Professional Documents
Culture Documents
5 (2003) 449452
A simple, fast and accurate method of designing directional couplers by evaluating the phase difference of local supermodes
Qian Wang and Sailing He
Centre for Optical and Electromagnetic Research, State Key Laboratory of Modern Optical Instrumentation, Zhejiang University, Yu-Quan, 310027 Hangzhou, Peoples Republic of China, Joint Laboratory of Optical Communications of Zhejiang University, Yu-Quan, 310027 Hangzhou, Peoples Republic of China and Division of Electromagnetic Theory, Alfven Laboratory, Royal Institute of Technology, S-100 44 Stockholm, Sweden E-mail: qian@coer.zju.edu.cn
Received 30 April 2003, accepted for publication 16 June 2003 Published 1 July 2003 Online at stacks.iop.org/JOptA/5/449 Abstract The local supermode formalism and a three-dimensional beam propagation method (BPM) are combined in designing directional couplers. The accumulated phase difference of the two local supermodes in the input or output region is evaluated quickly with the assistance of the BPM. The design method is simple, fast and accurate. A comparison of the present method with some other design methods is given. The design results are veried both numerically and experimentally.
Keywords: Directional coupler, beam propagation method, waveguides, integrated optics
1. Introduction
A directional coupler is a fundamental planar lightwave circuit (PLC) performing the functions of splitting and combining [1]. Combined with optical delay lines, such couplers can be used to construct MachZehnder interferometers (MZIs) [1], interleavers [2] and wavelength division multiplexers/demultiplexers (WDMs) [3]. An actual PLC in a directional coupler should consist of the input region, the central coupling region and the output region (see gure 1) for connecting to bres (or bre arrays) or use in constructing MZIs or WDM devices. The effective index method can reduce the computation time greatly through converting the three-dimensional (3D) model to a two-dimensional model. However, this method is not accurate in the design of directional couplers. Therefore, we have to consider the 3D model in the present paper.
In the design and analysis of directional couplers, the supermode solution [4, 5] has been widely used to calculate the coupling length of two parallel waveguides. In the supermode solution the two parallel single-mode waveguides are treated as a whole, which allows two supermodes (one is symmetric and the other is asymmetric with respect to the middle plane of the two identical parallel waveguides). The coupling between the two waveguides is a result of the interference between the two supermodes as they propagate along the directional coupler. When a PLC is used in an optical communication system, it should be connected to some input and output bres. Therefore, a PLC of a directional coupler consists of the input waveguide region, the central coupling region and the output waveguide region. The separation between the two parallel waveguides (in the central coupling region) increases gradually to e.g. 250 mm in the input and output waveguide regions used to connect the PLC to bres or bre 449
Q Wang and S He
Rin Input 1
Rout
Output
2 A0 x axis z axis L B0
directional coupler as in a conventional local supermode method). This method combines the local supermode formalism with the 3D BPM and can give the length of the central coupling region accurately by simulating the light propagation in the input and output regions just once (if the input and output regions have the same structure, we only need to simulate the light propagation in the input region with a 3D BPM once). We take a 3 dB directional coupler as an example in describing the design procedure, as follows. 2.1. Step 1 Use a 3D BPM to simulate the light propagation in the input region. The initial eld distribution f in at plane 12 for the BPM can be approximated with the eigenmode of one input waveguide (e.g. waveguide 1). It can also be chosen s) (a ) (s) (a ) + )/ 2, where as f in = ( ( in in in and in are the eld proles of the two supermodes at the input position 12. We then use a 3D BPM to simulate the light propagation in the input region and obtain the eld distribution (denoted as f AA0 ) at plane AA0 (the end of the input region). 2.2. Step 2 Evaluate the accumulated phase difference between the two local supermodes in the input region. On the basis of the local supermode theory, one has the following formula at the end of the input region: f AA0 = +
(s) c exp (a ) c exp AA0
y axis
Figure 1. The schematic structure of a directional coupler with an input region, a central coupling region and an output region.
arrays. The coupling between the two parallel waveguides in the central region dominates the splitting ratio of the directional coupling as compared to the coupling in the input and output regions. However, if one neglects the coupling in the input and output regions and uses the coupling length obtained from the supermode solution (in accordance with the two parallel waveguides) directly, the actual splitting ratio will be wrong (as shown below at the end of section 2). Therefore, the local supermode formalism has to be used in order to consider the coupling in the input and output regions (which contain bending parts). In a conventional local supermode method, each small interval along the directional coupler is treated as two parallel waveguides and the corresponding phase difference between the two local supermodes is calculated and accumulated. The relative output powers of the directional coupler are then given by P3 = (1 + cos t )/2 and P4 = (1 cos t )/2, where t is the total phase difference between the two supermodes for the whole directional coupler (including the input region, the central coupling region and the output region). In such conventional methods, one has to calculate the propagation constants of the two supermodes for each small interval (by using e.g. a nite-difference method [6] for the cross-sectional prole of the two 3D waveguides) and this is very time-consuming. Another way to design a directional coupler (as used in most of the existing commercial software) is to obtain the length of the central coupling region through some manual adjustment with a beam propagation method (BPM) [79] around an approximate coupling length (without considering the couplings in the input and output regions) calculated from the supermode solution. With this method one can nd accurately the length of the central coupling region, but one needs to simulate the light propagation with a 3D BPM repeatedly, particularly in the output region.
j
12 AA0
(s) (z ) d z (1)
j
12
(a ) ( z ) d z ,
where (s) (z ) and (a )(z ) are the propagation constants of s) (a ) the two supermodes at position z , ( c and c are the eld proles of the two supermodes at position AA0 (the same anywhere in the central coupling region). From equation (1) and the orthogonality of the two supermodes, one sees that the accumulated phases of the two local supermodes in the input region satisfy
AA0
exp j
12
(s) (z ) d z
(s) c ]
= and
f AA0 [
dx d y
(s) c [
(s) c ]
dx d y
AA0
exp j
12
(a ) ( z ) d z
(a ) c ]
f AA0 [
dx d y
(a ) c [
(a ) c ]
dx d y ,
where the asterisk indicates the complex conjugate. Therefore, the phase difference between the two local supermodes, AA in = 12 0 [ (s) (z ) (a )(z )] d z , can be calculated using the following useful formula: in = arg f AA0 [
(s) c [ (s) c ] dx d y (s) c ] dx d y (a ) c [ (a ) c ] dx (a ) c ] dx
dy dy (2)
f AA0 [
where arg denotes the argument (polar angle) of a complex number. With the above formula we can obtain the accumulated phase difference between the two local supermodes in the input region without calculating the propagation constants of the two local supermodes at every position of the input region (certainly with the assistance of the 3D BPM simulation (one time) in step 1). If the structure of the output region differs from that of the input region, we can carry out a similar evaluation, i.e., simulate the light propagation from the output to position BB0 and obtain the accumulated BB phase difference out = 34 0 [ (s) (z ) (a ) (z )] d z for the output region with a formula similar to equation (2). 2.3. Step 3 Calculate the length of the central coupling region from
(s) (a ) L = (t in out )/(c c ),
(3)
where the total phase difference t is determined by the desired splitting ratio (note that P3 = (1 + cos t )/2 and P4 = (1 (s) (a ) cos t )/2) and c (z ) and c (z ) are the propagation constants of the two supermodes for the two parallel waveguides in the central coupling region. The above formulation is of scalar type (like the BPM used), which is appropriate for the weakly guided buried silica (step-index) waveguides considered in the present paper. In some polarization-sensitive cases, a vectorial (or semivectorial) BPM and supermodes of vector form should be used. However, the above design idea can still be employed. In order to test the above design method, we choose a 3 dB directional coupler (formed by silica-on-silicon buried waveguides) as a design example. The refractive indices of the core and cladding are chosen to be 1.454 and 1.445, respectively. The cross-section of the waveguide is 5 m 5 m and the bending radius for the curved part of the input and output waveguides is Rin = Rout = 20 000 m. The separation distance of the two parallel waveguides in the central coupling region is 7 m and the separation distance between the two input (or output) waveguides is 45 m. The wavelength of the input light is 1.55 m. The eld distribution at plane AA0 is shown in gure 2; it was obtained from a threedimensional BPM. From gure 2 one sees that the coupling has occurred in the input region. The accumulated phase difference between the two local supermodes in the input region is then obtained from equation (2) as in = 0.125 31 (=out ). Here the alternate-direction implicit (ADI) nite-difference BPM is used in the calculation. This numerical method is very efcient and an analysis of its accuracy can be found in e.g. [10]. Equation (3) then gives L = 2420 m for the length of the central coupling region. If one neglects the coupling in the input and output regions, the corresponding length s) of the central coupling region will be L 0 = /[2k0 (n ( e (a ) n e )] = 2880.5 m, where k0 is the wavenumber in vacuum; s) (a ) n( e = 1.448 392 89 and n e = 1.448 258 36 are the effective refractive indices for the two supermodes in the central coupling region. This length differs from our design result by 460.5 m. Thus, one sees that the coupling in the input and output regions cannot be neglected. If we neglect the coupling in the input and output regions in the design and use L 0 directly as the length of the central coupling region, the splitting ratio will become 37.6%:62.4% (instead of 50%:50%).
Relative output
-30
-25
-20
20
25
30
35
Lateral position(m)
Figure 3. The nal output eld prole at the central line (along the x -direction) on plane 34. Relative intensities below 0.5 are not shown in order to give a clearer comparison.
Q Wang and S He
Table 1. The insertion loss of a fabricated 3 dB directional coupler (bre-pigtailed) at each output (unit: dB). Input 1 4 Output 2 4.39 4.33 Output 3 4.55 4.26 Uniformity 0.16 0.07
Table 2. Convergence of in with respect to grid size. Grid size 1.0 0.8 0.5 0.3 0.1 z (m) Calculated value of in 0.125 01 0.125 03 0.125 31 0.125 38 0.125 38
PECVD) and cladding layer have thicknesses of 20, 5 and 15 m, respectively. The measured results for the insertion loss at wavelength 1.55 m at each output waveguide (with the input at waveguide 1 or 2) are shown in table 1. In our experiment, the coupling loss between a bre and a waveguide is about 0.60.7 dB. Thus, it is reasonable that the measured insertion loss for each output waveguide is about 1.21.4 dB larger than the calculated insertion loss (since there are two brewaveguide connections in each light path). From this table one also sees that the uniformity is good.
length of the central coupling region (as done in most of the existing commercial software), each adjustment takes about 190 min in order to simulate the light propagation along the directional coupler one time. If N times of adjustment are needed to nd an accurate length of the central coupling region, the design procedure will take 190 N min. However, with the present design method it takes only about 60 min (since the 3D BPM simulation only needs to be carried out once in the input region for the present symmetric case).
5. Conclusions
We have introduced a simple, fast and accurate method for the design of directional couplers. With the assistance of a 3D BPM, the phase difference of the two local supermodes in the input (or output) region is evaluated quickly and accurately using formula (2). The method is much faster than any conventional local supermode method and the manual adjustment method used in most of the existing commercial software; also it is more accurate than the approximate method mentioned in [1]. The present method can also be modied for the design of asymmetric directional couplers.
Acknowledgment
The support of the WennerGren Foundation is acknowledged.
References
[1] Li Y P and Henry C H 1996 IEE Proc. Optoelectron. 143 26379 [2] Oguma M, Jinguji K, Kitoh T, Shibata T and Himeno A 2000 Electron. Lett. 36 1299300 [3] Oguma M, Kitoh T, Shibata T, Inoue Y, Jinguji K, Himeno A and Hibino Y 2000 Electron. Lett. 37 51415 [4] Gnewuch H, Rom an J E, Hempstead M, Wilkinson J S and Ulrich R 1996 Opt. Lett. 21 118991 [5] Ladouceur F and Love J D 1996 Silica-Based Buried Channel Waveguides and Devices (London: Chapman and Hall) pp 14558 [6] Stern M S 1988 IEE Proc. 135 5663 [7] Li D, Brug H and Frankena H J 1995 Pure Appl. Opt. 4 4758 [8] Scarmozzino R, Gopinath A, Pregla R and Helfert S 2000 IEEE J. Sel. Top. Quantum Electron. 6 15062 [9] M arz R 1995 Integrated Optics: Design and Modeling (Boston, MA: Artech House Publishers) pp 87129 [10] Kawano K and Kitoh T 2001 Introduction to Optical Waveguide Analysis: Solving Maxwells Equations and the Schr odinger Equation (New York: Wiley) pp 165233
452