2006 INTERNATIONAL RF AND MICROWAVE CONFERENCE PROCEEDINGS, SEPTEMBER 12 - 14, 2006, PUTRAJAYA, MALAYSIA

Fabrication and Characterization of PZT Thin Film Capacitors for MMICApplications

Nor Fazlina Mohd Lazim', Zaiki Awang', Sukreen Hana Herman2, Uzer Mohd Nor', Mohd NizamOsman3, Ashaari Yusof3, Asban Dollah3, Mohamed Razman Yahya3 and Abdul Fatah Awang Mat3

Microwave Technology Center, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia, 2Faculty of Electrical Engineering, UniversitiTeknologi MARA, 40450 Shah Alam, Selangor, Malaysia, 3Telekom Research & Development Sdn. Bhd., Idea Tower, UPM-MTDC, Technology

Incubation Centre One, Lebuh Silikon, 43400 Serdang, Selangor, Malaysia.zaiki437 0salam.uitm.edu.my

Abstract - This paper reports on the use of thin leadzirconate titanate (PZT) films for monolithicmicrowave integrated circuit (MMIC) capacitors toreplace existing materials for better size-reduction. Thefilms were sputtered on Pt/Ti/SiO2 coated, undopedsilicon wafers. Square Pt electrodes with sides rangingfrom 10 ptm to 50 ptm were patterned on the PZT layersto form the capacitors. Results of this study show thatPZT thin films can be utilized for efficient sizereduction in MMIC. The linewidth obtained for a 50 Qtransmission line is merely 300 nm - this results in asize reduction of approximately five times compared toconventional MMIC. For a 50 x 50 [tm2 electrode area,capacitance values ranging from 5 to 20 pF wereobtained at frequencies up to 20 GHz. Suitable de-embedding of S parameters using Cascade microwaveprobes revealed films with relative permittivities of theorder of 100 to 500.

Keywords: thin dielectric films; MMIC capacitors

processing made them ideal candidates for thispurpose.

2. Theory

The MMIC can be analyzed as a thin film microstrip(TFMS) structure [2], which shares the sameconfiguration as a conventional microstrip shown inFig. 1. It employs a flat strip conductor suspendedabove a ground plane on one side by a low-lossdielectric substrate on the other.

Width (w) Strip conductork A

thickness (t)Dielectric substrate £,

I; height (h)

Ground planeFigure 1: Microstrip circuit structure.

1. Introduction

Demand for reduced circuit sizes in integrated circuittechnology gave the impetus for this project. Withincreasing use of microstrip circuits and MMICs inwireless technology for mobile telephony, high speedcomputing and bio-medical applications, coupled withcontinuing demand for smaller devices and lowerpower consumption, there is a serious need for MMIC-derived circuits which will allow effective sizereduction. Since the size of these circuits is dictated bythe permittivity of the dielectric, it follows that in orderfor successful circuit reduction, substrate materialswith high dielectric constants are therefore desired. Wepropose here PZT thin films as the dielectric in MMICto achieve better size reduction [1]. These tilms offervery high dielectric constants ranging from 100 to 1000and their compatibility with integrated circuit

The structure in Fig. 1 can also be analyzed as aparallel plate capacitor made up of two parallelelectrodes, each of area A, separated by a dielectric ofthickness d and relative permittivity c, [3]. Itscapacitance is given by the following relation

c= gogrAd (1)

The Pt/Ti ground plane in our sample isolates theSiO2 layer from PZT, hence the substrate is effectivelymade up of PZT alone. Therefore the above equation isequally valid for the structure we propose here. Thisnegates the necessity to use the more complicatedvariational method [4], [5] meant for compounddielectrics.

The wavelength of an electrical signal propagatingin microstrip, /Am, is given by

0-7803-9745-2/06/$20.00 ©2006 IEEE. 177

m = (2)

where A, is the free-space wavelength and c, is thedielectric constant of the microstrip substrate [6].Because h for PZT is small compared tOWm, quasi-TEMapproximation can be used for numerical analysis asgiven by [7], [8].

Si3N4 with a relative permittivity of 7.6 iscommonly used in MMIC design. Using PZT as thenew material, however, will reduce the sizesignificantly since its c, varies from 100 to 1000. Forexample, the width of a 50 Q transmission line at 10GHz for conventional Si3N4-based MMIC is of theorder of 1 ptm, compared to 0.07 ptm for PZT - thisrepresents a size reduction of 15 times. In order tofabricate such small structures on the PZT films in ourstudy, electron beam lithography technique was used togenerate the electrode patterns and footprints requiredfor wafer probing.

S-parameter measurements are often required tocharacterize the performance of high-frequencydevices. On-wafer probing techniques are commonlyused to accurately and quickly measure S-parametersof monolithic elements [9, 10]. Planar calibrationstandards are used for accurate calibration and de-embedding of microwave probes at the probe tips.Combination of microwave probes with a correctednetwork analyzer and on-wafer impedance standardstherefore allows on-wafer S-parameter measurementswith a higher level of accuracy.

In vector network analyzer measurements,calibration techniques are used to determine a two-porterror model which enables the device scatteringparameters to be de-embedded from the measurements.The popular calibration techniques used have been theshort-open-load-thru (SOLT), the thru-reflect-line(TRL) and line-reflect-match (LRM) techniques. TheSOLT technique is popular for wafer characterizationand was used in this work. The method uses aphysically small pair of trimmed parallel 100 Qresistors, each 50 ptm long, as the reference loadimpedance, while the short circuit standard shorts theprobe contacts on a 50 ptm wide conductor bar. Theopen circuit standard is obtained with the probe liftedin air.

Following the calibration process, a set ofverification standards was employed in our work toensure the calibration has been performed andcomputed properly. These verification elements consistof transmission lines of different lengths - a properlycarried out calibration would yield smooth curves

(representing delays) along the periphery of the Smithchart.

The capacitance is extracted from impedancevalues, which are converted from the measured S-parameter data, using the following relation [ 1]:

C-1

2#xIm(Z) (3)

where C is the measured capacitance and Z is thecorresponding impedance value converted from S-parameter using the equation:

Z=50 I+(S4

3. Methodology

PZT films of thickness 0.5 pim were deposited onPt/Ti/SiO2-coated Si substrates using RF sputteringtechnique, the results of which have been reportedelsewhere in [12]. In order to investigate the feasibilityof the new dielectric, Pt electrodes which form thecapacitor were patterned on top pf the PZT as shown inFig. 2. The electrode is connected to the signal pad viaa 50 Q, 300 nm long transmission line. The co-planarground-signal-ground (GSG) pads are required for on-wafer probing, they are designed to conform to theprobe footprints. The 50 Q line was required tominimize mismatch at the electrode so that theaccuracy is improved.

The bottom Pt layer was used as ground. Contactto this layer was however not realized through viaholes but was made direct to the film since the Pt layerwas exposed at the edge of the samples. As we shallsee later, ground contacts were made over the PZT/Ptstep at the sample edge - this was decided primarily forconvenience and to reduce the processing steps becausewith this method no chemical etching of the PZT wasrequired.

_Z 300 nm transmission line

L1[ z electrode

Figure 2: The capacitor layout, which consists of a50 x 50 fim2 electrode. A 50 Q transmission line connects the

electrode to GSG probe pads for probing.

178

(4)

On-wafer microwave measurements were thenperformed on the samples using a Wiltron 37269Avector network analyzer and Cascade Microtech probestation with Cascade Infinity probes.

4. Results and Discussion

Figure 3 shows a set of verification results carriedout after the SOLT calibration.

...................................3; ....-. ?..... -

SWh

(I~~~~~~~~~~~~~~~~~ F

Figure 3: Comparison of verification results between(a) simulated and (b) measured data for different Cascade line

lengths: (i) 450 ~tm, (ii) 900 ~tm,(iii) 1800W tm, (iv) 3500 gtm, and (v) 52501tm

The verification elements consist of fivetransmission lines of lengths 450 ptm, 900 ptm, 1800ptm, 3500 ptm, and 5250 ptm. These lines were

simulated using Genesys and their results werecompared to measurement. The results agree quite wellas shown in Fig. 3. The lengths of the traces agree verywell, but the measured response of the longest linespiral inwards slightly at higher frequencies. This is tobe expected because in reality the alumina substrateused for the verification elements inevitablydemonstrate higher amount of losses than that assumedin the simulation. In consideration of the goodagreement obtained with simulation we can concludethat the calibration was successful.

The capacitors were probed after successfulcalibration. Figs. 4 and 5 show the capacitor structurefabricated in this work. The probe pads, 300 nmtransmission line, and electrode were formed at theedge of the PZT-Pt step as mentioned earlier tonecessitate grounding without the use of via holes.With this arrangement, the ground pads are simplyconnected to the Pt layer through a thin line from thepads to Pt over the PZT-Pt step. This arrangement alsoguarantees that the ground pads are level with thesignal pad - this is essential in wafer characterizationswhich use co-planar probes.

In using this arrangement however we can expectsome difference in the results obtained for capacitorsdeposited at different locations from the edge, since thePZT films would tend to level off away from the step.The capacitance produced depends on the PZT filmthickness - it is therefore pertinent to ensure that thecapacitors are located far away from the edge to reducethis effect so that consistent results are observed. Fig. 5shows a capacitor located at a location further awayfrom the edge. The effect of this on the capacitorperformance is shown in Fig. 6.

Figure 4: The probe pads, 300 nm transmission line, andelectrode were formed at the edge of the PZT-Pt step, shown

as a white vertical line in the photo (Cap])

179

Figure 5: A capacitor deposited further away from the PZT-Ptstep (Cap2)

inspection of the structures using a scanning electronmicroscope confirmed the poor grounding.

Fig. 7 shows a close-up view of the 300 nmtransmission line, the smallest structure used in thiswork. The line was delineated using a lift-off processto give better resolution. The quality of the structuregenerated by this process can be seen clearly in thephoto. At microwave frequencies it is important toensure corners are formed correctly to avoidundesirable effects - such as increased reflection -which result from bends.

Cap]

Cap2

Figure 6: The impedance response of the capacitors probedover 0.1 - 20 GHz

From Fig. 6, we observe that all the traces arelocated in the lower half of the Smith chart, as expectedfrom capacitors. Thus we can conclude that thecapacitors were successfully demonstrated in this work.All the traces however exhibit uneven curvature - thismay have been caused by the film quality which mayvary with different deposition conditions. The bestresponse, Cap2 trace, was shown by a capacitor locatedfurthest away from the PZT-Pt edge, as expected.

The quality of traces for capacitors of Fig. 4(Cap]) however, are not satisfactory - their responsesare shifted towards the upper half of the Smith chart,implying an increase in the inductive part of the totalreactance measured. This could probably be explainedin terms of improper or non-contacting grounding ofone of the probe pads. For such cases a GSG probeeffectively behaves like a GS probe instead, giving riseto stray magnetic fields which then lead to an increasein the ground path inductance [13, 14]. A physical

Figure 7: A close-up photo of the 300 nm transmission linewhich connects the electrode to the signal pad of the probe.

The sharp edges and corners obtained from the lift-offprocess are apparent.

The capacitance values were then derived from S-parameter data using equations 3 and 4, and the valuesfor a 50 x 50 [tm2 are plotted in Fig. 8 below. From thefigure we can see that for such a small electrode,reasonably high capacitance values are possible withour design. This is due to the high permittivity offeredby the PZT films used.

25

-,200-

@15a

c310Q

c-) 5

1 3 5 7 9 11 13 15Frequency (GHz)

Figure 8: Measured

17 19

capacitance for a capacitor with an areaof 50 x 50,um2.

The capacitance values are seen to decreasegradually with frequency as expected. This is due to

180

polarization effects in a dielectric, giving rise toincreasing loss as the frequency is increased. Fromthese capacitance values, the relative permittivityvalues of the PZT were derived from (2), and areplotted in Fig. 9.

5004502400

350-00

2500 -

O 200*150

a) 100

1 3 5 7 9 11 13 15 17 19Frequency (GHz)

Figure 9: Relative permittivity of the PZT films over thefrequency range.

5. Summary

MMIC capacitors using a new dielectric thin film wasdemonstrated in this work. The capacitors employedthin PZT films sputtered on Pt/Ti/SiO2-coated Sisubstrates. The processing steps adopted in this workare compatible with MMIC technology, thus offering apossibility of integrating high permittivity ceramicfilms for effective size reduction and enhanced signalprocessing capabilities.

Acknowledgement

Two of the authors (N. F. M. Lazim and Z.Awang) would like to thank Dr. Abdul Fatah AwangMat and his team of researchers from theMicroelectronics and Nano Technology Group atTelekom Research & Development Sdn. Bhd for theassistance given in the fabrication process. Additionalthanks go to Mohd. Nizam Osman for the initialarrangement, Ashaari Yusof for his expertise inelectron beam lithography and Asban Dolah for hishelp in the metallization process.

applications," Proc. 2005 Asia-Pacific Conf onApp. Electromagnetics, Johor Bahru, Malaysia, pp.297 - 300, Dec. 2005.

[2] G. E. Ponchak and A. N. Downey,"Characterization of thin film microstrip lines onpolyimide," IEEE Trans. Components, Packagingand Manufacturing Tech., Part B, vol. 21, no. 2,pp. 171-176, May 1998.

[3] Bahl, I. J., Lumped Elements for RF andMicrowave Circuits, London: Artech House, 2003,p. 164.

[4] E. Yamashita and R. Mittra, "Variational methodfor the analysis of microstrip lines," IEEE Trans.Microwave Theory and Tech., vol. MTT-16, no. 4,pp. 251-256, April 1968.

[5] E.Yamashita, "Variational method for the analysisof microstrip-like transmission lines," IEEE Trans.Microwave Theory and Tech., vol. MTT-16, no. 8,pp. 529-535, August 1968.

[6] E. H. Fooks, and R. A. Zakarevicius, MicrowaveEngineering Using Microstrip Circuits, NewJersey: Prentice Hall, 1990, p. 286.

[7] R. P. Owens, "Accurate analyticaldetermination of quasi-static microstrip lineparameters," The Radio and Electronic Engineer,vol. 46, no. 7, pp. 360-364, July 1976.

[8] E. 0. Hammerstad, "Equations for microstripcircuit design," in 5th European Microwave ConfDigest, pp. 268-271, Sept. 1975.

[9] E. W. Strid and K. R. Gleason, "A DC - 12 GHzmonolithic GaAsFET distributed amplifier," IEEETrans. Microwave Theory and Tech., vol. MTT-30, no. 7, pp. 969-975, July 1982.

[10] K. R. Gleason, et. al., "Precise MMIC parametersyielded by 18-GHz wafer probe," MicrowaveSystem News, pp. 55-65, May 1983.

[11] JEDEC Solid State Technology Association,"Procedure for measuring input capacitance usinga vector network analyzer," JEDEC PublicationJEP147, October 2003.

[12] S. Hana, et al.,"A new PZT thin film preparationtechnique using solid oxygen-source target by RFreactive sputtering," in Proc. IEEE Int. Conf onSemicon. Elect. 2002, ICSE2002, Penang,Malaysia, pp. 378-382, Dec.2002.

[13] Cascade Microtech, Summit 900 Probe StationUser's Guide Appl. Note PN 103-698, 2002, p. 68.

[14] S. Wartenberg, RF measurements of die andpackages, Artech House, Boston, pp. 67-69, 2002.

References

[1] N. F. M. Lazim, Z. Awang, S. H. Herman, and U.M. Nor, "Modeling and simulation of sol-gel thinfilms for monolithic microwave integrated circuit

181