A long InGaAsP/InP waveguide section with small dimensions

(1)

1112 IEEE PHOTONICS TECHNOLOGY LE?TERS, VOL. 4. NO. 10, OCTOBER 1992 h N 0

F

m -9 c -1 b -1 x v, a, 0 m 0 a, Q CI) c .-

-

t h

i

Frequency Offset (Hz)

Fig. 4. AM noise spectra of the FRASL operating in (a) Non-SSFS-sup pression state at 0.4 W pump level and (b) SSFS-free state at 0.18 W pump level at the fixed oscillation wavelength of 1.418 p m . Trace (c) is the noise spectrum of Nd:YAG pump laser.

state is determined by the requirement of SSFS suppres- sion in the fiber cavity [7], and low-noise 400 fs optical pulses with a white AM noise level of -120 dBc/Hz has been generated from our compact FRASL system.

REFERENCES

J . D. K a k a and T. Baer, “Fiber Raman soliton laser pumped by a Nd:YAG laser,” Opt. Lett., vol. 12, pp, 181-183, 1987.

A. S . Gouveia-Neto, A. S. L. Gornes, and J. R. Taylor, “Femtosec- ond soliton Raman generation,” IEEE J. Quantum Electron., vol.

M. N. Islam, G. Sucha, I. Bar-Joseph, M. Wegener, J. P. Gordon, and D. S. Chemla, “Broad bandwidths from frequency-shifting solitons in fibers,” Opt. Lett., vol. 14, pp. 370-372, 1989. U. Keller, K. D. Li, M. Rodwell, and D. M. Bloom, “Noise characterization of femtosecond fiber Raman soliton lasers,” IEEE

J. Quantum Electron, vol. QE-25, pp, 280-288, 1989.

E. J. Greer, D. M. Patrick, P. G. J. Wigley, J. I. Vukusic, and J. R. Taylor, “Tunable, femtosecond soliton generation from amplified continuous-wave diode-laser signals,” Opt. Lett., vol. 15, pp.

133-135, 1990.

J. P. Gordon, “Theory of the soliton self-frequency shift,” Opt.

Lett., vol. 11, pp. 662-664, 1986.

M. Ding and K. Kikuchi, “Realization of femtosecond soliton oscillation in all-fiber Raman laser with soliton self-frequency shift supprcssion,” IEEE Photon. Technol. Lett., vol. 4, Aug. 1992. M. Ding and K. Kikuchi, “Analysis of soliton transmission in optical fibers with the soliton self-frequency shift being compen- sated by distributed frequency dependent gain,” IEEE Photon.

Technol. Lett., vol. 4, pp. 497-500, May 1992.

R. H. Stolcn and E. P. Ippen, “Raman gain in glass optical waveguides,” Appl. Pliy.~. Lett., vol. 22, pp. 276-278, 1973. QE-24. pp. 332-340, 1988.

A Long InGaAsP/InP Waveguide Section

with Small Dimensions

P.

Verboom,

Y. S.

O e i ,

E.

Pennings,

M.

Smit,

J.

van Uffelen,

H.

van Brug,

I.

M o e r m a n ,

G.

Coudenijs, a n d

P.

Demeester

Y I

type waveguide with 12.4 m m length on a device area of 1 x 1 mm’. Insertion loss was measured to be 4 dB for TE-polariza- tion and 4.5 dB for TM polarization (at

dB/cm. Beaumont e t i l . [2] reported at the same confer- ence a single-turn folded arsenic-doped silica waveguide with a length of 2 cm and device dimensions 8 X 13 mm2, CLm wavelength).

INTRODUCTION

I

with a potential to integrate 240 mm on the same surface. The device measured 2.2 dB insertion loss. In this letter we present a 12.4 mm long InGaAsP/InP waveguide spirally folded on a device area of 1 X 1 mm2, with an insertion loss of 4 dB for TE-polarization and 4.5 dB for TM-polarization, measured at l .55 p m wavelength. ONG waveguide sections find applications in ex-

tended cavities and Mach-Zehnder interferometer structures. Raybon et al. [l] recently reported a mode- locked laser with a monolithically integrated 4.2 mm long

L

DESIGN Manuscript received May 4, 1992; revised July 6, 1992. G. Coudenijs

P. Verboom. Y. S. Oei. E. Penninps. M. Smit. and J. van UffeIen art: was supported by IWONL.

Fig. 1 shows an optical micrograph of the device. The

_-

_{- -}

with the Department of Electrical Engineering, Delft University of

Technology, 2600 GA Delft, The Netherlands.

A. van Brug is with the Department of Applied Physics, Delft Univer- sity of Technology, 2600 GA Delft, The Netherlands.

diameter of the outer loop is 1 mm. The spacing between the concentric waveguides in the is lo pm’ The waveguide structure is shown in Fig. 2. It consists of a 0.5 I . Moerman, G; Coudenijs, and P. Demeester are with the Laboratory

of Electromagnetics and Acoustics, University of Gent, 9000 Gcnt, Belgium.

p m Q1.3 layer and a 0.4 p m In@ top layer, in which a ridge is etched with CH,/He reactive ion etching. A n important design parameter is the lateral effective index IEEE Log Number 9202927.

(2)

VERBOOM et al.: InGaAsP/lnP WAVEGUIDE SECTION 1113

Fig. 1. Microscope photograph of the spirally folded waveguide. The diameter of the outer loop is I mm. The photograph also shows two of

the U-shaped waveguides used for determining propagation loss and facet reflection coefficients.

Fig. 2. SEM photograph of the waveguide structure.

contrast, which is controlled by the etch depth. Radiation loss in bent waveguides decreases with increasing index contrast whereas scattering loss due to waveguide edge roughness increases. Further, the bending loss increases rapidly with decreasing bending radius, the curves with the smallest radius dominate the total radiation loss. If at the center of the spiral the waveguide is folded back by 180" in order to leave the spiral into the opposite rotation direction without crossing other waveguides we need two 180" bend sections with R = 200 p m at the center of the spiral. Based on calculations made by Agrawal [3] we expect the crosstalk at waveguide crossings with a crossing angle of 70" to be negligible. Therefore, we decided to apply the structure as shown in Fig. 1, which allows for a much greater bending radius inside the spiral. If a crossing angle of 70" is applied the bending radius of the inner bends can be increased from 200 to 320 p m . With this radius the etch depth required in order to keep radiation loss sufficiently small is considerably reduced and the scattering losses will decrease correspondingly.

We applied a conformal transformation [4] in order to transform the problem of the curved slab waveguide into that of an equivalent straight waveguide with a transformed refractive index profile. The equivalent problem is then solved by approximating the smoothly varying transformed index distribution by a series of piecewise uniform

regions (staircase approximation) and using a transfer matrix method to determine the complex propagation constant in the transformed domain. The angular propa- gation constant of the curved waveguide is directly in- ferred from the latter, the radiation loss follows from its imaginary part. Fig. 3 (solid curve) shows the computed radiation loss for TE-polarization, as a function of the etch depth. Results are presented for the waveguide structure shown in Fig. 2, with 2.2 p m width and a bending radius of 320 p m . This is the smallest radius occuring in our design. Predicted TM-polarized losses are lower due to the higher lateral effective index contrast for this polarization.

The additional scattering loss due to edge roughness was calculated based on the assumption [SI that it is proportional to the normalized field intensity at the waveguide edge E & , / ( / E 2 &), and to the square of the effective dielectric contrast

(N:

- N $ ) 2 . Fig. 3 (dashed curve) shows the predicted dependence of the scattering loss on the etch depth, with an (etch-depth independent) film-loss contribution of 0.5 dB/cm. We calibrated the curve on experimental loss data obtained for straight waveguides as shown in Fig. 2. From Fig. 3 we see that the radiation losses should be negligible if the ridge is etched slightly into the quaternary layer. The corresponding straight-guide scattering loss is expected to be between 2 and 3 dB/cni.

In the curved waveguides the field pattern shifts to the outer edge, and becomes narrower than in an equivalent straight waveguide. In order to reduce field mismatch losses and excitation of higher-order modes, we adapted the straight-waveguide width (2.0 p m instead of 2.2 p m for the curved waveguides), and applied a lateral offset between the waveguide axes in order to compensate for the outward shift of the mode profile in the curved waveguides. At the junctions between (a) straight/R = 320 p m , (b) Straight/R = 500 p m , and (c) R = 320 pm/R = 450 p m waveguides we applied offsets of (a) 0.4 p m , (b) 0.2 p m , and (c) 0.1 p m . Predicted conversion losses at the junctions are negligible.

FABRICATION

The waveguide structure (InP buffer 1.2 p m , Q1.3-layer

0.5 p m , InP top layer 0.4 p m ) was grown on a semi- insulating InP substrate with LP-MOCVD. TE-propagation loss, as measured with the FP-method on waveguides with a very shallow ridge, amounts to 0.5 dB/cm, which is indicative for the quality of the grown layers. Details are described by Moerman et al. [6]. The wave- guide mask was produced with an optical pattern genera- tor, and transferred into image-reversal photoresist with a 4 x reduction mask aligner. The ridge was formed by reactive ion etching with 4 sccm CH, in SO sccm He at a

pressure of 60 mtorr and 0.4 W/cm2 RF power. It was etched approximately 20 nm into the Q1.3 layer (etch depth 0.42 pm). The low CH, contents of the gas mixture employed is chosen in order to reduce polymerization and, consequently, waveguide edge roughness. This im-

(3)

1114 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 4, NO. 10, OCTOBER 1992

I \

0.0 1

0.30 0.40 0.50 etch depth [pml

Fig. 3. Radiation loss (for a R = 320 F m bend) and scattering loss as a function of the etch-depth. Results are computed for the waveguide structure shown in Fig. 2.

provement is paid for with an increased surface roughness, as can be seen from Fig. 2 which, however, has little effect on the mode under the ridge.

EXPERIMENTAL RESULTS

The transmission of our devices was measured with a Fabry-Perot measurement setup. Reflection coefficients were computed according to Buus [7]. For TE-polarization a reflection coefficient of 0.35 was calculated and for TM-polarization 0.21. In order to verify these data experi- mentally for our waveguide structure, we included a series of U-shaped waveguides “chicanes,” as described earlier by Verbeek et al. [8], with in-line input and output guides in our mask design. All U’s had the same bending radii, but different lengths of the straight sections. The total length within a series varied from 9 to 15 mm in steps of

1 mm. Further, a series of straight waveguides were included.

Waveguide attenuation was measured to be approximately 2 dB/cm both for TE- and TM-polarized light, in good agreement with results obtained previously. From the measurement results of the U-bends we inferred the reflection loss by extrapolating the regression line through the measured loss data to zero waveguide length. For TE-polarized light we found a reflection value of 0.34, for TM-polarized light 0.23. These results are close to the predicted values. The reproducibility of the loss measurement data is

*

0.1 dB. The accuracy of the optical attenuation data as inferred from the F P measurement results is estimated to be kO.3 dB.

Based on the computed reflection coefficients we found a spiral loss of 5.5 dB for TE-polarization, and 6 dB for TM-polarization. High loss values measured on defect spirals indicate negligible direct transmission from input

to output waveguides. After subtraction of 1.5 dB loss occuring in the 7 mm long input and output leads, we obtain the following loss figures: 4 dB for TE-polarization and 4.5 dB for TM-polarization. From this loss approximately 2.5 dB may be contributed to scattering loss (1.24 cm x 2 dB/cm). The additional 1.5 dB for TE- and 2 dB for TM-polarization are due to additional bending losses. The radiation loss due to bending is expected to be largest for the TE-polarized mode, because the lateral effective index-contrast is smaller for TE-polarized modes than for TM-polarized ones. The fact that we did not find this dependence in our measurement results indicates that the contribution of the radiation loss to the total loss is small, and that most of the excess loss is due to scattering at the bend edge.

CONCLUSIONS

We realized a spirally folded InGaAsP/InP ridge-type waveguide with 12.4 mm length, corresponding to 4 cm free-space length, on a device area of 1 X 1 mm2. Inser- tion loss was measured to be 4 dB for TE-polarization and 4.5 dB for TM-polarization, most of which is caused by normal propagation loss (2 dB/cm for straight waveguides). Excess bending loss is estimated to be 1.5 dB for TE and 2 dB for TM-polarization.

ACKNOWLEDGMENT

The authors wish to thank Prof. R. Baets and Prof. B. Verbeek for coordinating and stimulating cooperation.

REFERENCES

G. Raybon, P. B. Hansen, U. Koren, B. I. Miller, M. G. Young, M. Chien, C. A. Burrus, and R. C. Alferness, “A monolithic extended-cavity laser with an integrated Bragg reflector for active mode-locking at 8.3 GHz,” in Proc. 17th Eur. Conf Opt. Comm. ECOC’91, Paris, Sept. 9-12, 1991, postdeadline papers, pp. 41-43.

C. J. Beaumont, S. A. Cassidy, D. Welbourn, M. Nield, and A. Thurlow, “Integrated silica optical delay line,” in Proc. 17th Eur.

Conf Opt. Comm. ECOCPI, Paris, Sept. 9-12, 1991, regular

papers, vol. 1, pp. 241-244.

N. Agrawal, L. McCaughan, and S. R. Seshadri, “A multiple scattering interaction analysis of intersecting waveguides,” J. Appl. Phys., vol. 62, no. 6, pp. 2187-2193, 1987.

M. Heiblum and J. H. Harris, “Analysis of curved optical wave- guides by conformal transformation,” ZEEE J. Quantum Electron.,

H. G. Unger, Planar Optical Waveguides and Fibres. Oxford: Clarendon.

I. Moerman, G. Coudenijs, P. Demeester, B. Turner, and J. Craw- ley, “Influence of gas mixing on the lateral uniformity in horizontal MOVPE reactors,” J. Crysf. Growth, no. 107, pp. 175-180, 1991.

J. Buus, “Analytical approximation for the reflectivity of double- heterostructure injection lasers,” ZEEE J. Quantum Electron., vol.

B. H. Verbeek, E. C. M. Pennings, J. W. M. van Uffelen and P. J. A. Thijs, “Fabrication and analysis of low-loss InGaAsP/InP opti- cal waveguides with extremely small bends,” in Proc. 15th Eur. Conf Opt. Comm. (ECOC’89), Sept. 10-14, Gothenborg, Sweden,

vol. 3, paper PDB-9. vol. QE-11, pp. 75-83, 1975.