Three-dimensional coupled-wave

analysis for square-lattice photonic

crystal surface emitting lasers with
transverse-electric polarization:
finite-size effects
Yong Liang,1 Chao Peng,1,∗ Kyosuke Sakai,1 Seita Iwahashi,1 and
Susumu Noda1,2
1 Department of Electronic Science and Engineering, Kyoto University,
Kyoto-Daigaku-Katsura, Nishikyo-ku, Kyoto 615-8510, Japan
2 [email protected]

∗ [email protected]

Abstract: We develop a coupled-wave model that is capable of treating

finite-size square-lattice photonic crystal surface emitting lasers with
transverse-electric polarization. Various properties of interest including
threshold gain, mode frequency, field intensity envelope within the device,
far-field pattern, as well as polarization and divergence angle of the output
beam for the band-edge modes are calculated. Theoretical predictions of the
lowest threshold mode and the output beam profile are in good agreement
with our experimental findings. In particular, we show that, contrary to the
infinite periodic case, the finite length of the device significantly affects
surface emission and mode selection properties of the laser device.
© 2012 Optical Society of America
OCIS codes: (140.3430) Laser theory; (160.5298) Photonic crystals.

References and links

1. M. Imada, S. Noda, A. Chutinan, T. Tokuda, M. Murata, and G. Sasaki, “Coherent two-dimensional lasing action
in surface-emitting laser with triangular-lattice photonic crystal structure,” Appl. Phys. Lett. 75, 316–318 (1999).
2. S. Noda, M. Yokoyama, M. Imada, A. Chutinan, and M. Mochizuki, “Polarization mode control of two-
dimensional photonic crystal laser by unit cell structure design,” Science 293, 1123–1125 (2001).
3. M. Imada, A. Chutinan, S. Noda, and M. Mochizuki, “Multidirectionally distributed feedback photonic crystal
lasers,” Phys. Rev. B 65, 195306 (2002).
4. I. Vurgaftman and J. R. Meyer, “Design optimization for high-brightness surface-emitting photonic-crystal
distributed-feedback lasers,” IEEE J. Quantum Electron. 39, 689–700 (2003).
5. D. Ohnishi, T. Okano, M. Imada, and S. Noda, “Room temperature continuous wave operation of a surface-
emitting two-dimensional photonic crystal diode laser,” Opt. Express 12, 1562–1568 (2004).
6. E. Miyai, K. Sakai, T. Okano, W. Kunishi, D. Ohnishi, and S. Noda, “Lasers producing tailored beams,” Nature
441, 946 (2006).
7. M. Kim, C. S. Kim, W. W. Bewley, J. R. Lindle, C. L. Canedy, I. Vurgaftman, and J. R. Meyer, “Surface-emitting
photonic-crystal distributed-feedback laser for the midinfrared,” Appl. Phys. Lett. 88, 191105 (2006).
8. G. Xu, Y. Chassagneux, R. Colombelli, G. Beaudoin, and I. Sagnes, “Polarized single-lobed surface emission in
mid-infrared, photonic-crystal, quantum-cascade lasers,” Opt. Lett. 35, 859 (2010).
9. L. Sirigu, R. Terazzi, M. I. Amanti, M. Giovannini, and J. Faist, “Terahertz quantum cascade lasers based on
two-dimensional photonic crystal resonators,” Opt. Express 16, 5206–5217 (2008).
10. Y. Chassagneux, R. Colombelli, W. Maineult, S. Barbieri, H. E. Beere, D. A. Ritchie, S. P. Khanna, E. H. Linfield,
and A. G. Davies, “Electrically pumped photonic-crystal terahertz lasers controlled by boundary conditions,”
Nature 457, 174–178 (2009).

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15945
 11. L. Mahler and A. Tredicucci, “Photonic engineering of surface-emitting terahertz quantum cascade lasers,” Laser
Photon. Rev. 5, 647–658 (2011).
12. H. Matsubara, S. Yoshimoto, H. Saito, Y. Jianglin, Y. Tanaka, and S. Noda, “GaN photonic-crystal surface-
emitting laser at blue-violet wavelengths,” Science 319, 445–447 (2008).
13. Y. Kurosaka, S. Iwahashi, Y. Liang, K. Sakai, E. Miyai, W. Kunishi, D. Ohnishi, and S. Noda, “On-chip beam-
steering photonic-crystal lasers,” Nat. Photonics 4, 447–450 (2010).
14. M. Plihal and A. A. Maradudin, “Photonic band structure of two-dimensional systems: the triangular lattice,”
Phys. Rev. B 44, 8565–8571 (1991).
15. S. Fan and J. D. Joannopoulos, “Analysis of guided resonances in photonic crystal slabs,” Phys. Rev. B 65,
235112 (2002).
16. H. Y. Ryu, M. Notomi, and Y. H. Lee, “Finite-difference time-domain investigation of band-edge resonant modes
in finite-size two-dimensional photonic crystal slab,” Phys. Rev. B 68, 045209 (2003).
17. M. Yokoyama and S. Noda, “Finite-difference time-domain simulation of two-dimensional photonic crystal
surface-emitting laser,” Opt. Express 13, 2869–2880 (2005).
18. H. Kogelnik and C. V. Shank, “Coupled-wave theory of distributed feedback lasers,” J. Appl. Phys. 43, 2327–
2335 (1972).
19. M. Toda, “Proposed cross grating single-mode DFB laser,” IEEE J. Quantum Electron. 28, 1653–1662 (1992).
20. K. Sakai, E. Miyai, and S. Noda, “Coupled-wave model for square-lattice two-dimensional photonic crystal with
transverse-electric-like mode,” Appl. Phys. Lett. 89, 021101 (2006).
21. K. Sakai, E. Miyai, and S. Noda, “Coupled-wave theory for square-lattice photonic crystal lasers with TE polar-
ization,” IEEE J. Quantum Electron. 46, 788–795 (2010).
22. Y. Liang, C. Peng, K. Sakai, S. Iwahashi, and S. Noda, “Three dimensional coupled-wave model for square-
lattice photonic-crystal lasers with transverse electric polarization: a general approach,” Phys. Rev. B 84, 195119
(2011).
23. K. Sakai, E. Miyai, T. Sakaguchi, D. Ohnishi, T. Okano, and S. Noda, “Lasing band-edge identification for a
surface-emitting photonic crystal laser,” IEEE J. Sel. Areas Commun. 23, 1335–1340 (2005).
24. W. Kunishi, D. Ohnishi, E. Miyai, K. Sakai, and S. Noda, “High-power single-lobed surface-emitting photonic-
crystal laser,” Conference on Lasers and Electro-Optics (CLEO), CMKK1, Long Beach, May, 2006.
25. W. Streifer, D. R. Scifres and R. D. Burnham, “Coupled wave analysis of DFB and DBR lasers,” IEEE J. Quantum
Electron. 13, 134–141 (1977).
26. C. Peng, Y. Liang, K. Sakai, S. Iwahashi, and S. Noda, “Coupled-wave analysis for photonic-crystal surface-
emitting lasers on air-holes with arbitrary sidewalls,” Opt. Express 19, 24672–24686 (2011).
27. M. J. Bergmann and H. C. Casey, “Optical-field calculations for lossy multiple-layer AlGl-1N/InxG1-xN laser
diodes,” J. Appl. Phys. 84, 1196–1203 (1998).
28. D. H. Sheen, K. Tuncay, C. E. Baag, and P. J. Ortoleva, “Parallel implemantation of a velocity-stress staggered-
grid finite-difference method for 2-D poroelastic wave propagation,” Comput. Geosci. 32, 1182–1191 (2006).
29. E. Miyai and S. Noda, “Phase-shift effect on a two-dimensional surface-emitting photonic-crystal laser,” Appl.
Phys. Lett. 86, 111113 (2005).
30. A. Yariv and P. Yeh, Photonics: Optical Electronics in Modern Communications, 6th ed. (Oxford University
Press, 2007).
31. Fundamentally, radiation fields emitted from the center of the laser cavity have similar properties to those emitted
from an infinite periodic structure described in Ref. [22]. Unlike CC air holes, Fourier coefficients (ξm,n ) of
the dielectric function ε (r) for ET air holes are complex numbers. Therefore, the radiation field intensity is
proportional to |ξ−1,0 Rx + ξ1,0 Sx |2 [see Eq. (A4) in Appendix A and Eq. (B11) in Appendix B]. These complex
Fourier coefficient terms multiplied to basic waves may change the phase difference of the waves diffracted
vertically, resulting in a suppression of the destructive interference.
32. H. A. Haus, “Gain saturation in distributed feedback lasers,” Appl. Opt. 14, 2650–2652 (1975).
33. S. H. Macomber, “Nonlinear analysis of surface-emitting distributed feedback lasers,” IEEE J. Quantum Electron.
26, 2065–2074 (1990).

1. Introduction
Two-dimensional (2D) photonic-crystal surface-emitting lasers (PC-SELs) are becoming in-
creasingly important due to their promising functionality and improved performance com-
pared to conventional semiconductor lasers. A number of successful demonstrations underpin-
ning this promise have already been made [1–13]. By utilizing the band edge of the photonic
band structure, single longitudinal and transverse mode oscillation in two dimensions has been
achieved with a large lasing area, enabling high-power, single-mode operation [1, 5]. The out-
put beam of such devices is emitted in the direction normal to the 2D PC plane and has a small
beam divergence angle of less than 1◦ due to the large area of coherent oscillation [5, 6]. Fur-

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15946
thermore, both the polarization and pattern of the output beam can be controlled by appropriate
design of the PC geometry [2,6]. Recent developments of 2D PC lasers have allowed the lasing
wavelength to be extended from the near-infrared regime to the mid-infrared, terahertz, and
blue-violet regimes [7–12]. In addition, we have recently demonstrated the operation of a PC-
SEL with entirely new functionality: on-chip dynamical control of the emitted beam direction,
achieved by using a composite PC structure [13].
Despite the experimental advances that have recently been made in the field of 2D PC-SELs,
theoretical studies on these types of lasers have thus far been limited. An important but unre-
solved issue concerns the mechanisms by which the PC structure determines the output char-
acteristics of the device, thereby limiting progress in optimizing the structural design of de-
vices. Computer simulations based on the 2D plane-wave expansion method (PWEM) [3, 14]
or the finite-difference time-domain (FDTD) method [15–17], can provide valuable information
about the lasing properties of the PC laser cavity. However, there are some inherent limitations
to these computational approaches. The 2D PWEM is only applicable to infinite structures,
and the FDTD method requires substantial computational resources in order to model finite
structures with realistically large areas. A coupled-wave theory (CWT) approach developed by
Kogelnik and Shank [18] can circumvent such limitations. However, extending this approach to
accurately model PC-SELs is a rather challenging task, since realistic PC-SELs require three-
dimensional (3D) analysis. Very recently, we developed a 3D CWT model that affords an exact
analytical treatment of the full 3D structure of typical laser devices and achieves a very accurate
and efficient analysis of the surface emission properties. This theory incorporates the key issues
in modeling surface-emitting-type PC lasers, i.e. the surface emission in the surface normal di-
rection and the in-plane higher-order coupling effects [22], neither of which was appropriately
described in the previous works [4, 19–21].
Nevertheless, in our previous study, we restricted our analyses to infinite periodic structures
for simplicity [22]. To predict and improve the performance of the practical PC-SEL device,
however, it is essential to consider a finite-size structure. For example, following the arguments
of our previous study, two antisymmetric band-edge modes of PCs with circular air holes should
be excited simultaneously without emitting a laser beam because both of their radiation con-
stants (i.e. parameters quantifying the surface radiation loss) are shown to be zero. This is in
contrast to the experiments where a laser with a doughnut-shaped beam pattern operating stably
at one of the two antisymmetric modes was demonstrated [6, 23, 24]. The physical mechanism
of this lasing behavior cannot be explained without considering the finite-size effects of the
laser device, as we will show later on. This is our motivation for the present work, where we
develop a 3D CWT model capable of treating the finite-size laser device by extending our pre-
vious work [22]. Our objective in this paper is not to exhaustively quantify the dependence on
the large number of parameters that determine device behavior. Rather, we focus on clarifying
the underlying physical mechanism of the effects caused by finiteness of the device.
The remainder of this paper is organized as follows. Section 2 describes derivations of the
coupled-wave equations for finite systems. Section 3 presents analysis results and discusses the
finite-size effects. Section 4 concludes with our findings.

2. Coupled-wave equations for finite-size structures

In this section, we will present the derivations of the coupled-wave equations for finite systems.
We shall not give a complete description of the formulation, but highlight only the major dif-
ferences from our previous study. For a more detailed derivation and discussion, please refer to
our previous paper [22].
The schematic structure of PC-SELs is shown in Fig. 1(a), in which the electric field can be
expressed as E(r,t) = E(r)eiω t . By eliminating the magnetic field from Maxwell’s equations,

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15947
 (a) surface emitting region n-electrode
(b)
z 0.300

Frequency (c/a)
y mode D

mode C
x CC 0.295 mode B
n-GaAs sub
n-clad mode A
Active 0.290
p-clad PC
Insulator ET
p-electrode X ← Γ → M

Fig. 1. (a) Schematic structure of square-lattice PC-SEL device with circular (CC) and equi-
lateral triangular (ET) air holes (inset: scanning electron microscope images). (b) A typical
photonic-band structure calculated by 2D-PWEM. There exist four band-edge modes in the
vicinity of the second-order Γ point, which we refer to as modes A, B, C and D, in order of
increasing frequency.

we obtain
∇ × ∇ × E(r) = k02 ñ2 (r)E(r), (1)
where k0 (= ω /c) is the free-space wavenumber, ω is the angular frequency, c is the ve-
locity of light in free space and ñ is the refractive index (a complex number) satisfying
k02 ñ2 (r)  k02 n2 (r) + 2ik0 n0 (z)α̃ (z) [22, 25], where n2 (r) is a periodic function, n0 (z) repre-
sents the average refractive index of the material at position z, and α̃ (z) represents the gain
(α̃ > 0) or loss (α̃ < 0) in each region. Here, we still focus our analysis on the transverse elec-
tric (TE) polarization because the lasing mode has been identified as being a TE mode [23]. For
TE polarization, E(r) = (Ex (r), Ey (r), 0) can be expanded according to Bloch’s theorem

E j (r) = ∑ E j,m,n (x, y, z)e−imβ0 x−inβ0 y , j = x, y, (2)

m,n

and the periodic function n2 (r) can be expanded as

n2 (r) = n20 (z) + ∑ ξm,n (z)e−imβ0 x−inβ0 y . (3)
m=0,n=0

Here, β0 = 2π /a, a is the lattice constant, m, n are arbitrary integers, and ξm,n (z) is the high-
order Fourier coefficient term. We note that ξm,n (z) is zero outside the PC region. For simplicity,
we assume that air holes within the PC region have perfectly vertical sidewalls (tilted case is
discussed in Ref. [26]) such that n20 (z) represents the average dielectric constant of PC and that
ξm,n is independent of z within the PC region. By substituting Eqs. (2) and (3) into Eq. (1) and
collecting all terms that are multiplied by the factor e−imβ0 x−inβ0 y , we obtain
∂2 ∂ Ex,m,n ∂ 2 Ex,m,n
[ + k 2 2
n (z) + 2ik 0 n0 (z) α̃ (z) − n β
2 2
]E x,m,n − 2inβ0 + + mnβ02 Ey,m,n
∂ z2 0 0 0
∂y ∂ y2
∂ 2 Ey,m,n ∂ Ey,m,n ∂ Ey,m,n
− + iβ0 (m +n ) = −k02 ∑ ξm−m ,n−n Ex,m ,n , (4)
∂ x∂ y ∂y ∂x m =m,n =n

∂2 ∂ Ey,m,n ∂ 2 Ey,m,n
[ + k 2 2
n (z) + 2ik 0 n0 (z) α̃ (z) − m β
2 2
]E y,m,n − 2im β0 + + mnβ02 Ex,m,n
∂ z2 0 0 0
∂x ∂ x2
∂ 2 Ex,m,n ∂ Ex,m,n ∂ Ex,m,n
− + iβ0 (m +n ) = −k02 ∑ ξm−m ,n−n Ey,m ,n , (5)
∂ x∂ y ∂y ∂x m =m,n =n
∂ ∂ Ex,m,n ∂ Ey,m,n
[( + ) − iβ0 (mEx,mn + nEy,mn )] = 0. (6)
∂z ∂x ∂y

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15948
Here, as we are considering a finite system, we have retained the spatial derivative terms of Ex
and Ey with respect to x and y. As defined previously [22],
√ the wavevectors can be classified
√ into
√to their in-plane wavenumber, m + n β0 : basic waves ( m + n =
three groups according 2 2 2 2

1), high-order waves ( m + n > 1), and radiative waves (m = n = 0). By assuming a separable
2 2

form of solutions for the fields, we can solve Eqs. (4)–(6) analytically. In the resonant case at
the second-order Γ point shown in Fig. 1(b), the four basic waves can be expressed as [22]

Ex,1,0 = 0, Ey,1,0 = Rx (x, y)Θ0 (z), (7)

Ex,−1,0 = 0, Ey,−1,0 = Sx (x, y)Θ0 (z), (8)
Ex,0,1 = Ry (x, y)Θ0 (z), Ey,0,1 = 0, (9)
Ex,0,−1 = Sy (x, y)Θ0 (z), Ey,0,−1 = 0. (10)

Here, Rx (x, y) and Sx (x, y) represent the amplitudes of basic waves propagating in the +x and
−x directions, respectively, and likewise, Ry (x, y) and Sy (x, y) represent the amplitudes of waves
propagating in the +y and −y directions, respectively. These four basic waves are assumed to
have identical field profiles in the z-direction, denoted by Θ0 (z), which is the same as the field
profile of the fundamental guided mode for a multilayer structure without PCs [25]. We express
the wave equation for the fundamental guided mode in terms of Θ0 (z) as

∂ 2 Θ0
+ [k02 n20 (z) − β 2 ]Θ0 = 0, (11)
∂ z2
where β is the propagation constant, which satisfies β  β0 in the vicinity of the second-order
Γ point. In this work, we calculate β and Θ0 (z) in Eq. (11) by employing the transfer matrix
∞
method [27] and normalize Θ0 (z) as −∞ |Θ0 (z)|2 dz = 1.
In order to obtain the coupled-wave equations, Eqs. (7)–(10) are substituted into Eqs. (4)–(6)
for (m, n) = {(1, 0), (−1, 0), (0, 1), (0, −1)}. Here, without loss of generality, we focus on the
case for which (m, n) = (1, 0). We then only need to consider Eqs. (7) and (5). Substitution of
Eq. (7) into Eq. (5) gives

∂ 2 Θ0 ∂ Rx
[ + (k02 n20 (z) + 2ik0 n0 (z)α̃ (z) − β02 )Θ0 ]Rx − 2iβ0 Θ0
∂z 2 ∂x
= −k02 ∑ ξ1−m ,−n Ey,m ,n . (12)
m =1,n =0

Here, the basic wave Rx is assumed to vary slowly compared to exp(−iβ0 x) so that its second
spatial derivative terms in Eq. (5) can be neglected. Then Eq. (11) is substituted into Eq. (12)
to yield

∂ Rx
(β 2 − β02 )Rx Θ0 + 2ik0 n0 (z)α̃ (z)Rx Θ0 − 2iβ0 Θ0 = −k02 ∑ ξ1−m ,−n Ey,m ,n . (13)
∂x m =1,n =0

The term Ey,m ,n on the right-hand side of Eq. (13) represents all the waves that may couple
to Rx , including basic, high-order, and radiative waves as described above. Specifically, we
express the radiative waves [for which (m , n ) = (0, 0)] as

Ex,0,0 = ΔEx (z), Ey,0,0 = ΔEy (z). (14)

Finally, we can obtain the coupled-wave equation for (m, n) = (1, 0) by multiplying Eq. (13)
by Θ∗0 (z) on both sides and integrating over (−∞, ∞) along the z direction. Three more coupled-
wave equations for (m, n) = {(−1, 0), (0, 1), (0, −1)} can be derived in analogous fashion. We

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15949
write the four coupled-wave equations in the following form:

∂ Rx k2
−i + (δ + iα )Rx = κ2,0 Sx − 0 ξ1,0 ΔEy (z)Θ∗0 (z)dz
∂x 2β0 PC

k2
− 0
2β0 √ ∑ ξ1−m,−n
PC
Ey,m,n (z)Θ∗0 (z)dz, (15)
m2 +n2 >1

∂ Sx k02
i + (δ + iα )Sx = κ−2,0 Rx − ξ−1,0 ΔEy (z)Θ∗0 (z)dz
∂x 2β0 PC

k02
−
2β0 √ ∑ ξ−1−m,−n
PC
Ey,m,n (z)Θ∗0 (z)dz, (16)
m2 +n2 >1

∂ Ry k02
−i + (δ + iα )Ry = κ0,2 Sy − ξ0,1 ΔEx (z)Θ∗0 (z)dz
∂y 2β0 PC

k02
−
2β0 √ ∑ ξ−m,1−n
PC
Ex,m,n (z)Θ∗0 (z)dz, (17)
m2 +n2 >1

∂ Sy k02
i + (δ + iα )Sy = κ0,−2 Ry − ξ0,−1 ΔEx (z)Θ∗0 (z)dz
∂y 2β0 PC

k02
−
2β0 √ ∑ ξ−m,−1−n
PC
Ex,m,n (z)Θ∗0 (z)dz. (18)
m2 +n2 >1

Here, δ = (β 2 − β02 )/2β0  β − β0 = ne f f (ω − ω0 )/c is the deviation from the Bragg condition,
ω0 is the Bragg frequency, ne f f is the effective refractive index for the fundamental guided
∞
mode, and α is the mode loss given by α = βk0 −∞ n0 (z)α̃ (z)|Θ0 (z)|2 dz. κ±2,0 and κ0,±2 are
0
the backward coupling coefficients defined as

k02
κ±2,0 = − ξ±2,0 |Θ0 (z)|2 dz, (19)
2β0 PC

k2
κ0,±2 = − 0 ξ0,±2 |Θ0 (z)|2 dz, (20)
2β0 PC

respectively, which are equivalent to the conventional one dimensional (1D) coupling coeffi-
cients [18]. The second and third terms on the right-hand sides of Eqs. (15)–(18) represent
the out-of-plane and 2D optical couplings, respectively. It should be noted that integrals in
Eqs. (15)–(18), as well as those in Eqs. (19)–(20), extend only over the PC region because the
Fourier coefficient ξmn = 0 outside that range.
As the fields of the radiative waves (ΔEx (z), ΔEy (z)) and the high-order waves (Ex,m,n (z),
Ey,m,n (z)) are unknown, we cannot yet evaluate the right-hand sides of Eqs. (15)–(18). In order
to determine these fields, Eqs. (4)–(6) must be solved for these waves. Details concerning the
solutions of ΔEx (z), ΔEy (z), Ex,m,n (z), and Ey,m,n (z) are shown in Appendix A. Finally, the
coupled-wave Eqs. (15)–(18) can be rewritten in matrix form as
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
Rx Rx ∂ Rx /∂ x
⎜ Sx ⎟ ⎜ Sx ⎟ ⎜ −∂ Sx /∂ x ⎟
(δ + iα ) ⎜ ⎟ ⎜ ⎟ ⎜
⎝ Ry ⎠ = C ⎝ Ry ⎠ + i ⎝ ∂ Ry /∂ y ⎠ .
⎟ (21)
Sy Sy −∂ Sy /∂ y

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15950
Here, C is a 4 × 4 matrix (see Appendix A). By considering a finite laser cavity
area with 0 ≤ x, y ≤ L (L: laser cavity length in both x and y directions) and defin-
ing appropriate boundary conditions, Eq. (21) can be discretized using the staggered-
grid finite-difference method [28]. This creates an eigenvalue problem with eigenvectors
(Rxj,k , Sxj,k , Ryj,k , Syj,k , Rxj+1,k , Sxj+1,k , Ryj+1,k , Syj+1,k , ...)t and normalized eigenvalues (δ + iα )L.

3. Analysis results and discussions

In this section, we will show numerical results of the coupled-wave equations derived above. By
solving the coupled-wave Eq. (21), we can obtain various properties of interest, including mode
frequency, threshold gain, field intensity envelope profile within the device, far-field patterns
(including the polarization profile) of the band-edge modes. The effects of finite size on these
properties will be discussed.
The structural parameters of the laser structure [see Fig. 1(a)] to be studied are shown in
Table 1. The PC layer is embedded inside the multilayer structure, which is assumed to support
only a single guided mode. The average refractive index of the PC layer is given by nav =
f n2a + (1 − f )n2b , where na is the refractive index of air, nb is the refractive index of the
background dielectric material (GaAs), and f is the air-hole filling factor (i.e., the fraction of
the area of a unit cell occupied by air holes). In this work, we focus our analyses on two typical
air-hole shape designs [see the inset of Fig. 1(a)]: circular (CC) and equilateral triangular (ET),
both of which have been studied experimentally in our previous works [6, 24]. We consider a
boundary condition that is defined as

Rx (0, y) = Sx (L, y) = Ry (x, 0) = Sy (x, L) = 0, (22)

which includes fixing the field amplitude to zero at only one of two sides, generalizing a similar
concept for 1D distributed feedback (DFB) laser structrue originally proposed by Kogelnik and
C. V. Shank [18]. Unless otherwise noted in the following examples, the lattice constant a = 295
nm and the air-hole filling factor f = 0.16, the device length L=70 μ m (all these values were
inferred from experimental data). Here, it is important to note that L=70 μ m approximately
corresponds to a 240a × 240a lasing area, which is almost impossible to use 3D FDTD method
to compute.

Table 1. Structural Parameters of the 2D PC-SEL Device

Layer Thickness (nm) Refractive index
n-clad (AlGaAs) 1500 3.307
Spacer (GaAs) 30 3.524
Active (3QW, InGaAs/GaAs) 64 3.553
Spacer (GaAs) 40 3.524
Cladding (AlGaAs) 40 3.279
GaAs 22 3.524
PC 118 nav
GaAs 20 3.524
p-clad (AlGaAs) 1500 3.307

3.1. Threshold gain behavior and field intensity envelope

In finite systems, mode frequency and threshold gain correspond to the real part and imaginary
part of the normalized eigenvalue of Eq. (21), (δ + iα )L, respectively. As an example, we show

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15951
in Fig. 2(i) a plot of the normalized threshold gain (α L) as a function of deviation from the
Bragg condition (δ L) for a PC-SEL with ET air holes. Though a large number of modes exist
due to the numerical calculation, we identify the four band-edge modes (A, B, C and D) using
the techniques described in Ref. [21]. First, we evaluate the field intensity envelope of the
individual modes throughout the laser cavity, which modulates the fast-varying Bloch waves in
the PC lattice and can be determined by [21]

I(x, y) = |Rx (x, y)|2 + |Sx (x, y)|2 + |Ry (x, y)|2 + |Sy (x, y)|2 . (23)

We extract the solutions of interest that have a singled-lobed profile throughout the laser cavity.
Then, we further identify the four band-edge modes by plotting their field distribution patterns
inside a unit cell located at the center of the cavity and comparing these patterns with those cal-
culated by 2D-PWEM. Figure 2(ii) shows the field intensity envelope profiles of the individual
band-edge modes, indicating that energy can be well confined inside the laser cavity. Figure
2(iii) shows field distributions of each mode inside a unit cell located at the center of the cavity.
Modes A and B correspond to the antisymmetric (nonleaky) mode of 1D DFB lasers, whereas
modes C and D correspond to the symmetric (leaky) mode of 1D DFB lasers [29].

(i) 1
A (ii) B A (iii) B
rel. intensity

3
0.5

2.5 C
D
0

2 +
αL

1.5 0
1
C D C D
B -
rel. intensity

1
A 0.5

0.5 0
0 0
y
0 1 1
−10 −5 0 5 10 0.5
0.5
0.5
0.5
δL y (L ) 1 0
x (L ) y (L ) 1 0
x (L ) x

Fig. 2. (i) Normalized threshold gain (α L) as a function of normalized mode frequency

deviation (δ L) for a PC-SEL with ET air holes. The four band-edge modes A-D are in-
dicated by arrows. (ii) Field intensity envelopes of the individual band-edge modes. (iii)
Field distributions inside a unit cell located at the center of the cavity, in which colors and
arrows represent H- and E-fields, respectively. Thick black triangles indicate the air holes.
In the calculations, we use the structural parameters shown in Tab. 1, the air-hole filling
factor f = 0.16, the lattice constant a=295 nm, and the device length L=70 μ m.The laser is
divided into 14 sections for which the eigenvalues (α L and δ L) converge well.

Table 2. Normalized Threshold Gain (α L) of the Four Band-Edge Modes A-D for CC and
ET Air-Hole Shapes ( f = 0.16, a=295 nm, and L=70 μ m)
Shape αA L αB L αC L αD L
CC 0.23 0.52 2.18 2.18
ET 0.43 0.57 2.08 2.01

The normalized threshold gains of modes A-D are shown in Table 2 in which the values for
the CC air-hole case were obtained in the same manner. We note that modes C and D exhibit a
much higher threshold gain compared with modes A and B, which is attributed to the symmetric
nature of their electric-field distributions with respect to the air holes as shown in Fig. 2(iii). As
the lasing action occurs at the mode with the lowest threshold gain (loss), Table 2 indicates that
mode A is favored for lasing for both CC and ET air holes when the device length L = 70 μ m,

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15952
with a large threshold-gain discrimination of over 20 cm−1 against mode B. This is consistent
with the experimental observations [23, 24] where stable single-mode operations at mode A
were demonstrated.
It is important to note that the single-mode lasing behavior at mode A cannot be explained
based on the infinite solutions [22]. For an infinite periodic system with the vertical structure
shown in Table 1, the radiation constants αin f of modes A and B are αin f ,A = αin f ,B = 0 cm−1
for CC and αin f ,A = 12.35 cm−1 , αin f ,B = 0.82 cm−1 for ET air holes, respectively. From
these results, one might expect that both modes A and B are favored for lasing for the CC air
holes because both of their radiation constants (losses) are equal to zero; and for the ET air
holes, mode B lases due to its much smaller radiation constant compared with that of mode
A. Apparently, these predictions are in contrast to our finite analysis results and experimental
findings. We suggest that the in-plane loss of finite structures might have a great impact on the
mode selection of PC-SELs, which will be discussed in detail in the following sections.

3.2. Far-field pattern and polarization

The far-field pattern and polarization of the output laser beam is one of the most unique features
of 2D PC-SELs. As demonstrated in Refs. [2, 6], both far-field pattern and polarization can be
tailored by appropriately engineering the PC geometry. Theoretically, it is possible to calculate
the far-field patterns of 2D PC-SELs based on our 3D CWT model since the the radiation
field emanating from the surface of the finite-size laser structure is formulated explicitly (see
Appendix A). The time-dependent amplitude of the far field Fj=x,y (subscripts denote the x
and y components, respectively), a function of the deflection angle (θx , θy ) from the surface
normal, can be found by taking the Fourier transform of the radiation field and multiplying by
an obliquity factor (cos θx + cos θy − 1) [30],
 L
Fj (θx , θy ,t) ∝ (cos θx + cos θy − 1) ΔE j (x, y, z = dPC /2)eiω t
0
−ik0 (tan θx x+tan θy y)]
·e dxdy, j = x, y, (24)

where ΔE j (x, y, z = dPC /2) is the complex radiation field just above the PC surface [for further
detail, see Eq. (A4) in Appendix A], dPC is the depth of the PC layer, and z = 0 is defined at the
center of the PC layer. Then, the time-averaged far-field intensity (i.e. beam pattern) is given as

B(θx , θy ) = Bx (θx , θy ) + By (θx , θy ) , (25)

where
 T
1
B j (θx , θy ) = |Re[Fj (θx , θy ,t)]|2 dt, j = x, y, T = 2π /ω . (26)
T 0

By calculating the x and y components of the far field using Eq. (26), we can directly evaluate
the effect of the air holes on the polarization of the output beam.
Figures 3(a) and 3(b) show the calculated far-field patterns of mode A (i.e. the lowest thresh-
old mode) for the two air-hole shapes. A doughnut-shaped profile is obtained for CC, whereas
a single-lobed profile is obtained for ET air holes. The insets show the x and y components of
the far field, indicating that CC air holes exhibit an azimuthal polarization, and ET ones exhibit
an almost linear polarization in the y direction. These semi-analytical results closely match our
experimental observations shown in Figs. 3(c) and 3(d) [6, 24]. The beam divergence angle of
all of the far-field patterns is around 1◦ , reflecting the large area of coherent oscillation. The
small side lobes of the calculated far-field patterns are caused by the abrupt termination of the
radiation field at the edges of the laser cavity.

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15953
 Fig. 3. Calculated far-field patterns (FFPs) of mode A (i.e. the lowest threshold mode) for
(a) CC and (b) ET air holes and experimentally observed FFPs for (c) CC and (d) ET air
holes. The insets in (a) and (b) represent the x and y components of the far field. Parameters
used for the calculations are the same as those shown in Fig. 2. The yellow arrows in (c)
and (d) indicate the directions of polarization.

It is interesting to note that the quantity of surface emission for the symmetric CC air-hole
shape is zero when an infinite periodic structure is considered [22], whereas it has a finite value
for a finite periodic structure. This difference can be understood by considering the effect of
finite device length on the electromagnetic field distribution within the device. For an infinite
periodic structure, the light from mode A cannot be coupled to the external system because
perfect destructive interference occurs everywhere due to the antisymmetric field distribution
pattern with respect to the perfectly symmetric air holes. On the other hand, the mode field is
spatially restricted for a finite system, leading to a small shift of the electromagnetic field with
respect to the air holes. Figure 4(a) shows the electromagnetic field distribution patterns at (i)
the center of the device and toward (ii-vi) the edges of the device. Although the mode field is
kept antisymmetric with respect to the air holes in the center region, the field in the regions
away from the center is no longer antisymmetric with respect to the air holes, thereby resulting
in an imperfect destructive interference.
To examine the resultant interference effect in further detail, we plot in Fig. 4(b) the intensi-
ties of basic waves propagating in the ±x directions (i.e. |Rx |2 and |Sx |2 ) and the radiation field
intensity along the line y = L/2. The radiation field intensity can be represented by |Rx + Sx |2 ,
which reflects the interference effects of the two counter-propagating basic waves Rx and Sx
[see Eq. (A4) in Appendix A and Eq. (B11) in Appendix B by noting that Fourier coefficient
terms (ξm,n ) for CC air holes are real numbers]. From Fig. 4(b), we can clearly see that the in-
terference is indeed perfectly destructive at the center but becomes imperfect toward the edges
of the device. The situation for ET can be understood in the same manner except that even at
the central region the perfect destructive interference is suppressed because of the asymmetric
air-hole effects [31].
The divergence angle of the far-field pattern is closely related to the finite size of the device.
When the device is infinitely large, the surface emission is strictly in the surface normal di-
rection. However, the realistic device has a finite length L, and therefore allows a wavenumber

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15954
 y
(a) (b)
L |R |2
x
1.2
|S |2
x
2
|Rx+Sx|

Relative intensity (a.u.)

1

+
0.8

0
L/2 0.6

0.4

- 0.2

0
0
x 0 0.2 0.4
x (L)
0.6 0.8 1
L/2 L

Fig. 4. (a) Illustrative field distribution patterns (plotted in a single unit cell) of mode A
for CC air holes at the center of the cavity (i): ( L2 , L2 ) and toward the edges (ii): ( 6L7 , 2 ),
L

(iii): ( 7 , 2 ), (iv): ( 2 , 7 ), and (vi): ( 2 , 7 ). Colors and arrows represent H- and E-fields,
L L L 6L L L

respectively. Thick black circles indicate the locations of air holes with respect to the unit
cell. (b) Field intensities of basic waves propagating in the ±x directions, |Rx |2 and |Sx |2
(red and blue curves), and the radiation field intensity represented by |Rx + Sx |2 (black
curve) along the axis y = L/2. Note that Rx and Sx are zero at x = 0 and x = L, respectively,
which corresponds to the boundary condition described by Eq. (22).

fluctuation of δ k  2π /L from the Γ point, which was demonstrated by 3D FDTD simulations

for finite structures [16, 17]. In fact, this wavenumber fluctuation corresponds to the slightly
shifted electromagnetic field distributions discussed above [see Fig. 4(a)]. Due to this fluctua-
tion, the emitted beam deviates from the surface normal direction with a deflection angle δ θ .
The relation between δ θ and the wavenumber fluctuation can be expressed as

k0 sin(δ θ ) = δ k, (27)

where k0 = 2π /λ (λ : the resonance wavelength in free space). For example, when λ = 960 nm,
L=70 μ m, δ θ  0.8◦ , which is a good approximation of the results shown in Fig. 3.

3.3. Device length dependence of threshold gain

The analysis for an infinite system corresponds to the case when L → ∞. In this case, thresh-
old gain and mode selection properties are determined only by the surface radiation loss (for
simplicity, in this work we do not consider absorption loss). However, for a finite structure,
influence of the in-plane loss (i.e. the power escaping from the edges of the laser cavity) has
to be taken into account. To clarify the effects of the in-plane loss and highlight the difference
between finite and infinite analyses, we investigate the threshold dependence on device length.
We plot in Fig. 5 the normalized mode frequency and threshold gain as a function of the
device length L for CC and ET air holes. Also included as dashed lines are the infinite solutions
(δ and α are first calculated for infinite periodic structures and then normalized by using (δ +
β0 )/(ne f f β0 ) [22] and α L, respectively). From Figs. 5(a) and 5(b), we note that mode frequency
increases with respect to L and converges quickly to the infinite solutions (dashed lines). The
threshold gain, in contrast, shows a quite different behavior: as shown in Figs. 5(c) and 5(d), in
both cases, the laser threshold gain (solid curves) initially decreases with increasing L and then
begins to be asymptotic to the infinite solutions (dashed lines). This behavior can be intuitively
understood as follows. For small values of L, the dominant loss mechanism in the laser is the

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15955
 (a) 0.301
B (finite)
(b) 0.301
B (finite)
0.3 A (finite) 0.3 A (finite)
B (infinite) B (infinite)

Fr equency (c/a)

Fr equency (c/a)
0.299 A (infinite) 0.299 A (infinite)

0.298 0.298

0.297 0.297

0.296 0.296

0.295 0.295
0 100 200 300 400 500 0 100 200 300 400 500
(c) 2 L (µm) (d) 2 L (µm)
Nor malized thr eshold (αL )

Nor malized thr eshold (αL )

B (finite) B (finite)
1.5 1.5
A (finite) A (finite)
B (infinite) B (infinite)
1 A (infinite) 1 A (infinite)

0.5 0.5

0 0
0 100 200 300 400 500 0 100 200 300 400 500
L (µm) L (µm)

Fig. 5. Normalized mode frequency (a, b) and threshold gain (c, d) as a function of the
device length L for CC and ET air holes ( f = 0.16 and a=295 nm). The solid curves and
dashed lines are calculated for finite and infinite structures, respectively.

in-plane loss. In particular, at very short lengths, the optical power exits the gain region at
the edges of the laser cavity without being diffracted out of the PC surface, leading to a high
threshold gain. However, for larger values of L, the in-plane loss steadily decreases and the
surface radiation loss becomes dominant. In fact, infinite solutions can correctly predict the
lasing mode (i.e. the lowest threshold mode) of the device only when L is large enough, e.g.
L > 400 μ m (see Fig. 5). For a relatively small L, e.g. L < 100 μ m, the influence of the in-plane
loss on threshold behavior should not be neglected.
To evaluate the influence of the in-plane loss quantitatively, we derive a formula (Appendix
B) to describe the power flow in 2D PC-SELs by extending the energy conservation theorem
developed for 1D DFB lasers [32], which can be expressed as

Pstim = Pedge + Prad . (28)

Here, Pstim , Pedge , and Prad represent the stimulated emission power, the power escaping from
the edges of the laser cavity (i.e. the in-plane loss), and the surface radiation power, respectively.
Therefore, the percentage of the in-plane loss power with respect to the total stimulated power
is given by

Pedge /Pstim = 1 − Prad /Pstim . (29)

Table 3 shows the calculated Pedge /Pstim of modes A and B for CC and ET air holes with
different device lengths. We can clearly see that, for both air-hole shapes, when L = 70 μ m,
the in-plane loss is indeed noticeably larger than the surface radiation loss, indicating that in-
plane loss plays a very important role in the mode selections of the laser device. While when
L=400 μ m, the in-plane loss drastically decreases and its influence on mode selections might
be neglected. Here, it is important to note that, in the case of ET air holes, when the device
length L is increased from 70 μ m to 400 μ m, the lasing mode likely switches from mode A to
mode B which has a smaller radiation constant [see Fig. 5(d)].

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15956
 Table 3. Percentage of the In-Plane Loss with Respect to the Total Stimulated Power,
Pedge /Pstim for CC and ET Air Holes with Different Device Lengths ( f = 0.16 and a=295
nm)
Shape/mode L=70 μ m L=400 μ m
mode A 67% 26%
CC
mode B 69% 26%
mode A 58% 2%
ET
mode B 71% 19%

4. Conclusion
In conclusion, we have developed a CWT model for finite-size square-lattice PC-SELs with TE
polarization by extending our previous 3D CWT framework. Using this model, we calculated
various properties of 2D PC-SELs, including threshold gain, mode frequency, field intensity en-
velope within the device, and the profile of the output beam (far-field pattern and polarization).
The theoretical predictions of the lowest threshold mode and the output beam profile showed
good agreement with our experimental findings, thereby affirming the validity of our analyses.
Our finite-size analysis results demonstrate that finite device length indeed strongly influ-
ences surface emission and mode selection properties of the PC-SEL device. Phenomena that
cannot be explained by the infinite CWT analysis are clarified by considering the finite-size
effects of the device. We showed that the finite device length may lead to a small shift of the
electromagnetic field with respect to the air holes. This slightly shifted field suppresses the
otherwise perfect destructive interference and thus enhances surface emission from the regions
away from the center of the cavity, which accounts for the finite surface radiation loss emitted
from the non-radiative (antisymmetric) mode of the CC air holes. By investigating the device
length dependence of threshold gain, we find that, at a very large device length L (e.g. L > 400
μ m), the finite-size analysis results become asymptotic to the infinite solutions, while at a rel-
atively small L (e.g. L < 100 μ m), the predicted lasing mode obtained using the finite-size
analysis may be quite different from that based on the infinite structures. This is attributed
to the influence of the in-plane loss, which can be noticeably large at short device lengths.
The in-plane loss can provide a strong mode selection mechanism with a large threshold-gain
discrimination of over 20 cm−1 , which allows us to explain the stable single-mode operation
observed in our previous experiments.
The insights obtained in this work are essential for understanding the performances and
physics of a practical PC-SEL device. Moreover, this theory also forms the basis for more com-
plicated analyses of the properties of the device, including slope efficiency and the nonlinear
effects [33] such as gain saturation [32], non-uniform gain spatial distribution, and hole burn-
ing. We believe that further theoretical work based on this semi-analytical 3D CWT framework
will enable efficient optimization of the structures of 2D PC-SELs for a range of applications
as well as facilitating a comprehensive understanding of the physics of PC-SELs.

Appendix A: Solutions of radiative and high-order waves

We assume, following Ref. [25], that: (1) Only basic waves are important in generating radiative
and high-order waves; (2) For radiative and high-order waves, their derivatives in Eqs. (4)–(6)
are negligible compared to the others; (3) α̃ is small and thus may be neglected. Under these

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15957
assumptions, Eqs. (4)–(6) are reduced to

∂2
[
∂ z2
+ k02 n20 (z) − n2 β02 ]Ex,m,n + mnβ02 Ey,m,n = −k02 ∑ ξm−m ,n−n Ex,m ,n , (A1)
m =m,n =n

∂2
[
∂ z2
+ k02 n20 (z) − m2 β02 ]Ey,m,n + mnβ02 Ex,m,n = −k02 ∑ ξm−m ,n−n Ey,m ,n , (A2)
m =m,n =n
∂
[mEx,mn + nEy,mn ] = 0, (A3)
∂z
which are in the same form as that of those derived for the infinite periodic structures [22].
Then Ex,m,n and Ey,m,n can be solved in a similar manner as that described in Ref. [22], thus we
shall only show the final solutions of radiative and high-order waves.

Radiative waves
The radiative waves [for which (m, n) = (0, 0)] can be expressed as

ΔEx (x, y, z) = [ξ0,−1 Ry (x, y) + ξ0,1 Sy (x, y)]k02 G(z, z )Θ0 (z )dz ,
PC

ΔEy (x, y, z) = [ξ−1,0 Rx (x, y) + ξ1,0 Sx (x, y)]k02 G(z, z )Θ0 (z )dz . (A4)
PC

Here, Green’s function G(z, z )  − 2βi z e−iβz |z−z | and βz = k0 n0 (z); It should be noted that ΔEx
(or ΔEy ) is a function of x and y. As a consequence, the second terms of the right-hand sides
of the coupled-wave Eqs. (15)–(18) can be replaced by terms only associated with the basic
waves.

High-order waves
√
The integrals of high-order waves (for which m2 + n2 > 1) included in the third terms of the
right-hand sides of the coupled-wave Eqs. (15)–(18) can be expressed as

E
PC x,m,n
(z)Θ∗0 (z)dz
∗
PC Ey,m,n (z)Θ0 (z)dz

(1,0) (−1,0) (0,1) (0,−1)
1 n m −mμm,n −mμm,n nμm,n nμm,n
= (1,0) (−1,0) (0,1) (0,−1) · V,
m2 + n2 −m n nνm,n nνm,n mνm,n mνm,n

(1,0) (−1,0) (0,1) (0,−1)
ςx,m,n ςx,m,n ςx,m,n ςx,m,n
≡ (1,0) (−1,0) (0,1) (0,−1) · V, (A5)
ςy,m,n ςy,m,n ςy,m,n ςy,m,n

where
 t
V = Rx , Sx , Ry , Sy , (A6)

(r,s)
μm,n = k02 ξm−r,n−s (z)Gm,n (z, z )Θ0 (z )Θ∗0 (z)dz dz, (A7)
PC

(r,s) 1
νm,n = − ξm−r,n−s (z)|Θ0 (z)|2 dz, (A8)
PC n20 (z)
1 
Gm,n (z, z ) = e−βz,m,n |z−z | , βz,m,n = (m2 + n2 )β02 − k02 n20 (z). (A9)
2βz,m,n

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15958
The elements of matrix C
Finally, the matrix C in Eq. (21) can be written as

C = C1D + Crad + C2D , (A10)

and
⎛ ⎞
0 κ2,0 0 0
⎜ κ−2,0 0 0 0 ⎟
C1D = ⎜ ⎟, (A11)
⎝ 0 0 0 κ0,2 ⎠
0 0 κ0,−2 0

⎛ (1,0) (−1,0)
⎞
ζ1,0 ζ1,0 0 0
⎜ (1,0) (−1,0) ⎟
⎜ ζ−1,0 ζ−1,0 0 0 ⎟
Crad = ⎜ ⎟, (A12)
⎜ 0 (0,1)
ζ0,1
(0,−1)
ζ0,1 ⎟
⎝ 0 ⎠
(0,1) (0,−1)
0 0 ζ0,−1 ζ0,−1

⎛ (1,0) (−1,0) (0,1) (0,−1)

⎞
χy,1,0 χy,1,0 χy,1,0 χy,1,0
⎜ (1,0) (−1,0) (0,1) (0,−1) ⎟
⎜ χy,−1,0 χy,−1,0 χy,−1,0 χy,−1,0 ⎟
C2D = ⎜ ⎟,
⎜ χ (1,0) (−1,0)
χx,0,1
(0,1)
χx,0,1 χx,0,1 ⎟
(0,−1)
⎝ x,0,1 ⎠
(1,0) (−1,0) (0,1) (0,−1)
χx,0,−1 χx,0,−1 χx,0,−1 χx,0,−1
(A13)

and

(r,s) k04
ζ p,q = − ξ p,q ξ−r,−s G(z, z )Θ0 (z )Θ∗0 (z)dz dz,
2β0 PC
(A14)
k2
∑
(r,s) (r,s)
χ j,p,q = − 0 ξ p−m,q−n ς j,m,n , j = x, y. (A15)
2β0 √
m2 +n2 >1

Note that the three square matrices: C1D , Crad and C2D correspond to the conventional 1D
feedback coupling, out-of-plane coupling via radiative waves, and 2D optical coupling via high-
order waves, respectively. It should also be noted that C1D and C2D are Hermitian matrices,
whereas Crad is not an Hermitian one.

Appendix B: Derivation of energy conservation theorem for 2D PC-SELs

The coupled-wave Eq. (21) can be rewritten as

− i(δ + iα )Rx = ∂ Rx /∂ x − i(C11 Rx +C12 Sx +C13 Ry +C14 Sy ), (B1)

−i(δ + iα )Sx = −∂ Sx /∂ x − i(C21 Rx +C22 Sx +C23 Ry +C24 Sy ), (B2)
−i(δ + iα )Ry = ∂ Ry /∂ y − i(C31 Rx +C32 Sx +C33 Ry +C34 Sy ), (B3)
−i(δ + iα )Sy = −∂ Sy /∂ y − i(C41 Rx +C42 Sx +C43 Ry +C44 Sy ), (B4)

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15959
where Ci j (1 ≤ i, j ≤ 4) = C1D,i j +Crad,i j +C2D,i j [see Eqs. (A10)–(A13)]. Multiplying Eq. (B1)
by R∗x , Eq. (B2) by Sx∗ , Eq. (B3) by R∗y , and Eq. (B4) by Sy∗ , and adding the four equations and
their conjugates, we obtain

∂ ∂
2α (|Rx |2 + |Sx |2 + |Ry |2 + |Sy |2 ) = (|Rx |2 − |Sx |2 ) + (|Ry |2 − |Sy |2 )
∂x ∂y
+2κv.i (|ξ−1,0 Rx + ξ1,0 Sx |2 + |ξ0,−1 Ry + ξ0,1 Sy |2 ),
(B5)

where the C1D,i j and C2D,i j terms have vanished due to their Hermitian property, and only the
k4 
term κv.i = −Im{ 2β0 PC G(z, z )Θ0 (z )Θ∗0 (z)dz dz} which is closely associated with the out-of-
0
plane coupling remains [see Eq. (A14)]. In order to model the power flow in the 2D PC-SELs,
we integrate Eq. (B5) over the whole laser cavity area, i.e., 0 ≤ x, y ≤ L. Then we have
 L
2α (|Rx |2 + |Sx |2 + |Ry |2 + |Sy |2 )dxdy =
0
 L
∂ ∂
[ (|Rx |2 − |Sx |2 ) + (|Ry |2 − |Sy |2 )]dxdy
0 ∂x ∂y

+ 2κv.i (|ξ−1,0 Rx + ξ1,0 Sx |2 + |ξ0,−1 Ry + ξ0,1 Sy |2 )dxdy.
(B6)

For a laser structure with boundary condition given by Eq. (22), the field amplitudes of basic
waves at the four edges can be expressed as

Rx (0, y) = 0, Sx (0, y) = Sex (y),

Rx (L, y) = Rex (y), Sx (L, y) = 0,
Ry (x, 0) = 0, Sy (x, 0) = Sey (x),
Ry (x, L) = Rey (x), Sy (x, L) = 0. (B7)

Then Eq. (B6) can be written as

Pstim = Pedge + Prad , (B8)

if we define
 L
Pstim = 2α (|Rx |2 + |Sx |2 + |Ry |2 + |Sy |2 )dxdy, (B9)
0
 L  L
Pedge = (|Rex |2 + |Sex |2 )dy + (|Rey |2 + |Sey |2 )dx, (B10)
0 0
 L
Prad = 2κv.i (|ξ−1,0 Rx + ξ1,0 Sx |2 + |ξ0,−1 Ry + ξ0,1 Sy |2 )dxdy. (B11)
0

Here, Pstim describes the total stimulated power inside the laser structure, Pedge represents the
power escaping from the edges of the laser cavity (i.e. the in-plane loss), and Prad represents the
radiation power emitted from the device surface [note that this quantity is proportional to the
intensity of the radiative waves described by Eq. (A4)], respectively. Equation (B8) states the
fact that the power generated inside the laser structure is equal to the sum of the power escaping
from the edges of the structure and the surface radiation power.

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15960
Acknowledgments
The authors are grateful to Dr. K. Ishizaki, Dr. M. Yamaguchi, Dr. A. Oskooi, and Mr. T.
Okino for helpful discussions and valuable suggestions. This work was partly supported by
the Core Research for Evolution Science and Technology program of the Japan Science and
Technology Agency (CREST-JST) and by the Global COE Program. Three of the authors (Y.
Liang, C. Peng, and S. Iwahashi) are supported by research fellowships of Japan Society for
the Promotion of Science (JSPS).

#164112 - $15.00 USD Received 6 Mar 2012; revised 22 May 2012; accepted 10 Jun 2012; published 28 Jun 2012
(C) 2012 OSA 2 July 2012 / Vol. 20, No. 14 / OPTICS EXPRESS 15961