debated and current problems in understanding the tectonics of eastern Gondwana (e.g., Katz and Premoli, 1979; Collins and Windley, 2002; Braun and Kriegsman, 2003; Ghosh et al., 2004; Collins and Pisarevsky, 2005; Collins, 2006; Ashwal et al., 2013; Gibbons et al., 2013; Ishwar-Kumar et al., 2013; Rekha et al., 2013, 2014; Ratheesh-Kumar et al., 2014). Katz and Premoli (1979) suggested that crustal-scale tectonic lineaments/shear zones/suture zones across the continental fragments are one of the major tools that can be used for paleo-geographic reconstruction, based on which they proposed a correlation between India andMadagascar.
Fig. 3. Sample locations and structural lineaments overlayed on the geological map of the Karwar block (modified after Geological Survey of India, 1993, 1996 and 2005; structural lineaments were extracted from Landsat ETM+ satellite imagery and ASTER digital elevation model).
Following this study, several workers attempted to place India and Madagascar into the Gondwana reconstruction, based on regional lithological, structural, geochronological, geochemical and geophysical evidence (e.g., Agarwal et al., 1992; Windley et al., 1994; Windley and Razakamanana, 1996; Collins and Windley, 2002; Raval and Veerasamy, 2003; Ghosh et al., 2004; Raharimahefa and Kusky, 2006, 2009; Ishwar-Kumar et al., 2013b; Rekha et al., 2013, 2014; Rekha and Bhattacharya, 2014; Ratheesh-Kumar et al., 2014). However, the many and varied models have led to considerable inconsistency, mismatch and disagreements about specific correlations. Some of the reasons for this inconsistency arevariation in age and exact position/extent of the shear/suture zones, problems with bathymetry, and the variation or distortion in scale, among other factors.
III. ANALYTICAL TECHNIQUES
A. Electron Microprobe Analysis
Chemical analyses of all the minerals were obtained using an electron microprobe analyser (JEOL JXA-8900R microprobe) housed at the Okayama University of Science, Japan and by JXA-8530F at the University of Tsukuba, Japan. The analyses were performed under conditions of 15-20 kV accelerating voltage, 10-12 nA sample current and a spot size of 3 µm. The data were regressed using the oxide-ZAF correction method. Natural and synthetic silicates and oxides were used for calibration. All the major minerals were analysed for SiO2, TiO2, Cr2O3, Al2O3, FeO, MgO, MnO, CaO, Na2O, and K2O.
B. SHRIMP U-Pb zircon geochronology and zircon Lu-Hf isotopic analysis
Zircon U-Pb dates were obtained using a sensitive high-resolution ion microprobe (SHRIMP II) at Curtin University, Australia. Analytical procedures for the U-Pb analysis followed those outlined by Williams
(1998). Zircons from each sample were mounted together with standards in an epoxy resin disc that was polished to obtain cross-sections through the grains. Prior to the U-Pb analyses, the mount was cleaned and coated with 40Å thickness of gold. To determine the internal structure of individual grains and to identify suitable analytical sites, backscattered electron (BSE) and cathodoluminescence (CL) images were obtained using a scanning
Fig. 4. P-T phase diagram modeled using the calualted bulk composition of quartz-phengite schist (IK-2302L) in the SiO2-TiO2-Al2O3-FeO-MnO-MgO-CaO-Na2O-K2O-H2O chemical system.
electron microscope (SEM; JEOL JSM-5900 LV) at Curtin University. To obtain the SEM images a 0.2nA electron beam current and a 15kV acceleration voltage was used on the gold coated mounts. For SHRIMP analyses, an O2− primary ion beam of 2 nA intensity was utilized to sputter the analytical spot of 25-30µm diameter. The zircon standard BR266 with a 206Pb/238U age of 599 Ma (Stern, 2001) was used as external standard to correct instrumental mass bias and isotopic fractionation. A correction for common Pb was made on the basis of the measured 204Pb and the model for common Pb compositions was that proposed by Stacey and Kramers (1975). The U-Pb data were reduced using the SQUID 2 and Isoplot 3 software (Ludwig, 2003, 2005). Uncertainties reported for
individual analyses are at the 1σ level, and for pooled ages they are at the 95% confidence level. In situ zircon hafnium isotopic analysis was carried out at the Institute of Geology
Fig.5.a. Representative cathodolu-minescence images of quartz-phengite schist (IK-2302L) illustrating zircon structures. b. SHRIMP U-Pb zircon concordia plot. Inset shows representative CL images for zircons for each sample, probability density graph for quartz-phengite schist and bar chart.
and Geophysics, Chinese Academy of Sciences, Beijing. Sites previously dated by SHRIMP were selected for Lu-Hf analyses, and the measurements were performed using a Neptune MC-ICPMS, fitted with a 193 nm ArF laser, with spot sizes 40-50µm and laser repetition rate of 8 Hz at a laser power of 100 mJ/pulse. Analytical and data correction procedures were as described by Wu et al. (2006). To calculate the 176Lu/ 177Hf ratio, interference of 176Lu on 176Hf was corrected by measuring the intensity of the interference free 175Lu, by using the recommended 176Lu/175Lu ratio of 0.02669 (DeBievre and Taylor, 1993). To calculate the 176Hf/177Hf ratio isobaric interference of 176Yb on 176Hf was corrected by using the 176Yb/172Yb ratio of 0.5886 (Chu et
al., 2002). Each ten analysis were bracketed by analyses of standards Mud Tank and GJ-1.
IV. RESULTS
A. P-T evolution A metamorphic pressure –temperature
estimation has done for quartz-phengite schist from the Kumta suture (Fig. 3). The mineral pair phengite ± quartz is stable in all fields (Fig. 10) and kyanite is stable on the high-pressure/temperature side. Chlorite appears when the pressure drops and chloritoid disappears at an even lower pressure. Lawsonite is stable only in the low-temperature/high-pressure corner of the phase diagram (Okamoto and Maruyama, 1999). There are no significant variations in the composition of phengite in Kumta suture samples. Based on the assemblage (phengite-quartz-rutile-chlorite-chloritoid-H2O) and the XMg-XNa composition of phengite, the quartz-phengite schist is calculated to have been stable at c. 13 kbar at 525ºC (Fig. 4). The presence of fine-grained alluminosillicates (sillimanite/ kyanite) in the phengite matrix (Fig. 4) (revealed by X-ray elemental mapping using a field emission microprobe) extends the P-T segment into the kyanite stability field. Thus, based on the present composition, the rock exhumed from a P-T condition of c. 18 kbar and 550ºC (Fig. 4) in the eclogite facies and underwent re-equilibration in the amphibolites facies (Liou et al., 1998; Oh and Liou, 1998). Such a P-T evolution provides confirmation that the Kumta suture is sited on a subduction zone, which was capable of such subduction and exhumation.
B. Timing of metamorphism Zircons from quartz-phengite schist
(IK-110123-02L) sample have rounded to sub-rounded morphology (grain size <100 µm) and no metamorphic overgrowths (Fig. 5a). These are detrital zircons with no metamorphic rims. A total of 32 zircon grains were analyzed, 14 spots out of 32 spots in 14 zircon grains were
used. The zircons from quartz-phengite schist (IK-110123-02L) sample gave mainly four age populations, ca. 2993 Ma, ca. 3101 Ma, ca. 3126 Ma, and ca. 3280 Ma (Fig. 5b). The lower intercept ages ranges from 1000-1200 Ma and used to broadly constrain the timing of metamorphism (Ishwar-Kumar et al., 2013).
Fig. 6. Concordia LA-ICPMS U-Pb plots of zircons. a.. KR23-20C- Garnet-biotite-kyanite-gedrite-cordierite gneiss. b. KR23-20K- Garnet-biotite-kyanite gneiss. c. KR23-20F- Garnet-biotite-hornblende gneiss.
Metasedimentary rocks from the Mercara suture show prominent Paleoarchean to Mesoproterozoic sources. The lower intercept of zircons in all rocks suggest a metamorphic event during the mid- to late
Fig. 7. A ɛHf (t) vs. U-Pb zircon age (207Pb/206Pb age) plot for the metasedimentary samples from the Kumta and Mercara suture zones.
Mesoproterozoic (Fig. 6). Surprisingly, all the zircons from samples KR23-20C, KR23-20K and KR23-20F fall on a common discordia line, suggesting source rocks of a similar age. However the spread in ɛHf (t) values indicates arrange from juvenile to ancient protolith materials, suggesting either magma mixing or a variety of rock-types with a similar age. The hafnium isotopic compositions suggest that the sediments were derived from mixed sources including both juvenile and recycled continental crust. The ɛHf (t) vs. U-Pb zircon age (207Pb/206Pb age) plot defines a major distribution along the 3.0 to 4.0 Ga Archean crustal growth line (Fig. 7).
V. DISCUSSION
Based on the present results from western India and eastern Madagascar, integrating with published results, we propose a close-fit correlation between India and Madagascar. The Betsimisaraka suture zone is mainly composed of paragneisses (with common augen, cataclastic and mylonitic fabrics) associated with pelitic mica schists
containing garnet, staurolite, kyanite, sillimanite and graphite) (Hottin, 1969; Collins and Windley, 2002). The metasedimentary rocks enclose abundant mafic-ultramafic lenses. These rocks and relations can be correlated with those of the Kumta and Mercara suture zones that include quartz-phengite schist, chlorite schist, garnet-biotite schist, marble and amphibolite- to granulite-facies garnet-, kyanite-, sillimanite- and quartz-, feldspar-bearing paragneisses with metagabbro and calc-silicate granulites. The Betsimisaraka suture zone of northeastern Madagascar (Kröner et al., 2000; Collins and Windley, 2002; Collins et al., 2006) correlates with the Kumta suture zone of southern India (Ishwar-Kumar et al., 2013). The southern end of the Betsimisaraka suture of Madagascar extend into the Mercara suture zone (Reported as Coorg suture zone in Ishwar-Kumar et al., 2013 and same suture mentioned as Mercara suture zone in Santosh et al., 2014. But since the block is named as Coorg block, suture is hereafter named as Mercara suture zone (McSZ)) of southern India (Ishwar-Kumar et al., 2013). The Palghat-Cauvery shear zone of southern India has been correlated with the Neoproterozoic Angavo shear zone of Madagascar, and the Tranomaro shear zone of the southern Madagascar has been correlated with the Achankovil shear zone of southern India.
VI. CONCLUSIONS AND FUTURE WORK
Structural and regional geological evidence suggest the presence of shear zones in western India. Textural evidence and metamorphic P-T estimations for the metasedimentary rocks suggest that the sediments were subducted to depths of about 40-50 km depth, and were later exhumed.
The zircon SHRIMP and LA-ICPMS U-Pb geochronology of metasedimentary rocks from the Kumta and Mercara suture zones suggest a common metamorphic event in the Mesoproterozoic.
Integration of results with published data indicates that the Archean blocks in western Peninsular India were sutured during the Mesoproterozoic. We therefore propose that the Kumta and Mercara sutures are the eastern extension of the northern and southern parts of the Mesoproterozoic Betsimisaraka suture of north-eastern Madagascar.
There are several challenges still exist in this field and needs more detailed study in future.
A. Variation in metamorphic grade of Karwar and Coorg region
The Kumta suture contains amphibolite- to greenschist-facies schistose rocks and the Mercara suture contains amphibolite- to granulite-facies rocks. The Karwar block on the western side of the Kumta suture contains ca. 3200 Ma TTG gneisses with enclaves of amphibolite (Ishwar-Kumar et al., 2013b). The Coorg block on the western side of the Mercara suture consists of ca. 3200 Ma gneisses and ca. 3200 charnockites with enclaves of amphibolite and mafic granulite (Santosh et al., 2014). The Coorg block can be considered as a high-grade equivalent of the Karwar block, and the Mercara suture as the high-grade equivalent of the Kumta suture, with deeper crust exposed in the Coorg area.
B. Extension of the Betsimisaraka-Kumta-Mercara suture zone
The northern and southern ends of Betsimisaraka suture in eastern coast of Madagascar are correlated with the Kumta and Mercara suture respectively. However, the further continuation of Mercara suture zone is unclear. The N-S trending eastern end of Mesoproterozoic Mercara suture zone is structurally discordant with NNW-SSE trending Moyar shear zone. The Moyar shear zone is proposed as early Paleoproterozoic suture by Plavsa et al. (2014). The detailed study including structures and geochronology are required to understand this problem.
ISTC/CCE/MS/302
Estimation of soil hydraulic properties in a catchment using agro-hydrological models and microwave remote sensing
K. Sreelash, M. Sekhar, S. K. Tomer, S. Bandyopadhyayand M.S. Mohan Kumar
Abstract–This paper illustrates with results the methodology of estimating multilayered soil hydraulic properties by inversion of a crop model using surface soil moisture and crop bio-physical parameters from microwave remote sensing data. Crop biophysical parameters such as leaf area index (LAI) and above ground biomass are sensitive to the properties of the root zone soil and hence these observations can be useful for estimating the properties of deeper soil layers. We used the radar vegetation index (RVI) and a parametric growth curve of LAI to estimate LAI from radar data. The surface soil moisture (SSM) for vegetated areas is estimated using the water cloud model (WCM). The LAI and SSM estimated from RADARSAT-2 data are then used to estimate multilayered soil hydraulic properties by inversion of STICS crop model. The methodology is tested with marigold crop at AMBHAS field site (www.ambhas.com) for different soil and climate combinations, theresults showthat the proposed approach is a promising one in estimating multilayered soil hydraulic properties.
I. INTRODUCTION
Good estimates of soil hydraulic parameters and their distribution in a catchment is essential for crop and hydrological models. Measurements of soil properties by experimental methods are expensive and often time consuming, and in order to account for spatial variability of these parameters in the catchment, it becomes necessary to conduct large number of measurements.
K. Sreelashis a CNES post-doctoral research fellow at INRA, Avignon, France. [email protected] M. Sekhar, Associate Professor in the Department of Civil Engineering, IISc, Bangalore. [email protected] S. K. Tomer is a post-doctoral research fellow at CESBIO, Observatoire Midi-Pyrenees, Toulouse, France. M.S. Mohan Kumar is a Professor in the Department of Civil Engineering, IISc,Bangalore. S. Bandyopadhyay is a scientist at EOS, ISRO headquarters at Bangalore.
Estimation of soil parameters by inverse modelling using observations on either surface soil moisture or crop variables has been successfully attempted in many studies, but difficulties to estimate root zone properties arise for heterogeneous layered soils. Inverse modeling has been extensively used at column scale (Kumar et al., 2010) but the spatial variability of the parameters and insufficient data sets restricted its applicability at the catchment scale (Vereeckenet al. 2008). Montzkaet al. (2011) demonstrated the possibility of estimating the soil hydraulic parameters using remotely-sensed surface soil moisture measurements by applying a sequential filtering technique to the mechanistic soil-water model HYDRUS 1-D.However, this approach is not applicable for multilayered soils without soil moisture measurements of the deeper horizons. Ines and Mohanty (2008) used the assimilation of near-surface soil moisture in a crop model and concluded that additional soil moisture data from deeper depths may be needed to better estimate the hydraulic properties of layered soils. Vereeckenet al. (2008) indicated that the use of surface soil moisture measurements alone is not sufficient to provide unique and physically reasonable estimates of hydraulic properties at the field scale.
Crop variables such as LAI and biomass are sensitive to the properties of the soil horizons belonging to the root zone and hence these observations could be useful for estimating soil properties of deeper horizons. Recently, Jana and Mohanty (2011) showed that using remote sensed crop variables (LAI) in addition to classical soil texture data improved estimation of soil hydraulic parameters (SHPs) by pedo-transfer functions. Remotely sensed data of crop variables have been used to estimate soil parameters of crop models. Varellaet al. (2010) performed the inversion of a crop model using LAI and yield at harvest to estimate soil parameters in a two-layer soil system and concluded that the estimation of parameters related to soil water content and soil depth could be significantly improved as
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
ISTC/CCE/MS/302
compared to prior information especially when water stress. Charoenhirunyingyos et al., (2011) and Sreelash et al., (2012) showed that a combination of crop and soil moisture variable can provide good estimates of soil water content parameters. Availability of remote sensed information of surface soil moisture and leaf area index makes this approach applicable at a larger scale.
Microwave methods have extensively been used to characterize biophysical crop parameters. Many studies have shown that there is a clear interdependence between biomass, LAI and observed backscattering coefficients from active microwave systems. Leaf area index and biomass can be estimated from microwave remote sensing data using the radar vegetation index (RVI). RVI has been found to one of the promising approaches in estimating the crop biophysical parameters and our initial studies showed good correlation between RVI and LAI.A number of studies have been carried out to investigate the relationship between radar backscattering and soil moisture for different study areas. Various theoretical and empirical models have been developed to retrieve the soil moisture from active microwave data. The major challenge to the theoretical and empirical models is the modeling of backscatter behavior under the vegetation canopy. A model which is frequently used in the modeling of soil moisture in vegetated areas is the water-cloud model (WCM) (Attema and Ulaby, 1978). The variations in the canopy descriptors used in the models that describe canopy backscattering are due to the complexity of vegetation structure and relative simplicity of the models. The parameters of the WCM depend on the crop and soil type, and it is necessary to calibrate the WCM for site specific conditions using suitable vegetation descriptors.
In this study we illustrate and test the methodology of estimating multilayered soil properties by inversion of a crop model using the generalized likelihood uncertainty estimation (GLUE) approach. The approach of estimating LAI from RVI using the parametric growth curve approach is also discussed. Calibration of WCM is carried for site specific crops. Estimability of multilayered soil properties is tested for marigold crop in different soil and climate combinations.
II. EXPERIMENTS, DATA AND MODEL
The experiments were carried out on several agricultural plots of approximately one hectare in size in the Berambadi watershed, 84 km2area (AMBHAS Research Observatory, www.ambhas.com) located in the Kabini river basin in South India, which is an experimental watershed for carrying out agro-hydrological, remote sensing and hydrological investigations. The main crops are turmeric, sunflower, maize, marigold, sugarcane, finger millet, groundnut etc. Soil and crop related measurements were performed during the cropping season from May- 2011 to Dec-2013. Surface soil moisture, profile soil moisture (PSM), leaf area index (LAI) and above-ground biomass were measured at ten day frequencyfrom the sowing to harvest and yield was measured at harvest date.
Table 1.Activities in the experimental plots.
Duration of crop growth (in days): 100 (May to August)
No of plots monitored: 26
Measures Variables: SSM, PSM, LAI, biomass, yield.
Frequency of measurements: 10 day frequency
Satellite Data: RADARSAT-2, 4 images in 2011
3 images in 2012 and 3 images in 2013
STICS crop model (Brisson et al., 2008) is used in this study. STICSis a dynamic, daily time-step model which simulates the functioning of a soil-crop system over a single or several successive crop cycles. STICS simulates the daily carbon balance, the water balance (evaporation and transpiration) and the nitrogen balance in the system, which makes it possible to calculate both agricultural and environmental variables in a variety of agricultural situations. STICS model has been calibrated for the crops specific to the study area.
II. LAIFROM RADAR DATA
RVI (Kim and Van Zyl, (2009) has been proposed as a method for monitoring the level of canopy growth variables like LAI, biomass and VWC. RVI is near zero for a smooth bare surface soil and increases as a crop grows. In this study we used the parametric
ISTC/CCE/MS/302
growth curve described in the previous section to develop the RVI growth curve, since RVI represents the LAI growth characteristics of the crop in some form.
The RVI is given by
𝑅𝑅𝑅𝑅𝑅𝑅 = 8𝜎𝜎𝑜𝑜𝐻𝐻𝑅𝑅
𝜎𝜎𝑜𝑜𝐻𝐻𝐻𝐻 + 𝜎𝜎𝑜𝑜𝑅𝑅𝑅𝑅+ 2𝜎𝜎𝑜𝑜𝐻𝐻𝑅𝑅 (1)
whereσHV is the cross-polarization backscattering cross section and σHH and σVV are the co-polarization backscattering cross sections represented in power units. RVI generally ranges between 0 and 1 and is a measure of the randomness of the scattering. RVI is near zero for a smooth bare surface and increases as a crop grows (up to a point in the growth cycle).
We used the standard “Parametric Growth Curve” (Baret, 1986) which reproduces the time variations of canopy variables like LAI, biomass and vegetation water content (VWC). The growth curve for the vegetation variable LAI is based on five parameters (K, a, b, Ti and Tf) and is given by
𝐿𝐿𝐿𝐿𝑅𝑅(𝑑𝑑) = 𝐾𝐾[1
1 + exp−𝑏𝑏(𝑇𝑇(𝑑𝑑) − 𝑇𝑇𝑖𝑖)
− exp𝑎𝑎(𝑇𝑇(𝑑𝑑) − 𝑇𝑇𝑓𝑓)) (2)
Where K is related to the amplitude of the LAI variations, a and b are responsible for the trend of the crop growth and senescence and the time parameters Ti and Tfcorresponds to the temporal characteristics of the growth curve. The time related parameters Ti and Tfto be constant for all plots in a particular year. The LAI growth curve for each field plots are then generated using the parametric growth curve by estimating the values of a, b and K using the field observed LAI values.
The RVI growth curve can be expressed as,
𝑅𝑅𝑅𝑅𝑅𝑅(𝑑𝑑) = 𝐾𝐾′[1
1 + exp−𝑏𝑏′(𝑡𝑡 − 𝑇𝑇𝑖𝑖′)− exp𝑎𝑎′(𝑡𝑡 − 𝑇𝑇𝑓𝑓′)) (3)
In this equation 𝐾𝐾′ represents the amplitude of the RVI variations similar to the parameter K in the LAI growth curve. The parameters𝑎𝑎′,𝑏𝑏′, 𝑇𝑇𝑖𝑖 ′ and 𝑇𝑇𝑓𝑓′ of the
RVI growth curve corresponds to parameters a, b, 𝑇𝑇𝑖𝑖 and 𝑇𝑇𝑓𝑓 of the LAI growth curve. These parameters can be estimated using the time series of RVI data.The shape of the RVI growth curve will be similar to that of the LAI growth curve, since the parametric growth curve is used in both cases. The relation between the LAI and RVI growth curve now depends on the amplifying factor K and 𝐾𝐾′, and the relation can be established as
𝐿𝐿𝐿𝐿𝑅𝑅(𝑑𝑑) = 𝑘𝑘′ ∗ 𝑅𝑅𝑅𝑅𝑅𝑅(𝑑𝑑) (4)
Where 𝑘𝑘′ defines the relation between K and 𝐾𝐾′ .
The value of K, depends on the maximum LAI of the crop and hence the 𝐾𝐾′can be expected to be a function of the maximum value of RVI.
IV. SSM FROM RADAR DATA
In the general form the WCM (Attema and Ulaby 1978) for a given incidence angle θican be expressed as a linear combination of backscattering by vegetation and backscattering by soil and is given by the following equation,
𝜎𝜎𝑝𝑝𝑝𝑝𝑜𝑜 = 𝜎𝜎𝑣𝑣𝑣𝑣𝑣𝑣𝑜𝑜 + 𝜏𝜏2𝜎𝜎𝑠𝑠𝑜𝑜𝑖𝑖𝑠𝑠𝑜𝑜 (5)
Where 𝜎𝜎𝑝𝑝𝑝𝑝𝑜𝑜 is the co-polarized total backscatter coefficient, 𝜎𝜎𝑠𝑠𝑜𝑜𝑖𝑖𝑠𝑠𝑜𝑜 is the backscatter contribution of the soil surface, 𝜎𝜎𝑣𝑣𝑣𝑣𝑣𝑣𝑜𝑜 is the backscatter contribution of the vegetation cover, expressed as,
𝜎𝜎𝑣𝑣𝑣𝑣𝑣𝑣𝑜𝑜 = 𝐿𝐿𝑅𝑅1𝑐𝑐𝑜𝑜𝑠𝑠𝜃𝜃(1 − 𝜏𝜏2) (6)
and𝜏𝜏2 is the two attenuation by the vegetation, expressed as,
𝜏𝜏2 = exp(−2𝐵𝐵𝑅𝑅2𝑠𝑠𝑣𝑣𝑐𝑐𝜃𝜃) (7)
Where A and B are the vegetation parameters, 𝑅𝑅1 and 𝑅𝑅2 are the vegetation descriptors.
Following Lievens and Verhoest (2011), the combination 𝑅𝑅1 = 𝐿𝐿𝐿𝐿𝑅𝑅 and 𝑅𝑅2 = 𝐿𝐿𝐿𝐿𝑅𝑅 is used in this study primarily because of the interest to estimate soil moisture and LAI for each vegetation and experimental plots can be retrieved using RVI as discussed in the previous section. The parameters A
ISTC/CCE/MS/302
and B are obtained through nonlinear regression or least square fitting approach.
The underlying soil contribution 𝜎𝜎𝑠𝑠𝑜𝑜𝑖𝑖𝑠𝑠𝑜𝑜 can be evaluated from a surface scattering model or, more simply, from a linear function of its soil moisture𝑚𝑚𝑠𝑠.
𝜎𝜎𝑠𝑠𝑜𝑜𝑖𝑖𝑠𝑠𝑜𝑜 = 𝐶𝐶 + 𝐷𝐷𝑚𝑚𝑠𝑠 (8)
The parameters C and D are obtained with linear model fitting.
V. SOIL PARAMETERS FROM RADAR DATA
Parameter estimation by inversion of a dynamic crop model like STICS is a complex process, since such models involve parameter interactions and hence obtaining a single optimum soil parameter set is not realistic. Generalized Likelihood Uncertainty Estimation (GLUE) (Beven, 2008) an informal Bayesian method using prior information about parameter values for estimating model parametersis used for the parameter estimation process. The sum of absolute errors (SAE),is used in this study as the likelihood measure. The likelihood SAE is calculated for each variable considered in the inversion for each model run as,
𝑆𝑆𝐿𝐿𝑆𝑆𝑖𝑖𝑘𝑘 = y𝑖𝑖 ,𝑗𝑗
𝑘𝑘− 𝑦𝑦𝑖𝑖𝑘𝑘
𝑀𝑀𝑖𝑖
𝑗𝑗=1
𝑖𝑖 = 1 𝑡𝑡𝑜𝑜 𝑛𝑛 (9)
where, 𝑆𝑆𝐿𝐿𝑆𝑆𝑖𝑖𝑘𝑘 is the sum of the absolute errors for any parameter set k, k=1,…N (N being the number of ensembles, 20000 in this study), variable i, i = 1 to n corresponding to each variable considered with n being the total number of variables, and measurement date j, j=1,…,Mi (Mi being the total measurements for the variable I) 𝑦𝑦is the simulated value of the variable for the (i,j)th parameter set and y is the measured value of the variable. The parameters related to soil water storage characteristics such as field capacity, wilting point and thickness of soil layer are estimated in this study.
V. RESULTS AND DISCUSSIONS
The parameters of Eq. (2), (3) and (4) are estimated using the GLUE approach using the observed values
of LAI and RVI values retrieved from RVI. It was observed that the slope parameters and temporal characteristics of LAI and RVI curves modeled by Eq.(2) and (3) were close to each other, whereas the amplification factor K and K’ varied with plot. The k’values for each field plots are estimated. The values of k’ varied from 3.8 to 5.2 for field. The observed LAI growth curve and the RVI growth curve modeled using Eq. (2) is shown in Fig.1 along with the corresponding observed values (median and standard deviation of all plots.
Figure 1: RADARSAT-2 retrieved RVI and the RVI growth curve, observed LAI and LAI growth curve for Marigold crop (2011).
Figure 2: Scatter plot of observed LAI and LAI estimated from RVI for different days after sowing (DAS).
ISTC/CCE/MS/302
Using Eq. (2) and Eq. (3) in combination will provide a time series of LAI from a time series of RVI data. Fig. 2 shows the scatter plot of observed LAI and the LAI obtained using the RVI and growth curve approach for 10, 20, 30, 40, 50, 60, 70, 80 and 90 days after sowing for a marigold crop.
The water cloud model was calibrated for the marigold crop using LAI and the vegetation descriptor. The parameters of the WCM A and B are found to be 4.52 and 0.024 respectively. The parameter C and D depends on soil type and are estimated separately for each soil class. The calibrated WCM was first used to estimate the σHH and then using inversion method the surface soil moisture was estimated. The estimated σHH and the RADATSAT-2 retrieved σHHagreed closely. The scatter plots of the modelled σHHand retrieved σHHshown in Fig. 5(a) and the scatter plots of surface soil moisture is shown in Fig. 5(b).
Figure 3: Scatter plot of (a) σHH and (b) surface soil moisture.
The LAI and SSM estimated from RADARSAR-2 data were used to estimate the soil hydraulic properties of two-layered vertically heterogeneous soils in the experimental plots. The scatter plot of the estimated field capacity of both horizons and the observed values are shown in Fig.4 (a) and (b). The SHP’s obtained from model inversion agrees closely with the SHP’s obtained from field experiments.The close agreement between the field experiments and the estimates obtained from inversion of satellite data shows that this approach has the potential to estimates soil hydraulic properties at a catchment scale, since remote sensing data of LAI and surface soil moisture are available at that scale.
Figure 3 shows the scatter plot of the field capacity and wilting points of different layers (2 layers in this study) estimated from field based experiments and estimates obtained using inversion of satellite LAI and surface soil moisture.
VI. SUMMARY AND CONCLUSIONS
ISTC/CCE/MS/302
The approach of estimating multilayered soil hydraulic properties using remote sensing data is a promising approach as demonstrated in this study. In this paper we examined the applicability of the approach usingmarigold crop. The suitability of the method for different climatic conditions (low and high rainfall) is examined using data from three seasons with different rainfall and water stress conditions. The approach of estimating LAI from RVI using the growth curve approach is a promising method for application in areas with data gaps. The calibrated water cloud model was found to be suitable for estimating the surface soil moisture under vegetated areas.
Acknowledgement: We thank Dr. Laurent Ruiz (INRA, Rennes), Dr Martine Guerif&DrSamuel Buis (INRA, Avignon) andDrSamuel Corgne (University of Rennes)for their technical support and advice at various stages of the work.
REFERENCES
[1] Attema, E.P.W. and Ulaby, F.T. (1978). Vegetation modeled as a water cloud. Radio Science, 13 (2), 357–364.
[2] Baret, F. (1986). Contribution au suive radiométrique de cultures de céréales. Phd Thesis Submitted to University of Paris South (Université Paris Sud).
[3] Beven, K. J., Smith, P., and Freer, J. 2008. So just why would a modeller choose to be incoherent? J. Hydrol., 354 (1-4), 15–32.
[4] Brisson, N., Mary, B., Ripoche, D., Jeuffroy, M.H., Ruget, F., Nicoullaud, B., Gate, P.,Devienne-Baret, F., Antonioletti, R., Durr, C., Richard, G., Beaudoin, N., Recous, S.,Tayot, X.P.D., Cellier, P., Machet, J.-M., Meynard, J.-M. andDelecolle, R. 1998. STICS:a generic model for simulating crops and their water and nitrogen balances.I. Theory and parameterization applied to wheat and corn. Agronomie 18,311-346.
[5] Charoenhirunyingyosa., S, Hondaa., K., Kamthonkiatb., D. and Ines., A. V.M. (2011). Soil moisture estimation from inverse modeling using multiple criteria functions. Journal of Computers and Electronics in Agriculture, 75(2), 278–287.
[6] Ines, A. V. M. and Mohanty, B. P. 2008. Near-Surface soil moisture assimilation for quantifying effective soil hydraulic properties under different
hydroclimatic conditions. Vadose Zone Journal, 7(1), 39-52.
[7] Jana, R. B. and Mohanty, B. P. 2011. Enhancing PTFs with remotely sensed data for multi-scale soil water retention estimation. Journal of Hydrology, 399, 201-211,
[8] Kim Y. J., and Van Zyl, J. (2009) “A time-series approach to estimate soil moisture using polarimetric radar data,” IEEE Trans. Geosci. Remote Sens., vol. 47,no. 8, pp. 2519–2527, Aug. 2009.
[9] Kumar, S., Sekhar, M., Reddy, D. V. and Mohan Kumar, M. S. 2010. Estimation of soil hydraulic properties and their uncertainty: comparison between laboratory and field experiments. Hydrological Processes, 24, 3426-3435.
[10] Lievens, H. and Verhoest, N.E.C. (2011). On the retrieval of soil moisture in wheat fields from L-band SAR based on water cloud modeling, the IEM, and effective roughness parameters. IEEE Geoscience and Remote Sensing Letters. 8(4): 740-744.
[11] Montzka, C., Moradkhani, H., Weihermuller, L., Franssen, H.H., Canty, M., Vereecken, H., 2011. Hydraulic parameter estimation by remotely-sensed top soil moisture observations with the particle filter. Journal of Hydrology 399, 410-421.
[12] Sreelash, K., Sekhar, M., Ruiz, L., Tomer, S.K., Guerif, M., Buis, S., Durand, P. and Gascuel-Odoux, C. 2012. Parameter estimation of a two-horizon soil profile by combining canopy and surface soil observations using GLUE. Journal of Hydrology. 456-457, 57-67.
[13] Varella, H., Guerif, M., Buis, S. and Beaudoin, N. 2010b. Soil properties estimation by inversion of a crop model and observations on crops improves the prediction of agro-environmental variables. Europ.J.Agronomy 33, 139-147.
[14] Vereecken, H., Huisman, J.A., Bogena, H., Vanderborght, J., Vrugt, J.A. and Hopmans, J.W. 2008. On the value of soil moisture measurements in vadose zone hydrology: a review. Water Resources Research 44.
Abstract—pumps, vconventiocryogenicpoints of lubricantmolybdenpolymerscarried ostudy thtemperatudevelopedwear typ(ASTM Gpins wabrasive load of 1and cryotemperatuof friction
In msurfaces valves, sconventiocryogenicpoints ocritical generatiolike tefloamorphoTribologicryogenicpresent ecarbon fi
PTFEand rettemperatuuse. The much hipolymersabsorb dimensio
D.S. Nafor Cryogenadig@ Dr. PauTribology paulpgISTC/CC
— Many cryovalves, seals
onal means wic temperature
these lubricans like po
num disulphid, etc. are used
on plain and cheir tribologure. A cryotrid which workspe tribometer G 99-95).Bothwere made tosurface rotatin
10N. Experimeogenic temperure on the wean are presented
I. IN
many cryogeniwith relativeseals etc. Thonal means lic temperaturef these lubriin respect t
on. For theseon (PTFE), mus carbon, ical experimc temperature
experimental silled PTFE maE exhibits a vtains useful ures from 13crystalline migher than ms. PTFE is cwater and
onal stability.
Tribolog
adig is a Principenic Technology,@ccf.iisc.eul George is a Scie
Division, LPSC,eorge@yahoCT/DSN/309
genic applicatetc cannot
ith oils and ge range is farnts. For these
olytetrafluoroede (MoS2), amd. In this studycarbon filled P
gical propertiibometer has bs on the princ
according toh plain and ca
slide for 10ng at 400 rpments were carratures. The ear, frictional fod.
NTRODUCTION
ic systems, th motion like hey cannot ike oils and ge range is faicants. Theseto wear and
e applicationsolybdenum di
polymers, mental data es are hardly study, the focaterial. [1] very low coef
mechanica3K upto 533Kelting point ismost of thechemically in
hence exhi
gical prop
al Research Scien, IISc., Bangaloreernet.in entist at Rotor dy, ISRO, Valiyamaoo.com
tions like bearbe lubricated
grease becauser below the applications,
thylene (PTmorphous cary, experimentPTFE materiaies at cryogbeen designed
ciple of pin ono ASTM stanarbon filled P minutes ove
m under the apried both at reffect of cryogorce and coeffi
N
ere are interacbearings, pu
be lubricatedgrease becausear below the e componentsd frictional s, solid lubricisulphide (Moetc. are uof materials
available. Inus is on plain
fficient of fricl propertiesK for contins at 600K whie semi-crystanert and doesibits very g
perties of te
D.S.Nadig, V
ntist in the Centre.
ynamics & ala
rings, d by e the pour solid
TFE), rbon, s are
als to genic
d and n disc ndard PTFE er an plied room genic icient
cting umps, d by e the pour
s are heat
cants oS2), used. s at n the n and
ction s at nuous ich is alline s not good
Polytetraemperatur
V.K.Pavan, P
re
The stru
critical factobehaviour of on the percenthe temperatuthe compositstudy, experifilled PTFE properties bot
The crydeveloped batype tribomeapplications. is made to slithe load for tperiod, the preduction in pin before anBased on thisdeveloped whof ASTM stmaterial is aspeed of the dto centre of contacting sushown in the
Figure 1. Schem
This conthe cryotribopin and thtemperature schematic offigure 2.
afluoroethres
Paul P Georg
ucture of the or at low
f the PTFE contage composure decreases,tes change iniments are ca
materials tth at room and
II.CRYOTRI
yotribometer ased on the preters suitable The test speide over the rthe fixed time
pin wears whilength. The d
nd after the tess concept, thehich is in accotandards (ASa function of disc, track radthe disc) and
urfaces. The figure 1.
matic of the pin on
ncept is extenometer where he disc is
using liquif the cryotrib
hylene at c
ge
polymer comtemperatures
omposite strucsition of carbo, the materialndividually. Inarried on plaito study thed cryogenic te
IBOMETER
has been drinciple of pin
for roomecimen in the rotating abrasie duration. Duich is clearly
difference in lst is a measure pin on disc ordance with tTM G99-95)the applied
dius of the pind the frictionschematic of
n disc tribometer
nded to the dethe interface
maintained d nitrogen bometer is s
cryogenic
mposites is as. The wearcture dependson within. Asl properties ofn the presentin and carbone tribologicalemperatures.
designed andn on disc wearm temperature
form of a pinive disc underuring this test
y indicated bylengths of there of the weartribometers isthe guidelines
).Wear of theload, rotating
n (with respectn between thef the same is
evelopment ofe between theat cryogenic(LN2). The
shown in the
c
a r s s f t n l
d r e n r t y e . s s e g t e s
f e c e e
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Figure 2. Schematic of the cryotribometer
A motor drives a shaft through a V belt. This shaft has an abrasive grinding wheel at the bottom. The test specimen (pin) is made to slide against the rotating wheel under the effect of applied load of dead weights. The top end of the pin is connected to an LVDT. As the pin wears, its length decreases which is sensed by LVDT. The final displacement of the LVDT is the actual wear of the test sample. Frictional force between the pin and disc is measured with the help of a load cell.
This system can be used to study the tribological properties at low temperatures by creating a cryogenic environment with liquid nitrogen. To facilitate this, the interfacing pin and disc region is housed within a double walled insulated chamber produced out of stainless steel (SS304L). Liquid nitrogen from a pressurised container is made to flow over the interfacing zone to maintain the cryogenic temperature. The motorised jack lifts up the insulated cryochamber upwards so as to form a leak proof assembly with the top cover of the tribometer.
III.DEVELOPMENT OF CRYOTRIBOMETER
The cryotribometer has been designed and developed by extending the concept of conventional pin on disc tribometer to suit cryogenic environment.
The entire system is housed in a metallic frame of size 700X600X920mm. This tribometer assembly is mounted to the top cover plate. The frame is assembled with sufficient stiffness to absorb dynamic forces developed during tests. The entire assembly is supported on table top with pads.
A 0.4kW AC motor mounted below the top cover drives a spindle assembly with a V belt. The spindle made out of hylam is made to rotate at a constant speed of 400rpm with the help of reduction
gear box. Hylam is specifically used since it is compatible with cryogenic environment and exhibits low value of thermal conductivity and good mechanical strength properties. This shaft is mounted on the cover plate with a labyrinth seal housing to avoid cold leakages. The shaft has circular platform at the bottom. Abrasive discs or papers of 100mm diameter can be rigidly held on this platform.
Parallel to this spindle assembly, a plunger is mounted on the cover plate at an axial distance of 30mm. This plunger made of hylam extends to both top and bottom of the cover plate and it is housed within a Teflon labyrinth seal which prevents cold leaks. The plunger moves up/down freely inside two stroke bearings enclosed within the bearing housing. A circular platform is located on the top to hold dead weights. These dead weights impart pressing force on the pin against the rotating abrasive disc.
At the bottom of this plunger, test pins of 6mm diameter and 15mm length can be vertically clamped on to a holder which is connected with the plunger. The pin is located at a constant track radius of 30mm from the axis of the spindle. As the disc rotates at a constant speed of 400 rpm for the selected time duration, the pin continuously wears due to the friction and its length reduces. This reduction is the total wear.
Wear of the test pin is measured with the help of an Linear Variable Differential Transducer (LVDT) which has the maximum sensing limit of 4 mm. LVDT is fixed vertically in inverted position on the top cover plate with its spring loaded plunger pressing against a horizontal reference plate which is connected with the load carrying plunger. As the pin wears, the reference plate moves downwards pushing the LVDT plunger inwards. The continuous inward movement of the plunger is a measure of wear.
The frictional force between the pin and abrasive disc is measured with the help of a button type load cell which has the maximum sensing limit of 50N. The frictional force sensor is firmly secured on a stationary bracket mounted on the plunger. The frictional force between the pin and disc is directly transmitted to the sensor through the bracket. Since the load cell pin is pressed against the sensor, the frictional force is transmitted to the load cell. The output readings of the calibrated load cell are the readings of the frictional force.
This tribometer can be used to study the tribological properties at cryogenic temperature using liquid nitrogen and a cryogenic chamber. The cryogenic chamber made of SS304 is a double walled insulated cylindrical unit. The space between the inner and outer vessel is filled with polyurethane
foam (PUF) to minimise heat in leaks from the ambient atmosphere.[4]This chamber encloses the test area comprising of pin and disc. The chamber can be raised with the help of a motorised jack so that its top flange butts against the bottom surface of the cover plate. An O ring is placed between them to prevent leakage of LN2.The top cover consists of two ports for inlet and exhaust of LN2, pressure relief valve, coupling for electrical connections and a pressure gauge. The motor of the jack is operated by a portable remote switch.PT 100 sensor is used to measure the temperature of the test region.
The tribometer is operated and monitored by an electronic controller. The front panel displays the wear, frictional force, temperature, speed and time. Various experimental parameters like duration, speed, applied load etc can be initialised before start of the experiment. The graphical and digital output readings of the tribometer like wear, frictional force, speed, temperature and time can be continuously observed on the monitor with a Data Acquisition System (DAQ) connected with the controller. The photographs of the cryotribometer assembly and the pin on disc assembly are shown in figures 3 and 4.
Figure 3. Experimental system
Figure 4. Pin on disc assembly
IV.EXPERIMENTS
In this experimental work, tribological tests were carried on two types of materials namely plain and PTFE with 30% carbon composition. Trials were conducted both at room and cryogenic temperature using the developed cryotribometer. Large of test specimens (test pins) with 7 mm diameter and 27mm length were fabricated from plain and carbon filled PTFE cylindrical rods. In order to obtain reliable test results, it was planned to conduct many repeated trials for various combinations of applied load and various grades of abrasive emery paper for fixed speed of 400 rpm, time duration of 10 minutes and track diameter of 60mm respectively. Initial trials were conducted with various combinations of 10N, 15N and 20N load and abrasive papers of grade 320, 600, 800, 1000 and 1200. Consistent results were obtained with applied load of 10N and emery paper of grade 1000.
The test pins were pre worn with the abrasive paper with the same experimental parameters to ensure parallelism between the mating surfaces of the pin and the paper. This pre run ensures identical rough surfaces for all the pins before the actual test run.
Before start of the test run, the pin was securely clamped on to the pin holder and the abrasive paper was rigidly held on the top of the circular platform. After placing the selected dead weight of 10N on the weight platform, the pin surface was made to come in surface contact with the abrasive paper. The initial digital outputs of displacement (wear) and frictional force were set to zero through the control panel. Ensuring these set parameters, the tribometer was switched on for the test duration of 10 minutes. As the test progressed, continuous data of wear, fictional force and coefficient of friction displayed on the monitor were monitored. After the test run for 10 minutes, the output data was stored both in digital and graphical form for further study and analysis.
The same tribometer was used to carry out the wear tests in cryogenic environment with the help of pressurised LN2. The wear takes place within the space of a cryochamber. The cryochamber was partially filled with LN2 and allowed to cool and stabilise for some time. The chamber was raised up with the motorised jack till its top flange seals with the bottom surface of the cover plate with an O ring. Pressurised LN2 was supplied into this chamber allowing liquid to flow over the pin-disc interface. A stabilised cryogenic temperature of 98K could be obtained within the chamber. Wear tests were carried out following the same procedure adopted for room temperature tests. These results were compared with the room temperature data and analyzed.
V.RESULTS
The tribological behaviour of a material depends on various factors like wear, coefficient of friction, frictional force between contact surfaces, roughness surface, applied load, rotational speed, duration, etc. As the temperature changes, the material properties also tend to change and in turn influence the tribological behaviour. As mentioned earlier, the emphasis was laid on the influence of cryogenic temperature on the wear properties of PTFE.
Frictional force is a measure of the actual force between the rubbing surfaces. This depends on the applied load and is measured with the help of load cell. Coefficient of friction is the ratio of the frictional force to the applied load. This can range between 0 to1 in the extreme cases. Higher values of coefficient of friction indicate higher transmission of applied load to the interfacing surfaces. The total wear of the material in a wear test depends on the frictional force and the coefficient of friction.
The wear test results at room and cryogenic temperature are shown in the following figures 5, 6 and 7.
Figure 5. Wear at room and cryogenic temperature
PTFE is relatively a soft material. It exhibits low coefficient of friction and high wear rate. This limitation of high wear rate is overcome by impregnating PTFE with carbon since carbon particles exhibit low wear properties. As the temperature is lowered to cryogenic zone, strength and hardness of both PTFE and carbon filled PTFE increase. [5, 6] Due to these combined effects, the wear is expected to reduce at cryogenic temperature. As per the test results, the fall in total wear was significant for plain PTFE at cryogenic temperature. Carbon filled PTFE exhibited lower wear at room temperature as compared with plain material due to the inherent wear resistance characteristics of carbon particles. The reduction in wear at cryogenic temperature was less as compared with the data of plain PTFE.
Wear property of a material depends on the frictional force and the coefficient of friction in the interface between the pin and disc assembly. The results of both coefficient of friction and frictional force at room and cryogenic temperature are shown in the figures 6 and 7.
Figure 6. Frictional force at room and cryogenic temperature
Figure 7. Coefficient of friction at room and cryogenic temperature
Both frictional force and coefficient of friction are temperature dependent. As compared with plain material, carbon filled PTFE exhibits lower coefficient of friction and frictional force at room and cryogenic temperature. Since both these values are comparatively lower at cryogenic temperature, wear should be expected to reduce at cryogenic temperatures. This effect has been observed by the reduction in wear at low temperature zone.
VI.CONCLUSION Experiments have been have been carried to
study the tribological properties of plain and carbon filled PTFE material at room and cryogenic temperatures. The results indicated reduction in wear properties for both plain and carbon filled PTFE at cryogenic temperature.
There is scope for further research to study the tribological properties of PTFE at cryogenic temperature with various percentage compositions of carbon. It is also worth to extend studies on
tribological properties of cryotreated PTFE and its composites at cryogenic temperatures.
REFERENCES
[1] G.Theiler, “PTFE and PEEK Matrix composites for tribological applications at cryogenic temperature and in Hydrogen”,Federal Institute for Materials Research and Testing,Berlin (2005)
[2] G. Theiler, “ Friction and wear of PTFE composites at cryogenic temperature”, Tribology International Vol 35,pp 449-458 (2002)
[3] S.K.Biswas, “Friction and wear of PTFE – a review”, Wear Vol 158,pp 193-211(1992)
[4] T.Gradt et al,” Low temperature tribometers and the behaviour of ADLC coatings in cryogenic environment”, Tribology International Vol 34,pp 225-230, (2001)
[5] “Properties Handbook-PTFE”, Dupont [6] Randall F Barron “Cryogenic systems”, Oxford University
Press (1985)
Portable Imaging Flow Analyzer for Biological Research Applications
in Space
Veerendra Kalyan Jagannadh and Sai Siva Gorthi
Abstract—Cell viability assays are performed to assess theresponse of a population of biological cells to an external stimuli.Conventional optical microscopy is one of the standard techniquesused to perform cell viability assay. Due to their bulky nature,microscopes are not ideal candidates for studying the behaviourof biological specimen in space. In this article, we report theuse of a portable imaging flow analyzer based on a mobilephone to perform an automated cell viability assesment of agiven population of yeast cells. With the use of a cellphoneaugmented with off-the-shelf optical components and customdesigned microfluidics, we demonstrate a portable optofluidicdigital microscope. We demonstrate an enhanced throughput ofabout 450 cells/second, by implementing a multiple microfluidicchannel geometry.
Keywords—Portable Microscopy, Microfluidics, Flow Cytometry,High-throughput Imaging, Cell viability assessment.
I. INTRODUCTION
In the recent past, there has been a significant growth ofinterest in studying and understanding behaviour of biolog-ical cells in space. Researchers have attempted to performcell-culture and tissue engineering experiments in space tounderstand the behaviour of biological cells in low-gravityenvironments [1]–[3]. Cell viability assay is a fundamentalmethod used to study proliferation of cells in a given pop-ulation subjected to any given external stimuli.
Morphological changes or changes in membrane perme-ability are used to asses the viability of cells [4]–[6]. Viabilityassays can be loosely classified into two types: analysis ofwhole population and analysis of individual cells in a givenpopulation. Assessment of viability on a single cell levelprovides a more detailed result, when compared to bulkmeasurements carried out on the whole population [5]. Themembrane permeability and/or physiological state is inferredby assessment of exclusion of certain types of dyes or uptakeand retention of other types of dyes. One of the widelyused viability assessment techniques is the trypan blue dyeexclusion test. It is based on the principle that live cellswith an intact plasma membrane exclude trypan blue; whereasnon-viable cells take up trypan blue as their cell membraneis not intact and hence, unable to control the passage ofmacromolecules through the cell membrane.
In general, optical detection techniques: flow cytometryand microscopy are employed to asses the dye exclusion orretention in cells and thereby infer their viability [7]. Both
Veerendra Kalyan Jagannadh is a graduate student at the Department ofInstrumentation and Applied Physics, Indian Institute of Science
Dr. Sai Siva Gorthi is an assistant professor at the Department of In-strumentation and Applied Physics, Indian Institute of Science. [email protected]
STC Project Code - ISTC/IAP/SSG/299/2013
flow cytometry and microscopy are gold-standard techniquesemployed in medical diagnostics as well as basic biologi-cal research. Flow cytometry enables automated high-speedquantitative multi-parameter analysis of large population ofcells. In contrast to flow cytometry, microscopy has a simplerarchitecture and enables viability assessment of cells viaexamination of images of cells smeared on a slide. Whileboth flow cytometry and microscopy are widely used standardtechniques, their implementation in space / on flight is quitedifficult owing to their bulkyness.
While it is possible to perform cell viability assesment post-flight, the availability of compact, low-weight and automatedimaging systems would enable on flight cellular studies andfurther the progress of bio-medical / biological research inspace. Some of the recently reported portable microscopes [8]–[13] are relevant in this context. However, these portable mi-croscopes inherit some limitations of conventional microscopy:requirement of skilled personnel for sample handling andpreparation, need for focus adjustments and manual scanningof the slide through multiple field of views to acquire imagesof larger number of cells on the smear/slide etc.
On the other hand, there has been recent trend of enhancingthe throughput of conventional microscopy with the synergisticuse of optics and microfluidics. These imaging systems are ingeneral referred to as Imaging Flow Cytometry (IFC) systems[14], [15]. Most of the reported IFC systems in literatureemployed high-speed cameras, which tend to be bulky andhence are not so ideal for realization of imaging flow analyzers,which can be used on flight.
Following a similar approach, there have been a fewattempts to develop portable imaging flow analyzers [16],which employ low weight consumer electronic devices (likecellphones) for imaging and microfluidics for sample handling.However, throughputs in these devices are limited by the lowframe rates of acquisition. The possibility of enhancing thethroughput of portable imaging systems with the use of customdesigned microfluidic devices has not been investigated uponpreviously.
In this paper, while employing the principles of imagingflow cytometry, we demonstrate a microfluidics based portablebright-field imaging system with the use of a cell-phoneand low cost off-the shelf optical components. In this work,with the use of a multiple microfluidic channel geometry, wedemonstrate the enhancement of throughput in the contextof imaging flow analyzer based on low-frame rate imagingdevices. With the frame rate of acquisition, being only 30frames per second (fps), we demonstrate a throughput of about450 cells per second. Further, we demonstrate an importantapplication of such a portable imaging system by performinga cell viability assay on a given population of yeast cells.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Fig. 1. Schematic of the cellphone based portable microscope.
II. CELL-PHONE BASED PORTABLE DIGITALMICROSCOPE
As mentioned earlier, there have been several demonstra-tions of portable microscopy systems. Most of the previouslydemonstrated systems predominantly fall into two distinctcategories, they are lens-based [8], [10]–[12], [17] and lens-free systems [9]. In this work, we have adopted a lens-basedsystem, wherein we have used an eye-piece (10X) and ageneric objective (40X, N.A = 0.65) to turn a cellphone intoa microscope. A 3W LED augmented with a ground glassdiffuser (Grit = 600, Diameter = 1 inch) and an asphericcondenser lens (f = 20 mm, Diameter = 1 inch) have beenused to provide a uniform illumination on the sample plane.The schematic of the imaging system is shown in figure 1.
A. Characterization of the Imaging System
A 1951 USAF Negative resolution target (Edmund Optics55-622) has been used to characterize the imaging system. Thesystem field of view was estimated to be about ∼ 180 µmin diameter. The system is easily able to resolve the smallestfeature present on the target, which is a 0.78 µm feature(shown in figure 2). The line profiles across the images of1st, 2nd, 3rd elements of the 9th group have been shown (figure2 (c)).
III. INTEGRATED MICROFLUIDIC PORTABLE IMAGINGSYSTEM
Custom-fabricated microfluidic devices were integratedinto the portable microscope system so as to enable automatedimaging as in the case of a typical IFC system. Conventionalphoto-lithography and soft-lithography [18] techniques wereused to fabricate microfluidic devices compatible with thecompact digital microscope. The microfluidic devices werebonded to microscope cover slips (thickness = 0.13 mm) usingoxygen-plasma treatment, in order to seal the microfluidicchannels. Each device had parallel channels with a width =20 µm, depth = 6 µm and an inter-channel spacing of about60 µm . At the inlet of the microfluidic device, a spatialfilter was incorporated so as to filter out particles bigger thandimensions of the narrow portions of the channel to avoid
Fig. 2. Imaging system characterization using USAF 1951 negative resolutiontarget. (a)-(c) Images of 1st, 2nd and 3rd elements of the 9th group and theircorresponding line profiles. Length of scale bar is 5 µm
clogging of the channels/device. The representative schematicof the microfluidic device is shown in figure 3.
Fig. 3. Design of the microfluidic device. Device consisted of straightchannels which were 20 µm wide in the central portion of the device and48 µm at other locations.
An attachment to the cell-phone, for housing the optics(LED, diffuser screen, condenser and objective lens) and thereplaceable microfluidic device (cartridge) was fabricated. Theattachment consists of a slot (opening), so that inserting the mi-crofluidic cartridge into this slot positions it at the sample planeof the imaging system. The positioning of the microfluidicdevice in the sample plane, enables us to image the biologicalspecimen flowing through the microfluidic channels of thedevice. The image of the fully integrated automated portablemicroscope is shown in figure 4.
IV. SYSTEM THROUGHPUT CHARACTERIZATION IN FLOW
The achievable throughputs using the portable optofluidicmicroscope were investigated by imaging a suspension ofHuman Red Blood Cells (RBCs) in flow. The suspensionwas flown through the microfluidic device at a flow rate ofabout 40 µlh−1 and the videos of the flow stream wererecorded. Images of different cells were extracted from thevideos using background subtraction and morphological imageprocessing toolbox of MATLAB. Using the presented system,
Fig. 4. Image of the integrated opto-fluidic imaging flow analyzer.
Fig. 5. [Color Figure] Imaging system field of view, while imaging themicrofluidic device with lesser inter-channel spacing. Field of View - 196µm× 180 µm. Length of scale bar is 20 µm
we have been able to achieve imaging throughputs of upto3000 cells per minute. Higher flow rates can be used toensure more particles in the field of view and thereby higherpossible imaging throughputs. However, it was observed thatfurther increase in flow rates resulted in loss of fidelity inthe acquired images due to motion-blur. In order to enhancethe throughput of the system, a multiple channel geometrywith inter-channel spacing about 9.8 µm was employed. Animage of the field of view of the optofluidic microscope,while imaging the device with closely spaced channels isshown in figure 5. Multiple channels were used in order tosimultaneously image cells flowing across different channelsand thereby obtain an optimally high throughput. Using thismultiple channel geometry and optimizing the concentrationand pumping speeds for the given frame rate and exposure,we have been able to enhance the throughput of the system toabout 27000 cells per minute.
V. CELL VIABILITY ASSESSMENT OF YEAST CELLS
In order to demonstrate the relevance of the presentedsystem to perform cellular studies in space / on flight settings,
Fig. 6. [Color Figure] Quantitative assessment of staining in the given yeastcell population. Color and saturation level images of cells corresponding todifferent gated regions of the histogram have been shown. Length of scale baris 8 µm.
we have performed an automated cell viability assay of agiven population of yeast cells. Yeast species Saccharomycescerevisiae was cultured in YPD (yeast extract, peptone anddextrose) (Himedia G037) liquid medium overnight at 37Cin orbital shaking incubator. The cells were harvested afterincubation by centrifuging at 500g for about 10 minutes.Harvested cells were resuspended in distilled water. The yeastsuspension of appropriate dilution was divided into two partsand one part of it was subjected to heating at 55C for 5minutes to prepare dead cells. The dead cells were stainedwith 5% methylene blue for 10 minutes at room temperatureand were separated by centrifuging at 500g for 10 minutesand supernatant was discarded. The stained dead cells and theviable cells were mixed in distilled water. It is known fact thatonly dead cells would take up and retain the methylene bluestain [19].
The suspension of yeast cells was assessed using theportable optofluidic imaging flow analyzer system. The sus-pension was flown through the microfluidic device at a flowrate of about 40 µlh−1 and the videos of the flow stream wererecorded. A snapshot of the field of view (196 µm× 180 µm),while imaging the suspension of yeast cells is shown in figure5. By computationally assessing the color content of the imageof a given cell, we determine its viability status. Since, boththe process of image acquisition and assessment of viabilityare automated, errors / variations due to subjective perceptionwould be reduced to a minimum.
The frames of the video were extracted and processedusing the background subtraction to extract the images of cellsin MATLAB (R2013b, Mathworks Inc.). Images of the cellsextracted from videos were further processed to quantitativelyassess the relative amount of dye retention/exclusion in a givencell. The saturation (color content) in a given image of cellwas quantitatively assessed by counting the total number ofpixels above the saturation level of the background (of theimage). To get an insight into the distribution of the stainingwithin the population, the histogram of total number of stained
pixels for different images has been plotted (shown in figure6). The histogram corresponds to images of about 4006 cellscorresponding to data recorded for about 9 seconds.
As it can be understood from the histogram, the cellspresent in the gated region (a) are very feebly stained andhave their cell membranes intact, while the cells present inthe other regions allowed the dye to permeate through the cellmembrane and hence, their cell membranes were disrupted tosome extent. Automated assessment of dye exclusion or uptakeand retention provides a quantitative insight into the stage ofcell death over a large population of cells.
VI. CONCLUSION AND FUTURE SCOPE
In this work, we have presented a cellphone based imagingflow analyzer for applications in basic biological researchfor on flight cellular studies in space. We have presented aninvestigation into achievable throughputs with microfluidicsbased system-level integrated portable imaging systems, whichemploy inexpensive low frame-rate cameras (in our case 30fps). In addition, we have demonstrated an enhanced through-put of about 450 cells/second with the use of closely spacedmultiple channel geometry. With the use of the presented sys-tem, we have performed an automated cell viability assessmentof a given population of yeast cells. Cell viability assessmentis an important technique for use in cytotoxicity and othercell-based biological research.
Future work in this approach would involve improvingthe design of our microfluidic catridge to perform on-chipmixing of the stain / dye with the bio-sample [20] to enableautomatation of sample preparation needed for performingvariety of diagnostics and biological assays with the presentedoptofluidic imaging analyzer.
ACKNOWLEDGMENT
This work was carried out under a project funded by SpaceTechnology Cell(ISRO-IISc).
REFERENCES
[1] L. E. Freed, R. Langer, I. Martin, N. R. Pellis, andG. Vunjak-Novakovic, “Tissue engineering of cartilage in space,”Proceedings of the National Academy of Sciences, vol. 94,no. 25, pp. 13 885–13 890, Dec. 1997. [Online]. Available:http://www.pnas.org/content/94/25/13885
[2] W. M. Saltzman, “Weaving cartilage at zero g: the reality of tissueengineering in space,” Proceedings of the National Academy ofSciences, vol. 94, no. 25, pp. 13 380–13 382, Dec. 1997. [Online].Available: http://www.pnas.org/content/94/25/13380
[3] C. A. Nickerson, C. M. Ott, S. J. Mister, B. J. Morrow, L. Burns-Keliher, and D. L. Pierson, “Microgravity as a novel environmentalsignal affecting salmonella enterica serovar typhimurium virulence,”Infection and Immunity, vol. 68, no. 6, pp. 3147–3152, Jun. 2000.[Online]. Available: http://iai.asm.org/content/68/6/3147
[4] W. Strober, “Trypan blue exclusion test of cell viability,” Curr ProtocImmunol, vol. Appendix 3, May 2001.
[5] M. J. Stoddart, “Cell viability assays: introduction,” Methods Mol. Biol.,vol. 740, pp. 1–6, 2011.
[6] S. Johnson, V. Nguyen, and D. Coder, “Assessment of cell viability,”Curr Protoc Cytom, vol. Chapter 9, p. Unit9.2, 2013.
[7] A. Kummrow, M. Frankowski, N. Bock, C. Werner, T. Dziekan, andJ. Neukammer, “Quantitative assessment of cell viability based on flowcytometry and microscopy,” Cytometry, vol. 83A, no. 2, pp. 197–204,Feb. 2013.
[8] A. Skandarajah, C. D. Reber, N. A. Switz, and D. A. Fletcher,“Quantitative imaging with a mobile phone microscope,” PLoSONE, vol. 9, no. 5, p. e96906, May 2014. [Online]. Available:http://dx.doi.org/10.1371/journal.pone.0096906
[9] M. Lee, O. Yaglidere, and A. Ozcan, “Field-portable reflection andtransmission microscopy based on lensless holography,” BiomedicalOptics Express, vol. 2, no. 9, pp. 2721–2730, Sep. 2011. [Online].Available: http://www.opticsinfobase.org/boe/abstract.cfm?URI=boe-2-9-2721
[10] J. S. Cybulski, J. Clements, and M. Prakash, “Foldscope: Origami-basedpaper microscope,” PLoS ONE, vol. 9, no. 6, p. e98781, Jun. 2014.[Online]. Available: http://dx.doi.org/10.1371/journal.pone.0098781
[11] A. Arpa, G. Wetzstein, D. Lanman, and R. Raskar, “Single lens off-chipcellphone microscopy,” in 2012 IEEE Computer Society Conference onComputer Vision and Pattern Recognition Workshops (CVPRW), 2012,pp. 23–28.
[12] S. Schaefer, S. A. Boehm, and K. J. Chau, “Automated,portable, low-cost bright-field and fluorescence microscope withautofocus and autoscanning capabilities,” Applied Optics, vol. 51,no. 14, pp. 2581–2588, May 2012. [Online]. Available:http://ao.osa.org/abstract.cfm?URI=ao-51-14-2581
[13] D. Shin, M. C. Pierce, A. M. Gillenwater, M. D. Williams, andR. R. Richards-Kortum, “A fiber-optic fluorescence microscope usinga consumer-grade digital camera for in vivo cellular imaging,” PLoSONE, vol. 5, no. 6, p. e11218, Jun. 2010. [Online]. Available:http://dx.doi.org/10.1371/journal.pone.0011218
[14] K. Goda, A. Ayazi, D. R. Gossett, J. Sadasivam, C. K. Lonappan,E. Sollier, A. M. Fard, S. C. Hur, J. Adam, C. Murray,C. Wang, N. Brackbill, D. D. Carlo, and B. Jalali, “High-throughput single-microparticle imaging flow analyzer,” PNAS, vol.109, no. 29, pp. 11 630–11 635, Jul. 2012. [Online]. Available:http://www.pnas.org/content/109/29/11630
[15] E. Schonbrun, S. S. Gorthi, and D. Schaak, “Microfabricatedmultiple field of view imaging flow cytometry,” Lab on a Chip,vol. 12, no. 2, pp. 268–273, Dec. 2011. [Online]. Available:http://pubs.rsc.org/en/content/articlelanding/2012/lc/c1lc20843h
[16] H. Zhu, S. Mavandadi, A. F. Coskun, O. Yaglidere, and A. Ozcan,“Optofluidic fluorescent imaging cytometry on a cell phone,” AnalyticalChemistry, vol. 83, no. 17, pp. 6641–6647, Sep. 2011. [Online].Available: http://dx.doi.org/10.1021/ac201587a
[17] Z. J. Smith, K. Chu, A. R. Espenson, M. Rahimzadeh, A. Gryshuk,M. Molinaro, D. M. Dwyre, S. Lane, D. Matthews, andS. Wachsmann-Hogiu, “Cell-phone-based platform for biomedicaldevice development and education applications,” PLoS ONE,vol. 6, no. 3, p. e17150, Mar. 2011. [Online]. Available:http://dx.doi.org/10.1371/journal.pone.0017150
[18] Y. Xia and G. M. Whitesides, “Soft lithography,” Annual Reviewof Materials Science, vol. 28, no. 1, pp. 153–184, 1998. [Online].Available: http://dx.doi.org/10.1146/annurev.matsci.28.1.153
[19] A. Bonora and D. Mares, “A simple colorimetric method fordetecting cell viability in cultures of eukaryotic microorganisms,”Current Microbiology, vol. 7, no. 4, pp. 217–221, Jul. 1982. [Online].Available: http://link.springer.com/article/10.1007/BF01568802
[20] A. P. Tan, J. S. Dudani, A. Arshi, R. J. Lee, H. T. K.Tse, D. R. Gossett, and D. D. Carlo, “Continuous-flowcytomorphological staining and analysis,” Lab on a Chip,vol. 14, no. 3, pp. 522–531, Dec. 2013. [Online]. Available:http://pubs.rsc.org/en/content/articlelanding/2014/lc/c3lc50870f
ISTC/MET/PK/0305
PROCESSING OF METALLIC THERMAL INTERFACE MATERIALS USING LIQUID
PHASE SINTERING FOLLOWED BY ACCUMULATIVE ROLL-BONDING
Deepak Sharma1, Rajesh Kumar Tiwari
2, Ramesh Narayan P.
3 and Praveen Kumar
1
1 Department of Materials Engineering
Indian Institute of Science, Bangalore 2Department of Manufacturing
National Institute of Foundry and Forge Technology, Ranchi 3MCD / MMG,
VSSC, ISRO P.O., Trivandrum
Abstract — This study reports on the processing and
characterization of composite solders, which comprise Cu, a
high-melting phase (HMP) embedded in In matrix, a low-
melting phase (LMP), produced by liquid phase sintering (LPS)
followed by accumulative roll-bonding (ARB). These solders
combine high electrical and thermal conductivities with high
mechanical compliance, and are suitable for a range of next-
generation thermal interface material (TIM) and interconnect
applications. Following the previously published work, an In-
40%Cu was found to possess the requisite combination of
compliance and conductivity, we start with effect of different
short sintering periods (30 and 60 s at 160 ºC each) on
compressive strength and electrical conductivities for both as-
sintered samples and samples processed through 0-10 passes of
ARB. Furthermore, rolling to a high compressive strain (50%)
increases the conductivity without significant increase in flow
stress; these findings prove to be promising for producing novel
TIM for new generation, advanced microelectronic packages.
I. INTRODUCTION
A continuous increase in computational power and the
miniaturization of the modern microelectronic devices, lead a
drastic increase in requirements for heat dissipation via TIMs
which cannot be met by the existing TIMs. Hence, the
packaging of advanced next generation microelectronic
devices demands development of new alloys with (1) high
thermal conductivity, (2) high shear compliance with
moderate compressive stiffness under creep conditions, and
(3) low melting temperature. To simultaneously attain the
above characteristics in a material, a novel composite
architecture, as shown in Fig. 1, composing of a highly
conductive HMP (e.g., Cu, Sn, etc.) uniformly distributed in
highly compliant LMP (e.g., In, Bi, etc.) matrix, has been
proposed and produced via LPS [1-5].
Fig. 1: Schematic of idealized microstructure with a uniform distribution of
HMP and LMP [1].
LPS of a HMP and a LMP above the melting temperature
of LMP has been suggested as a method to produce a
composite with the architecture shown in Fig. 1 [1,6]. A
composite system with Cu as the HMP and In as LMP was
chosen, one of the major advantages associated with In-Cu
system is that Cu is one of the materials with highest (thermal
as well as electrical) conductivity. As indicated in survey [1,
3, 6], current work involves an In-40vol.% Cu composition,
which possess the requisite combination of compliance and
conductivity. Preliminary experiments on Cu-In showed that
thermal conductivity of the composite solders were
significantly greater than that of In, although the flow stress
was high, most probably because of reactions between Cu and
In, and the formation of a solid skeleton structure of the HMP
and intermetallic compounds (IMCs) [6]. However, a critical
observation of the microstructures produced in the previous
works [1, 3, 6] reveals that: (i) the dispersion of Cu particles
was non-uniform in In matrix, resulting in agglomeration of
Cu particles in certain regions; this resulted in non-uniform
compliance (and equivalently, stiffness) and hence pre-mature
failure of the material, and (ii) the minimum thickness of the
Cu-In composites cannot be arbitrarily small, as the thickness
of such a composite is controlled by the diameter of Cu
particles and their agglomerates. These make these materials
unsuitable for the advanced microelectronic packages, which
have space only for few hundred microns thick TIM.
This work addresses the above two challenges by
suggesting implementation of a novel processing route
consisting of a combination of LPS and ARB, which is a
severe plastic deformation (SPD) technique. It is speculated
that ARB will not only flatten the Cu particulates but will also
increase the uniformity in mixing of Cu particles in Indium
matrix. Not only such an approach shows potential to resolve
the above two issues with Cu-In based TIMs, but it also
improve the thermal conductivity of such a composite, in
corroboration with the preliminary finite element analysis [7].
II. EXPERIMENTAL
Spherical Cu powders with average particle sizes of 10 µm
and 99.9% purity, were etched using 10% HCl solution for 10
min to remove the native oxide layer. The etched powder was
thoroughly rinsed using deionized water and isopropyl
alcohol (IPA) and dried by storing in vacuum (10-4
torr) for
12 h. An appropriate amount of Cu powder (40% volume
fraction) was mixed with10-40 µm diameter, 99.99% pure In
powder by vigorously shaking in a glass-vial inside a glove-
box under N2 atmosphere. Subsequently, the mixed powder
was uniaxially compressed by using hydraulic press with a
load of 4 MPa for 45 seconds, in a lubricated hardened steel
die to produce green pellets of 10.1 mm diameter and 1.5
mm height with a relative density of 88 ± 1%. The density
measurement for green pellets was conducted by using
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
geometry method, which was confirmed to be consistent with
that of Archimedes principle.
The green pellets were encapsulated in Al foil, and LPS
was done at 160 ºC by dipping them in a pre-heated, properly
agitated (using a magnetic stirrer for establishing a uniform
temperature distribution) 400 mL silicone-oil bath for
different lengths of time (30 and 60 s), followed by quenching
in water so as to inhibit the formation of IMCs. Sintered
relative density measurement was performed using
Archimedes principle so as to avoid the errors caused by
possible deviation in cylindrical shape of pellets during LPS.
ARB, ranging from 1 to 10 numbers of passes, and each
imposing a strain of ~50%, were conducted on the sintered
samples. A sintered sample was rolled, sectioned in two equal
halves along the “original” longitudinal direction, degreased
by generously applying acetone, stacked in such a way that
the surfaces that were originally perpendicular to the
cylindrical axis were lying adjacent to each other, and again
rolled for next cycle.
A few samples were prepared where reduced graphene
oxide (rGO) was placed in between the two surfaces of the
sample prior to the ARB processing. Electrical conductivity
and hence thermal conductivity of these samples were
measured.
Microstructures of the samples were characterized (both of
as-sintered and ARBed samples). The as-sintered pellet was
cross sectioned along transverse direction while the ARBed
samples were cross-sectioned along both transverse and
longitudinal directions. The cross-sections were observed
under a scanning electron microscope (SEM), using
backscattered and secondary electrons (BSE and SE,
respectively), as well as energy dispersive X-ray (EDX)
spectroscopy (FEI Quanta model 200).
The stress-strain behaviour of the LPS solder was
characterized by using an Instron Testing Machine (Instron
5967), which uses a 5 kN load cell to record the applied load.
Finally, the electrical resistivity of the sintered samples were
measured by using a custom designed 4-point Kelvin probe in
conjunction with a nano-voltmeter (Keithley Multi-meter,
Model 2010) to eliminate any contact resistance. Since in
metallic materials, electronic conduction is responsible for
current and heat flow, the electrical and thermal
conductivities are proportional to each other. Thus the
electrical resistivity measurements were used to calculate
thermal conductivity of the composite using Wiedemann-
Franz Law, which is given as follows:
K
LT
(1)
where K and σ are the thermal and electrical conductivities,
respectively, L is Lorenz number, and T is the ambient
temperature, which was room temperature (i.e., ~303 K) in
the present study.
III. RESULTS AND DISCUSSION
Fig. 2 shows the effect of green density and sintering time
on sintered density of In-40%Cu composite; density values
are given in terms of relative density (calculated by taking
percentage of fully dense or theoretical density of composite
that is 7.962 g/cm3) and densification was calculated by
taking the increase in density relative to initial density. As
shown in Fig. 2, a densification of > 10% occurred when
green density was small (< 85%) and the densification
monotonically decreased with the green density. Higher
densification in these low aspect ratio samples can be
attributed to high heat transfer rate due to large surface area.
A decrease in densification with green density can be
attributed to the decrease in the volume fraction of open pores
which, unlike closed pores, are removed during sintering
process.
0
2
4
6
8
10
12
84 85 86 87 88 89 90
604530
Den
sific
atio
n (%
)
Relative Green Density (%)
In - 40 Vol. % CuSintered @160 0CSintering time (s)
(b)
Fig. 2: Variation of densification over LPS with relative green density for
different LPS periods.
Fig. 3 shows the effect of ARB on the relative density of
Cu-In composite. Interestingly, ARB did not monotonically
affect the densification of the sintered samples: At first, the
density of the sample decreased; however, then it increased
monotonically approaching almost the as-sintered density. A
decrease in relative density of sample was not expected and
this might be attributed to collapse of “apparent” well bonded
microstructure of as-sintered sample.
84
86
88
90
92
94
0 1 2 3 4 5 6
In-40Vol%Cu
Rel
ativ
e de
nsity
(%)
Number of ARB passes
Sintered @1600 for 60 s
Fig. 3: Effect of ARB processing on the relative density of sintered samples.
Fig. 4 shows representative SEM micrographs of as-
sintered samples. As shown in Fig. 4a, a low magnification
micrograph of as sintered sample, the sample consisted of a
few agglomerated regions of cavities or pores (black regions)
and also the distribution of Cu and In was not uniform. The
latter is shown better in Fig. 4b, which is a high
magnification micrograph. A close inspection of Fig 4b also
shows that the bonding between Cu and In may not be very
strong following sintering for 60 s; this may explain the initial
drop in density following the first ARB pass (Fig 3).
(a) (b)
Fig. 4: SEM micrographs of as-sintered samples produced after LPS for 60
seconds (a) low and (b) high magnification. Images were taken using back-
scattered electron where dark and bright regions are Cu and In, respectively, whereas black regions denote pores.
Fig. 5 shows the effect of ARB processing on the
microstructure of In-Cu samples. A comparison of Figs 4 and
5 readily reveals the following: (i) overall the pore density did
not change in the beginning of the ARB processing, (ii) the
distribution of Cu and In became more uniform with an
increase in ARB passes, (iii) In truly became continuous
matrix where Cu was spread following ARB processing, and
(iv) the interface between Cu and In appears to be free of
pores indicating better contact between Cu and In. While the
outcome (i) is not dersired but the other aforementioned
outcomes of ARB were intended effects.
(a) (b)
(c)
Fig. 5: High magnification SEM micrographs of samples processed through ARB for (a) 1, (b) 2 and (c) 5 passes.
The impact of sintering time on stress-strain (σ−ε)
behaviour at strain rates varying from 10-3
s-1
to 10-1
s-1
is
shown in Fig. 6. All samples were tested at room temperature
with relative green densities of ~85% and an aspect ratio of
~0.5. Fig. 6 reveals that flow stress of composite increased
with sintering time (this may be attributed to an increase in
volume fraction of IMCs) and also with applied strain rate.
We can speculate some strengthening due to Cu-In interface,
which may act as dislocation source [3]. With few exceptions,
all samples showed significant softening after limited strain
hardening. This indicates that dislocation generation rate was
significantly lower than dislocation recovery. This is due to In
as test temperature is close to 0.7Tm, where Tm is melting
temperature of In, and this enables fast recovery. To further
analyse the effect of strain rate on mechanical behaviour,
yield strength (σYS) and strain rate sensitivity was calculated
as follows:
m
0.2% YS 1K (2)
where m is strain rate sensitivity and K1 is a constant.
0
5 106
1 107
1.5 107
2 107
2.5 107
0 0.25 0.5 0.75
True
Stre
ss (P
a)
True Strain
60 sec 45 sec 30 sec
10-310-210-1
Strain Rate (s-1)
Fig. 6: True Stress-Strain Plot of as-sintered In-40%Cu having relative sintered density of 87%. For plotting stress-strain data for 45s and 30s
sintered samples, strain axis was shifted by 0.25 & 0.5, respectively.
Fig. 7 shows the effect of sintering time and strain rate on
σYS of as-sintered samples. As shown in Fig. 7 that Eq. (2)
aptly captures the effect of strain rate on σYS and the as-
sintered In-Cu composites showed significant strain rate
sensitivity. Fig. 7 also reveals that ‘m’ monotonically
increased with decreasing sintering period. ‘K1’ seemed to be
independent of sintering period, this implies that σYS,
irrespective of sintering period, would be same if loading rate
was 1, and at very small strain rate, σYS would be widely
different.
6 106
7 106
8 106
9 106107
2 107
10-5 10-4 10-3 10-2 10-1 100
30 sec45 sec60 sec
y = 1.47e+07 * x^(0.066638) R= 0.72987
y = 1.6489e+07 * x^(0.054552) R= 0.98435
y = 1.6872e+07 * x^(0.022814) R= 0.94306
YS
(Pa)
Strain Rate (s-1)
In - 40 Vol. % CuSintered @160 0C
Fig. 7: Dependence of YS on applied strain rate.
Study of stress-strain behaviour of ARBed samples was
done at room temperature, with varying strain rate of 10-4
to
10-1
s-1
; Figs 8a and 8b show representative stress-strain and
effect of strain rate on σYS for samples processed through 1
pass of ARB. Comparison of Fig. 8 with Figs 6 and 7 shows
that the general stress strain behaviour of samples following
ARB processing qualitatively remained the same, i.e., the
samples showed some strain hardening as well as the strain
rate sensitivity.
0
5 106
1 107
1.5 107
2 107
2.5 107
0 0.05 0.1 0.15 0.2 0.25
10-1
10-3
True
Stre
ss (P
a)
True Strain
In-40Vol% CuSintering @ 160
0C for 30s ARB Pass: 1 (Strain : 50%)
(a)
6 106
7 1068 1069 106
107
2 107
3 107
10-5 10-4 10-3 10-2 10-1 100
30 sec45 sec
y = 2.3912e+07 * x^(0.10075) R= 0.93768
y = 2.3741e+07 * x^(0.080788) R= 0.97864
Y
S (Pa)
Strain Rate (s-1)
In-40 Vol.%CuSintering @ 160
0C, 1 pass, strain = 50 %
(b)
Fig. 8: (a) True stress-strain behaviour and (b) σYS versus strain rate plots of
single pass rolled In-40Cu composites.
Fig. 9 shows the effect of ARB passes on the stress-strain
response of the In-Cu samples at the strain rate of 10-3
s-1
.
Since, as shown in Fig. 3, the density of samples following
ARB processing did not show a monotonous variation in the
density of the sample and the density of a sample affects the
flow stress of a material, the stress endured by the samples
were normalized by the square of the density of the sample.
Such normalization with respect to the density was conducted
following standard results on the cellular structure. Fig. 9
readily reveals that the flow stress of the material did not
change with the ARB passes. This can be attributed to the fact
that ARB proffered attainment of continuous channel of In,
which would then accommodate most of the applied strain, as
envisioned for such material system.
0
0.001
0.002
0.003
0.004
0.005
0.006
0 0.05 0.1 0.15 0.2 0.25
In-40Vol%Cu
0 pass1 pass2 pass5 pass
True
stre
ss/(D
ensi
ty)2
True strain
Strain rate: 10-2 s-1
Sintered@1600C
ARB strain per pass: 50%
Fig. 9: True stress-strain behaviour of sintered In-40Cu composites processed
through various ARB passes, each pass imposing a strain of 50 %.
Electrical resistivity of composites was measured using a
custom design 4-wire Kelvin probe shown in Fig. 10.
Electrical conductivity and derived thermal conductivity
using Wiedemann-Franz Law (Eq. (1)) was calculated by
electrical resistivity for as-sintered and ARBed samples
(shown in Fig. 11b). Fig. 11a shows that electrical resistivity
decreased with increase in sintering time and number of ARB
passes. It appears to be against the observation of Cu-In IMC
for higher sintering period; however, it indicates the existence
of good contact or wetting between Cu and In; the latter may
be responsible for reducing the interface contact resistance
between the two phases of the composite. Fig. 11b shows that
thermal conductivity of as-sintered samples were actually still
considerably small compared to pure In and Cu, and this can
be attributed to poor relative density and existence of pores as
shown in Figs 4. However, the conductivity of the In-Cu
composites significantly increased with an increase of the
ARB passes. As shown in Fig. 11b, thermal (as well as
electrical) conductivity of In-Cu composites became greater
than that of pure In and Ni after 10 passes of ARB. Although
the maximum value attained in this study is smaller than that
of Al or Cu, it clearly shows that ARB can greatly improve
the thermal conductivities of samples.
Fig. 10: Custom made stage for mechanically fixing the sample
(parallelepiped shape) and probes.
Fig. 12 establishes the effect of relative density on the
electrical (and hence thermal) conductivity of In-Cu
composites following ARB processing. The linear variation
between these two shows that density of the sample is an
important parameter determining the conductivity of the
sample. Also, the improvement of Cu-In interface or
significant reduction in the interfacial contact resistance, if
any, occurred only just after the first pass of ARB.
1 10-7
2 10-7
3 10-7
4 10-7
5 10-7
6 10-7
25 30 35 40 45 50 55 60 65
As-sintered1 Pass2 Passes
Res
istiv
ity (
m)
Sintering time (s)
In - 40 vol. % CuSintered @160 0C
(a)
0
100
200
300
400
500
0 1 2 3 4 5 6 7Pb
SnIn
Ni
Al
Ag
Cu
Au
Ther
mal
Cond
uct
ivit
y (
W-K
)
Electrical Conductivity (x 107/ -m)
Wiedemann-Franz LawIn - 40 vol. % Cu
Sintered @160 oC for 45 sec
13 Pass
(b)
Fig. 11: Variation of (a) electrical resistivity with sintering time, and (b)
thermal conductivity of as-sintered and ARBed samples of In-Vol.40% Cu with electrical conductivity. The broken line in (b) represents Eq. (1).
7 106
8 106
9 106
1 107
1.1 107
1.2 107
1.3 107
1.4 107
1.5 107
84 86 88 90 92 94
In-40Vol. % Cu
3060
Ele
ctric
al c
ondu
ctiv
ity(/
m)
Relative density after ARB
Sintered@1600CSintering time (s)
Fig. 12: Variation of electrical conductivity with density of the In-40 vol. %
Cu samples processed through ARB.
Fig. 13 shows the variation of electrical (and hence thermal)
conductivity of ARBed In-Cu composite as function of their
yield strength; the variation is monotonous. However, Fig. 13
reveals a striking result: an increase in the conductivity by
almost of factor of 2 was observed in In-Cu samples
following ARB processing while the yield strength increased
only by ~10% in between the same range. This actually
satisfies the most stringent criteria of the desired process of
attaining high conductivity without compromising
compliance.
8 106
1 107
1.2 107
1.4 107
1.6 107
39 40 41 42 43 44 45
In-40Vol%Cu
Ele
ctric
al c
ondu
ctiv
ity(/
m)
Yeild stress (MPa)
Sintering@1600C for 30 sec
Fig. 13: Variation of electrical conductivity with yield strength of the In-40
vol. % Cu samples processed through ARB.
Further microstructural study was conducted to quantify
the homogenisation and breaking of agglomerates of the Cu
particles, as these are the probable changes that are affecting
the conductivity. Size distribution plots shown in Fig. 14
clearly reveal a narrowing of the range of Cu-particle size
with number of ARB passes as well as consistent with Fig. 5,
the distribution of Cu and In became homogenous with ARB
passes. This supports the proposed hypothesis of
homogenization with ARB.
0
10
20
30
40
50
3 5 7 9 11 13 15 17 19
Cou
nt
Diameter range (m) (a)
0
20
40
60
80
100
3 5 7 9 11 13 15 17 19
Cou
nt
Diameter range (m) (b)
0
20
40
60
80
100
120
3 5 7 9 11 13 15 17 19
Cou
nt
Diameter range (m) (c)
Fig. 14: SEM micrographs of sintered In-Cu samples processed through
ARB for (a) 1 pass, (b) 2 passes, (c) 5 passes. The right column histograms
show Cu-particle size in μm. SEM micrographs were taken using back-scattered electrons wherein dark and bright regions are Cu and In,
respectively, and black regions represent pores.
Fig. 15 shows the variation of the mean Cu particle size as
function of the ARB passes; Fig. 15 shows Cu particle size as
measured along the thickness (longitudinal) and rolling
direction (transverse) directions. In each direction, the Cu
particle size decreased with increase in the ARB pass.
8
8.5
9
9.5
10
0 1 2 3 4 5 6
LongitudinalTransverse
Cu-
size
(m
)
Number of ARB passes
Cross sectional view
In - 40 vol. % Cu
Sintered @ 160 oC for 45 s
Fig. 15: Variation of Cu-particle size with number of ARB passes for In-Cu samples sintered for 45sec.
Fig. 16 shows a representative microstructure of In-Cu
composites processed through 1 and 2 passes of ARB wherein
a thin sheet of rGO was placed in between the “mating”
surfaces of the two In-Cu pieces. Fig. 16 clearly shows that
highly conductive materials, such as graphene, carbon
nanotubes, etc., can easily be embedded in In-Cu composites
through ARB processing of sintered sample.
(a) (b)
Fig. 16: Representative micrographs showing the incorporation of reduced
graphene oxide (rGO) in In-Cu composites. The black elongated region
shows rGo layer.
Fig. 17 shows the effect of rGO on the thermal
conductivity of In-Cu composites. As expected, Fig. 17
clearly indicates a dramatic increase in the total thermal
conductivity of the In-Cu composites if rGO is placed
embedded into the In-Cu composites.
0
100
200
300
400
500
600
700
800
1 2
In-40Vol%Cu + rGO
Without GrapheneWith Graphene
Ther
mal
Con
duct
ivity
(W/m
k)
Number of ARB passes
Sintering @ 160 oC for 30 s
Fig. 17: Variation of Cu-particle size with number of ARB passes for In-Cu samples sintered for 45sec.
IV. CONCLUSIONS AND FUTURE WORKS
This study reports on the processing and characterization
of LPS In-Cu solders, starting with a green pellet, which
consists predominantly of particles of low melting In (60 %
by vol.) and smaller amount of particles of high melting Cu.
Short term LPS was conducted, which showed densification
dependence on sintering time. Once the metal compact
densified by LPS, it was processed through 1 to 10 passes of
ARB wherein each pass of rolling imposed a compressive
strain of 50%. Although the ARB did not cause a monotonous
increase in the density of the sample, the distribution of Cu
and In became uniform. ARB processing actually led to
uniform distribution of Cu (the HMP) into continuous matrix
of In (the LMP) which was never attained just after liquid
phase sintering. ARB also led to decrease in the Cu particle
size and hominization in the particle size distribution. The
electrical and thermal conductivities of composite were
observed to be increasing with number of passes of ARB and
with sintering period. Furthermore, the electrical and thermal
conductivities of In-Cu composite solders were increased by
more than an order of magnitude if a thin layer of rGO was
embedded into In-Cu composite through ARB processing.
Electrical Conductivity measurements confirm that
combining LPS with ARB is a promising route to meet the
requirements of a TIM.
Future work involves the mechanical characterization of
In-Cu/rGO samples. Also, mechanical response of all
processed In-Cu composite solders (with or without rGO)
under shear loading and creep conditions will be evaluated.
ACKNOWLEDGMENT
Authors would like to thank IISc-ISRO STC for the
financial support.
REFERENCES
[1] P. Kumar, I. Dutta, R. Raj, M. Renavikar, and V. Wakharkar, Proc.
Conf. on Thermal Issues in Emerging Technologies (ThETA 2), Cairo,
Egypt, IEEE, 17-20 Dec.(2008) pp. 339-346.
[2] “Solder Thermal Interphase Materials”. Available online:
http://www.indium.com/TIM/solutions/index.php.
[3] J. Liu, P. Kumar, I. Dutta, R. Raj, M. Renavikar and V. Wakharkar, J Mater Sci 46 (2009) 7012.
[4] J. Liu, P. Rottmann, S. Dutta, P. Kumar, R. Raj, M. Renavikar and I.
Dutta, Proceeding of the 12th Electronics Packaging Technology Conference (EPTC) (IEEE, Singapore) (2009) pp. 506-511.
[5] J. Liu, P. Kumar, I. Dutta, and M. Renavikar, Inter PACK, San
Francisco, USA (2009). [6] I. Dutta, R. Raj, P. Kumar, T. Chen, C. M. Nagaraj, J. Liu, M.
Renavikar and V. Wakharkar, J Electron Mater 38 (2009) 2735.
[7] P. Kumar and S. Awasthi, J Comp Mater. 48 (2014) 1391
Development of Nanoparticle Based Radiation Detectors for Space Applications
B.H.M. Darukesha1, V. Radhakrishna1, M. Ravindra1 and K. Rajanna2
`Abstract: This paper contains the details of the work carried out for the development of nanoparticle based radiation detectors for space applications. The design of experiments and fabrication of detection medium are explained. Detailed testing and characterization work planned are indicated. Key words: Radiation Detectors, Nanoparticles, Scintillators
I. INTRODUCTION
Main idea of this research is to use nanoparticles as energy conversion sites for the detection of ionising radiation. Nanoparticles embedded in a plastic scintillating medium would convert radiation energy into fast electron and the later would scintillate the medium. Conventional techniques using photomultiplier tubes and photodiodes would be used to detect the light output. The Motivation of this work is to enhance the efficiency, resolution and range of radiation detection. Flexibility in ‘form factor’ is a significant advantage in Space Applications. The design of experiments involves optimization of species of nanoparticles, their size, and percentage of loading and thickness of the medium. Fabrication of 150 numbers of detection medium covering these parameters is completed. The First cut results are expected soon. 1ISRO Satellite Centre, Bangalore-560017, 2Department of Instrumentation and Applied Physics, Indian Institute of Science, Bangalore-560012 Project Number: ISTC/PIN/KR/300
II. THEORY Upon interacting with the nanoparticle in the medium, part of the energy of radiation is converted as fast electrons. The resin chosen to embed nanoparticles is such that these fast electrons scintillate to release light in the suitable wavelength of around 425 nm. This range is suitable because most of the light counting apparatus viz. photomultiplier tubes and photodiodes have the sensitivity peak around it. Efficiency of a nanoparticle in energy conversion is proportional to Z4
, where Z is the effective
atomic number. Since nanoparticles are smaller in size than the wavelength of the light scintillated, light would escape from the medium.
III. EXPERIMENTS Nanoparticles of Lead oxide, Tungsten oxide, Silicon oxide and Gadolinium oxide are chosen in this study. Commercial plastic scintillator which has a peak emission around 425 nm has been used. Plastic scintillator used has 3 parts. Catalyst and the solvent are mixed first and then with the resin in the specified ratios. Vacuum settling is carried out at 10 mbar for about 20 minutes to get rid of air-bubbles. As control samples, without nanoparticles, the compound is poured into moulds prepared by machining the Teflon. Vacuum settling was again performed at 10 mbar for 10 minutes. Whole of the mould was kept in de-ionised water at 47 ºC for 14 days. Temperature was monitored twice a day using a thermocouple and whenever required, de-ionised water pre-heated to 47 ºC was poured to ensure that the cast is always immersed. In order to form a thin film for the detection medium, on a molded Teflon, the compound was spun. Curing as mentioned above was performed for this also.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
After 14 days of curing, the mold was taken out and dried. The medium were baked at 80 ºC for 8 hours in inert Nitrogen atmosphere. A dedicated oven was used for this purpose. Subsequently, the detection medium was separated from the mould using a surgical blade as shown in Fig. 1.
Fig. 1 Control samples, without any nanoparticles It is found that separation becomes difficult when Teflon mold is used. Surface of the mold showed roughness complementing to the surface of the Teflon. Cleaning the surface of the mold was tried with alcohol and acetone. While neither of them was able to clean the surface, acetone showed an adverse effect to soften the mold. Slicing of top surface with blade was found to leave the surface clean. It was noted that there would be reduction of the thickness of mold so as to form a curved top surface. This reduction was considered for calculating the percentage of loading of nanoparticles during the fabrication of samples with nanoparticles. Fabrication of nanoparticle based medium followed the similar sequence. After the first vacuum settling, nanoparticles with the chosen species & size, percentage of loading were mixed with resin and were cast as shown in Fig. 2 (a) and (b).
Fig. 2 (a) Mixing of nanoparticle and the resin
Fig. 2 (b) poring the compound into mold
Molds were prepared out of glass. Glass tubes of 10 mm diameter were cut in barrel shape. Size of 10 mm is chosen to suit the typical diameter of the photomultiplier tube. Teflon tubes with 9 mm diameter and 12.5 mm diameter were also used. These molds are shown in Fig. 3.
Fig. 3 Different types of molds used Percentage of loading was set to 0.5%, 1%, 1.5%, 5%, 10%, 20% and 40%. Last percentage was to develop radiation shield for protecting the electronic devices used in space applications. Different thicknesses of medium were controlled by pouring the compound to different heights of these molds. One-quarter, on-half, three-fourth and full heights were filled.
Curing of the mold was performed in an oven as shown in Fig. 4.
Fig. 4 Curing of mold at 47 ºC for 14 days in oven The Glass tubes were broken easily and Teflon tubes were cut vertically. The Surfaces were smooth. The detection media thus obtained are shown in Fig. 5.
Fig. 5 Nanoparticle based radiation detection medium
IV. TESTING & CHARACTERISATION
Nanoparticle based radiation detection medium are being subjected to SEM study to ascertain the uniformity of distribution. The test set-up as shown in Fig. 6 is being arranged at ISITE, ISAC, ISRO. Co-60 Gamma source is chosen for the initial testing. Further to testing, effect of various environmental tests viz. high temperature storage, thermal cycling, and vibration and thermo-vacuum tests would be performed. The performance of samples before and after the tests will be compared.
Optical grease
Fig. 6 Test set-up
V. CONCLUSIONS Fabrication of nanoparticle based radiation detectors is completed. Fabrication steps are discussed in detail. Experiments for the evaluation of uniform distribution of NPs in the medium are being performed. Characterization of detection medium is expected to be completed soon.
REFERENCE
[1] A multichannel nanoparticle scintillation microdevice with integrated waveguides for Alpha, Beta, Gamma, X-ray and Neutron Detection, Scott M Pellegrin, Chad Whitney and Chester G. Wilson, Journal of Microelectromechanical Systems, Vol.19, No. 5, October 2010, p 1207 – 1214. [2] Luminescence and Scintillation Properties at the Nanoscale; Christophe Dujardin, David Amans, Andrei Belsky, Chaput, Gilles Ledoux and Anne Pilonnet, IEEE Transactions on Nuclear Science, Vol., 57, No. 3, (June 2010), P 1348 – 1354. [3] Synthesis of tungsten oxide nanoparticles by acid precipitation method; Sitthisuntorn Supothina, Ceramics International, Volume 33, Issue 6, August 2007, Pages 931–936
NP Radiation Detector
Si Photodiode/PMT
Front-end Electronics
Processing Electronics
Data Acquisition
Spatial Coherence of Tropical Rain
R. Ratan1, V. Venugopal∗1, J. Sukhatme1 and R. Murtugudde2
1Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science, Bangalore, India.2Earth System Science Interdisciplinary Center, University of Maryland, Maryland,USA.
Abstract - We characterise the spatial coherence of tropical
rain from observations (TRMM) and assess if models (CMIP5)
are able to reproduce the observed features. Based on 15 years
(1998-2012) of TRMM 3B42 (V7) 1-degree, daily rainfall, we
estimate the spatial decorrelation scale (e-folding distance) of
rain at each location in the tropics. A ratio of zonal to merid-
ional spatial scales clearly illustrates that while rain patterns
tend to be anisotropic (ratio of 3-4) over tropical ocean re-
gions, over land regions, rain tends to be mostly isotropic (ra-
tio of 1-1.5). This contrast between ocean and land appears to
be reasonably well captured by CMIP5 models, although the
anisotropy (ratio) over ocean is much higher than in observa-
tions.
1 Introduction
In a recent study (Ratan and Venugopal, 2013), we sta-
tistically analysed and documented the wet and dry spell
characteristics of tropical rainfall, using Tropical Rainfall
Measurement Mission (TRMM) satellite-based (3B42
V6) daily 1-degree rainfall. Using an Intensity-Duration-
Frequency (IDF)-like approach (common in hydrology, but
seldom used in meteorology), they found that while both
ocean and land regions with high seasonal rainfall accumu-
lation (humid regions; e.g., India, Amazon, Pacific Ocean)
show a predominance of 2-4 day wet spells, those regions
with low seasonal rainfall accumulation (arid regions; e.g.,
South Atlantic, South Australia) exhibit a wet spell duration
distribution that is essentially exponential in nature, with
a peak at 1 day. The behavior that is observed for wet
spells is reversed for the dry spell characteristics. In other
words, the main contribution to the non-rainy part of the
season comes from 3-4 day dry spells in the arid regions, as
opposed to 1-day dry spells in the humid regions. The total
rainfall accumulated in each wet spell was also analyzed,
and we find that the major contribution to seasonal rainfall
for arid regions comes from 1-5 day wet spells; however,
for humid regions, this contribution comes from wet spells
of duration as long as 30 days. We also explored the role
of chance as well as the influence of organized convection
in determining some of the observed features. Specifically,
we showed that while the 2-4 day mode might be present
in random realizations of rainfall in a humid region as well
as rainfall observations, the differences in contribution of
∗venucaos.iisc.ernet.in; STC 320
longer duration wet spells to the total number of rainy days
or seasonal accumulation separate reality from chance.
As a follow-up, in this work, we have begun to explore
the role of organised convection in determining some of
these observed features. Specifically, the recent availability
of high-quality satellite observations of rain proxies or
derived observations of rain over the globe provided an
ideal opportunity to study the spatial scales associated
with convection and rainfall. Among previous studies of
rainfall proxies which focus on spatial scales of convection,
Roca and Ramanathan (2000) used infrared imagery (from
INSAT 1-B) to analyse the scale-dependence of convective
systems in the tropical Indian ocean. They found that
cloud systems span a wide range of spatial scales from
100 to 500 km2. Nesbitt et al. (2000), using brightness
temperature measurements from the TRMM satellite, anal-
ysed mesoscale convective systems (MCSs) and identified
differences in the size of the systems over land and ocean
regions. Ricciardulli and Sardeshmukh (2002) have perhaps
provided one of the more comprehensive analyses on local
time and space scales of organized convection over the
global tropics. They analysed cloud imagery data and found
that typical spatial autocorrelation of deep convection is
of the order of 130km and mean duration of convective
events lasted approximately 5 to 6 hours over the global
tropics. More recently, Wood and Field (2011) approached
the cloud size distribution problem from a scaling point
of view, and find that the horizontal size distributions of
clouds are shown to be well captured by a unique power-law
relationship over a range of spatial scales from 0.1 to 1500
km. They also find geographic and seasonal variations, in
that the largest clouds are predominant over west Pacific
and Indian Oceans, as well as land regions influenced by
monsoons, while the smallest clouds are found over arid
regions and the trade wind zones of the tropics.
Moving on to studies which analysed rainfall observations,
Smith et al. (2005) reported that the mean spatial scale of
precipitation for land and ocean is around 90 and 122 km,
respectively. Dai et al. (1997) found that spatial correlation
scale of precipitation is higher over southern tropics then
over the northern tropics (200 km and 550 km for norther
and southern tropics, respectively). Baigorria et al. (2007)
showed that the spatial correlation is higher in winter than
summer in mid latitudes regions, with the type of precipita-
1
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
tion (frontal in winter versus convective in summer), cited
as the main reason for the observed differences. In this
work, we systematically analyse 15-years of TRMM 3B42
(V7) daily gridded (1-degree) rainfall observations (e.g., see
Kummerow et al., 1998; Adler et al., 2000; Huffman et al.,
2007) to estimate the spatial scale of rainfall and assess if the
coherence has any anisotropy associated with it. The paper
is structured as follows: Section 2 describes the methodol-
ogy used in estimating the spatial scale, followed by a dis-
cussion of the main results in Section 3, and a summary in
Section 4.
2 Methodology
We use the decorrelation or e-folding length as a measure
of the spatial scale of precipitation. Specifically, the cross-
correlation between the daily rainfall time series at a refer-
ence grid and its neighbouring grids is estimated, and the
distance at which the cross correlation becomes less than 1e
is considered to be the decorrelation length. To illustrate the
procedure, Figure 1a shows the daily rainfall (for 2005) at
a reference point over the Pacific Ocean (10N, 130W; red)
and its neighbouring grids in the x-direction, one to the east
(blue) and one to the west (grey) of the reference grid. Fig-
ure 1b shows the cross correlation between the daily rainfall
at the reference grid (10N, 130W) with its neighbours, 20
lags (degrees) to the east and west. The decorrelation or e-
folding length from Figure 1b is ∼5 degrees in the east as
well as west direction. This is repeated for each year, and
a climatological correlation curve is estimated for +x and
−x directions for the reference grid. Noting that the decor-
relation in the positive (east) and negative (west) directions
are mostly similar in magnitude (or decay; e.g., see Figure
1b), a climatological decorrelation length is estimated based
on the average of these two climatological curves (for +xand −x). The spatial scale estimated in this manner yields a
zonal (x) spatial scale at the reference grid. A similar pro-
cedure is followed to estimate a meridional (y) spatial scale
by considering lags in the north and south directions.
3 Results and Discussion
To begin with, we consider four locations - two over ocean
(Pacific and Atlantic) and two over land (India and Ama-
zon). Figure 2 shows the climatological cross-correlation
curves for these four locations It is quite evident from
the decay rates of these curves and the e-folding lengths,
that rain over ocean regions is likely to be more spatially
coherent, at least zonally, than that over land regions.
Specifically, comparing Figures 2 a, b and Figures 2c,
d, we note that zonal scale of rain over oceans is larger
than that over land. This is perhaps not surprising given
the spatial structure of annual mean rainfall over tropical
convergence zones. The meridional spatial scales over land
appear to be more than over ocean. A comparison of the
two scales suggests that perhaps rain over land tends to be
more small-scale (at least zonally) and more isotropic (i.e.,
(a)
J F M A M J J A S O N D0
20
40
60
80
100
Rai
nfal
l (m
m/d
)
(10N,129W)(10N,130W)(10N,131W)
(b)
−20 −15 −10 −5 0 5 10 15 20
0
0.2
0.4
0.6
0.8
1
Lag (in degrees)
Cro
ss C
orre
latio
n
1
e
Figure 1: (a) TRMM-based daily rainfall in 2005 at a refer-
ence grid (10N, 130W) and at neighbouring locations. (b)
Cross-correlation curve as a function of spatial lag (in de-
grees). The solid black line represents 1/e, and the two filled
red circles show the e-folding lengths in east (+ve lag) and
west (-ve lag) directions.
comparable zonal and meridional spatial scales) compared
to that over ocean.
As mentioned earlier, repeating this procedure for each
1-degree grid for 15 years, we estimate the zonal and merid-
ional climatological decorrelation (e-folding) lengths for
rain over the tropics. Figure 3a, b show the climatological
zonal and meridional decorrelation lengths (also called spa-
tial scale), respectively, for tropical rain. The higher zonal
spatial scale over ocean regions compared to land, which
was partially evident from Figure 2, is now quite clear in
Figure 3a. Specifically, rainfall over most of the ocean
basins, especially over the core of the tropical convergence
zones near the equator, has a spatial scale of 6 degrees. On
the other hand, over many of the monsoon hotspots (India,
northern Australia, South America), the zonal spatial scale
is of the order of 3-4 degrees. Interestingly, the equatorial
0 2 4 6 8 10−0.2
0
0.2
0.4
0.6
0.8
1(a)
Cro
ss C
orre
latio
n
0 2 4 6 8 10−0.2
0
0.2
0.4
0.6
0.8
1(b)
0 2 4 6 8 10−0.2
0
0.2
0.4
0.6
0.8
1(c)
Spatial Lag (in degrees)
Cro
ss C
orre
latio
n
0 2 4 6 8 10−0.2
0
0.2
0.4
0.6
0.8
1(d)
Spatial Lag (in degrees)
ZonalMeridional
Figure 2: Climatological zonal (red) and meridional (blue) cross-correlation curves based on 15 years (1998-2012) of
daily, 1-degree TRMM 3B42 (V7) rainfall for two ocean and two land locations: (a) Pacific (10N, 130W) (b) Atlantic (5N,
30W) (c) Amazon (10S, 60W) and (d) India (20N, 76E). and represent the +ve (x and y) and -ve (x and y) directions,
respectively. The dashed line corresponds to 1/e, based on which decorrelation lengths are estimated.
Africa and South America show a substantially smaller
zonal spatial scale (< 1-degree) suggesting that rain over
those regions is highly localised, even though they receive
rain throughout the year much like the equatorial oceans.
The maritime-continent region also has a very low zonal
spatial scale, despite the fact that it receives high amount of
precipitation throughout the year. One possible reason can
be that this region gets mostly convective type of rain, which
lasts typically for a few hours, suggesting that moisture
transport is not from afar. The high zonal spatial scale over
southern Africa (6 − 8 degrees) could be attributed to the
presence of the intertropical convergence zone (ITCZ) in
southern hemisphere during the boreal winter. Moving
on to the meridional (y) spatial scale of rain (Figure 3b),
over most land regions it (2 − 3 degrees) is lesser than or
comparable to the zonal counterpart. Remarkably, over the
tropical convergence zones (except over equatorial Indian
Ocean) it ( 1-degree) is much lesser than the zonal spatial
scale ( 6 degrees).
The dominance of zonal or meridional scale over each
region can be assessed by their ratio, which can be seen to
be a measure of isotropy (or lack thereof) of rain, shown in
Figure 3c. The most striking aspect seen here is that, over
most land regions, rain tends to be small-scale (lower zonal
and meridional scales) and mostly isotropic (ratio of 1 to
1.5 in Figure 3c). Interestingly, the southern part of Africa,
which showed high zonal scale, also has a ratio between 1
and 2, suggesting that moisture transport is likely to be from
afar in both zonal and meridional directions. On the other
hand, oceanic rain, especially over the Pacific and Atlantic
convergence zones, is highly anisotropic (a ratio of 3 and
higher). This is generally consistent with the climatological
view of the (east-west elongated) spatial structure of rain
over the tropical convergence zones.
Following this, we analysed a set of 4 climate models to see
if they reproduce the feature reported above from observa-
tions. The models selected were from the suite of models
used in the Coupled Model Intercomparison Project (CMIP)
Phase 5. The four CMIP5 models used are CSIRO (Aus-
tralia), MPI (Germany), NCAR-CCSM (USA) and NOAA-
GFDL (USA) with spatial resolutions of 1.87 × 1.86 ,
1.87 × 1.86 , 1.25 × .9 and 2.5 × 2 . At the outset,
it is worth mentioning that the spatial resolution of 3 out
of the 4 models is around 2-degrees and thus a compari-
son between the zonal and meridional scales with observed
(at 1-degree) would not be fair or appropriate. However,
a meaningful assessment, at least qualitatively, can possi-
bly be gleaned from the ratio, i.e., the measure of isotropy.
For completeness sake, Figure 4 shows the estimates of the
zonal (first column) and meridional (second column) spa-
tial scales from model precipitation, using the procedure de-
scribed in Section 2. The over-estimation of zonal spatial
scale is more pronounced than the meridional counterpart,
when compared to observations. Their ratio (shown in the
third column), on the other hand, suggests that the isotropy
of rain over land and anisotropy over ocean (especially, the
Figure 3: Climatology (1998-2010) of the (a) zonal and (b) meridional spatial scales, based on daily, 1-degree rainfall.
Their ratio is shown in panel (c). The white area in the panels corresponds to regions which receive an annual mean rain
less than 2 mm/d.
Pacific and Atlantic convergence zones) are captured partic-
ularly well in 3 of the four models, even though the magni-
tude is 2 to 3 times higher than in the models. Apart from the
fact that models have a coarser resolution, the well-known
issue of the models not knowing when to stop raining (i.e.,
persistent drizzle) could also be a contributing factor to the
discrepancies shown here.
4 Summary
In this study we document the spatial scale characteristics
of tropical rainfall using 15 years (1998-2010) of daily, 1-
degree TRMM-based rainfall. Using decorrelation or e-
folding lengths in zonal and meridional directions as a mea-
sure of the spatial coherence of rain, we find that (i) the zonal
spatial scale of ocean rain is much higher than land rain; (ii)
the meridional counterparts are comparable between ocean
and land; surprisingly, the meridional spatial scale of rain in
the Pacific and Atlantic convergence zones, is much smaller.
The dominance of zonal or meridional scales is captured by
their ratio; it clearly shows that while most land rain tends to
be small-scale and isotropic (a ratio between 1 and 1.5), rain
over ocean tends to be mostly anisotropic, with the ratio of
4 in the Pacific and Atlantic oceans. Preliminary investiga-
tion into the ability of climate models (albeit at coarser res-
olution) to reproduce some of these observed features sug-
gests that even though the two scales are significantly over-
estimated, they are able to capture the qualitative feature of
anisotropy over ocean and isotropy over land.
References
Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and
E. J. Nelkin (2000), Tropical rainfall distributions deter-
mined using TRMM combined with other satellite and
rain gauge information, J. Appl. Meteorol., 39, 2007–
2023.
Baigorria, G. A., J. W. Jones, and J. J. O’Brien (2007), Un-
derstanding rainfall spatial variability in southeast USA
at different timescales, International Journal of Climatol-
ogy, 27(6), 749–760, doi:10.1002/joc.1435.
Dai, A., I. Y. Fung, and A. D. Del Genio (1997), Sur-
face Observed Global Land Precipitation Variations dur-
ing 190088, Journal of Climate, 10, 2943–2962, doi:
10.1175/1520-0442(1997)010¡2943:SOGLPV¿2.0.CO;2.
Huffman, G. J., R. F. Adler, D. T. Bolvin, G. Gu, E. J.
Nelkin, K. P. Bowman, Y. Hong, E. F. Stocker, and D. B.
Wolff (2007), The TRMM multi-satellite precipitation
analysis: Quasi-global, multi-year, combined-sensor pre-
cipitation estimates at fine scale, Journal of Hydrometeo-
rology, 8, 38–55.
Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simp-
son (1998), The tropical rainfall measuring mission
(TRMM) sensor package, J. Atmos. Oceanic Technol.,
15(3), 809–817.
Nesbitt, S. W., E. J. Zipser, and D. J. Cecil (2000), A cen-
sus of precipitation features in the tropics using TRMM:
Radar, ice scattering, and lightning observations, Journal
of Climate, 13, 4087–4106.
Figure 4: Same as Figure 3, but for rainfall from 4 CMIP-5 models (a) CSIRO (b) MPI (c) NCAR-CCM (d) NOAA -
GFDL. The first two columns show the zonal and meridional spatial scales, respectively, and their ratio is shown in the
third column.
Ratan, R., and V. Venugopal (2013), Wet and dry spell char-
acteristics of global tropical rainfall, Water Resources Re-
search, 49(6), 3830–3841, doi:10.1002/wrcr.20275.
Ricciardulli, L., and P. D. Sardeshmukh (2002), Local time-
and space scales of organized deep convection, J. Clim.,
15, 2775–2790.
Roca, R., and V. Ramanathan (2000), Scale dependence of
monsoonal convective systems over the Indian Ocean, J.
Clim., 13, 1286–1298.
Smith, D., A. Gasiewski, D. Jackson, and G. Wick (2005),
Spatial scales of tropical precipitation inferred from
TRMM microwave imager data, Geoscience and Remote
Sensing, IEEE Transactions, (7), 1542–1551.
Wood, R., and P. R. Field (2011), The distribution of cloud
horizontal sizes, J. Clim., 24, 4800–4816.
Amorphous Silicon Carbide thinfilms by Pulsed DC Magnetron Sputtering
for Micro Electro-Mecanical Systems (MEMS) applications
Habibuddin Shaik, Sumesh M.A*, Mohan Rao G
Department of Instrumentation and Applied Physics Indian Institute of Science, Bangalore
*LEOS, ISRO, Bangalore
Abstract- We are presenting here the synthesis and feasibility studies of a-SiC thin films for fabrication of a micro machined structure which will be used for development of a Micro-bolometer structure. In relation to this the micro machined structures are demonstrated in different designs on crystalline silicon substrate. Amorphous silicon carbide (a-Si1−xCx) films were deposited on silicon (1 0 0) and quartz substrates by pulsed DC reactive magnetron sputtering of Silicon in methane (CH4)–Argon (Ar) atmosphere. Optical, compositional and mechanical properties were investigated. We were able to synthesize a-Si1−xCx with x=0.5, and 100% Si-C bonding. We used X-ray photoelectron spectroscopy (XPS) to confirm this. These films, with 100% Si-C bonding, are used for the fabrication of micro structures. We studied the influence of substrate temperature and target power on the composition, carbon bonding configuration, band gap and hardness of a-SiC thin films. Increase in substrate temperature favors silicon–carbon (Si-C) bonding with increased hardness. The deposited films are chemically inert.
I. INTRODUCTION
The demand for sensors that can operate at high temperatures and often in severe environments such as aerospace, automotive, combustion processes or gas turbine control and oil industry is always increasing. Amorphous silicon carbide (a-SiC) is an attractive material from a technological point of view due to its excellent bandgap tunability, ability to withstand at high temperatures, good thermal conductivity, good mechanical strength and its inertness to wear and corrosive environments. Each or the combination of above properties have found
applications in different areas like MEMS, photovoltaics and protective coatings etc [1-5].
In the present investigation, the material aspect of a-SiC will be studied giving emphasis on its applications as structural material for the fabrication of MEMS systems, in particular for micro bolometer. We present here the study of the effect of the process parameters on stoichiometry and Si-C bonding configurations of the films and which in turn influence their optical and mechanical properties of the films.
II. EXPERIMENTAL DETAILS
Vacuum chamber with a base pressure of 2×10−6 mbar (with the combination of diffusion pump, cryo-sorption pump and rotary pump) was used to deposit a-SiC films. Single crystal silicon target of 10 cm diameter was sputtered in methane (CH4)–argon (Ar) ambient with a fixed CH4 partial pressure of 1.0 × 10−4 mbar and with working pressure of 2.0×10−3 mbar. Target to substrate distance was maintained as 7 cm. Substrate heater with k-type thermocouple was used. Pulsed (100 kHz) DC voltage (HUTTINGER Electronics, GmBH) was applied to the target.
The deposited films were characterized with DEKTAK profilometer for thickness measurement. Elemental composition and binding energy was analyzed using X-ray photoelectron spectroscopy (SPECS GmbH spectrometer (Phobios 100MCD Energy analyzer)). FTIR spectroscopy (Bruker Tensor 27) was used for identifying the fundamental vibrational modes and UV–VIS spectroscopy (HITACHI 3000) for bandgap calculations (using Tauc’s plot). Mechanical properties of the films were investigated using a nanoindenter (Triboindenter of Hysitron, Minneapolis, USA).
Two sets of films were deposited. The first set contains the films deposited at three different
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
substrate temperatures (473, 673, and 873 K) with a constant target power of 1200 mW cm−2 and the second set contains the films deposited with four different target powers (360 (110 mA), 540 (150 mA), 740 (200 mA), and 1200 (300 mA) mW cm−2) at a constant substrate temperature of 473 K.
Hard a-SiC films were used for the fabrication of microstructures. Microstructures in different designs, defined by mask and are transferred to the a-SiC coated substrate by using optical lithography. Microstructures were realized by micromachining of the deposited a-SiC films on Si by reactive ion etching (RIE from OXFORD Instruments).
III. RESULTS AND DISCUSSION
In order to check the composition and of Si-C bonding configuration in the films X-ray photo electron spectroscopy (XPS) was carried out. Al Kα (1486.6 eV) was used as X-ray source. It was observed that, as the substrate temperature is increased from 473 to 873 K, the C concentration in a-Si1−xCx films decreased from 0.3 to 0.2.This is because at higher substrate temperature, the adatoms of C and H reaching the substrate having less kinetic energy will get desorbed from the substrate and hence the percentage of C and H decreases. This reduction in C and H concentration has also been confirmed by the decrease in the bandgap (Fig. 1a). Similarly the C content in the films reduces from 0.54 to 0.3 with increase in the target power density from 360 to 1200 mW cm−2. This is due to the obvious reason that, at higher target powers more Si atoms sputter-out from the target and hence the intake of Si atoms into the film is more and dominates the intake of C. This decrease/increase in C/Si concentration can also be confirmed by (i) the decrease in the bandgap (Fig. 1b). Fig. 2 shows the deconvoluted fine spectra of C 1s core level into C-Si, C-C/C-H peaks for the first set of samples and it indicates that, though the C content is decreasing with increase in the substrate temperature (for a fixed target power), the percentage of C-Si bonding is increasing and C-C/C-H bonding is decreasing. The observed decrease in the bandgap, though the Si-C bonding is increasing with the substrate temperature is due to the fact that, the decrease in the C concentration is dominant than the
increase in Si-C bonds. This may be due to the fact that at high substrate temperatures the weak C-H and Si–H bonds break and the dangling bonds of Si and C combine to form strong Si-C bonds. Basa and Smit [6] had also suggested that the breaking of the Si-H and C-H bonds and the increased tendency towards Si-C bonding.
Fig 1. Variation of bandgap with (a) substrate temperature (b) target power.
Similarly Fig. 3 shows the deconvoluted fine spectra of C 1s core level into C-Si, C-C/C-H peaks for the second set of samples and it indicates that, as the target power is decreased (for a fixed substrate temperature) the percentage of C-Si bonding is increasing and C-C/C-H bonding is decreasing. This is due to the fact that, as the target power is increased (for fixed CH4 partial pressure) more Si atoms sputter-out from the target so along with C-Si bonding there is a more tendency towards Si-Si, O-Si-C, SiOx and Si-Hn bonding. Therefore the Si-C bond density is less at high target powers.
Fig. 2 Deconvoluted fine spectra of C 1s peak for the a-SiC thin film deposited with the target power of 1200 mW cm−2 at substrate temperature of (a) 473 K (b) 673 K and (c) 873 K.
Fig. 3 Deconvoluted fine spectra of C 1s peak for the a-SiC thin film deposited at a substrate temperature of 473 K with a target power of (a) 1200 mW cm−2 (b) 740 mW cm−2 and (c) 540 mW cm−2 (d) 360 mW cm−2.
Mechanical properties such as hardness and modules were calculated using the representative load–displacement curves (P-h curves) of both the sets of films. The maximum penetration depth of the indenter (hmax) at the peak load of 600 µN is less than 10% of the films thickness, which ensures that the substrate influence is negligible on the observed mechanical response. The mechanical properties such as hardness (H) and modulus (E) were calculated by fitting the unloading curve of the P-h curve using Oliver-Pharr method [7]. Fig. 4 and 5 show the variation of hardness and modulus with the substrate temperature and target power. Fig. 4 shows increasing hardness with increase in the substrate temperature for the first set of films. This can be confirmed by XPS analysis (Fig. 2) which indicates that, the Si-C bond density is increasing with substrate temperature leading to an increase in hardness and modulus. Increase in Si-C bond density will lead to enhancement of cross-linkage of Si and C atoms, resulting in a strengthened a-SiC material frame and ultimately improved amorphous network stiffness [8,9]. The same trend was observed for second set of films which were deposited at constant temperature and different target power densities as shown in Fig. 5.
Fig. 4. Variation of hardness and modulus with substrate temperature at constant target power of 1200 mW cm−2.
Fig. 5. Varconstant su We selecof 360 mfor the fa
Fig.6. Op
riation of hardneubstrate temperatu
cted a film whmW cm−2 at a abrication of m
ptical microscope
ss and modulus wure of 473 K.
hich is depossubstrate tem
micromachined
e images of microthe mask.
with target powe
ited with a pomperature of 4d structures.
ostructures define
er at a
ower 473K
ed on
The micrare shomicropatusing opoptimizetable 1.
PiraPhot-res
De-Spi
SoEx
Post ExDe
Post D
The chemby leavinlong timorpholoAfter detransfereashing stdevelopeparamete
PrTable T
ICPRF
After PRdetails oetching i
SFPre
Table TICPRF
After comcrystallin
ro patterns dewn in Fig.6tterns on to thptical lithogr
ed lithography
Lithoanha Clean sist (PR) used-hydration in coating oft Bake xposurexposure Bakeeveloper
Development Bake
Table 1. Lithmical inertnesng the sampleime. We dogical or thieveloping the
ed to RIE chtep to removeed region. Thers are shown
PR AO2
ressure Temperature P Power F Power
Table 2. RecipeeR ashing, we of the optimizis detailed in t
RIE a-SF6/O2 essure emperature
P Power Power
Table 3. Recmplete etchingne silicon etc
efined on the n6. We trans
he a-SiC coateraphy. The dy parameters
graphy 10 m
d AZ 15mins
4000rpm2mins
110mN
AZ 351
4mins
hography parametss of the a-SiCe in the piranhdid not obickness change PR, the shamber to pe the PR residhe details of t
in table 2. Ashing
50sc10 m
15150150
e for photoresist astarted a-SiC
zed recipee utable 3.
SiC Etch 18/9 sc
Valve posit15C
1000W400W
cipee for a-SiC etg of a-SiC, wching step to
negative masksferred thesed substrate by
details of theare shown in
mins 4562
s@250C m for 40sec
@110C mj cm-2
N/A 1 B (1:3)
@110C
ters C is confirmedha solution forbserved anyges of a-SiCsamples wereerform a PR
duals from thethe PR ashing
ccm mTorr 5C 00W 0W
ashing C etching. The
sed for a-SiC
ccm ion: 750
C W
W
tch e performed ao release the
k e y e n
d r y . e R e g
e C
a e
micrrecip
Tab
The micrFig structo leetcheoptimthe bgetti
The confstresdepo
rostructures. Tpee used for S
RSF6
Pressure le TemperaturICP Power RF Power
Table SEM image
ro machined st8 respectiv
ctures in fixedess defined bed out durinmized recipe obeams havingng etched-out
bent or curfiguration (Fisses developeosition.
The details Si etch is show
RIE Si Etch 10
Valve pre
9
4. Recipee for C-of fixed-fide
tructures are svely. The brd-fixed patterbeam width, ng Si etchingof Si etching,g a width let.
rled beams g 8) are dueed in the f
of the optimwn in table 4.
00 sccm position: 750
15C 900W 30W
-Si etch ed and fixedshown in fig 7roken beamsrn (Fig 7) are
which is geg. For the g, we observedess than 5µm
in the fixede to the resifilms during
mized
d-free 7 and s or e due etting given d that m are
d-free idual
the
IV.
a-SiC micdesigns ansubstrate. Thaving 100which is vbolometer atemperaturecorrosion rechemically process partarget powstructure ofmechanicalthat, the cawith increathe percentthe C-C/C-decrease incarbon conbonding isdecreasing.(i) target ptemperaturedemonstratemicromachmicro-bolom
[1] I. AfdiYunazKonagai, Sol. En[2] D. Pysch, MFilms 519 (2011[3] C.Y. Chang, Soc. 132 (1985) [4] L.S. Chang, 1882. [5] N. LedermanTellenbach, Surf[6] D.K. Basa, (1990) 439. [7] W.C. Oliver,[8] M.A. El KhaM.E. O’Hern, M96. [9] A. Jean, M.AKieffer, H. Pepi(1993) 2200.
CONCLUSION
cromachined nd dimensionThe synthesiz0% Si-C bon
very crucial foapplication ase and also esistant. The
inert. We srameters like wer on stof the films whl and optical parbon concen
asing the substage of C-Si b-H bonding isn target powncentration, ans increasing So we selecte
power=360 me= 473K for ed the fined structurmeter.
REFER
z, H. Nagashimanergy Mater. Sol.M. Bivour, M. H) 2550. Y.K. Fang, C.F. 418. P.L. Gendler, J
nn, J. Baborowskif. Coat. Technol.F.W. Smith, M
G.M. Pharr, J. Makani, M. Chake
M.F. Ravet, F. Ro
A. El Khakani, Mn, M.F. Ravet, F
NS AND FUTUR
structures wns were reazed a-SiC thnding with gor high temps it should be
it should ba-SiC films astudied the substrate tem
ichiometry ahich in turn inproperties. XPntration is getrate temperatbonding is incs decreasing. er causing a nd the percenand C-C/C-Hed the optimiz
mW.cm−2 and 100% Si-C
feasibility ores towards
RENCES
a, D. Hamashita, Cells. 95 (2011)
Hermle, W.G. St
Huang, B.S. Wu
.H. Jou, J. Mate
i, P. Muralt, N. X125 (2000) 246.
Mater. Res. Soc.
Mater. Res. 7 (199er, A. Jean, S. Boousseaux, J. Mat
M. Chaker, S. BoF. Rousseaux, App
RE WORK
with differentalized on Sihinfilms weregood hardnessperature micro
stable at highbe wear andre found to beeffect of the
mperature andand bondingnfluence their
PS data showsetting reducedture, howevercreasing while
Similarly therise in total
ntage of C-SiH bonding iszed conditions
(ii) substratebonding, and
of differentfabricating a
S. Miyajima, M 107. tefan, Thin Solid
u, J. Electrochem
er. Sci. 26 (1991)
Xantopoulos, J.M
Symp. Proc. 162
92) 1564. oily, J.C. Kiefferter. Res. 9 (1994)
oily, E. Gat, J.Cpl. Phys. Lett. 62
t i e s o h d e e d g r s d , e e l i s s e d t a
M.
d
.
)
M.
2
r, )
C. 2
Low temperature electrical transport studies on carbon nitride films prepared by chemical vapour deposition
K. Ramesh1, M. Prashantha1, R.Venkatesh1, N. Naresh1, M.V.N. Prasad2
1Department of Physics, Indian Institute of Science,
Bangalore 560 012. 2LPSC, ISRO, Bangalore 560 008.
E-mail: [email protected]
Abstract - The carbon nitride films have been prepared by chemical vapour deposition (CVD) at pyrolysis temperatures of 725, 750, 775, 800 and 825 oC. Electrical transport studies at low temperature (RT to 4.5K) show that the carbon nitride films exhibit Metal-Insulator (MI) transition. It is observed that the increase in pyrolysis temperature shifts the MI transition temperature to lower values. The transition temperatures for the samples prepared at 725 oC, 750 oC and 775 oC are 84.7 K, 67.7 K and 9.5 K respectively. The reduced activation energy indicates that the metallic regime of the samples prepared at pyrolysis temperatures > 800 oC lies at low temperatures. It is also observed that the activation energy decreases with the increase in pyrolysis temperature.
1. INTRODUCTION
The search for new materials for advanced application leads to the discovery of new materials with interesting electrical, physical, chemical and mechanical properties. The advancement of society and quality of human life also depends on these advanced materials. Carbon nitrides possess outstanding properties like high hardness [1-3], wide
1 Project Code: ISTC/PPH/KR/266.
band gap, excellent thermal conductivity, wear resistance, oxidation resistance, chemical inertness, high velocity of sound (~104 m/s), low dielectric constant (k~2), etc. These properties make them as a promising material for various applications such as organic semiconductors, fuel cells and photocatalysis, mechanical cutting tools, protective coatings, biomedical applications, electroluminescence devices, optical materials etc [4-6]. Medical devices such as bone joints, dental roots, heart valve etc which will be implanted into human body can be coated with carbon nitride films. The coating can reduce the wear and corrosion, which occurs due to the reaction of the device with body fluid. So coating the devices using carbon nitride can result in increase of life time. Apart from that, the coating is biocompatible [7]. In this article, the electrical transport properties of carbon nitride films have been reported. The films show the effect of disorderness and effect of incorporation of nitrogen on electrical transport properties. From earlier studies by Mott [8], some of the disordered materials show a transition from metallic to insulating behaviour at a particular condition of temperature, pressure, doping level etc. In view of this, the transition from metallic to insulator behaviour has also been observed in the prepared carbon nitride films.
2. EXPERIMENTAL
The samples were prepared by pyrolysis assisted chemical vapor deposition (CVD) method with the help of a home-built two zone furnace. 4-Azabenzimidazole (C6H5N3) of purity 99% was used as a precursor. About 0.1g of precursor was taken in a quartz tube of 85 cm length and 10mm diameter. The precursor was heated in the Zone I of a two zone furnace at 400 oC. The dense vapours of the precursor enter the zone II which was kept at a desired temperature where the pyrolysis occurs and the vapours get deposited on the quartz substrates and inner walls of the quartz tube. The samples were prepared at different pyrolysis temperatures of 725, 750, 775, 800 and 825 oC. As the C-N bond is significantly strong, even at high temperatures a considerable amount of the C-N bonding is retained. This method is simple and enables one to have control over the amount of nitrogen in the system by controlling the pyrolysis temperature, volume of the liquid and the process time.
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Sample at% of C at% of N at% of H
CN 725 70.70 25.98 3.31 CN 750 70.89 26.03 3.08 CN 775 70.54 26.30 3.16
CN 800 75.36 22.52 2.12 CN 825 77.81 20.85 1.34
Table 2. CHN data for samples prepared at different pyrolysis temperatures.
3. RESULTS AND DISCUSSION
The Raman spectrum of the prepared samples, shown in FIG.1, exhibits the D and G bands characteristic of amorphous carbon. The D (disorder) and G (graphite)
bands are observed in disordered polycrystalline and amorphous graphitic carbons. The G band is observed in the region 1530-1600 cm-1 and D band approximately at 1360 cm-1 respectively [9-11]. For single crystalline graphite only G band is observed. The G band arises due to the bond stretching of all
pairs of sp2 atoms in the ring structure. When nitrogen
is incorporated into this ring structure, the carbon-nitrogen bond breaks the symmetry in sp2 domains. As a result, Raman spectra exhibits a D band. Similarly,
graphitic materials with nanometric-size crystallites, impurities, doping, C≡N, imperfections, edges, exhibit a Raman D band. The relative intensity ratio of disorder band to graphite band (ID/IG) is a measure of the degree of order in the carbon nitride sample. Higher the value of ID/IG, higher the disorder in the sample. In the present work, Raman studies on the CNx film coated on the surface of quartz substrates show the D band at 1365 cm-1 and G band around 1575 cm-1. From FIG 2 it is seen that the ratio ID/IG increases with increase in preparation (pyrolysis) temperature and the result indicates that structural disorder increases in the samples with increase in pyrolysis temperature. CHN elemental analysis (Table 2) shows that the at% of nitrogen is about 26% in the samples from CN725 to CN775 and gradually decreases for the samples prepared at 800 and 825 oC.
3.1. Electrical transport study Electrical transport study was done from room temperature to 4.5 K. It is observed that the electrical resistivities of prepared carbon nitride film decreases with increase in preparation (pyrolysis) temperature. The samples CN725, CN750 and CN775 clearly show the metal-insulator transition at temperatures of 84.7K, 67.7K and 9.5K respectively. The results indicate that the disorderness and the preparation temperature of the films have a major role deciding the transition temperature. The plot of normalized resistivity versus temperature in FIG.3 shows a Metal-Insulator transition. In the figure, it can be observed that the transition temperature shifts to a lower value with increase in preparation (pyrolysis) temperature. Prasad et.al [12] reported that the electrical transport properties of amorphous carbon films prepared at 900 oC and above show Metal-Insulator transition. But our
Sample ID/IG
CN725 3.21 CN750 3.71 CN775 3.93
CN800 4.38 CN825 4.48
Table 1. Disorderness in the samples and crystallite size.
FIG.1. Raman spectra of the prepared Carbon nitride samples
FIG.2. Variation of ID/IG with pyrolysis temperature
FIG.3. Normalized resistivity vs Temperature in log scale for samples prepared at different pyrolysis temperatures showing the Metal-Insulator transition.
FIG.4. Reduced activation energy vs temperature. The positive slope shows that the samples are in metallic regime at low temperature
Sample Activation Energy (meV)
Resistivity at RT (Ω.cm)
CN725 94.36 4.89×10-1 CN750 60.22 1.22×10-1 CN775 24.22 3.65×10-3 CN800 18.61 3.13×10-3 CN825 8.58 1.19×10-3
Table. 3. Variation of activation energy and resistivity of samples with pyrolysis temperature.
study shows the Metal-Insulator transition for the CNx films prepared well below 900 oC. Moreover, the metallic regimes of CN800 and CN825 at low temperature are confirmed by the positive slope of the plot of reduced activation energy versus temperature. The reduced activation energy is calculated using the equation,
w(T) = -(∂lnρ/∂lnT) (1) From the plot of reduced activation energy (w) versus
temperature as shown in FIG 4 we can conclude that both CN800 and CN825 show metallic regime at low temperatures, the positive slope is the indication of metallic regime. Apart from the effect of preparation temperature, the impurity present in the amorphous network of carbon
also has an important role deciding the transition temperature. Some of the amorphous systems also show the metal-insulator transition because of the possible reason that the dopants can make the system more ordered at low temperatures [13,14] and so reducing the resistivity with decrease in temperature. Similar kind of metal to insulator transition at low temperature was observed in some disordered systems also [15,16]. A reason for the metal-insulator
transition at low temperature may be due to the formation of impurity band conduction, which occurs because of overlapping of electron wavefunctions and as a result a finite conductivity can be observed at 0K [13]. Decrease in resistance may also be due to weak phonon scattering of charge carriers. So, the samples show a decrease in resistivity with decreasing temperature below the transition temperature. Apart from that, the electrical conductivity depends on the density of states close to Fermi energy and the radius of localization of the charge carriers bound to the impurity states. From the results, it is observed that even though the amount of nitrogen is almost same for first three
FIG.5. Calculation of activation energy for CN825.
samples, the metal-insulator transition temperature gradually shifted to lower values with increase in preparation temperature. At higher temperatures, resistivity follows Arrhenius equation.
KT
Eaexp0 (2)
where, ρ0 is the pre-exponential factor, Ea is activation energy and K is the Boltzmann constant. From the values of activation energies (Table 3) it is observed that the activation energy decreases with increase in preparation temperature. Consequently, the resistivity of the carbon nitride samples at room temperature decreases with increase in pyrolysis temperature. The room temperature resistivity decreases gradually from 4.89×10-1 Ω.cm for the films prepared at 725 oC to 1.19×10-3 Ω.cm for the films prepared at 825 oC. FIG.5 shows the plot of resistivity versus temperature for the sample CN825 for the calculation of activation energy at high temperature (200-300K).
4. CONCLUSIONS
In conclusion, CNx films have been prepared using chemical vapour deposition at different pyrolysis temperatures using an organic precursor Azabenzimidazole. Disorderness of films increases with increase in pyrolysis temperature. Electrical transport studies show that these films exhibit metal to insulator transition at low temperatures.
REFERENCES [1]. A.Y. Liu and M.H. Cohen, Prediction of New
Compressibility Solids, Science, 245, 841-842, 1989.
[2]. D.M. Teter, and R.J. Hemley, Low-Compressibility Carbon Nitrides, Science, 271, 53-55, 1996.
[3]. A.Y. Liu and R.M. Wentzcovitch, Stability of carbon nitride solids, Phy Rev B 50 (14), 10362-10365, 1994.
[4]. X. Wang, K. Maeda, X. Chen, K. Takanabe, K. Domen, Y. Hou, X. Fu and M. Antonietti, Polymer Semiconductors for Artificial Photosynthesis: Hydrogen Evolution by Mesoporous Graphitic Carbon Nitride with Visible Light, JACS, 131, 1680–1681, 2009.
[5]. M. Zang, Y. Nakayama and M. Kume, Room temperature electroluminescence from hydrogenated amorphous carbon nitride film, Solid State Commun. 110, 679-683, 1999.
[6]. Y. Zheng, J. Liu, J. Liang, M. Jaroniec, S.Z.Qiao, Graphitic carbon nitride materials: controllable synthesis and applications in fuel cells and photocatalysis, Energy Environ. Sci, 5, 6717-6731, 2012.
[7]. F.Z. Cui and D.J. Li, A review of investigations on biocompatibility of diamond-like carbon and carbon nitride films, Surf and Coat Tech 131, 481-487, 2000.
[8]. N.F. Mott, Metal-Insulator Transition, Rev of Mod Phys, 40, 677-683, 1968.
[9]. R.O. Dillon, J.A. Woollam, and V. Katkanant, Use of Raman scattering to investigate disorder and crystallite formation in as-deposited and annealed carbon films, Phy Rev B, 29, 3482-3489, 1984.
[10]. A.K.M.S. Chowdhury, D.C. Cameron and M.S.J. Hashmi, Vibrational properties of carbon nitride films by Raman spectroscopy, Thin Solid Films. 332, 62-68, 1998.
[11]. D. Das, K.H. Chen, S. Chattopadhyay, L.C. Chen, Spectroscopic studies of nitrogenated amorphous carbon films prepared by ion beam sputtering, J. Appl. Phys, 91, 4944-4955, 2002.
[12]. V. Prasad and S.V. Subramanyam, Magnetotransport in the amorphous carbon films prepared from succinic anhydride, Physica B: Cond Mat, 369, 168-176, 2005.
[13]. B.I. Shklovskii, A.L. Efros, Electronic properties of doped semiconductors, Springer, 1984.
[14]. N.F. Mott, Electrons in disordered structures, Advances in Physics, 16:61, 49-144, 1967.
[15]. C. Uher and L.M. Sander, Unusual temperature dependence of the resistivity of exfoliated graphites, Phy Rev B, 27, 1326-1332, 1983.
[16]. I.L. Spain, K.J. Volin, H.A. Goldberg and I.L. Kalnin, Unusual electrical resistivity behaviour of carbon fibers, Solid State Commun, 45, 817-819, 1983.
1
Abstract— Generation of unmodulated microwave signals by
heterodyning optical signals is well established in literature. In
this paper it is shown that by modulating one of the optical
signals before heterodyning, a modulated microwave signal can
be synthesized. However, when the optical sources are free-
running, the resulting microwave signal lacks frequency stability
and exhibits poor phase noise. A novel approach for
implementing OPLL is proposed wherein angle modulated
microwave signals exhibiting high frequency stability can be
generated. Generation of BPSK and QPSK modulated microwave
signals by optical heterodyning is discussed in this paper. The
same principle can be extended to higher order modulation
schemes as well. Since optical modulators support very high bit-
rates as compared to microwave modulators, the proposed scheme
supports generation of high bit rate modulated microwave
signals.
Index Terms—Integrated optics, Microwave generation,
Microwave photonics, Optical modulation, BPSK, QPSK
I. INTRODUCTION
ENERATION of microwave signals using optical
techniques has been an active area of research. Several
heterodyne methods have been studied and proposed for the
optical generation of microwave signals. The optical signals
for heterodyning can be derived from two independent lasers
[1-2] or from one laser using the modulation-sideband
technique [3] or from a dual mode laser [4]. Optical
heterodyning using harmonic up-conversion in nonlinear lasers
[5], mode-locked lasers and pulsed lasers [6] have also been
reported. Techniques for improving the phase noise
characteristics have also been proposed [7]. These techniques
offer great flexibility in choosing the microwave frequency as
it is determined by the frequency spacing of the two lasers.
Generation of signals with frequencies ranging from few
megahertz up to terahertz is possible by optical heterodyning.
Most of the techniques based on optical heterodyning are
aimed at generating an unmodulated signal in microwave or
mm-wave region. Generation of angle modulated microwave
signals are discussed in this paper.
Section II of this paper discusses generation of unmodulated
microwaves signals by optical heterodyning. Optical Phase-
Locked-Loop (OPLL) for achieving frequency stability of
heterodyned output is discussed in Section III. Optical
synthesis of BPSK/QPSK modulated microwave signals is
discussed in Section IV. Limitations imposed by OPLL in
generation of angle modulated microwave signals are brought
out in Section V. A novel scheme for generation of BPSK and
QPSK modulated microwave signals that overcomes these
limitations is proposed in Section VI. Conclusions are
presented in Section VII.
II. GENERATION OF UNMODULATED MICROWAVE SIGNALS BY
OPTICAL HETERODYNING
A. Heterodyning Principle
Consider two monochromatic optical sources at frequencies
f1 and f2, where | f1 - f2| << f1, f2. Their electric fields can be
represented by:
where f1 and f2 are the optical frequencies of the two sources
and 1 and 2 are the phase variations of the two sources. If
the two sources are combined and given to a photodetector
(PD), which is a square-law device, the resultant signal can be
expressed as:
where R is the responsivity of the PD.
We then get:
From the above equation it can be seen that heterodyned
output contains a term that corresponds to the frequency
difference between the two sources. By choosing the optical
frequencies such that their difference lies in the microwave
region, a microwave signal can be generated by optical
heterodyning.
Fig. 1. Block schematic of an optical heterodyning system
The block diagram for generation of microwave signals
based on optical heterodyning is shown in Fig. 1. The optical
outputs of the laser modules LD1 and LD2 are combined using
an optical power combiner and fed to a photodetector (PD) to
obtain microwave signal at frequency determined by the
frequency difference between the lasers.
Generation of BPSK/QPSK Modulated
Microwave Signals Using Optical Techniques
K R Yogesh Prasad, T Srinivas, Gopal Hegde, Abdul Hameed
G
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
2
III. REALIZATION OF AN OPTICAL PHASE-LOCKED-LOOP FOR
ACHIEVING FREQUENCY STABILITY
When free-running lasers are used for optical heterodyning,
the resulting microwave signal, though centered at the
difference frequency, lacks frequency stability and hence
cannot be used directly for practical purposes. Figs. 2 and 3
show the heterodyned output of free-running lasers, captured
at two different instances. The figures clearly depict the
frequency variations of the generated microwave signal.
Frequency stability can be achieved by phase-locking one of
the lasers to a highly stable reference signal.
Fig. 2 Snapshot # 1 of spectrum of unlocked signal
Fig. 3 Snapshot # 2 of spectrum of unlocked signal
An Optical PLL, block schematic of which is shown in Fig.
4, can be implemented to achieve frequency stability.
Frequency-scaled heterodyned output and a reference signal
derived from a signal generator are compared using a phase
detector. The output of phase detector is passed through a loop
filter, which determines the dynamic performance of the
OPLL. The feedback voltage, which is proportional to the
phase difference between the heterodyned output and the
reference signal, is applied to the tune the laser output. Once
the optical PLL is locked, the frequency of the heterodyned
output is not allowed to deviate from that of the reference
signal. Also, the phase noise of the heterodyned output is
determined by that of the reference signal within the loop
bandwidth of the OPLL.
Fig. 4 Block diagram of OPLL
Phase noise values as low as -105 dBc/Hz at an offset of 10
KHz from the carrier frequency have been achieved.
Fig. 5 Spectrum of locked signal at 5 GHz exhibiting phase noise of
-105 dBc/Hz at an offset of 10 KHz
IV. OPTICAL SYNTHESIS OF ANGLE MODULATED MICROWAVE
SIGNAL
By extending the approach of generating unmodulated
microwave signal based on optical heterodyning, it is shown
that if one of the lasers is angle modulated then the output of
the photodetector contains a term corresponding to angle
modulated difference frequency. When an optical signal is
subjected to BPSK modulation, its phase is shifted by 0 or
radians depending on the modulating binary data being ’0’ or
‘1’ respectively. Such a signal can be expressed as:
where ‘θ’ is 0 or depending on binary data being 0 or 1
respectively.
If vbpsk is combined with an unmodulated laser output, v2, we
get:
where ‘θ’ is 0 or depending on binary data being 0 or 1
respectively.
3
If vc is passed through a square law device such as a
photodetector, we get current proportional to the square of
input voltage given by the expression:
where R is the responsivity of the photodetector.
The difference frequency term in the above expressions
suggests that the phase difference of ‘θ’ radians introduced by
the modulation is preserved in the process of down-
conversion. Thus we can obtain a BPSK modulated microwave
signal by choosing the laser wavelengths appropriately.
Similarly it can be shown that QPSK modulation introduced in
optical domain is preserved in the microwave domain.
V. LIMITATIONS IMPOSED BY OPTICAL PHASE-LOCKED-LOOP
IN GENERATION OF ANGLE MODULATED SIGNALS
In principle, angle modulation introduced in optical domain
will be preserved in the microwave domain by the
heterodyning process. However, when angle modulated
microwave signals are to be generated by optical heterodyning,
the presence of OPLL can prove to be a hindrance. As the
reference signal is unmodulated, angle modulation introduced
by the modulating signal will be tracked out by the OPLL if
the modulating frequency lies within the loop bandwidth.
Generation of BPSK/QPSK modulated microwave signals by
down-converting a modulated optical signal becomes
complicated with OPLL as BPSK/QPSK are suppressed
carrier modulation schemes with main lobe centered on the
carrier frequency. The heterodyned output within the OPLL
loop bandwidth is determined by the reference signal and will
not retain the modulation information as shown in Fig. 6.
Fig. 6 Effect of OPLL on QPSK modulated signal.
Full scale spectrum showing modulation being tracked out by OPLL.
VI. SCHEME FOR GENERATION OF BPSK/QPSK MODULATED
MICROWAVE SIGNAL
In this section, a simple scheme for generation of BPSK and
QPSK modulated microwave signals by heterodyning is
proposed wherein the problems due to phase-locking discussed
in the previous section are overcome. Figs. 7 and 8 show the
block schematics for generation of BPSK and QPSK
modulated signals respectively while still retaining the OPLL.
Fig. 7 OPLL scheme for generation of BPSK modulated microwave signal
Fig. 8 OPLL scheme for generation of QPSK modulated microwave signal
In the case of BPSK, the heterodyned signal containing
modulation information is self-heterodyned to extract the
carrier at the difference frequency of the lasers. The recovered
carrier is phase-locked to the reference signal by OPLL. Thus,
the OPLL tracks the carrier alone and the phase modulation
introduced in the optical domain is preserved in the microwave
domain.
In the case of QPSK, two levels of self-heterodyning are
required to extract the unmodulated carrier. The carrier thus
recovered is locked to the reference frequency as in the case of
BPSK. Optical heterodyne output, centered on difference
frequency of lasers, retains QPSK modulation.
Fig. 9 shows the QPSK modulated microwave signal in
which modulation remains unaffected by OPLL action by the
proposed approach.
Fig. 9 QPSK spectrum achieved by proposed approach.
Spectrum zoomed in as compared to Fig. 6 to clearly show the modulation.
4
VII. CONCLUSION
Generation of BPSK, QPSK modulated microwave signals
by optical heterodyning has been discussed. Since
BPSK/QPSK modulator design is based on electro-optic phase
modulators capable of being operated at frequencies of several
tens of GHz, it is possible to generate extremely high bit-rate
BPSK/QPSK modulated microwave signals by this approach.
REFERENCES
[1] R.P. Braun, G. Grosskopf, D. Rohde, and F. Schmidt, “Fiber optic mm-
wave generation and bandwidth efficient data transmission for 18-20
and 60 GHz-band communications”, in Proc. Int. Top. Meet.
Microwave Photonics, MWP’97, Germany, Sep. 1997, pp. 235-238,
paper FR2-5.
[2] J. B. Georges, J. Park, “Transmission of 300 Mbps BPSK at 39 GHz
using feedforward optical modulation”, Electron. Lett., vol. 30, no. 2,
pp. 160-161, Jan. 1994.
[3] H. Schmuck and R. Heidemann, “Hybrid fiber-radio field experiment at
60 GHz”, in Tech. Dig. 22nd Eur. Conf. Optical Communication,
Norway, Sept. 1996, vol. 4, pp. 59-62, paper ThC.1.2.
[4] D. Wake et al, “Optical generation of mm-wave signals for fiber-radio
systems using a dual-mode DFB semiconductor laser”, IEEE Trans.
MTT, vol.43, pp. 2270-2276, Sept. 1995.
[5] R.P. Braun et al, “Fiberoptic microwave generation for bidirectional
broadband mobile communications”, in Proc. IEEE MTT-S Int.
Microwave Symp., Denver, June 8-13, 1997, pp. 225-228, paper TU3E-
3.
[6] C.H.V. Helmholt et al, “A mobile broadband communication system,
based on mode locked lasers”, IEEE Trans. MTT, vol. 45, pp. 1424-
1430, Aug. 1997.
[7] R.P. Braun et al, “Low phase noise mm-wave generation at 64 GHz and
data transmission using optical sideband injection locking”, IEEE
Photonics Tech. Letters, vol. 10, No. 5, pp. 728-730, May 1998.
Project no. ISTC/MME/MSB/306
Development and characterisation of nano-porous aluminium
alloy surfaces
Arti Yadav, Subrata Chakarbarti, A. Manimaran, M.S Bobji
Abstract- We have obtained ordered porous
alumina nanostructures on aluminium alloy
surfaces of ISRO’s interest. After obtaining the
nanostructures hot sealing and heat
treatment experiments were conducted to test
the samples for any cracks that are observed
in the anodized samples generated according
to anodisation process carried out by ISRO.
We have observed that production of ordered
porous alumina structures reduced the
cracking of the Al 6061 sample after
anodisation, hot water sealing and heat
treatment. Al 7075 samples didn’t show crack
after anodisation and heat treatment but
cracked after hot water sealing.
I. Introduction:
Pure aluminium is a highly reactive metal and
develops a very thin oxide layer, of about 2-8 nm,
on the surface, whenever it is exposed to air. This
oxide layer acts as a protective layer on the
surface and prevents further oxidation of the
aluminium. In industry an anodic oxidation
process is used to produce a abrasive and
corrosion resistant protective oxide coating on
the surface. However, in the anodisation process
two different types of oxide layer forms, one type
is a porous oxide film and the other is a barrier-
type anodic oxide film [1]. However, under
certain controlled conditions anodic process can
form highly ordered nano-porous alumina
(figure 1) coatings on the aluminium surface.
Arti yadav is a graduate student in the Mechanical Engineering
Department, Indian Institute of Science, Bangalore, India
Subratha chakrabarti is Scientist F, LPSC, ISRO, Trivandrum, India.
A. Manimaran is Group Head, Components Production Group,
LPSC, ISRO, Trivandrum, India.
M.S.Bobji is an Associate professor at Mechanical Engineering
Department, Indian Institute of Science, Bangalore, India
Anodic nano-porous aluminium oxides are most
commonly formed in the solution of sulphuric
acid, oxalic acid and phosphoric acid by
anodisation of aluminium [2]. By changing the
anodisation parameters like pH of the solution,
anodizing voltage and temperature of
electrolyte, the diameter and pitch of the pores
in anodic aluminium oxide layer can be
controlled [1,2,3,4]. The pore diameter is
observed to be linearly proportional to the
anodising potential.
Figure 1: Schematic of nano-porous alumina
structure
Surface pretreatment processes like annealing
and electropolishing of the aluminium prior to
anodisation; help to improve the pore regularity.
Annealing helps relieve mechanical stress and
produces a coarse grained structure which
facilitates self-organization of the pores. Electro
polishing helps to produce a smooth surface
finish which plays an important role in regularity
of pore structure. Prior treatment of aluminium
is important because the roughness and other
parameters like voltage, type and concentration
of electrolyte, temperature, and impurities
present in aluminium including agitation of
31st Annual In-House Symposium on Space Science and TechnologyISRO-IISc Space Technology Cell, Indian Institute of Science, Bangalore, 8-9 January 2015
Project no. ISTC/MME/MSB/306
electrolytic bath influences the formation of self-
ordering porous alumina.
Fig. 2 shows a schematic diagram of pore
formation during the anodisation process. During
the process there is a balance between oxide
dissolution at electrolyte/oxide interface and
oxide formation at oxide/metal interface. This
balance is necessary for porous alumina
formation because it results in the formation of a
constant thickness barrier layer and hence allows
the steady state pore propagation into the
aluminium foil.
With regards to orderliness and pore uniformity,
the response of pure Aluminum to the above
mentioned condition is found to be extremely
good. However, the same cannot be said with
Fig.2. Schematic of porous alumina formation on
aluminum. 1. Formation of uniform barrier oxide.
2. Electric field distribution due to surface
fluctuation 3. Initiation of pores due to field
enhanced diffusion. (4) Stable pore growth.
certainty for structural alloys of aluminum which,
due to their high strength to weight ratio, finds
wide and extensive usage in the aerospace
industry. In this study we focus our attention on
the response of AA 7075 T7352 alloy and
compare it with our observations made on AA 3
6061 T652 alloy to uniform anodisation
parameters involving Sulphuric acid electrolyte.
The comparison is based on our observations
made on both these alloys after completion of
anodisation, dyeing and hot water sealing. The
chemical composition and physical properties of
AA 7075 T 7352 alloy are as follows [5].
This study was necessitated due to the high
rate of incidence of delayed leakage of high
pressure helium gases through O-Ring and plastic
seals involving anodized Aluminium alloys as was
observed in the Liquid Propulsion Systems Centre
of Indian Space Research Organisation. In order
to compare the anodisation response of this alloy
with other Aluminium alloys, an AA 6061 alloy in
its T652 temper condition was also taken and a
study was conducted on both these alloys
simultaneously in an attempt to understand the
differences and similarities.
II. Experimental Details:
We have obtained disc type specimens with AA
6061 T 652 alloy and AA 7075 T 7352 alloy
materials (Ø20 mm and 4 mm thick respectively)
from ISRO and anodised by two step anodisation
process at a constant voltage of 40 V at 18 0C in
0.3M oxalic acid solution in IISc.
Nano-porous alumina was formed by two step
anodisation process [6]. The pre-treatment of the
samples were performed before anodising.
Initially, to remove the impurity from the sample
after etching foil was rinsed with distilled water
for several times. Subsequently electropolishing
was performed in perchloric acid and Ethanol (in
the ratio of 1:4) solution at 10oC.
After pre-treatments the two step anodisation
begins, in the first anodisation process 0.3 M
oxalic acid solution was used to obtain porous
alumina layer. The applied constant voltage in
the anodisation process was 40V at temperature
of 180 C. The oxide layer created during the first
Project no. ISTC/MME/MSB/306
anodisation is chemically etched by using
phosphoric-chromic acid solution. Second
anodisation process and parameters were same
as used in first anodisation.
III. Results and Discussion
Characterisation of Standard anodisation
samples
To understand the existing industrial anodizing
process, we obtained samples from ISRO and
anodized as per their standard practices. SEM
images of the anodized samples show that for AA
7075 T 7352 has cracks on the surface. It is also
apparent that even after dying surface showed
cracks. However, for sealed and dyed AA 6061 T
652 alloy, there are no cracks present on the
sample.
Fig 3 (a) SEM image of sealed Al7075 alloy (b)
magnified image of fig (a), (c) SEM image of dying and
sealing Al7075 alloy (d) magnified image of fig (c), (e)
SEM image of dying and sealing Al6061 alloy (f)
magnified image of fig (e)
Controlled anodisation of Aluminium alloys.
Anodisation on alloy Al 6061
Fig 4 Time variation of current during first
anodisation of aluminium ally Al6061
A typical current–time relation (figure 4)
recorded during anodisation of aluminium in
oxalic acid at constant anodising voltage shows
that initially during the process current density
decreases rapidly with time and a minimum value
of current is achieved within first few seconds.
Current then slightly increases to give a local
maximum. After that it gradually reduced and
achieves a steady state current at around 3-4 min
of anodisation time. The SEM image (figure 5) of
the anodized sample showed non-uniform pores
formed on the surface
Fig 5 SEM image of Al6061 alloy after first
anodisation
Project no. ISTC/MME/MSB/306
Fig 6 Time variation of current during second
anodisation of aluminium ally Al6061
Fig 7 (a) SEM image of Al6061 alloy, anodised at
ISRO (b) SEM image of Al6061 alloy after second
anodisation (anodised at IISc)
Current–time relation during second anodisation
of aluminium alloy 6061 (figure 6) looked similar
to that observed for first anodisation. Fig 6 (a)
shows the SEM image of Al6061 alloy, anodised
at ISRO and Fig 6 (b) SEM image of Al 6061 alloy,
anodised at IISc by two step anodisation process.
From the images it is clear that the experimental
sample which was carried out at IISc has more
uniform porous structure compared to the
experimental sample from ISRO.
For Al 7075, current-time relation (figure 8) for
first anodisation is slightly different from that
observed for Al 6061 specially in the initial stages
of anodisation. Though current decreases rapidly
within first few seconds, it showed slight
oscillations in current not observed for Al 6061
alloy. However, after around 3-4 min a steady-
state current of the porous oxide is achieved
similar to that for Al 6061 alloy. SEM image of
Al7075 alloy after first anodisation shows that
the pores were formed on the surface, but it is
not uniform and pore growth is random
compared to Al 6061.
Anodisation on alloy Al 7075
Fig 8 Time variation of current during first
anodisation of aluminium ally Al7075
Fig 9: SEM image of Al7075 after first anodisation
From the graph it can be seen that, at the initial
period of the process current decreases quickly
Project no. ISTC/MME/MSB/306
with time similar to alloy 6061. A linear increase
then leads to a local maximum current. After
reaching the second maximum, the current start
decreasing slightly and a steady-state current is
achieved.
Fig 10 Time variation of current during second
anodisation of aluminium alloy Al7075
Fig 11 SEM image of Al7075 alloy, anodised at
ISRO and Fig13 (b) SEM image of Al7075 alloy
after second anodisation
From the Fig 11 (b), it is observed that Al 7075
alloy also have the uniform porous structure
compared to the process carried out at ISRO (Fig
10(b)).
Hot water treatment for 15 min on alloy Al 6061
& alloy Al 7075
After the two step anodisation process of both
the Al6061 and Al7075 aluminium alloys samples,
hot water treatment was performed on both the
surfaces for 15 min in boiling water. Figure 12 (a)
shows the SEM image of Al6061 alloy after
heating with boiling water and Fig 12 (b) SEM
image of Al7075 alloy after hot water treatment
using boiling water. It is found that there are no
cracks formed in the alloy Al6061 (figure 12(a)).
Fig 12 SEM images after hot water treatment
using boiling water for 15 min of (a) Al6061 alloy,
(b) Al7075 alloy
Hot water treatment for 1 hr on alloy Al 6061 &
alloy Al 7075
The hot water treatment on both the surfaces
Al6061 and Al7075 aluminium alloys samples
have been carried out for 1hr in boiled water.
Figure 13 (a) shows the SEM image of Al6061
alloy after hot water treatment using boiling
water and Fig 13 (b) SEM image of Al7075 alloy
after hot water treatment using boiling water.
We have found there are cracks formed in alloy
Al6061 but some cracks are observed in Al 7075
sample.
Project no. ISTC/MME/MSB/306
Fig 13 SEM images after hot water treatment
using boiling water for 1 hr of (a) Al6061 alloy,
(b) Al7075 alloy
Heat treatment for 3 hr on alloy Al 6061 & alloy
Al 7075 at 300oC
We carried out heat treatment experiments to
check whether the cracking is resulting from
formation of hydroxide during sealing or due to
exposure to high temperature. For this we have
heated the sample in absence of water to 300oC
in a furnace preventing hydroxide formation.
Figure 14a, b show that both aluminium alloy
samples did not develop any crack even after a
soaking time of 3 hr at 300oC. This probably
confirms that cracks in aluminium alloy 7075 are
formed due to hydroxide formation. Further
confirmation could come from low temperature
methods of sealing
I. Conclusion:
Standard anodisation procedures in ISRO
produce micro cracks on the surfaces of the Al
7075 alloy. Uniform and ordered nanostructures
in aluminium alloys of both Al 6061 & Al 7075 are
achieved using a lab scale procedure. It is found
that surfaces of the Al 6061 alloy showed no
signs of cracks after anodisation and sealing in
hot water. It is also found that surfaces of the Al
7075 alloy are not forming cracks after
anodisation but form cracks after hot water
sealing. It is also found that there are no cracks
formed on both alloys Al 6061 and Al 7075
anodized and heat treated at 300oC for 3 hr.
Fig 14 SEM images after heat treatment at 300oC
for 3 hr (a) A l6061 alloy, (b) Al 7075 alloy
References 1) Masuda H, Hasegwa F and Ono S 1997 Self‐Ordering of
Cell Arrangement of Anodic Porous Alumina Formed in
Sulfuric Acid Solution J. Electrochem. Soc. 144 L127–30
2) O’Sullivan J P and Wood G C 1970 The Morphology and
Mechanism of Formation of Porous Anodic Films on
Aluminium Proc. R. Soc. Lond. Ser. Math. Phys. Sci. 317
511–43
3) Masuda H and Satoh M 1996 Fabrication of Gold Nanodot
Array Using Anodic Porous Alumina as an Evaporation
Mask Jpn. J. Appl. Phys. 35 L126
4) Lei Y, Cai W and Wilde G 2007 Highly ordered
nanostructures with tunable size, shape and properties:
a new way to surface nano-patterning using ultra-thin
alumina masks Prog. Mater. Sci. 52 465–539
5) Rocca E, Vantelon D, Gehin A, Augros M and Viola A 2011
Chemical reactivity of self-organized alumina nanopores
in aqueous medium Acta Mater. 59 962–70
6) Masuda H and Fukuda K 1995 Ordered metal nanohole
arrays made by a two-step replication of honeycomb
structures of anodic alumina Science 268 1466–8