The use of gallium antimonide (GaSb) is increasing, especially for optoelectronic devices in the infrared wavelengths. It has been demonstrated in gallium nitride (GaN) devices operating at ultraviolet (UV) wavelengths, that surface textures increase the overall device efficiency. In this work, we fabricated eight different surface textures in GaSb to be used in enhancing efficiency in infrared wavelength devices. Through chemical etching with hydrofluoric acid, hydrogen peroxide, and tartaric acid we characterize the types of surface textures formed and the removal rate of entire layers of GaSb. Through optimization of the etching recipes we lower the reflectivity from 35.7% to 1% at 4 μm wavelength for bare and textured GaSb, respectively. In addition, we simulate surface textures using ray optics in finite element method solver software to provide explanation of our experimental findings.

In recent years gallium antimonide (GaSb) has become integral to many optical devices1 including light emitting diodes,2,3 photodetectors,4,5 and diode lasers.6,7 Efficient performance of GaSb-based devices remains elusive. For example, the power conversion efficiency in diode lasers based on GaSb has only reached 17.5%.8 However, efficiency in GaSb devices can be enhanced through the use of surface texturing. The use of chemical etching to create surface textures to enhance device efficiency has been proven on other materials for a wide variety of applications.9–12 The surface roughening of GaN-based LEDs has been demonstrated to double the output power when compared to a smooth surface LED.13 The increase in roughened GaN LED efficiency is due to increased scattering on the textured surface.14 

Snell’s law, also known as the law of refraction, defines the relationship between the angles of incidence and refraction, when referring to light or other waves passing through a boundary between two different isotropic media, such as semiconductor or air. According to Snell’s law, a large change in refractive index creates a small incident angle before the light exhibits total internal reflection. With a refractive index of 3.8 (at 4 μm wavelength), GaSb requires an incident angle smaller than 15 degrees in order to avoid total internal reflection. By texturing the surface, the likelihood of meeting the surface within the critical angle increases. In addition, the graded index due to the surface texture allows for a smaller change in the effective refractive index relative to untextured GaSb and air, reducing internal reflections at the interface. David et al.14 shows via numerical modeling that increase in rays scattered and decrease in the number that are totally internally reflected increases overall LED device efficiency.

In this work we study the surface textures in GaSb for two of the three atmospheric transmission windows represented by wavelengths ranging from 3 μm to 5 μm and 8 μm to 13 μm. The atmospheric transmission windows represent wavelength regions in which light can travel through the atmosphere with high transmission and without significant attenuation. The devices operating in different atmospheric transmission windows require different textures for increased performance. This is because a wavelength of light scatters the most due to surface feature sizes that correspond to the magnitude of the wavelength in the material. By characterizing the textures that induce scattering in the infrared transmission window we are able to predict the textures needed to increase device efficiency for devices operating at specific wavelengths in the infrared. We also develop the theoretical basis behind our experimental results using ray optics module in COMSOL finite element method solver software.

In order to characterize the impact of different surface textures on light extraction, we varied either the concentration of etch chemicals or the etch time. In all cases we used a Teflon mesh to suspend the GaSb samples and a magnetic stirring rod at 350 revolutions per minute (rpm) to ensure texture uniformity. The back of each sample was covered with photoresist to keep it unetched, similar to the way it would be for an actual device, where top of the substrate will be the emitter and the back of the substrate will be the textured surface. In addition, we also covered a portion of the front of the sample with photoresist to determine the etch rate of each recipe. Table I shows the etch chemistries whereby varying the volume of hydrofluoric acid (HF, ≥49% vol., Sigma Aldrich), hydrogen peroxide (H2O2, 30 % (w/w) in H2O, Sigma Aldrich), and tartaric acid (C4H6O6, 99% Sigma Aldrich) we created a range of surface textures on GaSb. The parametric sweep of etch chemicals allow for a systematic approach to determining how specific chemical volumes impact surface textures and substrate etch depth. The end goal was to develop chemistries that result in optimum surface texture with high scattering in infrared wavelengths as well as thinned substrates that will result in high light extraction.

The samples were imaged after etching using either a scanning electron microscope (SEM) for samples with greater than nanometer scale textures and an atomic force microscope (AFM) for samples with nanometer scale textures. Fig. 1 shows SEM images of the larger and deeper surface textures. The smaller surface textures required the use of an AFM, as shown in Fig. 2. Using image analysis of the surface textures,15 we were able to approximate the average surface feature size. Surface features range in size from around 10 μm in Fig. 1, Etch Recipe 7 to 0.065 μm in Fig. 1, Etch Recipe 3.

By changing each chemical’s volume at a time we determined the chemical’s impact on the surface texture. We began with characterizing the role of HF in the etch solution while keeping the volume of H2O2 at 60 ml and C4H6O6 at volume percent of 10% (4.8 g of C4H6O6/48 ml of H2O). Based on the parametric sweep on the HF volume, we found that HF has both a significant impact on surface features and on the etch rate of the substrate. At the highest volume of HF (15 ml) in Fig. 1, Etch Recipe 1 there are two distinct surface features. The larger features are around 1.15 μm in diameter while the smaller are around 0.35 μm. Lowering the volume of HF to 5 ml (Fig. 1, Etch Recipe 2) results in surface features that range from 6 μm to around 1 μm. With the smallest volume of HF (1 ml) no features were discernable under the SEM. However, using AFM the features were measured to be around 64.6 nm in diameter (Fig. 2, Etch Recipe 3). As shown in Fig. 3(a), the etch depth as a function of HF volume increases linearly, with 230 μm of the substrate etched in 5 minutes at HF volume of 15 ml.

We characterize the role of H2O2 to generate surface textures and etch depth while keeping the volume constant for HF at 5 ml and C4H6O6 at volume percent of 10% (4.8 g of C4H6O6/48 ml of H2O). We discover that the H2O2 volume in the etch solution has a large impact on surface texture of GaSb. The highest volume of 60 ml H2O2 (Fig. 1, Etch Recipe 2) results in features sizes ranging from diameter of 6.1 μm to 1 μm. Changing the volume of H2O2 to 45 ml (Fig. 1, Etch Recipe 4) results in feature sizes ranging from around 7.5 μm to 0.7 μm. Lowering the volume of H2O2 to 30 ml (Fig. 1, Etch recipe 5) produces an even larger range of feature sizes from 9.2 μm to 0.3 μm. The lowest volume of 15 ml H2O2 was measured with an AFM (Fig. 2, Etch Recipe 6) since the texture features were too small to be characterized using SEM imaging. Etch Recipe 6 sample exhibits features approximately 339 nm in diameter. Fig. 3(b) shows the dependence of etch depth of GaSb to H2O2 volume in the etch solution. The etch depth increases drastically as the H2O2 volume increases from 15 ml to 30 ml, and then remains constant as the H2O2 volume is increased to 60 ml.

Finally, we varied the volume percent of C4H6O6 in the etch solution, while keeping volume constant for HF at 15 ml and H2O2 at 60 ml, and characterized the resultant surface textures and etch depth. The highest volume percent of 10% C4H6O6 (Fig. 1, Etch Recipe 1) results in features sizes ranging from diameter of 4.2 μm to 0.2 μm. Changing the volume percent to 5% C4H6O6 (Fig. 1, Etch Recipe 7) results in feature sizes ranging from around 10.2 μm to 0.4 μm. Lowering the volume percent to 2% C4H6O6 (Fig. 1, Etch Recipe 8) produces small feature sizes ranging from 0.2 μm to 0.02 μm. Fig. 3(c) shows the dependence of etch depth of GaSb to C4H6O6 volume percent in the etch solution. The etch depth increases linearly as the C4H6O6 volume percent increases with 240 μm removed at 10% C4H6O6 for the etch time of 5 minutes.

By measuring the reflectivity of each surface we found how each would improve light extraction when integrated onto a device. Reflectivity measures the scattering within a range of wavelengths and reveals the wavelength region where a certain texture scatters light the most. For our measurements we used a Nicolet Magna 760 FTIR with mirrors arranged as shown in Fig. 4. Each sample was measured at a 45° angle of incidence and compared to a baseline measurement with a gold mirror. Fig. 5 shows five of the representative reflectivity curves. Etch Recipe 5 was found to result in the lowest reflectivity curve across the entire infrared wavelength regime. The average feature size is around 3.75 μm but the range of feature sizes from 9.2 μm to 0.3 μm allows the surface texture to be effective for a wide range of wavelengths. In addition, to determine the reproducibility of our chemical etching, we used the same etch recipe and etched two different GaSb samples using it. We then characterized each sample separately using the FTIR measurements. We found that the etch is quite reproducible and reflectivity measurements for each sample was found to be almost identical.

We modeled the textured surfaces using COMSOL finite element method16 solver software’s Ray Optics Module, which can be used to model the propagation of electromagnetic waves when the wavelength is much smaller than the smallest geometric entity in the model. To compare light extraction with reference to a planar GaSb substrate we modeled planar as well as a random textured substrate. We modeled for an emission wavelength of 3.8 μm vacuum wavelength which corresponds to 0.82 μm wavelength in GaSb with a refractive index of 4.6. Fig. 6 shows the results of the model, where for planar substrate the number of rays extracted is 88 while for the random texture with average feature size of 0.9 μm the number of rays extracted 504. This indicates that with random textures with feature size around the wavelength of light in GaSb results in higher than five times light extraction relative to planar substrates.

In conclusion, we have developed etching recipes to fabricate surface textures tailored to enhance efficiency at a certain wavelength in GaSb. As predicted in a COMSOL simulation, surface texture enhances scattering with the most effective scattering happening at a texture tailored to the wavelength. Using the textures presented the surface reflectivity of GaSb can be reduced to 1% at 4 μm wavelength relative to 35.7% for planar GaSb.

1.
P. S.
Dutta
,
H. L.
Bhat
, and
V.
Kumar
,
J Appl Phys
81
(
9
),
5821
-
5870
(
1997
).
2.
A.
Krier
,
M. K.
Parry
, and
D. S.
Lanchester
,
Semicond Sci Tech
6
(
11
),
1066
(
1991
).
3.
R. J.
Ricker
,
A.
Hudson
,
S.
Provence
,
D. T.
Norton
,
J. T.
Olesberg
,
L. M.
Murray
,
J. P.
Prineas
, and
T. F.
Boggess
,
Ieee J Quantum Elect
51
(
12
),
1
-
6
(
2015
).
4.
Y.
Huang
,
J. H.
Ryou
,
R. D.
Dupuis
,
V. R.
D’Costa
,
E. H.
Steenbergen
,
J.
Fan
,
Y. H.
Zhang
,
A.
Petschke
,
M.
Mandl
, and
S. L.
Chuang
,
J Cryst Growth
314
(
1
),
92
-
96
(
2011
).
5.
X. C.
Li
,
D. W.
Jiang
,
Y.
Zhang
, and
L. C.
Zhao
,
Superlattice Microst
91
,
238
-
243
(
2016
).
6.
K.
Merghem
,
R.
Teissier
,
G.
Aubin
,
A. M.
Monakhov
,
A.
Ramdane
, and
A. N.
Baranov
,
Applied Physics Letters
107
(
11
), (
2015
).
7.
H. Q.
Le
,
G. W.
Turner
,
S. J.
Eglash
,
H. K.
Choi
, and
D. A.
Coppeta
,
Applied Physics Letters
64
(
2
),
152
-
154
(
1994
).
8.
L.
Shterengas
,
G.
Belenky
,
M. V.
Kisin
, and
D.
Donetsky
,
Applied Physics Letters
90
(
1
), (
2007
).
9.
F.
Toor
,
H. M.
Branz
,
M. R.
Page
,
K. M.
Jones
, and
H.-C.
Yuan
,
Applied Physics Letters
99
(
10
),
103501
(
2011
).
10.
R. H.
Horng
,
S. H.
Huang
,
C. C.
Yang
, and
D. S.
Wuu
,
Ieee J Sel Top Quant
12
(
6
),
1196
-
1201
(
2006
).
11.
Y.-H.
Cheng
,
J.-L.
Wu
,
C.-H.
Cheng
,
K.-C.
Syao
, and
M.-C. M.
Lee
,
Applied Physics Letters
90
(
9
),
091102
(
2007
).
12.
G.
Horst
,
Jpn J Appl Phys
46
(
7R
),
4125
(
2007
).
13.
T.
Fujii
,
Y.
Gao
,
R.
Sharma
,
E. L.
Hu
,
S. P.
DenBaars
, and
S.
Nakamura
,
Applied Physics Letters
84
(
6
),
855
-
857
(
2004
).
14.
A.
David
,
J Disp Technol
9
(
5
),
301
-
316
(
2013
).
15.
K.
Nakagawa
,
T.
Yoshimura
, and
T.
Minemoto
,
Appl. Opt.
32
(
25
),
4898
-
4903
(
1993
).
16.
J. H.
Coggon
,
GEOPHYSICS
36
(
1
),
132
-
155
(
1971
).