microstructure analysis of pvd deposited thin copper films · 2011-12-08 · 4 2. goal before this...
TRANSCRIPT
Supervisors: Ir. J. Neggers Dr. ir. J.P.M. Hoefnagels 16 May 2011
Microstructure analysis of PVD deposited thin copper films Bergshoeff, N.D. MT 11.17
2
Content
1. Introduction ......................................................................................................................................... 3
2. Goal...................................................................................................................................................... 4
3. Literature ............................................................................................................................................. 5
Substrate temperature ........................................................................................................................ 5
Background pressure ........................................................................................................................... 5
Sample rotation ................................................................................................................................... 6
4. Setup .................................................................................................................................................... 7
TEER Deposition Machine ................................................................................................................... 7
Pumping system .............................................................................................................................. 7
Sputter Etching ................................................................................................................................ 8
Magnetron Sputtering ..................................................................................................................... 9
Scanning Electron Microscope .......................................................................................................... 10
Electron Backscatter Diffraction ........................................................................................................ 11
5. Results ............................................................................................................................................... 13
Sputter Deposition ............................................................................................................................ 13
Effect of changing rotation ............................................................................................................ 13
Effect of changing duration ........................................................................................................... 13
Effect of changing pressure ........................................................................................................... 14
Heat Treatments................................................................................................................................ 16
Effect of changing temperature .................................................................................................... 16
Effect of changing duration ........................................................................................................... 17
Effect of changing heating up time ............................................................................................... 18
Effect of changing purge gas ......................................................................................................... 19
6. Conclusion ......................................................................................................................................... 20
References ............................................................................................................................................. 21
3
1. Introduction Thin film materials are currently widely used in industry, because they form an essential ingredient for
advanced technological products. Common applications are electric semiconductors and optical coatings.
Examples are integrated circuit (IC) chips, micro electro mechanical systems (MEMS), hard disk drives and
flexible monitors. Some pictures of these products are shown in figure 1.
Figure 1: Examples of products made with the use of thin film technology: a flexible screen,
an IC-chip and a MEMS thermal actuator
Thin films are materials with a thickness ranging from nanometers till micrometers and made of different
materials, like silicon, oxides, nitrides, metals and polymers. The complexity of the application of thin films
requires that the mechanical behavior is fully understood. The mechanical properties of these thin films are
different from their bulk counterparts because they typically have one or more dimensions smaller than the
intrinsic length scale of the microstructure. Understanding the formation of the microstructure during
production is therefore very important.
The act of applying a thin film to a surface is called thin film deposition, a technique for depositing a thin film of
material on a substrate or previous layers. Most deposition techniques control layer thickness within a few tens
of nanometers. There are two categories of deposition techniques; chemical and physical. Magnetron
sputtering is a form of physical vapor deposition, which will be used in this project. Sputtering relies on a
plasma, using an inert gas like argon, to knock material from a target and deposit it onto a substrate. A
schematic view of this process can be found in figure 2.
Figure 2: Overview of a magnetron sputter device with a target, a substrate
and a plasma created by argon [http://www.reade.com]
In this report the microstructure of deposited layers is studied, especially the effect of deposition parameters
and different heat treatment methods on the thin film microstructure.
4
2. Goal Before this project another student worked with the bulge tester and the magnetron sputter device during his
bachelor end project. His mission was to investigate size effects of thin copper films by measuring the stress-
strain curve of thin Cu films [Ruybalid, p. 1]. He used the bulge test device to deflect a thin SiN membrane and
after that a membrane with a copper layer on top. The copper layers of various thicknesses are deposited using
the magnetron sputter device.
The first batch of deposited samples showed very brittle fracture behavior for samples with Cu layers thicker
than 500 nm. They also fracture without any visible plasticity. Looking at the samples with the SEM shows that
the grains were very small and the grain boundaries were poorly formed [Ruybalid, p.3]. New settings for
deposition are used which resulted in very similar microstructure. Bulge tests with the new samples result in
some plastic strain before fracturing, although most still fractured without yielding. The stress-strain curve
showing the problems can be found in figure 3.
Figure 3: Stress-strain data for bulge tests performed on the first batch of SiN/Cu bilayer membranes
for various thicknesses [Ruybalid, Figure D.2]
To get more useful results from the bulge tests, decent SiN-Cu samples are needed to study the yielding
behavior of Cu films at different thicknesses in detail. Cu layers with a grain size of the whole thickness of the
copper layer will probably give better results, so the goal of the project is to produce Cu layers with the new
microstructure and grains over the whole thickness of the layer by vary deposition settings with the TEER
magnetron sputter device.
5
3. Literature Many deposition parameters may affect the structure of thin films. Examples of these parameters are
background pressure, sample rotation, deposition rate, substrate material, deposition temperature, surface
cleanliness and film thickness [Machlin, p. 1]. Only the deposition parameters which are varied in this project to
reach the goal of large grains will be discussed in this chapter.
Substrate temperature The substrate temperature is the deposition parameter that has the greatest effect on the deposited thin film.
Substrate temperature will rise if the magnetron power will be increased and when the deposition without
rotation will last longer. Adatom (an atom which lies on a crystal surface) diffusion on the substrate surface,
the rate of adatom desorption and the rate of nucleation control the grain size at impingement and are all
exponentially dependent on temperature [Thompson, p. 182]. The grain size at impingement increases with
increasing substrate temperature. Movchan & Demchishin made a structure zone model of the relationship
between substrate temperature and structure evolution during deposition, it is shown in figure 4a.
Figure 4: Movchan & Demchishin zone models which shows a relation between temperature and structure evolution. The
right map is the zone model for FCC metals under ultra-high vacuum conditions [Thompson, p. 184]
The zone diagram of Movchan & Demchishin is a useful tool to illustrate the expectations for structural
evolution as a function of the process parameters. However, such diagrams are developed for a specific
material or class of materials and specific processing conditions. For materials like copper, which is mainly used
in this project, the structure zone model of figure 4b would be a better representation. Reason for this is the
higher atomic mobilities characteristic of FCC materials compared with BCC materials [Thompson, p. 183].
Background pressure The background pressure of the inert gas during deposition affects the film microstructure in several ways.
When the pressure is increased the material will result in a more open structure and it has been suggested that
it also decreases adatom mobilities [Ohring, p. 499]. A reduction in gas pressure results in increased energy of
particles which densifies the film.
6
In figure 5 the Thornton structure
zone model is shown. It does not only
show the effect of substrate
temperature on the film structure, but
also background pressure as variable.
Thornton noted that by sputter
deposition the adatom mobility is
affected by the energy with which the
atoms arrive at the film surface. This
energy depends on the number of
collisions of the atom which scales
with the gas pressure [Thompson,
p.183]. This is illustrated by adding
another dimension to the Movchan &
Demchishin structure zone model.
Sample rotation Sample rotation is a deposition
parameter which also affects the
microstructure of the deposited
material. When the rotation of the
samples is stopped during deposition
the substrate temperature will be
much higher due to the constant flux
of high energy atoms to the surface.
The effects of increasing temperature
can be found in the first part of this
chapter.
The line-of-sight travel of the
incoming particles is another aspect
which is affected by the rotation of the samples. In line-of-sight travel of particles that are deposit on a thin
film, projections on the surface can shadow other parts of the film surface from the incoming particles and a
network of voids separating fibers or columns of film material is formed in the shadowed regions. When all the
particles have the same direction, it is possible to change the crystallographic texture and the orientation of
columnar boundaries by changing the incoming angle of the depositing atoms [Machlin, p.4]. When the
substrate is right in front of the target, straight columnar boundaries will form.
Line-of-sight travel of atoms applies only when the chamber pressure is less than about 0.01 mbar. The mean-
free path between gas collisions exceeds about one centimeter at lower chamber pressure [Machlin, p.4].
When sputtering with the standard argon pressure of the used setup of 0.008 mbar this will be the case.
Figure 5: Thornton structure zone model with substrate temperature and
inert gas pressure as variable deposition parameters [Ohring, p.500]
7
4. Setup
TEER Deposition Machine
Figure 6: Picture of the TEER coatings Ltd. magnetron sputter device in W-Hoog 0.053 [Made by Jan Neggers]
Figure 6 shows a picture of the setup of the TEER magnetron sputter device which can be found in W-Hoog
0.053. The setup consists of a lot of different devices, at the left side the inflow of argon and nitrogen and the
cooling system, in the middle the chamber, the pumps and pressure sensors and on the right a cabinet with the
control unit and the AC and DC magnetrons. In the next paragraphs there is a short explanation about the most
important parts of the setup: using the turbo pump to create a vacuum, clean the specimens in the chamber by
sputter etching and the deposition process using the AC and DC magnetrons.
Pumping system
The pumping system of the deposition system consists of two pumps: a mechanical rotary piston pump and a
turbo molecular pump. Figure 7 shows a schematic view of the placement of the pumps and how they are
connected. It is very important that the turbo pump should never be in direct contact with atmospheric
pressure when on and also the backing pump should always be running when the turbo pump is on [Neggers p.
1]. When the turbo pump is exhausted directly to the atmosphere it stalls and breaks, so the rotary pump is
used to produce the minimum vacuum level required (about 10e-1 mbar) to let the turbo molecular pump
operate.
Figure 7: Schematic view of the placement of the rotary pump and the turbo pump and how they are connected with the
chamber and each other (a turbo molecular pump is used instead of a multistage diffusion pump) [Ohring, p. 81]
In the rotary piston pump gas is drawn into a space when
isolated from the inlet after one revolution, then compre
molecular pump work on the principle that gas molecules can be given molecular motion in a preferred
direction. This impulse is caused by impact with a rapidly rotating turbine rotor with rotating speeds
30.000 revolutions per minute [Ohring, p. 76
this type of pump.
Sputter Etching
When the samples are placed and there is a high vacuum in the chamber, the samples have to be cleaned first
before deposition. This is called etching, where atoms will be removed from the film or substrate surface in a
plasma to get a flat surface without pollution
cathodes during sputtering in low-pressu
sensitive to the magnitude of atomic bonding forces and structure. For this reason there is no selectivity in the
removal of surface atoms. Figure 8 shows the mechanism of sputter etc
Figure 8: Mechanism of sputter etching, where an ion strikes the surface
and removes atoms
8
view of the placement of the rotary pump and the turbo pump and how they are connected with the
(a turbo molecular pump is used instead of a multistage diffusion pump) [Ohring, p. 81]
In the rotary piston pump gas is drawn into a space when the piston rotates [Ohring, p. 71
isolated from the inlet after one revolution, then compressed and exhausted during the next cycle.
work on the principle that gas molecules can be given molecular motion in a preferred
direction. This impulse is caused by impact with a rapidly rotating turbine rotor with rotating speeds
lutions per minute [Ohring, p. 76]. Extreme low pressures below 1e-10 mbar
When the samples are placed and there is a high vacuum in the chamber, the samples have to be cleaned first
This is called etching, where atoms will be removed from the film or substrate surface in a
to get a flat surface without pollution. According to [Ohring, p. 238], sputter etching happens at
pressure plasmas. Because their energies are high, ions striking a surface are
sensitive to the magnitude of atomic bonding forces and structure. For this reason there is no selectivity in the
shows the mechanism of sputter etching.
: Mechanism of sputter etching, where an ion strikes the surface
and removes atoms from the surface [Ohring, p.239]
view of the placement of the rotary pump and the turbo pump and how they are connected with the
(a turbo molecular pump is used instead of a multistage diffusion pump) [Ohring, p. 81]
p. 71]. There the gas is
ssed and exhausted during the next cycle. The turbo
work on the principle that gas molecules can be given molecular motion in a preferred
direction. This impulse is caused by impact with a rapidly rotating turbine rotor with rotating speeds up to
10 mbar are achievable with
When the samples are placed and there is a high vacuum in the chamber, the samples have to be cleaned first
This is called etching, where atoms will be removed from the film or substrate surface in a
sputter etching happens at
re plasmas. Because their energies are high, ions striking a surface are
sensitive to the magnitude of atomic bonding forces and structure. For this reason there is no selectivity in the
: Mechanism of sputter etching, where an ion strikes the surface
9
Magnetron Sputtering
Sputtering is a mechanism by which atoms or molecules are ejected from a target material by high-energy
particle bombardment and condense on a substrate as a thin film. Magnetron sputtering is the most widely
used variant of DC sputtering, because deposition rates are much higher for the same applied voltage and the
operating pressures can be reduced [Ohring, p. 223].
Figure 9: Schematic view of the principle of DC magnetron sputtering [Institute of materials & machine mechanics, Slovak
academy of sciences, http://www.umms.sav.sk]
Figure 9 shows a schematic view of the DC magnetron sputter process. When the chamber is filled with
pressurized argon the magnetic field in combination with the voltage of the direct current creates a glow
discharge (purple color in figure 9). A glow discharge is a self sustaining type of plasma, creating free electrons
(red dots) within the discharge region. The glow discharge also generates high-energy particles (Ar+ ions) which
are directed at the target material. The ions sputter atoms from the target and they are transported (blue
arrow) through a region of reduced pressure to the substrate. When the atoms arrive at the substrate they
condense on it, forming a thin film.
10
Scanning Electron Microscope A scanning electron microscope (SEM) is a type of microscope which doesn’t work with light, as conventional
microscopes do, but it uses a beam of electrons to get an image of the specimen. It is primarily used to study
the surface or near surface structure of materials.
According to [Goodhew, p. 106], an electron gun produces
electrons and accelerates them to energy between about 2
keV and 40 keV. Two or three condenser lenses reduce the
electron beam to a diameter of only 2e-10 m. The beam is
scanned across the specimen while a detector counts the
number of radiations given off from each point of the
surface. This radiation could be secondary electrons,
backscattered electrons or X-rays. At the same time the
spot of a cathode ray tube (crt) is scanned across the
screen. The electron beam and the crt spot are both
scanned in a similar way to a television receiver, which
creates an image of the specimen on the screen to study
the surface. In figure 10 a schematic view can be found of
the main components of a scanning electron microscope.
Image processing
The copper samples from which the surface will be studied
at the SEM have a maximum thickness of about 1 µm and
some of them have a thickness of about 100nm (0,1 µm).
The region into which the electrons penetrate the
specimen is called the interaction volume and it’s shown in
figure 11. According to [Goodhew, p.110], electrons will
not be backscattered out of the specimen if they have
penetrated more than a fraction of a micron and
experiments with materials of medium atomic weight, like
copper, indicate this will be about 0.1 µm. For secondary
electrons this is even worse, the detected signal originates mainly from a region which is little larger than the
diameter of the incident beam, like the top of the teardrop-shaped volume in figure 11.
Conclusion for use with thin film copper in the range of
hundreds of nanometers will be that only secondary electrons
will give a reliable view of the surface of the sample. However,
this will not be a problem because secondary electrons give the
best spatial resolution caused by their small sampling volume.
Therefore the secondary electron signal is most used at
scanning electron microscopes.
Figure 10: Schematic view of the main components of
a scanning electron microscope
[http://www.purdue.edu]
Figure 11: Drawing of the interaction volume
and the regions from which secondary electrons,
backscattered electrons and X-rays may be
detected [Goodhew, p. 110]
11
Electron Backscatter Diffraction When studying the surface of a sample is not sufficient, it is also possible to get some information about crystal
orientation of the material with a scanning electron microscope. When the electron beam penetrates the
material the electrons undergo various interactions with the atoms in the crystal lattice. Some of these
electrons emerge from the sample and when they are collected on a phosphor screen a pattern is formed. This
diffraction pattern (also called Kikuchi pattern) can be formed because the intensity of the emerging electrons
varies with direction. To get enough contrast in this pattern the specimen must be tilted through a large angle
(>70 degrees). The diffraction pattern can be used to obtain information about crystal orientations and
measuring grain size, which is very useful for this project. An example of a very clear diffraction pattern of
copper at 20keV can be found in figure 12.
Figure 12: Very clear backscatter diffraction pattern from copper
at 20 keV [http://www.crystaltexture.com]
Image processing
When using electron backscatter diffraction at thin film copper layers the problem with the interaction volume
occurs again, because this time backscattered electrons instead of secondary electrons will be used. As stated
in the last chapter about the SEM, the source of backscattered electrons is about 1-2 µm under the surface.
When using copper layers with a thickness of under 1 µm, this could be a problem, in particular when the
specimen is tilted seventy degrees. When using high energies and beam size the beam will go through the
copper layer and backscattered electrons from the underlying silicon will be collected at the detector. To get
the best results a beam energy of 15 kV and a spot (indicator for beam size) of 4 is used.
To get an overview of the crystal orientation in an indicated area of the specimen, the electron beam scans
over the sample and measures the orientation of the diffraction pattern at each point. The scan area must be
big enough to scan multiple different grains to measure a reliable grain size, for example 10x10 µm. When the
scan is completed the dataset can be processed to get the right information from the material. To obtain a
reliable grain size from the dataset some parameters has to be set in the software: confidence index (CI), grain
tolerance angle and minimum grain size.
The confidence index indicates how reliable the patterns of the different grains at one point are. Therefore low
confidence indices has to be filtered out of the dataset to calculate grain size. In practice a confidence index
>0,1 is acceptable, so all points in the dataset with a CI <0,1 will be filtered out of the grain size calculation. The
grain tolerance angle is the angle between grains at which two adjacent grains will be different grains. In the
manual of the OIM software an angle of fifteen degrees is assumed to be the grain tolerance angle for most
12
materials. The minimum grain size is the minimum number of scan points a grain must include to be called a
grain. By looking at the inversed pole figures a number of ten points at one grain will be a good assumption.
With these settings the dataset will be filtered to calculate a reliable grain size.
13
5. Results
Sputter Deposition
Effect of changing rotation
To work on the goal of larger grains, one of the options is to stop sputtering with rotation of the mill which hold
the samples. As explained in chapter 3, this will have great influence on the microstructure of the material
because temperature during deposition will rise and the direction of impact of the copper particles will change.
The mill is positioned at an angle where the samples are right in front of the copper target to get the best
transport of copper through the chamber. The flow of copper atoms to the surface in this situation will be a lot
bigger compared to a rotating mill. To get samples with the same thickness, the duration of the sputter session
has to be shorter. The samples are at a length of 2/3π in the field of the floating copper particles, so they will
receive copper particles about one third of the time. To compare the samples with and without rotation, the
other deposition parameters will be the same, 300W DC, 50W AC and 8*10-6
mbar argon pressure.
Figure 13: ESEM pictures of sputtered copper samples, 300W DC, 50W AC
Left: 4 minutes deposition time, with rotation, layer thickness ~50 nm, grain size 20±5 nm
Right: 1,5 minutes deposition time, without rotation, layer thickness ~80 nm, grain size 22±5 nm
One sputter session was done with rotation and 4 minutes deposition time, while the next session was without
rotation and 1,5 minutes deposition time, about three times shorter. Results are shown in figure 13. Grain size
of both samples is almost the same and far from big enough. Conclusion is that only stopping the rotation will
not give large grains because the temperature increase is not high enough.
Effect of changing duration
Only stop the rotation will not give large grains, so the temperature has to rise in another way to let the grains
grow. One of these ways is to use a longer deposition time. At the first sessions 4 minutes with rotation and 1,5
minutes without rotation is used, but the next time the deposition time will be three times longer to get a quick
look of which time range will be good. Rotation will still be switched off because without rotation the
temperature will rise a lot faster.
14
Figure 14: ESEM pictures of sputtered copper samples, 300W DC, 50W AC
Left: 4,5 minutes deposition time, without rotation, layer thickness ~250 nm, grain size 28±5 nm
Right: 15 minutes deposition time, without rotation, layer thickness ~800 nm, grain size 2±0.5 µm
Result of the 4,5 minutes sputter session is shown in figure 14a. Grain size is still almost the same as 1,5
minutes deposition time. The deposition time of the next session will be again three times longer, so the
temperature is high enough to reach a point where the grains will grow substantially. Figure 14b shows the
copper surface measured by ESEM, showing that for extreme sputter times without sample rotation the
surface will reach temperatures where the copper layer will show grain growth or even recrystallisation.
Nevertheless, this method is not desired because the high temperatures causes too large residual stresses to
remain in the sample at room temperature. More importantly, this method of grain growth will only occur for
very thick films in the range of micrometers and is thus not possible over the entire film thickness range.
Effect of changing pressure
Another parameter during the sputter process which can be modified is the pressure of the argon in the
chamber. In chapter 3 the influence of this parameter on the microstructure is explained. To get large grains
without large residual stresses and over the entire film thickness range, the argon pressure has to go up instead
of deposition time.
In the previous sputter sessions an argon pressure of 8e-6 mbar is used, this time pressures of 1e-5 ,3e-5 and
5e-5 mbar will be used. Higher pressures are not possible due to the maximum argon flow of the Mass Flow
Control (MFC).
15
Figure 15: HRSEM pictures of the copper surface, 500W DC
Left: 5 minutes deposition, argon pressure 50e-6 mbar, with rotation, grain size100±20 nm
Right: 5 minutes deposition, argon pressure 30e-6 mbar, with rotation, grain size 50±10 nm
During sputtering the plasma shows a total different color. The plasma at 30e-6 mbar shows a light green color,
while the plasma at 50e-6 mbar shows a intense dark green color. Sputtering with standard argon pressure
gives a light purple color. Figure 15 shows the results of the sputter sessions with 50e-6 mbar and 30e-6 mbar.
With grain sizes of about 100 and 50 nm, the grains are significantly larger compared with the standard
pressure. They are still not big enough to reach the goal, but the grains are about four times larger than with
the standard argon pressure so larger grains will occur at lower temperatures with higher argon pressure.
16
Heat Treatments After several sputter sessions the choice is made to stop with varying parameters of the process and find
another way to produce copper layers with big grains. Problems as too much residual stress, too thick layers
and small grains at low temperatures don’t give the desired result. Heat treatment of the copper layers is
another way to let the grains grow.
Information about thin film heat treatment is not available, so typical heat treatment temperatures and
durations for bulk materials are used to start with. According to [Callister, p. 198], the recrystallisation
temperature Tr of copper is 120°C and the melting temperature Tm is 1200°C. Typical temperatures for heat
treatments for grain growth are about 0,5 Tm, a typical duration is one hour.
To compare the results of the different heat treatments every time the same samples will be used, sputtered at
300W DC and 50W AC and a deposition time of 45 minutes. Alternating current is used to get samples with
nicely formed grains and to get some small stress in the material to trigger grain growth. To avoid the forming
of oxidation at the surface of the sample the vacuum furnace will be used.
Effect of changing temperature
For the first heat treatment a temperature a bit below 0,5Tm will be used. According to [Callister, p.197], there
will be some recrystallisation and grain growth in bulk materials already above Tr. To look if anything happens
with grain size 400°C is chosen to start with. The duration will be one hour, a standard heat treatment time for
bulk materials.
Figure 16: ESEM images of copper layers after heat treatments
Left: One hour heat treatment at 400°C, grain size at surface 35±10 nm
Right: One hour heat treatment at 500°C, oxidation at the surface
An ESEM picture of the surface of the copper layer treated one hour at 400°C is shown in figure 16a. Grain size
is about 35 nm, which means almost no grain growth: from 30 nm to 35 nm. This setting of heat treatment will
not give large grains, so the next attempt will be with higher temperature: 500°C. Figure 16b shows the surface
of the sample heat treated at 500°C. It is a total different result compared with the surface of the previous
sample. The particles at the surface are significantly bigger, but there form are different than before.
Unfortunately the particles at the surface seems to be oxidation of copper caused by a little bit present oxygen
in the vacuum chamber.
17
At the surface the grain size can’t be measured anymore because the surface hardly changes during heat
treatment and the presence of oxidation, but with other electron microscopy techniques it is possible. One of
them is orientation imaging microscopy (OIM) which uses the principle of EBSD, more information about this
technique and how it is used can be found in chapter 4.
Figure 17 shows the inversed pole figure of the copper
sample heat treated one hour at 500°C. The different
colors in the figure shows the different directions of the
grains in the material. With another function in the OIM
software the grain size can be determined, which is about
250nm. This means notable grain growth with grains of
about half the thickness of the copper layer. It’s a quite
good result, but the goal of grains over the total thickness
(about 800 nm) of the layer is not reached.
When the temperature gets increased further, the
residual stresses in the material will be bigger. At 500°C
this was almost critical for the copper layer. To avoid the
same problem as during sputtering the temperature
cannot be increased further at the next heat treatment.
Effect of changing duration
Another way to get more grain growth without increasing temperature is to treat the material longer. One hour
is too short to let the grains grow to the desired size, so the next heat treatment will be with duration of two
hours at 400°C. At this temperature there was almost no grain growth at the surface, but to be sure there will
not be too much residual stress in the material afterwards the temperature can’t go up even more.
Figure 18: Inversed Pole Figures of copper samples
Left: Deposited half an hour with 300W AC/50W DC, heat treated two hours at 400°C
Right: Deposited half an hour with 300W AC/50W DC, heat treated ten hours at 400°C
Figure 17: Inverse Pole Figure of a copper sample,
deposited 300W AC/50W DC, heat treated one hour
at 500°C
18
An inverse pole figure of the copper sample heat treated two hours at 400°C can be found in figure 18a. The
grain size is about 200 nm, which is six times bigger than the grains without heat treatment. At the top layer
the grains are also measured by ESEM and the size of them are about 35 nm. Conclusion is that the grains at
the surface of the copper layer almost don’t grow. Still the grains are not large enough, so a longer heat
treatment will be the next step.
To get a quick view of which duration of the heat treatment at 400°C will give the desired grain size, the next
treatment will be ten hours long. Figure 18b shows the inverse pole figure of the copper sample heat treated
10 hours. Grain size from the software is about 300 nm. This grain size still not satisfy the goal of grains over
the thickness of the layer, so this ten hour treatment concludes that a lot more time will not give more grain
growth and the goal will not be reached at 400°C, no matter how long the heat treatment will be.
Effect of changing heating up time
One of the reasons for getting moderate grain growth with heat treatments could be the long heat up time of
the used furnace. It takes the furnace about half an hour to get from room temperature to the desired
temperature. Due to recrystallisation the stress in the material after deposition will disappear, which prevents
the grains from growth. Solution for this problem will be to get a really high temperature gradient in the
material so recrystallisation is minimized. Internal stresses from deposition will remain in the material and will
effectuate grain growth. The vacuum furnace is not suitable for this way of heat treatment, for this reason a
atmospheric furnace with nitrogen purge is used.
The goal of the heat treatments with shorter heating up time is to get large grain growth by putting the
samples in the furnace at the desired temperature. This will be 500°C and the duration will be one hour, to
compare the results with that from the previous heating up time.
Figure 19: SEM picture of oxidation at the surface of a copper layer, deposited half an hour with 500W DC and heat
treated one hour at 500°C
Result of this heat treatment is a lot oxidation or copper nitride (CuN) on the copper surface which is even
visible with the naked eye. Figure 19 shows a SEM picture of oxidation or CuN on the surface. As a result of the
oxidation it is impossible to get a good signal during the oim scan, so unfortunately the effect of this heat
treatment on the grain size can't be measured. Three treatments with higher nitrogen pressure (0.5 bar, 1.0
bar and 1.5 bar) do not help to expel the oxygen from the chamber and prevent oxidation.
19
Effect of changing purge gas
In practice inserting of the sample in a hot surface will work, but despite the nitrogen purge there is still too
much oxygen left in the chamber which causes oxidation on the copper surface. To research the effect of a high
temperature gradient during heating up on grain growth, a solution could be to purge the chamber with argon
instead of nitrogen. Argon is an inert gas, which means that it almost do not react with other substances and
will reduce the forming of oxidation.
Temperature and duration will again be 500°C and one hour to compare the results with previous treatments.
Figure 20: SEM picture of the surface of a broken copper layer, deposited half an hour with 500W DC and heat treated
one hour at 500°C
Unfortunately the heat treatment with argon purge did not work. Too much oxygen is present in the chamber
which can't be expelled by the argon flow. For this reason there is still too much oxidation on the surface of the
copper samples to get a usable signal for OIM scanning.
Another way to reduce the forming of oxidation is to speed up the cooling down process by laying the copper
sample on a block of aluminum directly after the heat treatment. When the temperature gradient in the
material is very high there is less time to oxidize during cooling down. Figure 20 shows the result of this fast
cooling down process; the stresses in the material during cooling down are getting too big and the copper layer
fractured. For thin film materials the conclusion is that the fast cooling down method will not work.
20
6. Conclusion Varying the background pressure and stopping the rotation of the substrate during deposition did not lead to
the desired grain size at a layer thickness of about 100 nm. However, the combination of longer duration times
and no rotation creates a higher substrate temperature which is necessary for producing copper layers with the
desired grain size. After fifteen minutes of deposition at medium background pressure a grain size of about two
μm was obtained, which achieves the goal of columnar grains at a layer thickness of about 800 nm. However,
after cooling the samples there was too much residual stress in the material and, additionally, this grain size
will only occur for very thick films.
Heat treatments of the copper layers after deposition are another way to increase the grain size of the small
grains created by sputtering. Heat treatments in vacuum gave some moderate grain growths to grains of
approximately 200-300 nm. Nevertheless, regarding the 750 nm thick copper layers, this is not large enough to
achieve the goal. An attempt was made to increase the grow speed by reducing the ramp-time. But, all of these
samples fractured due to too high internal stresses in combination with surface oxidation, making it impossible
to measure the grain growth.
The temperature during deposition was identified as the biggest influence on the microstructure of the
sputtered material and in the current setup the substrate temperature is completely unknown. Therefore it is
recommended to add a temperature control module to the current deposition setup to allow for temperature
controlled deposition. A recommendation for heat treatment is to use a furnace where a short ramp time can
be used in a vacuum. That would solve the problem of oxidation. Another way of heat treatment is possible in
the chamber of the sputter device, where the radio-frequency (RF) generator could be used to give the
material a heat treatment directly after deposition. Additionally, if possible, using a different metal material
could also be a solution to create the desired microstructure.
Finally, it can be concluded that while the desired microstructure could not be reached, using the equipment of
the MultiScaleLab, for these thin copper films, instead a lot of experience is gained about using the TEER
magnetron sputter device. Also, working with the electron microscopes has lead to useful insights into the
highly specialized settings needed when using orientation imaging microscopy for these thin film materials.
21
References
Books:
[Machlin] Machlin, E.S., Materials Science in Microelectronics I, Elsevier, 2005
[Ohring] Ohring, M., Materials Science of Thin Films; Deposition and Structure, Academic Press, 2002
[Goodhew] Goodhew, P.J. & Humpfreys, F.J., Electron Microscopy and Analysis; Second Edition, Taylor &
Francis, 1988
[Callister] Callister Jr., William D., Materials Science and Engineering; An Introduction Seventh Edition,
John Wiley & Sons, Inc., 2007
Articles:
[Thompson] Thompson, C.V., Structure Evolution During Processing of Polycrystalline Films, Annu. Rev.
Mater. Sci. 30:159-190, 2000
[Neggers] Neggers, J., Manual for the TEER PVD deposition machine located at wh0.053, May 2009
[Ruybalid] Ruybalid, A., Appendix D: Measuring size effects in thin Cu films, 2010
Websites:
[1] http://www.siliconfareast.com/sputtering.htm (Consulted march 2011)