Planar waveguides are basic elements for integratedoptics, optical communication devices, and biomedi-cal sensor applications. There are different tech-niques for thin-film deposition: rf magnetronsputtering,1–5 chemical vapor deposition, and solgeltechnology. Each technology has its advantages anddisadvantages with respect to the film quality andthe fabrication cost. The concept of material mixing iswell known in thin-film technology for providing anintermediate refractive index of the constituent ma-terials or to alter the crystal structure of the film.6,7
Engineering the waveguide optical constants (realand imaginary refractive indices) is rather an impor-tant aspect in that each application requires thatcertain specifications be met. The optical properties
of the waveguide affect the efficiency and the perfor-mance of the integrated optical devices in differentaspects. For example, the waveguide imaginary re-fractive index controls the optical losses and thuslimits the size and performance of the thin-film de-vice. The real refractive-index contrast, on the otherhand, determines the energy confinement inside thewaveguide and therefore the tolerance for thewaveguide bends to network the integrated optic de-vices. Mixing oxides, with different optical properties,can produce a relatively wide range of films’ opticalconstants, which provides the designer the flexibilityto meet the requirements for reaching the optimalthin-film-device performance. Engineering the opti-cal constants can also be done by chemical vapordeposition technology, as discussed in Ref. 8. Onemajor advantage of rf sputtering over chemical vapordeposition is that rf sputtering can provide a widerarray of material mixing.
In most applications the optical losses need to bereduced; it is well known that annealing reduces op-tical losses2 by oxidizing the nonoxidized sputteredmaterial and homogenizing the deposited thin film byenhancing the diffusion process. So, the effect of an-nealing on optical losses will be investigated for some
The authors are with The Department of Electrical and Com-puter Engineering, Wayne State University, Detroit, Michigan.The e-mail address of R. Rabady is [email protected].
Received 4 September 2003; revised manuscript received 23June 2004; accepted 12 October 2004.
of the fabricated waveguides and optical resonantfilters as well.
This paper presents results of a new and system-atic study of thin-film waveguides fabricated by themixing of silica, titania, and tantalum pentoxide dur-ing rf magnetron sputtering.
2. Experimental Research and Results
In this paper, rf magnetron sputtering was employedto deposit two-oxide-mixture film on standard BK7glass substrates (Fig. 1). Three mixing combinationswere realized by use of three targets (titanium, sili-con, and tantalum). Introducing oxygen to the vac-uum chamber during sputtering oxidizes thesputtered material, whereas changing the forward rfpower ratio of the two magnetrons controls the molarratio of the two oxides, thus engineering the opticalconstants of the deposited film. X-ray diffraction wasperformed after all depositions to verify that amor-phous films resulted in all cases. Figure 2 shows thex-ray diffraction for three film mixtures that are freeof Bragg peaks, which indicate a free-crystalline sig-nature of the films.
The deposition chamber was cylindrical (80 cm inheight and 70 cm in diameter) that was equippedwith two rf magnetrons. All depositions were per-formed at room temperature with a base pressure ofthe order of 10�7. The purity of the targets was99.94% for tantalum and 99.999% for titanium. Gasflow of 5 and 30 cubic centimeters per minute at STP(SCCM) for argon and oxygen, respectively, wereused for all depositions. The glass substrates werecleaned in an ultrasonic bath with acetone, thenmethanol, and finally with distilled water just beforedeposition. Deposited films were thick enough (400–1000 nm). First, thicker film provides more extremesin the transmission spectra and thus more accuracyto extract the film thickness and the real refractiveindex. Second, it ensures the waveguide’s first modeexcitation by passing the cutoff thickness for the632.8-nm wavelength. We achieved waveguidedmode excitation by etching a shallow holographicgrating on the substrate before the deposition and
designed the period of the gratings to ensure excita-tion with the 632.8-nm wavelength; Fig. 3 shows thewaveguided mode excitation. The main equation thatgoverns the mode excitation (also the spatial–spectral filtering processes for the optical resonantfilter) is given by9
�n* � m�
��2
� sin2 � � ���
4��2
, (1)
where � is the Bragg coupling coefficient and in-creases with the grating depth, � is the wavelength, �is the gratings’ period, � is the angle of incidence, m
Fig. 1. Rf magnetron sputtering setup for the two-oxide-mixturedeposition. MW, microwave.
Fig. 2. (a) X-ray diffraction for the first titania–silica sample inTable 1. (b) X-ray diffraction for the first titania–tantalum pentox-ide sample in Table 2. (c) X-ray diffraction for the first silica–tantalum pentoxide sample in Table 3.
is the order of diffraction, and n* is the waveguideeffective refractive index.
Equation (1) indicates a hyperbolic relation be-tween the spectral (the � parameter) and the spatial(the � parameter) filterings. The spectral gap be-tween the lower and the upper parts of the hyperboliccurve is a function of the Bragg coupling factor � andthe wavelength �. If � is relatively high, realized bydeep gratings, it is evident that at normal incidenceEq. (1) can be satisfied by two wavelengths. In con-trast, a shallow grating with relatively low � pro-duces one sharp resonance wavelength at normalincidence. This is obvious when � is set to zero, andthe excitation condition in Eq. (1) reduces to a linearone that is satisfied at normal incidence by a singlewavelength:
n* � m�
� sin �. (2)
To fabricate the gratings, we used the interfer-ence of a two coherent deep-ultraviolet beams froma Ar� laser with 257.2-nm-wavelength radiation toprint the gratings in SPR505 photoresist (from Shi-ply) film. Adjusting the angle of interference of thetwo deep-ultraviolet beams controls the gratings’ pe-riod; Fig. 4 shows the deep-ultraviolet exposure setupfor the gratings’ patterning. To obtain the gratings’formation we developed the exposed photoresist andthen hard baked it to increase its etching resistance.We noticed that, when preparing the gratings on theBK7 glass substrate and etching with CF4 to transferthe grating formation to the glass substrate, the etch-ing rate in the glass was very low (1–2 nm�min)compared with that rate in the hard-baked photore-sist (20–30 nm�min), which met our need to produceshallow gratings for coupling only and not to signifi-cantly perturb the waveguide film.
Microwave (MW) power was applied to the oxygenthat was fed to the vacuum chamber in the TiO2–SiO2
depositions. Using MW produces waveguides withrelatively lower optical losses, as indicated in Table1. This is not surprising because MW power pro-duces more ionization for the oxygen and, conse-quently, better oxidizing of the sputtered materialand less titanium content in the film. Systematicstudy of how MW ionization of oxygen affects thewaveguide quality will be the subject of separateresearch. After the film was deposited, the trans-
Fig. 3. Measurement setup for waveguide excitation and opticallosses.
Fig. 4. Computer-controlled setup for a period-controlled gratings’ patterning.
Table 1. TiO2–SiO2 Films’ Deposition Settings and the Resulting Optical Constants and Deposition Rates
mission spectra were measured and fitted, a UV–IRspectrometer was used to obtain the transmissionspectra, which was later fitted with respect to thereal refractive index and the thickness of the de-posited thin film. Figure 5(a) shows the fitting re-sults for the film of the tantalum oxide–titaniamixture; the 0.0051 error is the rms value of thedifference between the data and the fit. The smallmismatch between the data and the fit can be at-tributed to ignoring the effects of dispersion and thefilm optical losses when we performed the fitting.The wavelength range of the transmission-spectrameasurements was determined where the film op-
tical losses were negligible; the number of fittingparameters were thus reduced. For measuring thewaveguide optical losses at 623.8 nm, a waveguidedmode was excited (Fig. 3), and the power decay alongthe excited waveguide plane was measured and re-corded with the corresponding position by a slidingfiber. Figure 5(b) shows the power decrease along thepath of the excited Ta2O5–SiO2 waveguide (solidcurve) and the first-order exponential decay fit(dashed curve) that was made to extract the opticallosses. The rms difference between the data and thefit for this particular sample was 0.0022. The error inextracting the optical losses, real refractive index,and deposition rate (or the film thickness) was ex-pressed as the difference between the two points inthe fitting parameter(s) (i.e., optical losses, real re-fractive index, and the film thickness) where the min-imum error of fitting is doubled.
Tables 1–3 show an overview of all depositions’settings and the resulting film optical constants,thicknesses, and corresponding errors that were ex-tracted by the procedures described above. The op-tical losses shown here were considerably highbecause they were measured at the 632.8-nm wave-length. For the optical communication wavelengthof 1550 nm, the optical losses can be lower by afactor that is as great as 36 because the opticallosses are dominated by the scattering losses thatare proportional to ��4.
Figure 6 shows the normalized optical power de-tected in the fiber-sliding setup. The upper solidcurve corresponds to the annealed sample, thedashed–dotted curve corresponds to the exponen-tial decay fit of the annealed sample, and thedashed curve corresponds to the as-deposited sam-ple. The normalized power axis was represented ina logarithmic scale for comparison purposes. Table4 summarizes all the results of the optical lossesbefore and after annealing for five waveguide com-posites.
The annealing was done in air with a 10 °C/minramp up and held at 450 °C for 1 h, and then thesamples were left for cooling by our opening theoven’s door. As expected, annealing decreases theoptical losses for all annealed waveguides. The en-hancement of optical losses by annealing was ashigh as 5.3 dB/mm for the Ta2O5 waveguide and aslow as 0.4 dB/mm for the TiO2SiO2 waveguide. The
Fig. 5. (a) Transmission spectra (solid curve) and fit (dashedcurve) for titania–tantalum pentoxide film for extracting the film’sreal refractive index and thickness. (b) Optical losses’ power decay(solid curve) and exponential decay fit (dashed curve) for titania–tantalum pentoxide film.
Table 2. TiO2–Ta2O5 Films’ Deposition Settings and the Resulting Optical Constants and Deposition Rates
silica–titania mixture showed relatively low opticallosses from the beginning 1.4 dB/mm in that MWpower was used in those particular depositions. Thewaveguide optical losses were further improved byannealing to 1.0 dB/mm.
The effect of annealing can be observed clearlyfrom the highly sensitive shift of the resonance wave-length and the full width at half-maximum reductionof the optical resonant filters. A detailed descriptionof the optical resonant filters’ fabrication technologycan be found in Ref. 9. Figure 7(a) shows the opticalsetup that was prepared to characterize the trans-mission spectra of the fabricated filters. The spatialfilter that is made from an objective with 5-mm focallength and a fiber with a 60-�m core diameter pro-vides 0.006 rad of spatial-filtering width, which isacceptable to characterize the filter’s spectral behav-ior separately. Figure 7(b) shows the transmission
spectra for a resonant filter that is made of a titaniawaveguide. A clear decrease in the spectral-filteringwidth with a deeper resonance minimum in the trans-mission spectra after the filter was annealed at differ-ent temperatures can be attributed to decreasing theoptical losses, whereas, the resonance wavelength shiftcan be attributed to the change in the optical proper-ties of the waveguide film. A different optical resonantfilter that is made of a TiO2–SiO2 waveguide was char-acterized by our measuring the transmission spectrajust after fabrication, then after two weeks of 80%relative humidity exposure, and finally after anneal-ing, as shown in Fig. 7(c). No change in the transmis-sion spectra was observed after the change that isassociated with annealing. For this particular filter,the grating was relatively deep, which produces astrong Bragg coupling between the forward and thebackward waveguide propagating modes; thus tworesonance wavelengths appeared at normal incidenceas is evident from Eq. (1). A clear shift and degrada-tion of the filter performance can be attributed to anunstable composition of the waveguide film that isinherited from the sputtering process with the hu-midity exposure, resulting in changes in the opticaland physical properties of the waveguide film.
The change in the optical resonant filter’s trans-mission spectra after annealing can be attributed tothe reordering of the material atoms and moleculesby breaking of unstable bonds and establishment ofother stable ones. It may even slightly change thematerial composition of the film by releasing gas mol-ecules from the film that were trapped during depo-sition and by oxidizing the nonoxidized sputteredmaterial.
Fig. 6. Measurements of Ta2O5�SiO2 waveguide optical lossesbefore and after annealing; the dashed curve is the measuredoptical losses before annealing, the solid curve is the measuredoptical losses after annealing, and the dashed-dotted line is the fitof the measured optical losses after annealing.
Table 3. Ta2O5�SiO2 Films’ Deposition Settings and the Resulting Optical Constants and Deposition Rates
Total pressure � 7 mTorr, distance of the target from the substrate distance of the target from the substrate � 12.5 cm, MW power� 0 W, O2 flow � 30 SCCM, and AR flow � 5 SCCM.
Table 4. Enhancement of Optical Losses by Annealing
In this paper rf sputtering of three different two-oxidemixtures have been investigated for the optical thin-film technologies and proven to provide a relativelywide range (1.42–2.45) for the real refractive indexwhen one mixes silicon dioxide with either titania ortantalum pentoxide. The 1.42 real refractive index cor-responds to silica film, which is not a waveguiding filmwhen deposited on a BK7 glass substrate with index1.51, whereas the 2.45 real refractive index corre-
sponds to titania films. Titania–tantalum pentoxidefilms produced a smaller range with a relatively highreal refractive index in all mixtures. We also concludethat TiO2–SiO2 films produce, with the assistant ofMW power, both a wide range of real refractive indi-ces (1.42–2.45) and relatively lower optical losses0.5 –1.6 dB/mm at the 632.8-nm wavelength, whichare expected to be much lower at the 1550-nm com-munications operating wavelength.
Annealing, as expected, enhances the optical prop-erties of the deposited films. Measuring the waveguideoptical losses before and after annealing confirmed thiseffect. The enhancement was further confirmed whenone compares the resonance width in the transmissionspectra of the optical resonant filters before and afterannealing. Annealing did not show significant opticalproperty enhancement for films made of a TiO2–SiO2mixture in which MW power was used for ionizing theoxygen during deposition, but it did show a healingeffect against the degradation and the instability thatis inherited from the sputtering process and the ex-posure to relatively high humidity.
This research was funded by the National ScienceFoundation grant ECS00096800. The authors thankDaniel Durisin, Michel Krause, and Gregory Aunerfor their help and support.
References1. K. Okimura and A. Shibata, “Deposition of high quality TiO2
films by RF magnetron sputtering with an auxiliary perma-nent magnet,” Jpn. J. Appl. Phys. 36, 4917–4921 (1997).
2. C. Tosello, F. Rossi, S. Ronchin, R. Rolli, G. C. Righini, S. Pelli,E. Moser, M. Montagna, M. Ferrari, and C. Duverger, “Erbium-activated silica–titania planar waveguides on silica-on-siliconsubstrates prepared by rf sputtering,” J. Non-Cryst. Solids 248,230–236 (2001).
3. P. Alexandrov, J. Koprinarova, and D. Todorov, “Dielectricproperties of TiO2 reactivity sputtered from Ti in an RF mag-netron,” Vacuum 47, 1333–1336 (1996).
4. T. Yamagauchi, H. Tamura, S. Taga, and S. Tsuchiya, “Interfa-cial optical absorption in TiO2�SiO2 multilayer coatings pre-pared by rf magnetron sputtering,” Appl. Opt. 25, 2703–2706(1986).
5. K. G. Geraghty and L. F. Donaghey, “Preparation of suboxidesin the Ti-O system by reactive sputtering,” Thin Solid Films40, 375–383 (1977).
6. E. N. Farabaugh and D. M. Sanders, “Microstructure of dielec-tric thin films formed by e-beam coevaporation,” J. Vac. Sci.Technol. A 1, 356–359 (1983).
7. B. J. Pond, J. I. DeBar, C. K. Carniglia, and T. Raj, “Stressreduction in ion beam sputtered mixed oxide films,” Appl. Opt.28, 2800–2805 (1989).
8. A. V. Osinsky, R. A. Bellman, I. A. Akwani, P. A. Sachenik, S. L.Logunov, and J. W. McCamy, “Optical loss mechanisms inGeSiON planar waveguides,” Appl. Phys. Lett. 81, 2002–2004(2002).
9. R. Rabady and I. Avrutsky, “Reliable fabrication technologies foroptical resonant filters,” Appl. Opt. 42, 4499–4504 (2003).
10. R. Swanepoel, “Determination of the thickness and opticalconstants of amorphous silicon,” J. Phys. E 16, 1214–1222(1983).
11. E. E. Khawaja, “The determination of the refractive index andthickness of a transparent film,” J. Phys. D 9, 1939–1943 (1971).
12. L. Ward, The Optical Constants of Bulk Material and Films,2nd ed. (Institute of Physics, London, (1994).
Fig. 7. (a) Setup for the characterization of the optical resonantfilter. (b) Transmission spectra of an optical resonant filter that ismade of a TiO2 waveguide and the effect of annealing for twoannealing conditions. (c) Transmission spectra of an optical reso-nant filter that is made of a TiO2–SiO2 waveguide and the effectsof two weeks’ exposure to relative humidity of 80% and annealing.
