Wavelength Selective Switch Using Arrayed Waveguides with Linearly Varying Refractive Index Distribution

Transcription

1 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 (25) pp Wavelength Selective Switch Using Arrayed Waveguides with Linearly Varying Refractive Index Distribution Kazuhiko SHIMOMURA and Yasumasa KAWAKITA Department of Electrical and Electronics Engineering, Sophia University, 7-1 Kioi-cho, Chiyoda-ku, Tokyo , Japan A wavelength selective switch and demultiplexer using arrayed waveguides with linearly varying refractive index distribution is proposed. The thickness of the arrayed waveguides is varied linearly using selective metal-organic vapor phase epitaxy (MOVPE) growth. The dependence on growth conditions and mask pattern of the structural and optical characteristics of the arrayed waveguides are determined. Successful wavelength demultiplexing on four output ports with a wavelength separation of approximately 25 nm is realized using input and output star couplers based on Rowland circle geometry. Finally, wavelength selective switching by dynamic refractive index control of the arrayed waveguides is described. KEYWORDS: wavelength selective switch, deflector, demultiplexer, arrayed waveguides, refractive index variation, MOVPE, selective area growth, GaInAs/InP, MQW 1. Introduction Wavelength division multiplexing (WDM) is currently deployed in high-capacity, long-haul fiber-optic transmission systems to support multiple high-speed channels. WDM takes advantage of the enormous bandwidth offered by optical fiber while allowing individual wavelength channels to be utilized at bit rates suited to low-cost electronic components. In addition to the large bandwidth, WDM all-optical networks provide high-speed data transmission without need for intermediate-node electrical converters and are transparent with respect to the data formats used 1-3). Implementing WDM functions in an all-optical layer network requires versatile, reliable and highperformance optical devices, particularly wavelength selective switches such as optical cross connects (OXCs) and optical add drop multiplexers (OADMs). Such devices comprise a small number of components, including the multiplexer/demultiplexer (MUX/DEMUX), various switches and other functional devices. The arrayed waveguide grating (AWG) has become increasingly important in these types of channel-selective routing devices for WDM signals 4-7). The AWG is an imaging device that disperses the image field of an input waveguide onto an array of output waveguides. The desired diffraction properties are obtained through linear variation in the length of the arrayed waveguides. Waveguide-type selective switches using AWGs have been demonstrated on silica planar lightwave circuits (PLCs) 8-17) and InP semiconductor substrates 18,19). On PLCs, AWGs have been integrated with 2x2 switches 14), a phase shifter 15), a Mach-Zehnder interferometer (MZI) 16) and 1x2 switches 17). Integrated AWG and MZI switches have also been fabricated on InP substrates 18,19). While the AWG is the key component in the wavelength selective switches mentioned above, these devices consist primarily of a few AWGs and switches or phase shifters, making them costly and large. We recently proposed a novel demultiplexer using a GaInAs/InP multiple quantum well (MQW) arrayed waveguide in which the refractive index varies linearly across the device [P9,P4]. The operational principles of this device resemble a conventional AWG wavelength demultiplexer. However, conventional AWGs are designed so that the phase differences between adjacent waveguides are obtained by gradually varying the waveguide length across the array. As such, the dimensions of the device grow as the number of channels increase. In the design proposed in this study, phase differences between adjacent waveguides are obtained by varying the refractive indices of the waveguides. The output ends of the waveguides act as a diffraction grating. Different wavelengths of light are subjected to different phase shifts in the waveguides, resulting in positional differences in light diffraction peaks and allowing wavelength demultiplexing to be realized. The proposed device has straight waveguides. Differences in the lengths of the waveguide optical paths are not a function of the waveguide geometry, therefore, increasing the number of channels does not require a significant increase in device size. A further advantage of the proposed device is its potential for monolithic integration into semiconductor active devices, such as laser diodes (LD), photodetectors (PD) and semiconductor optical amplifiers (SOA), to compensate for propagation losses inherent in such devices. In particular, InP-based optical devices are well suited to photonic integration for optical communications at the 1.55 m wavelength. Additionally, because of its capability for dynamic control of the refractive indices of waveguides in the array through the application of an electric field to the MQW structures 2-22), this device can be applied to wavelength selective switches [P2,P9]. And because the proposed device itself can provide such dynamic functionality, the integration of other functional devices becomes unnecessary. To realize a linearly varying refractive index profile [P1,P3,P1], a selective area growth technique 23,24) was used to fabricate the GaInAs/InP MQaveguide array by metal-organic vapor phase epitaxy (MOVPE). Placing asymmetric widths of an SiO 2 mask on either side of the arrayed waveguide, the waveguide thickness could be varied gradually across the array, and an MQW structure with different band gaps could be formed in a single growth step. In this paper we present a wavelength selective switch and demultiplexer using arrayed waveguides with linearly varying refractive index distribution. We show experimental results demonstrating a four-channel demultiplexer with 341

2 342 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 25 nm wavelength spacing in a GaInAs/InP MQW waveguide. The arrayed waveguides are studied, with a focus on refractive index distribution, through a selective MOVPE growth technique to determine demultiplexing characteristics. Finally, we discuss applications of the proposed device as an optical deflector and wavelength selective switch by applying an electric field to the waveguides. This paper is organized as follows. Section 2 provides a background discussion of arrayed waveguide grating fundamentals for wavelength multiplexing/demultiplexing. Section 3 discusses the structure and operational principles of the proposed arrayed waveguide with linearly varying refractive index variation. Section 4 describes the selective MOVPE growth used with the proposed arrayed waveguides. Section 5 presents the demultiplexing characteristics of the proposed device, and Section 6 discusses an application of the device as a wavelength selective switch and deflector. In Section 7, we present our conclusions. 2. Background In this section, we explain the basic concepts of wavelength demultiplexing in a conventional AWG consisting of an input waveguide, an output waveguide, two focusing star couplers and a phased array of multiple-channel waveguides having a constant optical path length difference L between adjacent array waveguides, as shown in Fig. 1. Star Coupler Input Waveguide Arrayed Waveguides Star Coupler Output Waveguides Fig. 1 Schematic of a conventional AWG demultiplexer. In the conventional AWG, the incoming light wave propagates into the slab waveguide of the star coupler and diverges. It then propagates to the input aperture of the arrayed waveguides and is collected by, and propagates independently into, each arm of the array. Each successive array waveguide has a path length greater than its adjacent waveguide by an integer multiple of the center operating wavelength of the device in the effective medium of the waveguide. Thus, the array has a path length difference between each successive waveguide of m (1) L n a where m is the grating order of the array, is the center operating wavelength of the device and n a is the effective index of the arrayed waveguide. At the center wavelength, waves in each arm of the array arrive with the same relative phase, producing a mirror image at the second star coupler of the field in the first star coupler. Thus, when propagated into the second slab waveguide, waves at the center wavelength converge and focus at the center of the output side of the second star coupler. At other wavelengths, however, as the field propagates along the array waveguides, differences in the waveguide path lengths will cause the phase front to tilt. As a result, waves at the other wavelengths focus at an angle to the path of the center wavelength. Waveguides are positioned at the output of the second star coupler so that the focused fields can be properly confined and guided. The slab waveguides are designed within the context of the star couplers, and the star couplers in the AWG are based on Rowland circle geometry. The central phase matching equation resulting from this system is given by reference 6,7) in equation (2) as n L n d sin m (2) a s x i (3) where n s is the effective index of the slab in the star coupler, d is the separation between the array waveguides, i is the number of output waveguides, x is the separation between the output waveguides and L f is the focal length of the star coupler. To determine the angular dispersion of the system, we differentiate equation (2) with respect to. Assuming, we have L f na m n d L s and near the center frequency a m n d s a (4) n m (5) n n mn m a g dn 1 n (6) dn n s a s a where n g is the group index of the waveguide mode as follows. na n n (7) g a Thus, equation (6) gives us the dispersion relations in terms of wavelength. Channel spacing in terms of wavelength is given by the following.

3 Photonics Based on Wavelength Integration and Manipulation IPAP Books x xdn n s L L mn f From equation (2), we see that the response of the AWG is periodical. After each 2 change in, the field is imaged at the same position. In the frequency domain, this period is referred to as the free spectral range (FSR). The free spectral range can be solved using equation (2) for the order (m+1) channel. Thus, f g a (8) different refractive indices of the waveguides in accordance with the different waveguide thicknesses. Hence, in the proposed arrayed waveguide, a multiplexed optical signal can be separated in terms of each wavelength for wavelength demultiplexing. Arrayed waveguides (1) InP sub. Output waveguides FSR Nf where N is the number of channels and f is the channel spacing in terms of frequency. Using the phase matching equation with the (m+1) 2 channel, and assuming that f ( f FSR) f and n n ( n ), we find the following. g s s (9) c FSR (1) n L g 2 Using f ( c ) and solving for FSR in terms of wavelength, we then have the following. n m n a FSR (11) 3. Structure and Operational Principle g The proposed wavelength selective switch/demultiplexer using arrayed waveguides with linearly varying refractive index distribution is shown in Fig. 2. The device consists of an input waveguide, an input star coupler to distribute input light to each of the arrayed waveguides, an array of equally spaced straight waveguides in which the refractive index varies gradually across the array by the difference in waveguide thicknesses, an output star coupler and the output waveguides. In contrast to a conventional AWG, all waveguides in the array are straight, except in the areas near the star couplers. They are also equal in length and have different refractive indices. If all of the arrayed waveguides had equal refractive indices, the zero th -order diffraction peak would be located at the center of the output waveguide because the wavefronts from each waveguide focus at the center. However, with a variable refractive index arrayed waveguide, the phases of propagating light in each waveguide differ, and the zero th -order diffracted light shifts from the center because of the differing refractive index from one waveguide to another. In the case of conventional AWGs, the phase difference among waveguides is determined by the differences in waveguide lengths. In the proposed arrayed waveguide, the phase difference among waveguides is determined by the [11] (1) InP sub. Input waveguide Input star coupler Output star coupler Fig. 2 Schematic of the wavelength selective switch/demultiplexer using arrayed waveguides with linearly varying refractive index distribution. On the other hand, when an electric field is applied to the arrayed waveguide, each propagating light can be deflected. Even if a voltage is applied uniformly to each arrayed waveguide, the electric field present in the MQW layer differs in each waveguide due differences in the MQW thickness. It is therefore possible to control the refractive index of each arrayed waveguide through the application of an electric field to the arrayed waveguide. As a result, the diffraction angle, and therefore the output position for each wavelength, can be changed dynamically by controlling the refractive index variation. Accordingly, the arrayed waveguide is expected to have wavelength demultiplexing functionality and dynamic switching capability. We first calculated results for wavelength demultiplexing in the variable refractive index arrayed waveguides. The calculation procedure is as follows. In the thickness direction of the waveguide, the equivalent refractive index method is used. A two-dimensional model is employed for the analysis. We assume light is equally distributed from the input waveguide to the waveguides in the array. If there is a refractive index profile, a phase difference proportional to the refractive index difference occurs. We use the coordinate system shown in Fig. 3, where y is the axis perpendicular to the waveguide at the end of the array waveguide, z is the distance from the waveguide end, and y is the axis parallel to the y axis at a distance of z. If the field distribution at the waveguide end plane (z = ) is E(y,), then the field distribution on a plane at a distance z from the array waveguide end plane can be computed by the Fresnel-Kirchhoff diffraction integral as

4 344 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 r E E( y', z) C ( y,) exp( jkr) dy (12) where C is a constant, k is the wave number and 2 2 ( y' z) z. n Arrayed Waveguides y y' Channel spacing [nm] 1 1 N = 16, w = 3 m, d = 3 m n =.5 % 1 % 2 % 5 % 1 z Waveguide length L [m] n Fig. 4 Wavelength channel spacing as a function of arrayed waveguide length L for various n when N = 16, w = 3 m, d = 3 m. Fig. 3 Coordinate system used in calculating the diffraction pattern from arrayed waveguides with linearly varying refractive index distribution. In a practical device, a star coupler is used to focus output light from the array onto the output waveguides based on Rowland circle geometry. In this calculation, however, the design of the star coupler is not taken into consideration. Hence, there is some error when calculating device insertion loss or crosstalk using this procedure. Figure 4 shows wavelength channel spacing as a function of waveguide array length L for the refractive index differences between the two sides of an array of a number n of waveguides. is taken as the difference in wavelength between adjacent channels and is calculated for an overlap of those channels at an intensity of 1/e 2 maximum. In the calculation, the number of arrayed waveguides is N = 16, the width of each arrayed waveguide is w = 3 m and the spacing between arrayed waveguides is d = 3 m. From the figure we can see that can be decreased by increasing waveguide length L and refractive index difference n. As explained in the next section, the refractive index difference in arrayed waveguides, obtained by selective MOVPE growth, was around 1~2%. In this case, for a variable refractive index arrayed waveguide with n = 1.% and L = 5 m, we can expect a channel spacing of = 2 nm. The wavelength selective switching and optical deflection in arrayed waveguide with linearly varying refractive index distribution when an electric field is applied are discussed in Section Selective MOVPE Growth One of the most important processes associated with the proposed device is fabrication of the arrayed waveguides with linearly varying refractive index distribution. Desiring a simple process that minimizes the various waveguide losses such as scattering and radiation, we applied selective metal-organic vapor phase epitaxy (MOVPE) for the formation of the arrayed waveguides. Selective area growth of compound semiconductors by MOVPE on planar patterned substrates is a useful technique that can expand the flexibility of lateral device design and the fabrication of various devices on the same wafer. In selective MOVPE, an optical waveguide can be grown between a pair of SiO 2 mask stripes. The mask width controls the thickness and composition of the grown layer, and optical waveguides with various bandgap energies can be formed in a single growth step by changing the width of the mask stripe. Furthermore, with narrow stripe selective MOVPE, an optical waveguide can be produced directly without need for a subsequent mesa-etching process. Using a dielectric stripe mask oriented in the [11] direction on a (1)-oriented substrate, a low-loss passive optical waveguide surrounded by (1) and (111)B planes with mirror surfaces can be realized. This waveguide also has strong confinement, and radiation loss is lowered even for a curved waveguide. These excellent qualities that are characteristic of advanced waveguide devices can be attributed to the highly controllable and reproducible features of the selective MOVPE technique. In this section, we explain the growth conditions of selective MOVPE and show the structure and optical charac-

5 Photonics Based on Wavelength Integration and Manipulation IPAP Books teristics of arrayed waveguides with linearly varying refractive index distribution. 4.1 Growth conditions of selective MOVPE Figure 5 is a schematic view of the SiO 2 arrayed waveguide mask used in the experiment. The array region consists of eight or 16 waveguides in which the open stripe width is 3 m, the mask width is 3 m and the waveguide length is 1 m. The arrayed waveguide mask has an asymmetric structure. There is a wide-width SiO 2 mask at one side of the array region, hereafter referred to as the wide mask. The wide mask width Ww was varied from 3 to 2 m in the experiment. A 3 nm thick SiO 2 film was deposited by plasma enhanced chemical vapor deposition (PECVD) on a (1)-oriented InP substrate, and arrayed stripes, or windows, orientated parallel to the [11] direction were then formed by standard photolithographic techniques. d w L triethylgallium (TEGa), tertiarybutyl-arsine (tbas) and phosphine (tbp). The V/III ratios were 246 for InP and 2 for GaInAs. The GaInAs/InP MQaveguides consisted of an n- InP buffer layer, a 3-period i-gainas/inp MQW, and a p- InP cladding layer and p-gainas contact layer as shown in Fig. 5. The grown GaInAs layer was lattice-matched on an unmasked region. The cross-sectional profiles of selectively grown waveguides were observed by scanning electron microscopy (SEM). The PL measurements of the arrayed waveguide were carried out at room temperature using 3 m diameter focused laser beams of 68 nm wavelength AlGaAs/GaAs LD. 4.2 Growth enhancement factor As explained in the discussion of operational principle in Section 3, this device has (de)multiplexing functionality through the refractive index differences among the arrayed waveguides. If the refractive index differences are large, then there should be considerable differences in the thicknesses of the arrayed waveguides. Figure 6 shows a cross-sectional SEM image of the arrayed waveguides. From this figure, we can see that the thickness of the waveguides decreases with the distance from the wide mask. SiO 2 Mask p-gainas p-inp 3-period GaInAs/InP MQW SiO 2 n-inp buffer (1) n-inp Subs. Fig. 5 SiO 2 mask pattern used in selective MOVPE growth. Schematic structure of GaInAs/InP MQW waveguide. Selective MOVPE was performed in a vertical reactor under a growth pressure of 6-12 Torr and a temperature of 6-66 o C. The total H 2 carrier gas flow rate was 6 slm. The source materials were trimethylindium (TMIn), (c) Fig. 6 Cross-sectional SEM images of arrayed waveguides. Overview of 16 array waveguides where wide mask is placed at right side. Waveguides adjacent to wide mask. (c) Waveguides on opposite side of wide mask. As one index of the refractive index difference, we introduce a growth enhancement factor. We define growth

6 346 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 enhancement factor t 1 /t N as the ratio of thickness at the two sides of the arrayed waveguides, where t 1 is the thickness of the waveguide next to the wide mask, and t N is the thickness of waveguide at the other side of the wide mask. Table I shows the dependence of the growth enhancement factor on growth temperature and pressure when the number of array waveguides and wide mask width are, respectively, N = 8, = 12 m and N = 16, = 2 m. Under a growth pressure of 6 Torr, the growth enhancement factor increased slightly with a decrease in growth temperature. On the other hand, at the higher pressures of 8 and 1 Torr, the growth enhancement factor increased with an increase in growth temperature. At a low growth temperature of 6 o C, the growth enhancement factor increased as growth pressure decreased. On the other hand, at a temperature of 64 o C, the enhancement factor decreased as growth pressure decreased. Changes in the growth enhancement factor are due to variations in surface migration length and lateral vapor phase diffusion of the metal-organic species as affected by growth temperature and pressure. Table I. Dependence of the growth enhancement factor on growth temperature T and pressure P. N = 8, = 12 m, N = 16, = 2 m. T ( o C) P (Torr) mask width as small as 5 m, the waveguide thickness did not vary linearly. When the wide mask width was increased, however, the slope of waveguide thickness approached a linear distribution resulting in a large variation in waveguide thickness. The equivalent refractive indices in each array waveguide are calculated from the measured thickness of n-inp buffer layer, MQW layer and p-inp cladding layer. In this calculation, the refractive indices of InP and GaInAs are assumed to be constant, and strain effect in the MQW is not taken into consideration. Table II shows the dependence of the equivalent refractive index difference at either side of a waveguide n on wide mask width. We obtained 1.36 % refractive index difference in the array at = 2 m. Waveguide thickness t [m] t InP sub. =5 m Waveguide number =1 m =15 m =2 m T ( o C) P (Torr) Fig. 7 Waveguide thickness across the fabricated arrayed waveguide for various where N = 16, w = 3 m, d = 3 m, with a growth temperature of 64 o C and growth pressure of 1 Torr. When surface migration is short, at low temperature or high pressure, the quantity of growth species supplied from the wide mask decreases in the arrayed waveguide region and the thickness of the waveguide next to the wide mask becomes thinner. On the other hand, when surface migration is long, at high temperature or low pressure, the growth species supplied from the wide mask migrates over the arrayed waveguide region, and differences in thickness of the arrayed waveguides decrease. Figure 7 shows the thicknesses of waveguides across the array for various wide mask widths at a growth temperature and pressure of 64 o C and 1 Torr, respectively. It is clear from this figure that the slope of waveguide thickness was well controlled by the wide mask width. At a wide Table II. Dependence of the equivalent refractive index difference at either side of a waveguide n on wide mask width. (m) n eq (%) Photoluminescence measurement We next show the optical characteristics of the arrayed waveguides as determined by photoluminescence (PL) measurements. Figure 8 shows the PL peak wavelength across the array waveguides for various wide mask widths at a growth temperature and pressure of 64 o C and 1 Torr, respectively. In this figure, we can see that the PL peak wavelength for

7 Photonics Based on Wavelength Integration and Manipulation IPAP Books each waveguide varied in the lateral direction with each wide mask width. As wide mask width was increased, the PL peak wavelength next to the wide mask shifted to a longer wavelength, and the difference in the PL peak between the two sides of the array became larger. A wavelength shift of about 13 nm was obtained when = 2 m. Table III shows the dependence on growth conditions of the PL peak wavelength difference between sides of the waveguide array at various growth temperatures T and pressures P. Table IV shows the full width half maximum (FWHM) of PL for the waveguide array at various growth temperatures T and pressures P. PL peak wavelength [nm] =5 m =1 m =15 m =2 m Waveguide number At a growth pressure of 1 Torr, the PL peak shift decreased and the FWHM widened as the growth temperature was increased. At a growth temperature of 64 o C, the PL peak shift decreased by increasing the growth pressure. The PL peak wavelength difference is related to the refractive index difference of the arrayed waveguides, and the growth enhancement factor reveals the structural waveguide thickness ratio. On the other hand, the PL wavelength difference reflects the change in material composition of the array waveguide, including the strain effect. Hence, a larger difference results in a higher refractive index difference in the array waveguides. The FWHM of the PL indicates the crystal quality of the waveguides, and a narrower FWHM indicates a uniformity of well thickness in the MQaveguides. Considering the above mentioned growth enhancement factor and the optical characteristics of the arrayed waveguides, we chose a growth temperature of 64 o C and growth pressure of 1 Torr for the fabrication of our linearly varying arrayed waveguide device. [1]. [11] - [11] o.7 o Fig. 8 PL peak wavelength across the fabricated arrayed waveguide for various where N = 16, w = 3 m, d = 3 m, at a growth temperature of 64 o C and growth pressure of 1 Torr. 1.3 o 2. o Table III. PL peak wavelength difference between sides of waveguide array at various growth temperatures T and pressures P. T ( o C) P (Torr) nm nm nm nm nm - - (c) 4.7 o 5.4 o Table IV. Full width at half maximum (FWHM) of PL in waveguide array for various growth temperatures T and pressures P. T ( o C) P (Torr) meV 35.6meV 49.3meV meV meV - - Fig. 9 Cross-section of arrayed waveguides tilted with respect to the [11] direction. Waveguides with tilt angles of approximately º and.7º, 1.3 º and 2.º, (c) 4.7º and 5.4º. 4.4 Arrayed waveguides for connection to a star coupler To reduce the channel wavelength spacing associated with the demultiplexing characteristics, it is necessary to increase the number of arrayed waveguides and decrease

8 348 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 the spacing between the adjacent arrayed waveguides. However, because the waveguide arrangement and orientation of the waveguide axis (tilt) is governed by Rowland circle geometry in the connection to a star coupler, any further reduction of waveguide spacing in the array would cause adjacent waveguides to coalesce. We next investigate direct formation of tilted waveguides with respect to the [11] direction by selective MOVPE. Figure 9 (c) show SEM images of the tilted waveguides, cleaved perpendicular to the [11] direction, i.e., the (11) cleavage plane. The waveguide oriented parallel to the [11] direction had a trapezoidal profile that narrowed in the vertical direction. This waveguide is bound by sidewalls of remarkably smooth and well-defined (111)B planes. By increasing the tilt angle with respect to the [11] direction, a plane having an obtuse angle with respect to the (1) surface appeared and an apex was formed in the sidewalls of the waveguide at the point where the two planes intersect, as shown in Fig. 9. The profiles changed to an inverted trapezoid that widened in the vertical direction, as shown in Fig. 9(c), as the tilt angle was further increased, and adjacent waveguides began to coalesce creating a void between neighboring waveguides. The change in the profile is characterized by the waveguide top width W, the apex-to-apex width W apex (defined as the width from one apex to the apex on the other side of the waveguide), and the apex position h apex (the height from the (1) surface to the apex), as indicated in Fig. 1. W W/W apex Apex position h apex [m] Measured Fitting curve Waveguide angle [deg.] Measured Fitting curve Waveguide angle [deg.] Apex h apex W apex Apex Fig. 11 Dependence of W/W apex on stripe angle with respect to the [11] direction. Dependence on stripe angle with respect to [11] direction of apex distance from (1) surface (h apex ). Fig. 1 Schematic cross-section of tilted waveguide with respect to the [11] direction. Figure 11 shows W/W apex between the stripe angles from 1º. The switch to an inverted trapezoidal waveguide profile occurs at a stripe angle of around 1º. Figure 11 shows the variation for stripe angles from 1º. As the stripe angle increases, h apex increases gradually and eventually saturates as the apexes of the waveguide reach the waveguide top (1) plane. More conspicuous alterations by thicker waveguides are not expected, as the change in h apex was most rapid at the smaller angles. Therefore, the alteration of the waveguide profile is most significant at smaller tilt angles. x As mentioned before, the waveguide spacing at the interface to the star coupler is 2 m, corresponding to an obtuse angle between the waveguide sidewall and the (1) plane of approximately 125º in all cases (i.e., the (111)A plane). Therefore, when the apexes reach the waveguide top (1) plane, the apexes of neighboring waveguides merge creating a buried void between waveguides. Waveguide width enhancement x illustrated in Fig. 1 is defined as x h tan(125 apex 9 ), and the relation between tilt angle (deg.) and hapex is given as h apex (13) where is less than the stripe angle of 1º. To avoid coalescence of adjacent waveguides, the arrayed waveguide spacing d should satisfy the relation d 2x 1.4h.53 apex.46. (14)

9 Photonics Based on Wavelength Integration and Manipulation IPAP Books The tilt angle of the waveguide in selective MOVPE growth should therefore be designed in consideration of the waveguide spacing. 5. Wavelength Demultiplexer In this section, we describe the experimental wavelength demultiplexing characteristics of the fabricated variable refractive index arrayed waveguides. First, we show the fundamental diffraction patterns that are generated by the arrayed waveguides. The patterns were measured at a distance from the edge of the waveguide array in the Fresnel region. The measured sample was cleaved at the end of the waveguide array with no star coupler. The parameters of the arrayed waveguides were N = 16, length of waveguide L = 48 m, waveguide width w = 3 m and waveguide spacing d = 3 m. The input slab waveguide was tapered from 3 m to 96 m over a distance of 15 m in the propagation direction. The photoluminescence (PL) peak wavelength of the grown substrate without patterning was 1439 nm. Figure 12 shows the dependence of the diffraction angle on input wavelength for the arrayed waveguides using various values for wide mask width. The positions of the diffraction peaks differ for different values of since different phase shifts take place in each device. As increases, the dependence of the diffraction angle on wavelength increases because larger refractive index differences are formed in the arrayed waveguides. The difference in the diffraction angle for light at 152 nm and 158 nm for different arrays is summarized in Table V. We obtained 4.4 o diffraction angle difference at = 2 m. Diffraction angle [deg.] = 5 m = 2 m =15m = 1 m Wavelength [nm] Fig. 12 Dependence of diffraction angle on wavelength as measured in the Fresnel region for arrayed waveguides having N = 16, L = 48 m, w = 3 m and d =3 m using various as indicated. Table V. Dependence of the difference in diffraction angle for light at 152 nm and 158 nm on wide mask width. (m) ( o ) From the figure, we can confirm that the diffraction angle depends on the refractive index distribution of the array, which in turn depends on the waveguide thickness, or. This demonstrates that a waveguide array in which the refractive index can be varied across it allows the diffraction properties to be controlled. One important aspect in the design of a practical-use wavelength demultiplexer is the input/output slab waveguide and its connection to the arrayed waveguides. For our device, we need to consider the circular component to focus the light, the position of the arrayed waveguide for highly efficient coupling and the arrangement of the slab waveguide to reduce crosstalk from uncoupled light in the arrayed waveguide. The structure of fabricated wavelength demultiplexer is the same as that shown in Fig. 2. It consists of an input waveguide, an input star coupler, the arrayed waveguides in which the refractive index varies gradually from one side to the other, an output star coupler and four output waveguides. The number of arrayed waveguide is N = 16, the width of the waveguides is w = 3 m and the waveguide spacing was narrowed from d = 3 m to 2 m toward the star coupler. Four output waveguides with 6- m spacing were located at the end of the output star coupler. In the star coupler, incident lights are distributed to the array with an aligned phase front, and output lights from the array are focused on the output waveguides based on Rowland circle geometry. The star couplers are arranged in a curved alignment, and all of the arrayed waveguides between the input and output star couplers have equal lengths, however, the straight arrayed waveguides and star couplers are joined with tilted waveguides. Successful wavelength demultiplexing was demonstrated with this device. Figure 13 shows the near-field pattern (NFP) at wavelengths of = 1518, 1543, 1563, 1588 and 1618 nm, and their intensity profiles are shown in Fig. 13. Every wavelength was found to be spatially separated. Cyclic wavelength output patterns were obtained in = 1518 nm and 1618 nm with an approximately 1 nm free spectral range (FSR) in terms of wavelength. The diffraction orders of 16 was estimated from this FRS. As can be seen in this figure, although the demultiplexed wavelengths were not equally spaced, a wavelength separation of about 25 nm was obtained. In the NFP at = 1543 nm and 1618 nm, unexpected outputs were observed at adjacent ports and crosstalk was slightly worse than the other cases. The degraded performance originates in the nonlinearity of the refractive index distribution in the array and could be improved by optimizing the growth condition. Figure 14 shows the relative transmitted power spectrum in four channels for wavelengths of nm. In this

10 35 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 measurement, output light from channel 1 to 4 was coupled to a single mode fiber, and its power was measured with an optical spectrum analyzer. In the measured device, the PL peak wavelength distributions in the array varied almost linearly in the lateral direction from 1344 to 1469 nm. Every wavelength was spatially separated. Moreover, cyclic output patterns were observed, and an approximately 1 nm FSR in terms of wavelength was obtained. Adjacent crosstalk, defined as crosstalk between neighbouring channels with respect to wavelengths, was between 1 db and 22 db. The degraded performance originates in the non-linearity of the refractive-index distribution in the array and could be improved by optimizing the growth condition. As explained in Section 4, coalescense of the waveguides near the star coupler by the tilted waveguide is also one of the problem for the reduction of crosstalk nm 1543 nm proposed design is suitable for application as a wavelength demultiplexer. Relative Transmitted Power [db] ch1 ch2 ch3 ch Wavelength [nm] Fig. 14 Transmission spectrum of the arrayed waveguides wavelength demultiplexer with linearly varying refractive index distribution nm ch4 ch3 ch2 ch Position [m] nm 1618 nm 1518nm 1543nm 1563nm 1588nm 1618nm Intensity [a.u.] Fig. 13 Wavelength demultiplexing characteristics in arrayed waveguides with linearly varying refractive index distribution. Near field pattern of output waveguides from channel 1 to 4. Output intensity profiles from channel 1 to 4. The results show that although the demultiplexed wavelengths were not equally spaced, an approximately 25 nm wavelength separation was obtained. This proves that the 6. Wavelength Selective Switch and Deflector As explained in the previous section, the linearly varying refractive index arrayed waveguides exhibit a demultiplexing function in which a wavelength-multiplexed optical signal can be separated into its constituent wavelengths. We now consider operation of the proposed device as an optical deflector or wavelength selective switch when an electric field is applied to the arrayed waveguides. The core of the arrayed waveguides is the MQW structure in which the thickness of the quantum well varies. When an electric field is applied to the waveguides, the refractive index variation in the quantum well occurs through the quantum confined Stark effect (QCSE) 21). Because of the differing quantum well thickness, the bandedge wavelength in each waveguide is different. Even with the same voltage applied to all of the arrayed waveguides, the resulting electric field in the MQW layer differs at each waveguide. Hence, the phase change is different for each arrayed waveguide, and the position of the image formed by diffraction shifts according to the phase difference of the propagating light. The magnitude of the shift can be controlled continuously by the applied voltage, therefore, this device can be used as an optical deflector. In addition, light can be deflected through the application of an electric field, allowing the array to function as a switch. By varying the electric field applied to the array, the refractive index of the array waveguides can be controlled. As a result, the diffraction angle, i.e. the position of the diffraction peak for a given wavelength, changes dynamically with the applied electric field. The proposed device is thus expected to offer both wavelength demultiplexing and dynamic wavelength selective switching functionality.

11 Photonics Based on Wavelength Integration and Manipulation IPAP Books Optical deflector Figure 15 shows the variation of the equivalent refractive index of an n-i-p MQaveguide at a wavelength of 1.55 m for different applied voltages, where the quantum well thickness ranges from 6. to 6.6 nm. Interband and exciton transitions were considered in calculating the variations in the refractive index by the applied electric field. When no voltage is applied, the equivalent refractive indices vary linearly according to the thickness differences of the waveguides. When voltage is applied, the equivalent refractive index differences increase with well layer thickness because of the greater refractive index variation of the well. Figure 16 shows the diffraction intensity distribution of an optical deflector whose array waveguide structure has a waveguide width of 3 m, a waveguide spacing of 2 m, 16 waveguides and a waveguide length of 45 m. The calculation procedure is the same as that in Section 3. From the figure, we can see that this device can control the light position continuously by changing the applied voltage to the arrayed waveguides. In this calculation, no absorption variation accompanying the refractive index variation is taken into consideration. As can be seen from the inset in Fig. 15, the operation wavelength of 1.55 m is longer than that of exciton resonance wavelength, and the absorption change is considered to be small. Equivalent Refractive Index Absorption Coefficient (cm -1 ) -1 1 GaInAs/InP MQW Lz=6.6nm m Wavelength (m) Applied Voltage 5.V 2.5V Well Layer Thickness (nm) V = 1.55 m GaInAs/InP MQW Fig. 15 Dependence of equivalent refractive index on well layer thickness as a function of applied voltage. Intensity (a.u.) V 4.7V 4.3V 4.V 3.5V V 3.V w=3m d=2m N=16 L=45m 2.V Angle (deg.) Fig. 16 Deflected light power dependence on angle as a function of applied voltage. 6.2 Wavelength selective switching In the static state, meaning that no electric field is applied to the arrayed waveguides, the position of the output light for various wavelengths is spatially separated through dispersion by the arrayed waveguides due to the linearly changing refractive index of the waveguides. On the other hand, if we apply an electric field to the arrayed waveguides, we can change the position of the output light as explained for the optical deflector. In this calculation, only one input light wavelength is used, and the same voltage is applied to all array waveguides. We then consider the case in which the input lights are wavelengthmultiplexed, and the refractive index of each array waveguide is controlled individually. We can then easily predict the possible output positions of different wavelengths through control of the various refractive index patterns in the arrayed waveguides. We refer to this dynamic change in the spatial position of various wavelengths as wavelength selective switching. Here, we show the results from a numerical calculation of wavelength selective switching in the arrayed waveguides. In the numerical calculation for wavelength selective switching, we consider switching of three multiplexed wavelengths where the waveguide array consists of four waveguides with 3 m width, 3 m spacing and 34 m in length. The refractive index difference between the two sides of the waveguide array is n=.9%, and the refractive index variation in each waveguide changes individually. The calculation results suggest that the output light pattern by wavelength can be categorized as a round type or an exchange type. V Round type When the difference in refractive index variation between neighboring waveguides is constant, the refractive index distribution takes the staircase-like shape shown in Fig. 17. Here, the output position of one wavelength shifts either to the neighboring waveguide or to the other

12 352 Photonics Based on Wavelength Integration and Manipulation IPAP Books 2 side of the waveguide. Figure 17 shows the characteristics of round-type wavelength selective switching where the maximum refractive index variation within the waveguide is n =.5 %, n =.1 %, n =.15 %. As can be seen from the figure, the output position for each wavelength shifts to the neighboring waveguide one after another in steps of.5 % of the refractive index variation. A notable characteristic of the round type is that the refractive index variation is as small as.5 % and is positive. As is well known with QCSE, the field-induced refractive index variation occurs near the band-edge wavelength, and the sign of large refractive index variation is negative. At this wavelength, fundamental absorption is large and we cannot ignore the increased insertion loss of the device. At wavelengths longer than the band-edge, the field-induced refractive index variation becomes small and positive. But this wavelength range is located away from the bandedge, so the absorption effect in this case should be more remarkably reduced than that of a negative wavelength range. The round type characteristics are due to the periodical response characteristics of arrayed waveguide grating. Hence, if the field-induced refractive index variation in each arrayed waveguide is equal, then even if the number of multiplexed wavelengths increases we can obtain the characteristics of the round type in which the output light position shifts to the neighboring waveguide and the order of wavelength positions can be maintained Exchange type Next, we consider the case in which the refractive index distribution does not support the staircase-like shape. In the four arrayed waveguides, we consider the case when the field-induced refractive index variation of array waveguide #3 becomes larger than that of #2, as shown in Fig. 18. Figure 18 shows the exchange-type wavelength selective switching characteristics where the maximum refractive index variation within the waveguide is n =.5 %, n =.1 %, n =.15 %. Comparing the wavelength order in Fig. 17 and Fig. 18, we see that the output light wavelengths on both sides are exchanged with each other regardless of field-induced refractive index variation. Table VI shows the wavelength selective switching pattern for three wavelengths under refractive index variation in four array waveguides. The number in the column for refractive index variation represents the normalized fieldinduced refractive index variation in each array waveguide, and 3 is equal to.5 %. Because there are six combinations of output light patterns for three wavelengths, we can exchange the output light positions of different wavelengths by controlling the refractive index of each array waveguide using either the round type or exchange type. We can therefore obtain all switching patterns, but we must pay attention to the fact that crosstalk, the power in undesired wavelengths, is not consistent for the various switching patterns. In round-type switching, this is caused by the periodical response of the arrayed waveguide grating, hence, the amount of crosstalk is the same even if field-induced refractive index variation occurs. On the other hand, with exchange-type switching, the fieldinduced refractive index variation will destroy the periodicity of the response, hence, crosstalk will increase. As shown in Fig. 18, the undesired power in each port becomes large with an increase in field-induced refractive index variation. Especially for a.15 % refractive index variation, the undesired intensity rises to 1 db in the worst case while the static undesired field intensity is around 25 db. Table VI. Wavelength selective switching pattern with various refractive index variation in arrayed waveguides, where 3 is equal to.5 % refractive index variation. Refractive index variation in 4 arrayed waveguides [3,2,1,] 1.53 m 1.55 m 1.57 m [6,4,2,] 1.55 m 1.57 m 1.53 m [9,6,3,] 1.57 m 1.53 m 1.55 m [3,1,2,] 1.57 m 1.55 m 1.53 m [6,2,4,] 1.53 m 1.57 m 1.55 m [9,3,6,] 1.55 m 1.53 m 1.57 m 7. Conclusion Channel 1 Channel 2 Channel 3 A wavelength selective switch and demultiplexer using GaInAs/InP MQW arrayed waveguides with linearly varying refractive index distribution was presented. Linear variation of the arrayed waveguides was achieved by selective MOVPE growth using an asymmetric SiO 2 mask pattern, where a wide mask was placed at one side of the arrayed waveguides. Arrayed waveguides with a 1.4% refractive index difference and 2 nm bandgap wavelength shifts were obtained in 16 waveguides under a growth temperature of 64 o C and pressure of 1 Torr using a 2 m wide mask width. The transformation of crosssectional profiles of tilted waveguides with respect to the [11] direction in selective MOVPE growth was investigated. The ability to achieve a diffraction angle difference of 4.4 o between wavelengths of 152 nm and 158 nm was confirmed experimentally. Successful wavelength demultiplexing with a wavelength separation of around 25 nm in four output ports was realized with input and output star couplers based on Rowland circle geometry. Further development of dynamic refractive index control is expected to also yield wavelength selective switching and optical deflector functionality using the proposed linearly varying arrayed waveguides. Acknowledgments The authors would like to acknowledge former colleagues T. Kihara, K. Miki and Y. Moriguchi for important contributions to this project. The authors would also like to thank A. Kawai, S. Shimotaya and D. Machida for fabrication and measurements of the device.

13 Photonics Based on Wavelength Integration and Manipulation IPAP Books n n n/3 n/3 n/3 n n/3 n #1 #2 #3 #4 #1 #2 #3 #4 1 n =.5 % 1.53 m 1.55 m 1.57 m 1 n =.5 % 1.57 m 1.55 m 1.53 m -1-1 Intensity [db] Intensity [db] Diffraction angle [deg.] Diffraction angle [deg.] 1 n =.1 % 1.55 m 1.57 m 1.53 m 1 n =.1 % 1.53 m 1.57 m 1.55 m -1-1 Intensity [db] Intensity [db] Diffraction angle [deg.] Diffraction angle [deg.] 1 n =.15 % 1.57 m 1.53 m 1.55 m 1 n =.15 % 1.55 m 1.53 m 1.57 m -1-1 Intensity [db] Intensity [db] Diffraction angle [deg.] Diffraction angle [deg.] Fig. 17 Round-type wavelength selective switching in arrayed waveguides with linearly varying refractive index distribution. Schematic of refractive index change in arrayed waveguides. Switching characteristics for various wavelengths. Fig. 18 Exchange-type wavelength selective switching in arrayed waveguides with linearly varying refractive index distribution. Schematic of refractive index change in array waveguides. Switching characteristics for various wavelengths.