Analysis and fabrication of overlapping-electrodedesigns for poling and modulating channels inpolymer thin films
Robert Mustacich, Michael Gilbert, Ronald Finn, and Catherine Swann
A design for overlapping electrodes for use with thin-film nonlinear optical polymer waveguides isanalyzed. The suggested electrode structure serves both to pole birefringent waveguides in a polymer thinfilm and to apply rf modulation to the waveguide thus defined. This structure is also designed to minimizecapacitance, to maximize electro-optic overlap, and to not require a conducting substrate or surfacecoating of the substrate. A figure-of-merit is developed to optimize both the longitudinal and thetransverse geometry of this electrode structure. A fabrication approach to this general design is presented.Optical characterization data are presented for prototype optical phase modulators fabricated by thisdesign. Fiber coupling losses, propagation losses, and dependence of propagation losses upon claddingthicknesses, composition, and electrode composition are presented.
The purpose of the electrodes on an optical phasemodulator is to distribute the rf signal around a lightwave that is being guided through an electro-opticmaterial. Surface electrodes combined with conduct-ing substrates or films also typically provide theelectric fields required to pole the polymer thin films.These fields, applied near the glass transition temper-ature of the polymer, render the material noncen-trosymmetric; the resulting linear electro-optic effectthen scales with the applied field." Further, thepoling typically results in a birefringence, which iflocally confined, can be used to define channelwaveguides in thin films.5 This approach has beendemonstrated with lithographed fine-line electrodepatterns in surface films that are then used to polethe underlying polymer against a ground plane thatconsists of a conducting film or substrate. Whilemicrometer geometries can be readily lithographed inthe surface films, the fringing fields are large and ofbroad extent in the polymer films,6 especially whenlarge polymer and cladding thicknesses are used toreduce optical propagation losses and to match modesizes to conventional optical fibers. Further, electrode
The authors are with the General Research Corporation, 5383Hollister Avenue, Santa Barbara, California 93111.
geometries of this design for poling channel wave-guides are expected to have suboptimal designs for rfmodulation. For this reason, additional fabricationsteps are expected for fabrication of an optimal rfelectrode structure on the modulator. If a conductingsubstrate for poling the polymer is required, then therf electrode structure must further accommodate forthe presence of this underlying conductor.
For new integrated optics applications based onnonlinear-optical polymer thin films, standard fabri-cation methods for thin-film devices suggest newelectrode geometries better suited to these thin-filmdevices. Here we explore a novel electrode geometryin which the electrodes can be optimized for bothchannel-waveguide poling and rf modulation. Throughcontrol of film thicknesses and photolithographicpatterns above and beneath the polymer films, chan-nel waveguides of different sizes are possible, as areelectrodes optimized for maximum electro-optic over-lap and minimum capacitance. This approach relaxesthe requirement for a conducting substrate for polingthe polymer films. Since the proposed electrode struc-ture serves both poling and modulation purposes, thefabrication can largely be viewed as substrate indepen-dent as long as a suitable planarizable surface isavailable for the device fabrication.
We show an analysis of this generalized electrodegeometry, which includes the computation of fieldand transmission-line characteristics to aid in thedesign of electro-optic modulators that use overlap-
ping electrodes. A figure-of-merit (FOM) is intro-duced for optimizing the transverse geometry of theelectrodes. Both standing-wave and traveling-waveelectrode designs are considered. We further show thefabrication and optical characterization of prototypephase modulators that consist of single-mode channelwaveguides poled in polymer thin films by using thiselectrode geometry.
Optical Phase-Modulator Electrode Design
The long narrow electrodes of an optical phase modu-lator can be accurately modeled by a TEM-modetransmission line. The parameters characterizing thetransmission-line model can be conveniently dividedinto two parts: longitudinal and transverse parame-ters. The two subsections that follow describe theoptimizing of each of these in turn, with regard foruse in a narrow-bandwidth optical link.
Longitudinal Parameters
The longitudinal parameters characterize the lengthand the impedance loading of the electrodes. Thissubsection considers the optimizing of the length fortwo different types of loading: traveling-wave andstanding-wave designs, shown respectively in Figs.1(a) and 1(b).
The traveling-wave design7-10 launches the electri-cal wave into one end of the electrodes, where it isguided to a matched load at the opposite end. Afeature of this design is that when the velocities of theelectrical and optical waves are matched, the modula-tion is additive along the entire length of the elec-trodes. Although some electrical power is wasted inthe resistive load, the resistive input impedance mayprovide for the simplest wide-bandwidth match to theelectrical source.
The standing-wave design launches the electricalwave into the center of the electrodes, where it dividesand reflects off of the two open-circuited ends. Themodulation is not completely additive for this designsince the electrical standing wave changes phaseduring the time it takes the optical wave to pass the
ELECTRICALINPUT
MATCHINGNETWORK
X ~~~~~~~~ ~ ~~Z .Y * zLIGHTWAVE
|. ~ ~ ~~~~e (a)
ELECTRICALINPUT 0_X MATCHING
NETWORK
LIGHTWAVE
(b)
Fig. 1. Transmission-line models for the electrodes of an opticalphase modulator: (a) traveling-wave type, (b) standing-wave type.
electrodes. For narrow-bandwidth designs, however,this effect is more than compensated for by the moreefficient use of electrical power.
A FOM for optimizing the electrode design can beexpressed by
FOM = /A 2 MN TOWG,VP'MN"1WG (1)
where A is the optical phase shift in radians, Pm isthe power delivered to the modulator in watts, qOWG isthe efficiency of the optical waveguide over the lengthof the electrodes, and MN is the efficiency of thematching network. The square of the FOM appears asa direct factor in the electrical-electrical efficiency ofan optical link that uses phase modulation and directdetection.
The FOM given by Eq. (1) was evaluated numeri-cally and was plotted against electrode length for thetwo types of loading by assuming a priori parametersfor the transverse geometry. For simplicity of compu-tation the matching network efficiency was set equalto 1, making the results for the standing-wave casestrictly applicable only for narrow-bandwidth de-signs. Figure 2(a) shows the results for electricalfrequencies of 1 and 10 GHz applied to the traveling-
4 .0. . ...
1t1OGH
51 ' / t a-.;.
ELECTRODE LENGTH, mm
(a)
0n I\ / \ X1.0
,.0 10 100ELECTRODE LENGTH. mm
(b)Fig. 2. FOM versus electrode length at two frequencies: (a)traveling-wave electrode configuration and (b) standing-wave con-figuration. The FOM reaches a maximum at an optimum electrodelength for a given frequency. The maximum FOM for bothfrequencies is obtained with the standing-wave design. Attenua-tion a = 0.011 Np/mm.
wave electrode design. For this case the optimumelectrode length is seen to be approximately 20 mm,independent of frequency. The trade-off that providesfor this optimum is found to be between the increasein phase modulation with increasing electrode lengthand the decrease in optical efficiency with increasingelectrode length.
Figure 2(b) plots the results for the standing-waveelectrode design. For this case the optimum electrodelength is seen to be approximately 6 mm for the1-GHz frequency and approximately 2 mm for the10-GHz frequency. For both frequencies the trade-offthat provides for the optimum in electrode length isfound to be between the increase in phase modulationwith increasing electrode length and the increase inthe resistive loss of the electrodes as the lengthapproaches self-resonance.
The numerical evaluation of Eq. (1) has shown thestanding-wave electrode design to be superior to thetraveling-wave type for use in narrow-bandwidthoptical links. Not only is the computed FOM higher,but the optimum electrode lengths are shorter for thestanding-wave design. The next subsection considersthe optimizing of the transverse parameters.
Transverse ParametersThe transverse geometry of the electrodes can becharacterized by three parameters: a complex char-acteristic impedance Z0, a complex propagation con-stant y, and an electro-optic overlap integral F. Eachof these can be computed from the material proper-ties and transverse dimensions shown in Fig. 3 by anumerical solution to the corresponding electrostaticproblem.
The optimum design of the transverse dimensionsultimately involves a trade-off between the electro-optic overlap, the film thickness, and the electrodecapacitance. The functional relationship for thistrade-off can be deduced from Eq. (1). For example,consider the case of a standing-wave design that usesa wide-bandwidth matching network, in which thedominant dissipative electrical loss is in the film.Then Eq. (1) can be reduced to
FFOM -, hg
With the goal of optimizing the electrode overlap A,we developed a computer code for computing ce and F.Figure 4(a) shows the increase in capacitance ex-pected for increasing electrode overlap. Figure 4(b)shows the increase in electro-optic overlap expectedover the same range. The optimum trade-off betweenthese opposing effects is shown in Fig. 5, in which themaximum FOM is seen to occur where the electrodeoverlap is approximately equal to the film thickness.Also made apparent by this figure is the increase inthe FOM expected for decreasing film thicknesses.The lower bound of the film thickness is dictated bythe size of the optical waveguide that it contains.
Device Fabrication and Experimental Testing
Fabrication of channel waveguides with the overlap-ping-electrode design described above for both poling
0.7
0.6
0.5
E
To
0.4
0.3
0.2
0.1
0100
A (Am)
(a)
1.0
0.9(2)
where F is the electro-optic overlap, h is the filmthickness, and Ce is the electrode capacitance per unitlength.
Fig. 3. Generalized transverse-electrode geometry. Electrodes ofwidth w have an overlap A over an electro-optic polymer ofthickness h. Any cladding polymers are included in the polymerthickness h. The lower electrode is buried on the glass substrate
and the top electrode is at the device surface.
0.8
0.7
- ~~~~~~2
OPTICALBEAMWIDTH
(>m)
1 10 100
A (m)
(b)
Fig. 4. Components of the FOM for the transverse-electrodegeometry versus electrode overlap: (a) Capacitance per unit length,(b) electro-optic overlap; h = 10 pm, w = 1 mm, Er1 = 3.5, and Er2 =
Fig. 5. FOM for the transverse-electrode geometry versus elec-trode overlap for different film thicknesses h; w = 1 mm, Er1 = 3.5,and Er2 = 4.0. Broad maxima are observed with a value of theelectrode overlap approximately equal to the film thickness h.
and modulation fields was performed by using lowdielectric substrates and claddings with the HCC1232polymer developed by Hoechst-Celanese Corpora-tion.1-3 To obtain large mode sizes with single-modewaveguides, we used tunable index HCC1232A poly-mer4 as a cladding material. The HCC1232A indexwas formulated to be 0.006 less than the HCC1232guiding-layer index. The HCC1232 polymers formplanarizing films with extremely smooth surfaces byspin coating. Rms surface roughnesses of approxi-mately 10 A have been measured for films made withthese polymers.4
Planar waveguiding in the thin films was measuredto determine optical losses caused by claddings, sub-strates, metallization, and processing steps. Besidesproviding measurement simplicity, prism couping11
of planar waveguides also provides measurement ofthe refractive index, the refractive index changescaused by poling, the film thickness, and the modeindices. Film thickness and refractive indices weredetermined from mode coupling angles by using themethod described by Kirsch.12 Planar waveguideswere also used to calibrate the spin-coating process-ing for each polymer batch. Refractive-index changesthat resulted from poling were measured by spanningpoled and unpoled regions of the polymer with aprism and by measuring the relative change in modecoupling angles as the optical beam is translatedbetween the two regions.
The substrates were 3-in. (7.6-cm) ultraparallelglass wafers obtained from Hoya Electronics Corpora-tion (San Jose, Calif.). The optical properties ofplanar waveguiding films on these glass wafers werenot significantly different from those of films formedon prime quality silicon wafers. Aluminum or goldmetallization of substrates or deposited films wasperformed by vacuum evaporation. A photomask was
fabricated that consisted of a light-field comb patternof 1-mm-width parallel lines that were continuous atone end with the dark field border of the mask. Themask was designed for self-alignment after a 180°rotation and was offset to achieve the overlapping-electrode design shown in cross section in Fig. 3.Optical verniers were designed into the mask tocontrol the self-alignment and to vary the electrodeoverlap.
Poling of the samples was done according to theprocedures recommended by Hoechst-Celanese Corpo-ration. The sample was raised to a temperature abovethe glass transition temperature of the polymer, andthe poling field was applied. The field was monitoredat the film surfaces during the poling process. After abrief poling at high temperature, the sample tempera-ture was lowered with the field still being applied. Wecleaved the poled samples away from the waveguideswith a diamond scribe and extended the break acrossthe waveguides by applying uniform pressure fromthe back side of the substrate. A scanning electronmicrograph of a channel-waveguide endface is shownin Fig. 6. This waveguide consisted of an HCC1232film with UV-curing spin-coated epoxy cladding lay-ers on a glass substrate. The lower epoxy layer isclearly seen to form a planarizing layer over the0.2-pm-thick aluminum electrode. The boundary be-tween the bottom cladding and the HCC1232 film canbe observed since the HCC1232 film is deposited by asolvent evaporation process after curing of the epoxyfilm. The boundary with the top cladding layer is notobserved since the epoxy is covalently linked to theHCC1232 film in its cure process. The HCC1232 filmswere found to give glassy breaks that were suitablefor fiber end-firing.
Some early fabrications were done by using epoxycladdings (UV-curing epoxies for spin-coat processingfrom Master Bond, Hackensack, N.J.), but these clad-dings result in small single-mode waveguide dimen-sions because of the large 0.1 difference in refractiveindices. For single-mode channel waveguides of largersize suitable for matching to single-mode fiber coredimensions of approximately 6 lm, the HCC1232Acladding was used because of its small 0.006-indexdifference. However, the use of claddings with closely
Fig. 6. Scanning electron micrograph of a channel-waveguide endface. The lower epoxy cladding is visible as a planarizing layer overthe edge of the aluminum electrode. The top cladding layerpolymerizes to the waveguiding film during cure, and the boundarywith the top cladding is not observed. The relatively smooth endface was suitable for fiber end-firing.
matched indices requires thick cladding layers sincethe guided mode significantly penetrates the claddinglayer. To reduce the overall device thickness whilerealizing the mode size advantages of a closely matchedcladding, we fabricated an asymmetrically cladwaveguide design. Since a single small-delta claddingis sufficient to define a large single-mode size, theother cladding can have a large delta. This is analo-gous to the asymmetric index profiles of titaniumindiffused surface waveguides in lithium niobate,which have a small index gradient into the crystal anda large step at the crystal surface. A simulation of thisasymmetric design is shown in Fig. 7. A finite-element simulation with methods described earlier6was created for the multilayer dielectric. The dielec-tric layer boundaries and the electrodes are shown inan overlay of the finite-element solution for the singleTM mode guided by the channel. The finite-elementsolutions also give the lower effective index modesthat are planar guided but below the cutoff for thechannel mode. In this simulation the HCC1232 filmsare both 5 lm in thickness and the UV-15LV epoxylayer is 2 lm in thickness. For processing reasons theUV-15LV layer was placed as the final layer. Thesingle TM mode has a l/e width of approximately 6gim. Because it significantly penetrates the lowercladding, the optical propagation loss will be sensitiveto the thickness of this cladding layer.
Experimental measurements of the insertion lossfor channel devices that consist of a 4.3-gm-thicklayer of HCC1232 cladded with 5.3 glm of HCC1232Aand 2.3 gm of TV-15LV epoxy were performed. Theburied electrode was 0.2-gm-thick aluminum, andthe surface electrode was 0.04-gm-thick gold. Weused 1.32-gm-wavelength light to end-fire the devicewith Corning single-mode polarization-preservingfiber. The birefringence created by poling at 100V/gm results in a single-mode guide for the TM
polarization. Both single-mode fiber and step-indexmultimode fiber (50-gm core diameter) were tested asoutput fibers with no significant difference observed.The variation of insertion loss with channel length isshown in Fig. 8. The intercept gives coupling losses ofapproximately 1.2 dB/end face. This end-face loss isin approximate agreement with the single-mode-fiberto planar-waveguide coupling loss of 1.4 dB that wasmeasured by output coupling the planar waveguidewith a prism. These modest coupling losses areconsistent with the relatively smooth end face shownin Fig. 6. No attempt was made to minimize theselosses by polishing the end faces or by treating thesurfaces with fluids or gels.
The slope in Fig. 8 shows a substantial opticalpropagation loss of 4.4 dB/cm. As shown in thesimulation of a channel that has similar dimensionsin Fig. 7, a significant portion of the guided lightpenetrates the cladding, which can result in opticalloss owing to absorption by the electrode. Optical lossfrom the electrodes was confirmed by optical propaga-tion-loss measurements with polymer-clad planarwaveguides with and without metallization of thesubstrate. These measurements were made by chang-ing the spacing of the input and output couplingprisms and by measuring the change in the opticalthroughput. Large increases in propagation lossamounting to several decibels per centimeter weremeasured with the metallized glass substrates rela-tive to propagation losses of approximately 0.7dBm/cm for the polymer-clad waveguides on glass.
To analyze further the optical loss resulting fromthe electrodes and the insulation provided by thecladding, a ray-optics model for waveguiding in multi-layer dielectrics with metallic boundaries was used.This model was based upon published methods that
0.00 3.10 6.10 0.2 12.30 15.46 I. 21I 1-I I
1L71 I
13.10 I
104
7.
5.24
2.2
0.00 0.00 3.10 6.19 9.26 12.30 25.46
.67
_ 15.71
4 13.10
21
0_ 67
C:
0n -(W 4:
10.4a
7.16
5.24
2:2.2
18.57 21.07
pm
Fig. 7. Finite-element simulation of the guided mode in theasymmetrically clad channel waveguide that was fabricated. Tocalculate the intensity contours of the guided mode, 288 first-orderelements were used. The guided mode had a Ile width of approxi-mately 6 WLm1 and it shows significant penetration of the lowercladding.
0.0
A
0.5 1.0Channel Length
1.5 2.0(cm)
Fig. 8. Channel-waveguide insertion loss. The guiding-layer thick-ness was 4.3 pum, the lower cladding thickness was 5.3 pm, and thetop cladding thickness was 2.3 pLm. A propagation loss of 4.4 dB/cmwas observed with a combined coupling loss for both end faces of2.3dB.
used complex dielectric constants.'3" 4 We similarlyused handbook values for the complex dielectricconstants of the metals.'5
Calculations for three different cladding-metalliza-tion combinations were made by using complex refrac-tive indices for the metals estimated at 1.32 gim. Anapproximately logarithmic decay of the attenuationcoefficient was observed as a function of claddingthickness. This dependence upon cladding thicknessfor these different cladding-metallization combina-tions is shown in Fig. 9. Note that the plots are thelogarithm of an exponent. This shows that the pre-dicted propagation loss is a very sensitive function ofcladding thickness. These results show that theHCC1232A cladding thickness must be comparablewith or larger than the HCC1232 layer, especiallywith aluminum, in order to achieve propagationlosses below 1 dB/cm, because of the weaker confine-ment of the guided light by the HCC1232A cladding.In contrast, the UV15-LV epoxy cladding, a high-delta cladding, is only required to be approximately 2gim thick for insulating the guided mode from alumi-num.
The predictions of Fig. 9 were checked with thecorresponding planar waveguiding loss measure-ments. The loss over gold metallization was shown tobe less than aluminum, as predicted. The HCC1232Acladding thickness dependence of the optical loss byusing aluminized-glass substrates was investigated inmore detail. These results are shown in Fig. 10.Reasonable agreement is observed with the ray-opticstheory for the loss. For HCC1232 waveguiding overUV-15LV-clad aluminum, a loss of approximately 2dB/cm was measured.
These metallization losses can account for theoptical propagation loss measured with the channel
100
0m
Cl0.
10:
1
nI 0 l l l l l l l l l l l l l l l l l l l l l l l l
Cladding Thickness (pm)Fig. 9. Predicted planar waveguiding propagation losses for dif-ferent materials and cladding thicknesses. Minimizing propagationloss requires large thicknesses of the cladding with the closelymatched index. Aluminum also results in greater loss than goldelectrodes for a given cladding thickness.
100:
m
Ce
0.
10-
1-
0.1 2 I 6 8 1 . .I . . . . . . . . I
Cladding Thickness (pm)Fig. 10. Measured and predicted optical propagation loss forair-HCC1232 (4.3 m)-HCC1232A (variable thickness)-alumi-num-glass planar waveguides. Reasonable agreement betweenexperimental measurements and the approximate theory are ob-served. These results can explain the large 4.4-dB/cm propagationloss observed in the fabricated channel devices (Fig. 8).
waveguides (Fig. 8). The approximate ray-optics modelapplied to the same dielectric layer structure as in thechannel guides gives a nearly identical numericalresult of 4.4 dB/cm for the lowest planar TM mode.This corresponds closely to the predicted loss for5.3-Lm-thick HCC1232A cladding over aluminizedglass shown in Fig. 10. The upper cladding of gold-clad 2.3-gm-thick UV-15LV appears relatively insig-nificant in terms of its contribution to the optical loss.
Both the planar and channel waveguiding devicesdescribed above provided the expected phase modula-tion in laboratory test beds. Phase modulation wasmeasured by including either a planar device or achannel device in one branch of a bulk optics Mach-Zehnder interferometer. The modulation of fringemovement in the interferometer was either measuredvisually or with a fiber to a detector. The electro-opticr coefficients were in agreement with previouslyreported poling data for the HCC1232 polymer.4
Summary
A parametric study of the overlapping-electrode geom-etry has been conducted to aid in the design of anintegrated optics phase modulator that uses an electro-optic polymer. A figure-of-merit, selected for minimiz-inglink loss, leads to the followingresults: standing-wave electrodes were shown to have a higher figure-of-merit than traveling-wave electrodes for narrow-bandwidth design; the electrode overlap should equalthe film thickness; and the film thickness should beminimized.
A fabrication approach that involves photolithogra-phy of metallic films beneath and above the polymerfilm was developed and demonstrated. The electrodestructure was used to both pole and modulate channel-
waveguide phase modulators in the polymer films.The resulting single-mode waveguides were character-ized in terms of propagation loss and fiber couplingloss. The large propagation loss in the prototypechannel devices was accounted for by predictions andexperimental measurements of optical absorptionlosses by the electrodes for the cladding thicknessused. The required variation in the cladding thick-ness to minimize this loss was determined by usingplanar waveguiding measurements. Coupling to opti-cal fibers with cleaved waveguides was shown toresult in losses of approximately 1 dBm/end face.
References1. R. DeMartino, D. Haas, G. Khanarian, T. Leslie, H. T. Man, J.
Riggs, M. Sansone, J. Stamatoff, C. Teng, and H. Yoon,"Nonlinear optical properties of polymers," in Materials Re-search Society Symposium Proceedings (Materials ResearchSociety, Pittsburgh, Pennsylvania, 1988), Vol. 109, p. 65.
2. C. C. Teng, D. R. Haas, H. T. Man, and H. N. Yoon,"Guided-wave polymeric electro-optic modulators" in Inte-grated and Guided-Waue Optics, Vol. 5 of 1988 OSA TechnicalDigest Series (Optical Society of America, Washington, D.C.,1988), pp. 377-380.
3. D. R. Haas, H. N. Yoon, H. Man, G. H. Cross, S. Mann, andN. J. Parsons, "Polymeric electro-optic waveguide modulator;materials and fabrication" in Nonlinear Optical Properties ofOrganic Materials II, G. Khanasian, ed., Proc. Soc. Photo-Opt.Instrum. Eng. 1147, 222-232 (1989).
4. H. T. Man, K. Chiang, D. Haas, C. C. Teng, and H. N. Yoon,"Polymeric materials for high speed electro-optic waveguidemodulators," in Photopolymer Device Physics, Chemistry, and
Applications, R. A. Lessard, ed., Proc. Soc. Photo-Opt. In-strum. Eng. 1213, 7-16 (1990).
5. J. I. Thackara, G. F. Lipscomb, M. A. Stiller, A. J. Ticknor, andR. Lytel, "Poled electro-optic waveguide formation in thin-filmorganic media," Appl. Phys. Lett. 52, 1031-1033 (1988).
6. R. V. Mustacich, "Scalar finite element analysis of electro-optic modulation in diffused channel waveguides and poledwaveguides in polymer thin films," Appl. Opt. 27, 3732-3737(1988).
7. R. C. Alferness, "Waveguide electro-optic modulators," IEEETrans. Microwave Theory Tech. MTT-30, 1121-1137 (1982).
8. D. Marcuse, "Optimal electrode design for integrated opticsmodulators," IEEE J. Quantum Electron. QE-18, 393-398(1982).
9. F. Auracher and R. Keil, "Design considerations and perfor-mance of Mach-Zehnder waveguide modulators," Wave Elec-tron. 4, 129-140 (1980).
10. R. A. Becker, R. H. Rediker, and T. A. Lind, "Wide-bandwidthguided-wave electro-optic intensity modulator at = 3.39I'm," Appl. Phys. Lett. 46, 809-811 (1985).
11. P. K. Tien, R. Ulrich, and R. J. Martin, "Modes of propagatinglight waves in thin deposited semiconductor films," Appl.Phys. Lett. 14, 291-294 (1969).
12. S. T. Kirsch, "Determining the refractive index and thicknessof thin films from prism coupler measurements," Appl. Opt.20, 2085-2089 (1981).
13. A. Reisinger, "Characteristics of optical guided modes in lossywaveguides," Appl. Opt. 12, 1015-1025 (1973).
14. I. P. Kaminow, W. L. Mammel, and H. P. Weber, "Metal-cladoptical waveguides: analytical and experimental study," Appl.Opt. 13,396-405 (1974).
15. D. E. Gray, ed., American Institute of Physics Handbook(McGraw-Hill, New York, 1963).
