JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 14, NO. 1, JANUARY 1996

Narrow-Band Vertically Stacked Filters in InGaAlAdInP at 1.5 pm

Sang-Kook Han, Member, IEEE, Ramu V. Ramaswamy, Fellow, IEEE, and Robert F. Tavlykaev

Abstract- A narrow-band wavelength filter in InGaAlAanP has been modeled, fabricated, and tested. A highly asymmetrical, vertically coupled directional coupler operating near the band- edge is formed by a narrow ridge and a wide strip-loaded waveguide. The results of numerical simulation, performed by employing the spectral index method, effective-index method, and a modified coupled-mode theory, are used to fabricate a filter structure with a prescribed filter response. Operation at a center wavelength around 1.5 pm with a bandwidth of 18 d and transfer efficiency of -46-68 % is achieved. Excellent agreement between the designed and measured bandwidth is demonstrated. A multichannel filter device based on an array of individual filters that is capable of extending the usable spectral range is analyzed.

I. INTRODUCTION

AVELENGTH-division-multiplexing/demultiplexing (WDM) provides a powerful means of increasing

the aggregate transmission rate of optical communication networks. By taking advantage of the enormous information capacity offered by the optical fiber, effective rates in the Terabith range are feasible. The most essential component of high-density multichannel WDM networks is an optical cross-connect whose building block is a narrow-band wavelength-selective filter. In this regard, filter devices based on integrated-optic structures in semiconductor materials have been attractive due to their potential for deployment in practical communication systems. Efficient and compact, WDM filters in semiconductors are inherently compatible for on-chip integration with other photonic components (optical source, modulator, optical amplifier, receiver, etc.) as well.

Thus far, a substantial amount of effort has been devoted to developing high-performance filter devices operating at the preferred wavelengths 1.3 pm and 1.5 pm. Grating-assisted directional coupler filters with the center wavelength around 1.5 pm have recently been reported [l]. These filters are rather narrowband (2-3 nm), however, they require complex regrowth processes. On the other hand, an easier to grow vertically stacked filter configuration in InGaAsPAnP [2] has been shown to achieve a narrow bandwidth as well as high transfer efficiency at a center wavelength around 1.3 pm.

Parallel to the development of device schemes, the process of selecting and analyzing materials suitable for mass pro-

Manuscript received Apnl 10, 1995; revised September 22, 1995. S.-K. Han is with the Photonic Devices Laboratory, Hyundai Electronics

Industries Co., Ltd., Kyoungki-do, Korea. R. V. Ramaswamy and R. F. Tavlykaev arc with the the Department of

Electrical Engineering, Photonics Research Laboratory, University of Florida, Gainesville, FL 3261 1 USA.

Publisher Item Identifier S 0733-8724(96)007074.

duction of filters has continued. Recently, there has been an increasing interest in the quaternary InGaAlAs system on account of the feasibility of tailoring the bandgap over a broad range combined with the availability of large electro-optic effects as well as fairly strong material dispersion [3], [4]. This material system is relatively easy to grow lattice-matched to InP at an In content of 0.53 and represents an interesting alternative to InGaAsP. Moderate-loss waveguide structures and other associated devices, such as modulators [3]-[5] and lasers [6] , have already been demonstrated.

In this paper, we present detailed, theoretical and ex- perimental studies of a narrow-band wavelength filter and a filter array designed and fabricated in InGaAlAsAnP for WDM applications. The filters operate at -1.55 pm with a bandwidth of 1.8 nm and the measured transfer efficiency of -50-70%. A highly asymmetric vertically stacked, directional coupler structure comprising a narrow ridge and a wide strip- loaded planar waveguide, both being single-moded, is utilized to achieve the narrow filter bandwidth. An array of filters covering a 110 nm wavelength range is analyzed with ridge width varying from 2 to 6 pm. The structure considered does not require regrowth processes and can easily be integrated with other WDM components.

11. BASIC STRUCTURE

The principle of operation of the vertically coupled structure is illustrated in Fig. l(a). The coupled waveguides are phase- matched at a specific wavelength (center wavelength) so that efficient transfer of light between them occurs if they are close enough to each other. Light coupled into one of them can then be switched to the other after propagating a distance corresponding to the coupling length. The waveguides rapidly fall out of synchronism and the power exchange is suppressed as wavelength is tuned away from the center wavelength. The device spectral response therefore represents that of a notch filter at the bar-state channel and that of a channel-dropping filter at the cross-state channel. The filter bandwidth is inversely proportional to the interaction length and the differential dispersion of the propagation constants of the waveguides [7], the latter being determined by structural asymmetry and material dispersion.

To maximize the structural asymmetry, a narrow ridge waveguide and a wide strip-loaded planar waveguide (low waveguide dispersion) are used. Compared to the structure studied earlier in InGaAsP [2], the strip loading in the present work is much wider than the ridge waveguide, adding to a

0733-8724/96$05.00 0 1996 IEEE

78 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 14, NO. I , JANUARY 1996

(b)

Fig. 1. Wavelength switching in a vertically coupled filter (a) and the device cross section (b). All layers except the substrate are Ino,ssGq.47.yAlyAs with varying A1 composition y. To create waveguiding and to ensure structure asymmetry, the following inequality on the refractive indices of the layers must be met: n i > n3 > nz which leads to the relation y1 < y3 < 9 2 .

larger differential dispersion between the propagation con- stants of the two waveguides. In addition to the waveguide dispersion, the material dispersion of InGaAlAs is utilized to achieve a large differential mode dispersion. The composition of the InGaAlAs layer forming the ridge waveguide was chosen so that the operating wavelength of the filter occurred in the region of the strongest material dispersion, near the bandgap.

111. MODELING

Modeling of the basic structure is aimed at determining the structural parameters ensuring high efficiency at a center wavelength close to 1.55 pm. Numerical simulation comprises the following main steps.

The ridge and strip-loaded waveguides are considered separately, and the propagation constants of their funda- mental modes are evaluated, as a function of wavelength, for a trial set of such structural parameters as dimensions and compositions. The values of the parameters are optimized in order to increase the differential dispersion. Finally, a set of optimum values is determined that defines the two waveguides. Coupling interaction is analyzed by computing the cou- pling coefficients versus the separation between the waveguides for the set of parameters determined in the previous step. The separation is varied until the trade- off between the device length and bandwidth (can be

3)

4)

roughly estimated for a known coupling length and differential dispersion) is reasonable. Evolution of optical power along the structure is then computed in order to obtain more accurate estimates for the required device (coupling) length and the cor- responding center wavelength. The filter response is computed by varying the wave- length of light launched into the strip-loaded waveguide and determining the relative fraction of power trans- ferred to the ridge waveguide as the light paopagates throagh the structure.

To analyze the filter structure, a method applicable to ridge waveguides needs to be employed. For this purpose, we usedl the spectral index (SI) method, which is fairly fast and acceptably accurate [8]. The SI method is employed first to calculate the propagation constant and field profile of the fiindmental mode of the ridge waveguide. The modal properties of the strip-loaded planar waveguide are computed by using the effective index method [9] to separate variables in width and depth directions.

In our calculations, the adjustable parameters are the dimen- sions and compositions (A1 content) of the layers forming the filter structure. The necessary data on the refractive index of In0 s=,Gao 47.yAlyAs is calculated by using Sellmeier's relation with coefficients readily available in the literature [lo]

where

A(y) = 9.689 - 2 . 1 0 8 ~ B ( y ) = 1.590 - 0 . 7 8 3 ~ C(y) = 1102.4 - 1462.5~ + 14341~'

and X is in nanometers. It is essential to minimize the bandwidth for a given device

length. Therefore, we require that the two dispersion curves, crossing at 1.55 pm, have as large a differential dispersion as possible. This can be achieved by maximizing the dispersion of the ridge waveguide while ensuring low dispersion for the strip-loaded waveguide. It follows then that the composition of the ridge waveguide should be selected so that the correspond- ing banclgap wavelength is close to, yet shorter (to avoid strong mode attenuation due to absorption) than the desired operating wavelength. The differential dispersion is also increased by the disparity between the cross sections of the two waveguides. In this connection, it is preferable to use a very compact ridge waveguide; however, a realistic minimum width must also bela consideration. With these conditions in mind, we vary layer dimensions and compositions so as to maximize the diffserential dispersion. It is worth noting that a true optimizzttion aimed at finding the absolute maximum would be extrelmely complicated given the number of variables. Due

cross-relations among the parameters, changing would lead to having to adjust the others in the same center wavelength and coupling length.

Therefoj-e, use of the above conditions was important to reduce the nudber of variables by providing reasonable estimates for

HAN et al: NARROW-BAND VERTICALLY STACKED FILTERS 79

3.358 1.54 1.545 1.55 1.555 1.56

Wavelength (pm)

Fig. 2. Dispersion of the mode index of 3 pm x 0.66 p m ridge and 2.78 pm-thick strip-loaded waveguides.

such parameters as the width and composition of the ridge waveguide.

We arrived at the following set of structural parameters [see Fig. l(b)]: H = 0.66 pm, W = 3.0 pm, D = 2.78 pm, L = 20.0 pm, and G = 0.6 pm. The AI composition was selected to be y = 0.10 (the corresponding bandgap wavelength A, = 1.42 pm), 0.20 (A, = 1.24 pm), and 0.30 (A, = 1.08 pm) for the ridge waveguide, strip-loaded waveguide, and the cladding layer, respectively.

Fig. 2 shows the dispersion of the mode index in the ridge waveguide (3.0 pm wide and 0.66 pm high) and the strip- loaded waveguide (2.78 pm thick). The dispersion curves intersect at ~ 1 . 5 5 pm, the filter’s center wavelength. As seen from Fig. 2, the structure exhibits a significant differential dis- persion (difference of the curve derivatives at the intersection point) resulting from the material dispersion of InGaAlAs and strong structure asymmetry. Tuning of the center wavelength can be achieved via the electro-optic effect in semiconductors [l] which leads to a translation of the dispersion curves and, as a consequence, to a shift of their intersection point.

After the mode index and field profiles of the ridge and strip-loaded waveguides have been calculated, each of the fields is normalized to unit power. An IMSL routine is used to interpolate the mode field of the ridge waveguide between the nodes of the computational mesh. The field profile of the strip-loaded planar waveguide is modeled by cosine functions (of real and imaginary arguments) both in the width and depth direction. The noirnalized field profiles are then used to calculate the coupling coefficient of the coupled-waveguide structure in accordance with a modified coupled-mode theory u11.

Coupling coefficients Kab (coupling from strip-loaded to ridge waveguide) and K b a (coupling from ridge to strip-loaded waveguide) were calculated as a function of the separation S between the two waveguides with the wavelength being a pa- rameter. Fig. 3 shows the results computed for the wavelength 1.55 pm. Given these dependencies, the coupling length (L,) can be calculated using the following relation [l 11

h

r!

450

400

350

300

250

2 0 0

150

100

50 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2 . 1

Separation S (mm)

Fig. 3. Coupling coefficients at 1.55 pm vs. the separation between waveguides with structural parameters being: H = 0.66 pm, W = 3.0 D = 2.78 pm, L = 20.0 p m and G = 0.6 pm.

two Pm,

As seen from Fig. 3 and (2) , a waveguide separation S of -1.5 pm results in a coupling length L, of -5 mm which is reasonably short while ensuring, as shown below, a narrow bandwidth. If this value of S is fixed, the filter structure is uniquely determined. Next, power transfer characteristics between the ridge and strip-loaded waveguide are computed in order to determine the coupling (device) length more accu- rately so as to evaluate the filter bandwidth. Compared to using a simplified expression for the filter bandwidth [7], calculating the power transfer characteristics of the coupler structure as a function of wavelength represents a more rigorous analysis. We use Marcatili’s approach [ 121 to calculate the distribution of optical power between the waveguides as light traverses the structure. If light is input only into the lower strip-loaded waveguide, the optical power in the strip-loaded waveguide (Ps) and the ridge waveguide (Pr) can be expressed as a function of the propagation distance ( z ) as

P , = l - (3)

where c is defined as

c a b + C b a E = 2

C,, = 11: E,(‘) x Ht(p) . i dx dy (6)

and S = (pa - pb)/2(KabKba)li2 is a normalized parameter that characterizes the asynchronism between the guides.

Equations (3)-(6) and the determined spectral dependencies of the propagation constants (Fig. 2) and coupling coefficients provide the basis for computing power transfer between the waveguides. In order to validate the numerical routine devel- oped for this purpose, we verified whether the total power was conserved along the structure. The sum of powers in the two waveguides was computed along the propagation distance in 100 pm increments. The field at the output of each segment served the input field for the next. This process continued over

c I cd - 0.8 - 4 '$ - 5 c & a 0.4 c

0.6

-

0 2 4 6 8 10

Propagation distance (mm)

Fig. 4. direction with wavelength being a parameter.

Evolution of power in the ndge waveguide along the propagabon

a total propagation length of 10 mm. We observed that the total power was conserved with an accuracy of better than 0.1%, confirming the adequacy of the modified coupled-mode theory.

To illustrate the internal mechanism of wavelength switch- ing, the evolution of power in the ridge waveguide along the propagation direction is depicted in Fig. 4 for several values of wavelength. It follows that complete power transfer from the strip-loaded planar to the ridge waveguide can be achieved at X = 1.5504 pm with the approximate device length of 4.6 m. We used this value of the device length to evaluate the full- width half-maximum (FWHM) bandwidth of the filter. Fig. 5 represents the calculated spectral response of the filter, i.e., transfer (coupling) efficiency vs. wavelength. The coupling efficiency is defined as the ratio P r i d g e / [Prldge + Pstrlp-ioaded] where P r l d g e and Pstrlp-loaded are the optical power at the output of the ridge and strip-loaded waveguide, respectively. As seen, a -17 A FWHM bandwidth is expected at the center wavelength 1.5504 pm. As mentioned above, a longer length should result into a narrower bandwidth. For instance, the bandwidth can be narrowed to 14 A for a 5.7 mm-long device with an increased waveguide separation of 1.8 pm for complete switching while the other parameters remaining the same.

Besides the individual filter device described above, we considered an array of several filters formed in the same chip (Fig. 6). This structure would be advantageous, as it is capable of increasing the number of channels for a fixed tuning range of the individual filter. As seen, the width of ridge waveguides varies from filter to filter leading to a set of center wavelengths (channels) available. Since all filters have the same etching depth, this array structure can be fabricated with a single fabrication step. Integration of the filter array with an N x N star coupler would be essential to create a practically viable device for multichannel WDM systems. Fig. 7 shows the center wavelength of a channel as a function of ridge waveguide width. As seen, an array of filters with widths ranging from 2 to 6 pm covers a -110-nm spectral range. It should be noted that the bandwidth of an individual filter tends to increase as the center wavelength shifts to longer wavelengths (wider ridge widths). Thus, there is a trade-

J O U R " - OF LIGHTWAVE TECHNOLOGY, VOL. 14, NO. 1, JANUARY 1996

100

80 h

E

0 1.545 1 . 5 5 0 1 . 5 5 5

Wavelength (pm)

Fig. 5. Filter response centered at 15504 pm with a 17 A bandwidth calculatedl for the following parameters: H = 0 66 pm, W = 3 0 pm, D = 2.713 pm, L, = 20.0 pm, S = 1.5 km, and G = 0.6 pm.

off between the shift in the center wavelength and the filter bandwidth.

IV. EXPERIMENT

A sclhematic of the device structure used in experiments is illustrated in Fig. l(b). Since the ridge and strip-loaded waveguide are closely spaced (1.5 pm separation), care should be taken to avoid coupling of the light into the ridge waveguide when exciting the lower waveguide. Therefore, the ridge waveguides were set back about 2-3 mm away from the input endface to allow endfire coupling of light exclusively into the strip-loaded waveguide.

Undoped In0 s3Gao 47.yAlyAs layers were grown on a semi- insulating InP substrate by MBE. The designed values were used as input parameters for A1 compositions and layer di- mensions. To fabricate a vertically stacked structure, a con- ventional process involving wet etching was used. We used a mixture of H3P04 :H202:H20 (1:l:S) which resulted in a -1.0 p d m i n etching rate that was insensitive to A1 composition. To form the ridge and strip patterns, a two-step etching process was used. First, 20 pm wide and 0.6 pm deep strips were fabricated. Leaving the photoresist on top of the strips, 3 pm wide ridges were patterned and formed at the center of each strip. Even though the same photoresist had been exposed twice to delineate the strips and ridges, both patterns were well resolved. The interaction length of the cleaved sample (whose spectral response is provided below) was 5.0 mm, close enough to the computed value 4.6 mm of the coupling length.

Fig. El shows the experimental setup used in measurements. To char,acterize the fabricated samples, a tunable laser source was used in conjunction with a Ge photodetector. Light was modulated at 1 kHz and a lock-in amplifier was used to improve SNR. Before measuring the spectral response of the device, ]near-field measurements of the lower planar waveguide were performed. They confirmed that the presence of the strip loading ' over the lower waveguide provided two-dimensional mode confinement; single-mode in both the width and depth directions.

HAN et al: NARROW-BAND VERTICALLY STACKED FILTERS 81

Fig. 6. Schematic of an array of wavelength filters with varying ridge widths

1.6;

A 3 I.,,_ 2 1.56;

1.541 k 42 C 1.52; s

1.5- . 1 2 3 4 5 6 7 8 9

Width (pm)

Fig. 7. Center wavelength as a function of width of the ridge waveguide.

When measuring the filter response, the ultranarrow spectral line ( 4 0 0 kHz) of the laser was swept over the range of 1480-1578 nm. The output radiation of the filter was collimated onto the detector with a microscope objective. The optical output from the ridge waveguide was filtered through a small aperture to block the guided beam in the strip-loaded waveguide and the scattered radiation.

v. RESULTS AND DISCUSSION Fig. 9 shows the measured filter response for TE polar-

ization. As seen, a maximum transfer efficiency of about 46% was achieved at the center wavelength 1537.7 nm; for TM polarization, the response centered around 1517 nm. The transfer efficiency -68% was measured in one sample [4]; however, the center wavelength deviated from the designed value by a wide margin. In fact, the coupling efficiency at the center wavelength should be even higher, since the measured value was partly limited by the numerical aperture of the microobjective used to focus outcoupled light onto the photodetector. Due to the tight mode confinement in the ridge

40x

Tunable Laser Camera

\ \ \ \ \ \ \ \

Polarizer l i Aperture

U A Monitor \

U Photodetector - Optical path

Electrical connection - - - Fig. 8. Experimental setup.

waveguide, the measured value of Prldge and, hence, the cou- pling efficiency are more likely to have been underestimated. Another factor limiting the maximum transfer efficiency was the difference between the estimated value 4.6 mm of the coupling length and the 5 mm length of the fabricated device.

The measured spectral response clearly demonstrates the bandpass behavior of the asymmetric directional coupler filter. The measured bandwidth 18 8, is in excellent agreement with the designed value, namely, 17 A, validating the numerical simulation of the device. On the other hand, there is a blue shift of -12 nm of the measured center wavelength from the calculated one. There are a number of possible reasons for this shift: a slight lateral offset between the centers of the ridge waveguide and strip loading, deviations of the structural dimensions from the designed values, or compositional variation of A1 causing fluctuations in the refractive index of InGaAlAs layers, etc.

Among the factors that can be responsible for this shift, fluc- tuations of the refractive index and the structural imperfection

82 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL 14, NO 1, JANUARY 1996

l o 0 I 1530 1534 1538 1542 1546

Wavelength (nm)

Typical filter response with an 18 a bandwidth as measured at the Fig. 9. output of the ridge waveguide.

in fabrication are likely to affect the center wavelength of the filter response the most. If the refractive index of the ridge waveguide is lower than that used in simulation, the center wavelength will be blue shifted. Similar effect may also appear should the refractive index of the lower waveguiding region be larger than the designed value. Inaccuracy in geometrical sizes of the fabricated layers may also add to the observed blue shift due to decreasing propagation constant of the ridge waveguide and/or due to increasing propagation constant of the strip-loaded waveguide. A variation of the center wavelength of N i l nm was observed over several filters fabricated in the same sample. As a numerical example, we provide here the results of calculations showing the maximum allowable deviations in structural parameters that cause a 10 nm shift in the center wavelength

parameter nominal value variation (%) ridge width 3 Pm 0.2 pm (7%)

refractive index 3.459 0.0021 (0.06%) ridge height 660 nm 11 nm (2%)

As seen, a variation in the layer composition has the strongest influence on the center wavelength. It is, therefore, essential to use processes with better controls over the epitaxial growth of layers and subsequent device fabrication in order to reduce deviations of the center wavelength. Besides, they can be effectively compensated if we are able to tune the center wavelength either by current injection or the electro- optic effect. Tunability in the same range (30-60 nm) as that for InGaAsP filters is expected.

VI. SUMMARY In summary, a narrow-band wavelength filter based on a

vertically stacked directional coupler in InGaAlAsAnP has been designed, fabricated, and tested at 1.55 pm. The asym- metry of the structure and the large material dispersion of InGaAlAs were utilized in the design to obtain a narrow filter bandwidth at 1.55 bm. The attained bandwidth of 18

is in excellent agreement with the designed value. A multichannel structure based on an array of individual filters has been analyzed that can provide an extended usable spectral range and convenience in integration with other photonic components of WDM systems.

ACKNOWLEDGMENT

The authors gratefully acknowledge C. G. Fonstad and W. Y. Choi of MIT for providing the MBE-grown samples used in this work.

REFERENCES

R. (S. Alferness, L. L. Buhl, U. Koren, B. I. Miller, M. G. Young, T. L. Koch, C. A. Bums, and G. Raybon, “Broadly tunable InGaAsPAnP buried rib waveguide vertical coupler filter,” Appl. Phys. Lett., vol. 60, _ _ ~

pp. 98C982, 1692. C. \Nu. C. Rolland. F. Sheoherd. C. Larocaue, N. Puetz, and J. M. Xu, “InGa&P/InP vertical direkioni coupler fher with optimally designed wavelength tunability,” IEEE Photon. Technol. Let?. , vol. 4, pp. 457-459, 1993. S. IC. Han, R. V. Ramaswamy, W. Q. Li, and P. K. Bhattacharya, “EflTcient electro-optic modulator in InGaAlAsDnP optical waveguides,” IEEE Photon. Technol. Lett., vol. 5 , pp. 4649, 1993. S. 1C. Han, “Optical filters, modulators and interconnects for optical communication systems,” Ph.D. dissertation, Univ. Florida, 1994. K. ’Wakita, I. Kotaka, 0. Mitomi, H. Asai, Y. Kawamura, and M. Na- gamma, “High-speed InGaAlAslInAlAs multiple quantum well optical modulators,” J. Lightwave Technol., vol. LT-8, pp. 1027-1032, 1990. A. Kasukawa, R. Bhat, C. E. Zah, M. A. Koza, and T. P. Lee, “Very low threshold current density 1.5 pm GdnAs/AlGaInAs graded-index separate-confinement-heterostructure strained quantum well laser diodes gown by organometallic chemical vapor deposition,” Appl. Phys. Lett., vol. 59, pp. 2486-2488, 1991. R. C . Alferness and R. V. Schmidt, “Tunable optical waveguide direc- tional coupler filter,” Appl. Phys. Lett., vol. 33, pp. 161-163, 1978. P. 12. Kendall, P. W. A. Mcilroy, and M. S . Stem, “Spectral in- dex method for rib waveguide analysis,’’ Electron. Lett., vol. 25, pp. 107-108, 1989. R. V. Ramaswamy, “Strip loaded film waveguide,” Bell Syst. Tech. J., vol. 53, pp. 697-704, 1974; R. M. Knox and P. P. Toulios, “Integrated circuits for the millimeter through optical frequency range,” in Proc. MR,i Symp. Submillimeter Waves, 1970, pp. 497-517. M. J. Mondry, D. I. Babic, J. E. Bowers, and L. A. Coldren, “Re- fractive indexes of (AI, Ga, 1n)As epilayers on InP for optoelectrouic appllications,” IEEE Photon. Technol. Lett., vol. 4, pp. 627-630, 1992. S. 1,. Chuang, “A coupled mode formulation by reciprocity and a variational principle,” J. Lighmave Technol., vol. LT-5, pp. 5-15, 1987. E. Marcatili, “Improved coupled-mode equations for dielectric guides,” IEEE J. Quantum Electron,, vol. QE-22, pp. 988-993, 1986.

Sang-Kook Han (M’95) was born in Seoul, Korea, on September 3, 1963 He received the B S. degree in electronics engineering from Yonsei University, Seoul, Korea, in 1986, and the M.S and PhD degrees in electrical engineenng from the University of Flonda, Ganesville in 1988 and 1994, respec- tlvely.

Dunng h s graduate studies, he has been en- gaged in both experimental and theoretical research on several guided-wave ophcal devices in com- pound semconductors, includmg an efficient phase

modulatol, natrow bandwidth wavelength filter devices, and the monolithic integratioji of a laser &ode/intensity modulator. He is currently with Photonics Devices 4esearch Laboratory in Hyunda Electronics Ind. CO Ltd , Ichon-kun, Kyoungy-do, Korea, where he is workmg on hgh speed DFB laser diodes His research interests include the semconductor laser diodes and the WDM devices fdir the apphcahons of high bit rate optical communication systems and mulhwavelength ophcal networks.

HAN et al: NARROW-BAND VERTICALLY STACKED FILTERS

Ramu V. Ramaswamy. (M’62-F’90) received the M.S. and Ph.D. degrees in electrical engineering from Northwestern University, Evanston, IL, in ~ 1962 and 1969, respectively. His doctoral work consisted of wave-propagation studies in semiconductor plasmas.

From 1962 to 1965, he served as a Member of the Research Department of Zenith Radio Corporation, Chicago, IL, working on solid-state parametric amplifiers and microwave components. In 1969, he joined Bell Laboratories, Crawford Hill

Laboratory, Holmdel, NJ, where he was engaged in research on integrated- optic devices, polarization effects in single-mode fibers, and fiber-waveguide couplers. Since 1981, he has been a professor in the Department of Electrical Engineering, University of Florida, Gainesville, where his current interests include passive and active integrated-optical devices, wavelength-selective filters, and phenomena in ferroelecrtric crystals. He spent a year at the University of Tokyo’s Research Center for Advancement of Science and Technology (RCAST) during 1991-1992 as the NTT Chair Professor.

Dr. Ramaswamy is the founder of Advanced Photonics Technology, Inc. which specializes in R&D in the area of photonics and guided-wave optics.

83

Robert F. Tavlvkaev received the M.S. degree I

with honors from the Moscow Institute of Physics and Technology and the Ph.D. degree, both in physics, from the General Physics Institute of the Russian Academy of Sciences, Moscow in 1984 and 1993, respectively. His Ph.D. dissertation was on the optimization of performance of an integrated- optic Mach-Zehnder modulator for analog optical systems.

In 1993, he joined the Photonics Research Lab- oratory as a research associate in the Department

of Electrical Engineering, Univ&ity of Florida, Gainesville. Since 7994, he has also been employed by Advanced Photonics Technology Inc., Gainesville, where he is the director of technical development. His research interests in- clude integrated and fiber optics, optoelectronics, optical sensors, modulators, and filters. He has published approximately 30 publications and conference papers and holds two patents.