OFC/NFOEC 2008
a1342_1.pdf JThA23.pdf
InP-Based Arrayed-Waveguide Grating with a Channel Spacing of 10 GHz F. M. Soares1, W. Jiang1, N. K. Fontaine1, S. W. Seo1, J. H. Baek1, R. G. Broeke1, J. Cao1, K. Okamoto1, F. Olsson2, S. Lourdudoss2, and S. J. B. Yoo1 Department of Electrical and Computer Engineering1, University of California, Davis, 95616 Department of Microelectronics and Information Technology2, Royal Institute of Technology, Sweden email:
[email protected] Abstract: We realize a high-precision 10-channel InP-based Arrayed-Waveguide Grating (AWG) with a 10GHz channel spacing. The AWG showed approximately 10dB excess-loss, 10 dB crosstalk, and 8.2 × 6.8 mm2 dimensions. ©2008 Optical Society of America OCIS codes: (130.3120) Integrated optics devices; (130.5990) Semiconductors.
1. Introduction Ultrahigh-density WDM systems require high-resolution wavelength multiplexing and demultiplexing components. Silica-based AWGs have already demonstrated channel separations down to 10 GHz [1] and 1 GHz [2]. However, typical applications require discrete lasers for each channel, and the resulting ultrahigh-density WDM system will include individual temperature control of many independent lasers. For this reason, high-resolution InP-based AWGs are particularly attractive, because they can be monolithically integrated with a mode-locked laser [3] or a multi-wavelength laser array (and a multi-wavelength photodetector array) [4] that simultaneously generates (detects) multiple wavelengths on a single platform with a single TE cooler. Furthermore, InP-based AWGs offers integration with high-frequency electro-optical phase modulation and electro-absoprtive amplitive modulation necessary for optical arbitrary waveform generation (OAWG) [5]. In this paper, we report an InP-based AWG with a channel spacing of 10 GHz. 6.8 mm # channels
8.2 mm
# arms
5 inputs, 10 outputs 31
width of arrayed waveguides 4 µm channel spacing 0.08 nm free-spectral range ∆L
0.96 nm 719.4 µm
grating order 1497 size of free propagation region (FPR) 190 x 655 µm2 bend radius of arrayed waveguides 700 µm input/output wav. pitch at FPR
5.5 µm
arrayed wav. pitch at FPR
5.3 µm
gap between arrayed wav. at FPR
0.8 µm
arrayed-waveguide width at FPR
4.5 µm
a) b) Fig. 1 (a) Layout of the 10GHz channel-spacing AWG, and (b) a table of the main design parameters.
2. Design Fig. 1 (a) shows the layout of the 10-GHz AWG, and (b) a table of the most important design parameters. The AWG contains 5 input waveguides, 10 output waveguides, and 31 array arms with a path length difference ∆L of 719.4 µm between the adjacent arms. The diffraction order at λ = 1.55 µm is 1497, which results in a free spectral range (FSR) of 0.96 nm and a channel spacing of 10 GHz. The gap between the adjacent arrayed waveguides at the edge of the slab region is 0.8 µm. The challenge of realizing high-resolution AWGs lies in the ability to maintain the phase relation between the optical signal propagating through the different array arms, while the ∆L gets increasingly longer. In our case, the path-length difference between the inner- and outer array arm is 2.23 cm and, therefore, it is difficult to maintain the phase relation due to fabrication tolerances, such as variations in the waveguide width [6] as well as variations in the core-layer-composition and thickness. This disruption of the phase relation (i.e. phase errors 978-1-55752-855-1/08/$25.00 ©2008 IEEE
Authorized licensed use limited to: Univ of Calif Davis. Downloaded on March 28, 2009 at 16:00 from IEEE Xplore. Restrictions apply.
OFC/NFOEC 2008 a1342_1.pdf JThA23.pdf
in the array arms) degrades the transmission of high-resolution AWGs and increases the crosstalk. This paper discusses the design and fabrication of a relatively fabrication-error-tolerant 10-GHz AWG. The waveguides consist of buried heterostructure waveguides, which show a higher tolerance to core- width and thickness variations as compared to other types of high-contrast waveguides, such as high-mesa waveguides. In addition, we have chosen a waveguide width of 4 µm (as opposed to the conventional width of 3 µm), in order to decrease the dependence of the propagation constant on the waveguide width [7]. An additional advantage of choosing the wider waveguide width is the lower propagation loss. Although the 4-µm-wide waveguides support three guided modes (fundamental, first-, and second order) shown in Fig. 2a, the second-order mode is filtered when propagating through the curved waveguides (see Fig. 2c). The excitation of the first-order mode (when coupling the optical signal into the chip) can be suppressed by positioning the input fiber at the center of the input waveguide. We have used the box geometry design to move the FPR as far as possible to one side of the AWG and to minimize the amount of curved waveguides. The bends have a large radius-of-curvature (R=700 µm), to prevent first-order mode excitation from straight to curved waveguide transitions. An equal bending radius of R=700 µm was imposed for all the arrayed waveguides, such that any systematic phase errors in the bending waveguides will be constant for all arrayed waveguides. Furthermore, for the 10 GHz AWGs, the ∆L is so large that the two slab regions were crossed (shown in Fig. 1aFig. 1) and reduce the dimension of the AWG. Relatively large excess losses (1.5 dB) at the waveguide bends are possibly due to high-defect density of the waveguide grown on the semiinsulating InP substrates. 2.5
100
Fe-doped InP
y [µm]
1
300
400
500
600
700
800
900
1000
-0.5
0.5 0
InGaAs
-0.5 -1
Q(1.15)
-1.5
InP
-2
bend loss [dB/ 90o]
InP
1.5
-2.5 -3
200
0
2
-1
-1.5
-2
fundamental mode -2.5
1st-order mode 2nd-order mode
-2
-1
0
1
2
3
-3
x [µm] bend radius [µm] a) b) c) Fig. 2 (a) Buried heterostructure waveguide: cross section diagram and mode profiles of the supported modes (solid: fundamental mode, dotted line: first-order mode, and dashed line: second-order mode) (b) SEM picture, and (c) bend loss as function of bend radius for the different modes.
3. Fabrication The waveguide structure consists of a n-type 1.5-µm InP layer, an intrinsic 0.5-µm Q(1.15) waveguide core layer, a p-type 2-µm InP top cladding layer, and a p-type 0.1-µm InGaAs layer, grown on an InP substrate in a metal-organic vapour-phase-epitaxy (MOVPE) reactor. The InGaAs layer was added for testing purposes and does not influence the waveguiding properties. The 4-µm wide waveguides were etched in a Br2/N2 reactive ion beam etcher using a 550-nm SiO2 layer as mask. Subsequently, the SiO2 mask was selectively removed in a buffered hydrofluoric-acid solution. Finally, Fe-doped semi-insulating InP was regrown by low pressure hydride vapour phase epitaxy (HVPE) [8], resulting in the buried hetero-structure waveguide shown in Fig. 2b. 4. Results and Discussion The measurement setup is as follows. Light from a tunable laser source (TLS) is passed through a polarization controller, and launched into one of the inputs of the AWG by coupling in with a lensed fiber. At the output of the AWG, the light is collimated by a microscope objective and passed through a pinhole, and a TE polarizer. After the polarizer, the light is focussed onto a photo detector (PD). The spectral response of the AWG is obtained by measuring the optical power in the PD, while sweeping the wavelength of the TLS. The measurements are calibrated against straight reference waveguides passing through the chip. Fig. 3a shows the spectral response of the 10 GHz channel-spacing AWG for TE polarization. The excess loss of the AWG, with respect to a straight waveguide, is approximately 10 dB with a crosstalk level of 10 dB. The measured spectral response has high error bars due to the high precision required in the tunable lasers. We have also measured the spectral response of the same AWG with an optical vector network analyzer (OVNA) and the results are smoother, as shown in Fig. 3b (top). Furthermore, the OVNA measurement provides, in addition to the intensity response, the relative phase of each wavelength with respect to a certain reference, from which we can calculate the phase errors for each individual array arm. The
Authorized licensed use limited to: Univ of Calif Davis. Downloaded on March 28, 2009 at 16:00 from IEEE Xplore. Restrictions apply.
OFC/NFOEC 2008 a1342_1.pdf JThA23.pdf
intensity and the phase measurement of one output channel is equal to the transfer function of the AWG (for that particular output channel), and the time-domain impulse response of the AWG is calculated by doing an inverse Fourier transform of the transfer function [9]. Fig. 3b (bottom) shows the impulse response that we calculated from the transfer function. Ideally, the AWG phase response to an input impulse should be linear as a function of the arrayed waveguide number, and thus any deviations from linear is equal to the phase error. With this analysis, we have calculated that the root-mean-square (RMS) of the phase errors for the central array arms of our 10-GHz AWG is 0.5 radians (=29 degrees). The measured impulse response (Fig. 3b, bottom) also reveals that four of the central array arms have lower intensity, possibly due to optical losses due to defects in these arrayed waveguides. The phase error profile also shows a parabolic trend, possibly due to the concentric contour of the wafer thickness and composition of a MOVPE grown wafer, which rotates during the growth. -5 -20
Transmission (dB)
-25
e xc e s s lo ss [d B ]
-10
-15
-30
-35
-40
-45
-20
1548.1 1548.2 1548.3 1548.4 1548.5 1548.6 1548.7 1548.8 1548.9 1549 1549.1 Wavelength (nm)
-25
-30 1550.1 1550.2 1550.3 1550.4 1550.5 1550.6 1550.7 1550.8 1550.9
wavelength [nm]
1551
2
-20
0
-30
Phase (rad)
Transmission (dB)
0 -10
-2
-40 -100
-50
0 Time (ps)
50
100
150
a) b) Fig. 3 (a) Spectral response of the 10-GHz AWG for all 10 channels measured by sweeping light from a tunable laser, and (b) the spectral response (top) and the corresponding impulse response (bottom) of the AWG measured with an optical vector network analyzer.
4. Conclusions We realized an InP-based AWG with 10 GHz channel spacing using buried hetero-structure waveguides. The AWG design employed a relatively wide 4µm-waveguide-width and included 31 arrayed waveguides all with an equal bending radius of the same length to minimize phase errors. The size of the AWG was 6.8 mm × 8.2 mm, and the excess loss was 10 dB. Measurements of the transfer function of one of the outputs channels with an optical vector network analyser showed an RMS phase error of 5.0 radians (=30 degrees) for the 31 arrayed wavegude arms. 5. References [1] K. Takada, et al., “480 channel 10 GHz spaced multi/demultiplexer,” Electron. Lett. , vol.35, no.22, pp.1964-1966, 28 Oct 1999. [2] K. Takada, et al., “1-GHz-spaced 16-channel arrayed-waveguide grating for a wavelength reference standard in DWDM network systems,” IEEE J. Lightwave Technol., vol.20, no.5, pp. 850-853, May 2002. [3] C. Ji, et al., “Synchronized Transform-limited Operation of 10 GHz Colliding Pulse Mode Locked Laser,” IEEE Photon. Technol. Lett., vol. 18, no. 4, pp. 625-627, Feb. 2006. [4] R. Nagarajan, et al., “400 Gbit/s (10 channel × 40 Gbit/s) DWDM Photonic Integrated Circuits,” Electron. Lett. , vol.41, no.6, pp.347-349, 17 March 2005. [5] N. K. Fontaine, et al., “32 phase × 32 amplitude optical arbitrary waveform generation,” Opt. Lett., Vol. 32, No. 7, pp. 865-867, Apr. 2007. [6] C. D. Lee, et al., “The role of photomask resolution on the performance of arrayed-waveguide grating devices,” IEEE J. Lightwave Technol., vol.19, no.11, pp.1726-1733, Nov 2001. [7] W. Bogaerts, et al., “Compact Wavelength-Selective Functions in Silicon-on-Insulator Photonic Wires,” IEEE J. Select. Topics Quantum Electron., vol. 12, no. 6, 2006. [8] S. Lourdudoss, et al, “Hydride vapor phase epitaxy revisited,” IEEE J. Select. Topics Quantum Electron., vol.3, no.3, pp.749-767, Jun. 1997. [9] N. K. Fontaine, et al., “Determination of 20 GHz InP AWG Phase Errors by Measurement of AWG Pulse Train,” accepted to LEOS (2007).
This work was supported in part by DARPA/SPAWAR under agreement number N66001-02-1-8937, and by DARPA/DSO OAWG under agreement number HR0011-05-C-0155.
Authorized licensed use limited to: Univ of Calif Davis. Downloaded on March 28, 2009 at 16:00 from IEEE Xplore. Restrictions apply.
