Professional Documents
Culture Documents
ABSTRACT
We describe and demonstrate a method of decreasing the divergence angle of multi-mode VCSELs, and show how we
can obtain a low and stable divergence angle. We first explain the relationship between the lateral wave-vectors of
resonant modes and the divergence angle. Then we attempt to optimize the oxide aperture and the electrode structure.
Here, we calculate the electro-magnetic field of the VCSELs by the Finite Difference Time Domain (FDTD) method
and the far-field pattern by combining the diffraction integral and the FDTD. Finally, we compare the theoretical and
experimental results of the divergence angle of the VCSELs.
1. INTRODUCTION
Vertical Cavity Surface Emitting Lasers (VCSELs) have received much attention in the field of optical communications.
This is because VCSELs have beneficial properties, such as high-speed modulation, low threshold current and low
fabrication cost. Some manufacturers have started mass production of VCSELs, and thus they are gradually being used
in local area networks (LAN) and storage area networks (SAN). However, some issues remain to be solved. One of the
main issues is the coupling loss of light emitted from the VCSEL into the fiber core. VCSELs, which generally emit
multi-transverse modes, are very useful when using multi-mode optical fibers. However, when TO-CAN packages and
OSA (Optical Sub-Assembly) housings, which are both cost-effective and widely used, are employed, it is necessary to
reduce the divergence angle owing to the limited magnification of the focusing lens and the accuracy of optical
alignment. Of course, this is easily achieved when we employ a single-mode VCSEL structure, but this gives rise to an
unstable convergence length to the Equilibrium Mode Distribution (EMD) of the multi-mode fiber. So, multi-mode but
low divergence VCSELs are desirable.
So far, many research results on the far field pattern (FFP) and the divergence angle of VCSELs have been reported [1-
7]; e.g. the small divergence angle of single mode VCSELs [1,2], far field pattern control of a single high-order
transverse mode[3], beam quality factor of VCSELs (degree of deviation from a Gaussian beam) [4], and so on.
However, most of these are experimental or theoretical results for specific modes and thus we are unable to obtain
useful information on how to reduce the divergence angle of multi-transverse-mode VCSELs. In this paper, we discus
the divergence angle of multi-mode oxide-confined 850nm VCSELs associated with the lateral resonance determined by
the total reflection condition.
Physics and Simulation of Optoelectronic Devices XIII, edited by Marek Osinski, 201
Fritz Henneberger, Hiroshi Amano, Proc. of SPIE Vol. 5722 (SPIE, Bellingham, WA, 2005)
0277-786X/05/$15 · doi: 10.1117/12.590028
polyimide
Confinement layer
Al0.15Ga0.85As
/Al0.9Ga0.1As 85.6nm
10o 20nm
4pair 8nm
8nm
n-DBR: 37.5pairs 3MQW
Al0.15Ga0.85As (GaAs/Al0.3Ga0.7As)
/Al0.9Ga0.1As
25µm
Fig. 1: Schematic of our VCSEL (vertical section)
First, we calculate the photonic band gap (PBG) of DBR using a 1-dimensional plane-wave expansion method (1D
PWM). Then we estimate the resonant frequency of the VCSELs, which is called the “defect mode” in the photonic
crystal field, by combining the super-cell technique with PWM.
We calculate the electro-magnetic field of the device by a 2-dimensional Finite Difference Time Domain method (2D
FDTD). The Berenger’s perfect matching layer (PML) is utilized at the end of the calculation area. Here, we excite the
active region with a Gaussian pulse at around the resonant frequency estimated above. The metal electrode is assumed
to be a perfect electric conductor, and the surrounding region is calculated by the sub-cell method. In this calculation,
we should transform the electro-magnetic field of the emitted light to the frequency region in order to calculate the Far
Field Pattern (FFP) by the diffraction integral. The electro-magnetic field in the far field region is described as follows,
E(r , ω ) = cos θ ∫ {− Z 0 (n′ × H (ω )) − (E(ω ) × n′) × r0}exp{ik (r ′ ⋅ r0 )}ds, (1)
c
In this paper, we assume the refractive index of the oxidized layer is 1.6, the
polyimide region is 1.78, and that of the AlxGa1-xAs layer is given by the
following equation (4),
(
n( x ) = x 0.131E 2 − 0.194 E + 2.99 )
( )
(4) Oxide aperture
+ (1 − x ) 0.1625E 2 − 0.07629E + 3.386 , Fig. 3: The core and cladding region
where E is 1.239/λ, and λ is 0.85µm.
In the following section, we calculate the divergence angle and the transverse modes for various structures using these
methods and attempt to reduce the divergence angle of our VCSELs by optimizing the oxide aperture and the electrode
structure.
3. CALCULATION
Let us start with a band diagram analysis. Figure 4(a) shows the results obtained form calculating photonic band
structure of the DBR. Here, a is the thickness of the DBR, i.e. the lattice constant of 1-dimensional photonic crystal (1D
PhC), and c is the velocity of light. The solid lines seen in Fig. 4(a) are propagation modes of the DBR. We can see in
Fig. 4(a) that there are no propagation modes in the frequency region from 0.1458 c/a to 0.1599c/a. This is the PBG.
The thickness of each DBR layer is set to be 850nm/4n so that the resonant wavelength is about 850nm (a is set to be
129.981nm), and then the wavelength region corresponding to the PBG frequency is from 812.9 to 891.5nm (λ = c/f =
c/(f’c/a) = a/f’ = 129.981nm/0.1458 = 891.5nm). Figure 4(b) shows the band diagram of a VCSEL calculated by 1D
PWM using super-cell. We can see in Fig. 4(b) that a single defect mode appears in the PBG with a group velocity
(gradient) of zero. This means that this mode is a standing wave, and thus the resonant mode of the VCSEL
(longitudinal mode). The resonant frequency (wavelength) is 0.15264c/a (851.6nm).
0.30 0.170
0.25
Frequency [ c /a ]
0.20 0.160
Frequency [ c /a ]
PBG
0.15
0.150
0.10
0.05
0.140
0.00
0.0 0.1 0.2 0.3 0.4 0.5
0.130
Wave Vector [2π/a ]
Wave Vector
(a) Propagation mode of DBR (b) Resonant mode of VCSEL
Fig. 4: Band diagram of our VCSEL
Amplitude [a.u.]
1.0 9 1.0 9
0.5 7 0.5 7
0.0 5 0.0 5
-0.5 3 -0.5 3
-1.0 1 -1.0 1
1 11 21 31 41 51 61 71 32 36 40 44 48 52
z [a ] z [a ]
where k 0 // = k // and k 0 z = k 02 − k //2 ≅ k0 (rough approximation). But in actual fact, k// is small enough for its value to
be changed without affecting eq. (6), which restricts kzr, and thus the divergence angle can easily be increased following
eq. (8). (For instance, when we assume the aperture length to be 13µm and a = 130nm, kzr is about 100 times grater than
k// of the fundamental mode, and thus k// is quite small compared with kzr. However, the divergence angle for k//=π/130,
2π/130, … etc. is 4deg, 8deg, …etc., and easily increases, although this approximation is somewhat rough.) In the
identical mode, k// is mainly determined by the diameter of the resonant cavity, that is, the aperture diameter d, which
should be in inverse proportional to k//. So, the divergence angle of the individual mode calculated by 2D FDTD and the
Therefore, we can predict that the divergence angle of VCSELs can be decreased by the following method; [1] reducing
the effective refractive index of the resonant modes and [2] reducing the effective refractive index difference between
the core and cladding regions. In the following section, we attempt to realize these conditions by tuning the confinement
structure and the oxide layer thickness.
3.292
3.273
3.271 3.290
3.269 3.288
3.267 3.286 Low-index
x=0.3-0.9 confinement
β /k
3.265
3.284 Original
x=0.3-0.6 confinement
3.263
3.282
3.261
3.259 3.280
3.257 3.278
0 1 2 3 4 5
Oxide Appetture Diameter [µm]
LP11
to variations in the fabrication process can LP21
move the horizontal lines slightly. However, it 3.290
is not always necessary to select this region. LP02
Because, the effective refractive index LP31
difference, which determine the total refraction 3.288
LP12
condition, is not changed by the aperture LP41
diameter. In other words, the curved line in Fig.
3.286 LP22
12 defines the maximum divergence angle and
cannot be moved by fabrication varieties in the 0 4 8 12 16 20 24 LP03
diameter. So, even if another mode happens to Oxide Layer T hickness [nm] LP51
resonate owing to variations in the fabrication
process, the maximum divergence angle of the Fig. 12: Effective refractive index and resonant modes LP04
device is still kept narrow.
Normalized Intensity
0.8
set to be 11.0µm in this calculation. Figure 13 shows
the calculated FFP of the 1st order mode of a VCSEL 0.6 R=8.0um
with the electrode aperture diameter of 8.0µm and 0.4
R=13.0um
11.0µm. As can be seen in Fig. 13, sub-peaks exist near
the main peak and the intensity of these peaks become 0.2
large when we introduce an electrode with small 0.0
aperture diameter. Figure 14 is the calculated
-30 -20 -10 0 10 20 30
divergence angle of the VCSEL with respect to the
electrode aperture diameter. The horizontal lines show Angle [deg.]
the original divergence angle without a metal electrode. Fig. 13: Far field patterns of VCSELs
We can see in Fig. 14 that the divergence angle of each with different electrode aperture diameters
mode becomes extremely large when the diameter of
the metal electrode is smaller than that of the oxide aperture. Next, we investigate the electric field distribution in order
to clarify the reason for wide divergence. The calculated electric field distribution is shown in Fig. 15. We can see in
Fig. 15 that the electric field is concentrated at the edge of the electrode when the aperture diameter is smaller than
oxide aperture. This localized element of the optical field spreads out in all directions like an optical point source. This
is because the localized electric field has all wavenumber (spatial frequency) elements in a similar way to that in which
an impulse wave has all frequency elements. As a result, the total divergence angle of the device becomes extremely
large. This localization become notable when we introduce a metal electrode under relatively large electric field of the
core region. So, we should not introduce a metal electrode inside the core region. Therefore, the diameter of the metal
electrode should be set equal to the oxide aperture in order to introduce absorption losses without sacrificing FFP.
Electrode
25
Oxide Aperture Diameter 16.0
Divergence Angle [deg.]
12.0
20
0th mode 8.0
Amplitude [a.u.]
0 -8.0
5 7 9 11 13 15 17 -12.0
Electrode Diamet er [µm] -16.0
x
Fig. 14: Divergence angle of VCSELs
Fig. 15: Electric field distribution
with respect to electrode aperture diameter
4. EXPERIMENT
Having found a method for decreasing the divergence angle, we fabricated VCSELs with the structure mentioned above.
First, we tried to verify the effect of the confinement layer structure on the divergence angle. We employed 2 kinds of
composition for n-type confinement layer, Al0.3→0.6Ga0.7→0.4As and Al0.3→0.9Ga0.7→0.1As, where Xy→z means that the
component of X linearly alters from Xy to Xz. The composition for p-type confinement layer is both Al0.3→0.6Ga0.7→0.4As.
Normarized Intensity
(i) re=13.0µm < ro=15.0µm (ii) re= ro=15.0µm (iii) re= 17.0µm > ro=15.0µm
Fig. 18: Far field patterns of VCSELs with different electrode diameters
Finally, we fabricated the most promising VCSELs with a thin oxide layer. The thickness of this layer was 15nm, and
the diameter of both the electrode and the oxide aperture was about 11µm. The composition of the p- and n-type
confinement layers were Al0.3→0.6Ga0.7→0.4As and Al0.3→0.9Ga0.7→0.1As, respectively. The experimental results of the
divergence angles of VCSELs with 15 and 30nm thick oxide layers are shown in Fig. 19. We can see in Fig. 19 that the
divergence angle of VCSELs with a thin oxide layer is quite narrow compared with VCSELs with a 30nm thick oxide
layer and that a maximum divergence angle of less than 26 degrees is achieved. We think this characteristic is quite
important for optical alignment.
20
15nm
30nm
10
10 20 30 40
Divergence Angle (horizontal) [deg.]
Fig. 19: Divergence angle of VCSELs with thin
oxidized layer compared with original VCSELs
5. CONCLUSION
In this paper, we have presented a principle and methods for decreasing the divergence angle of VCSELs. First, we
found that [1] reducing the effective refractive index of the resonant modes in order to reduce the normal component of
the wave-vector in free space and [2] reducing the effective refractive index difference between the core and cladding
regions for limiting resonant modes with large lateral wave-vectors are significant in order to achieve a low divergence
angle. Secondly, we investigated theoretically the influence of both methods by tuning the refractive index of the
confinement layers and found that [2] is more important. Therefore, we tried to reduce the thickness of the oxide layer
in order to restrict the total reflection for lateral resonance and found that resonant modes with large transverse wave-
vectors can be drastically reduced. We also investigated the influence of the electrode on the divergence angle and
found that the concentration of the electro-magnetic field at the edge of the metal electrode can cause sub-peaks in the
FFP and affect the divergence angle when the diameter of the electrode was set to be smaller than that of oxide aperture.
So, when we consider the suppression of higher-order modes by absorption by the metal electrode, we should set the
diameter of the metal electrode equal to the oxide aperture. Finally, we confirmed experimentally how the effective
refractive index of the confinement layers and the diameter of the metal electrode influence the divergence angle, and
fabricated VCSELs with a thin oxide layer and ideal electrode diameter. As a result, we have obtained a low divergence
angle less than 26 degrees experimentally.
6. REFERENCE
1. C.Yan, Y. Ning, L. Qin, Y. Liu, L. Zhao, Q. Wang, Z. Jin, Y. Sun, G. Tao, G. Chu, C. Wang, L. Wang, H. Jiang,
“High-power vertical cavity surface emitting laser with good performances”, Electron. Lett., 40, pp. 872-874, 2004
2. A. Hsu, D. Leonard, N. Cao, F.M. Ahedo, C. LaBounty and M.H. MacDougal, “Low-divergence singlemode
1.55µm VCSEL with 1mW output power”, Electron. Lett., 39, pp. 59-91, 2003
3. S. Shinada and F. Koyama, “Single High-Order Transverse Mode Surface-Emitting Laser With Controlled Far-
Field Pattern”, IEEE Photonic. Technol. Lett., 14, pp. 1641-1643, 2002
4. Ihab Kardosh, “Beam Propaties and Quality Factor of VCSELs”, Annual Report 2003, Optoelectron. Dept. Univ. of
Ulm
5. Sulaiman S. Al-Sowayan and Kevin L. Lear, “Beam Divergence Angle Distribution Dependence on Data Pattern”,
Vertical-Cavity Surface-Emitting Lasers VIII, Proceedings of SPIE, 5364, pp. 138-145, 2004
6. Delai Zhou nad Luke J. Mawst, “High-Power Single-Mode Antiresonant Reflecting Optical Waveguide-Type
Vertical-Cavity Surface-Emitting Lasers”, IEEE J. Quantum Electron., 38, pp 1599-1606
7. Y.-G. Zhao, Y.-S. Zhang, X.-L Huang, L.-T. Zhang, W.-X Chen, L.-F. Cong, C.-Z. Jin, X.-W. Hu, and W. Wang,
“Far-field and beam characteristics of vertical-cavity surface-emitting lasers”, Appl. Phys. Lett. 69, pp 1829-1831,
1996