- Email: [email protected]
Modified AlAs epitaxial layers for use as pattern transfer masks
MICRO~I.g~nlOI~ ELSEVIER
Microelectronic Engineering 46 (1999) 327-330
M o d i f i e d AlAs epitaxial layers for use as pattern transfer m a s k s C.J.M. Smith ~, T.F. Krauss ~, S.K. Murad ~, C.D.W. Wilkinson", A. Boyd~, C.R. Stanley ", M. Dawson b and R. M. De La Rue" ~Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8LT, Scotland blnstitute of Photonics, University of Strathclyde, 106 Rottenrow, Glasgow G4 0NW, Scotland The viability of using a thermally oxidised AlAs layer as a robust mask for the transfer of nanometre scale features is reported. Plasma fluorinated AlAs layers were investigated also, but despite selectivities for fluorinated masks of 100:1 between the mask and GaAs, this technique is currently not suitable for pattern transfer as it produces robust, but non-uniform, intermediate masks. In contrast, aspect ratios of 10:1 or greater can be obtained with a selectivity of 100:1 between the oxidised layer and GaAs for sub-micron features. 1. I N T R O D U C T I O N The ability to fabricate sub-wavelength optical structures has introduced exciting new device possibilities, such as diffractive optical elements and photonic bandgap structures. Fabrication technology has now reached the stage where strong light-matter interactions are being observed in epitaxial semiconductor structures, which is opening up a whole new area of device engineering. The fabrication of these new optical devices requires the use of high-resolution lithographic and pattern transfer processes, including electon-beam lithography (EBL) and reactive-ion etching (RIE). Unfortunately, these technologies have limiting factors, with one such limitation being the poor resilience of polymethylmethacrylate (PMMA), the high-resolution resist used in EBL, to most RIE plasmas. Although much research has been done to tackle this issue, some of the new resists are too sensitive and push current electron beam projection systems past their maximum working frequency. Therefore, PMMA is still commonly used today. The problem of its poor resilience has been overcome by the use of an intermediate layer, such as S i O 2, Si3N 4 or certain metals. However, both dielectric and metal masks have their disadvantages: the need to deposit these thin films, which can damage the underlying material; the need to etch the dielectric layer after initial patterning and the difficulty of reliable sub-micron lift-off. An alternative strategy, which addresses some of these
points is the inclusion of the intermediate mask layer at the growth stage and its modification after growth to give improved masking properties. The example of an "in-situ" epitaxial layer that we have investigated is AlAs and its modification by wet thermal oxidation or treatment in a fluorinebased RIE plasma. The wet, thermal oxidation of A1GaAs layers (x>0.8) has been used widely in recent years [1] after Holonyak et al.[2] first demonstrated the process. They showed that at temperatures of approximately 400°C, in the presence of steam, A1GaAs can be transformed into a durable oxide; this oxide has almost the same density as the original A1GaAs layer[3], but is significantly more robust when exposed to RIE plasmas[4]. The treatment of AlAs in fluorine-based plasmas has not been reported yet, to the authors knowledge, but fluorine addition to chlorine plasmas has been shown to provide selective etching of GaAs over A1GaAs[5], by the formation of an A10×Fy film at the A1GaAs surface. Sputtered metal halide films have also been used as an electron-beam resist[6] and are used for IR coatings[7]. The use of plasma treatment to create an amorphous A1Fx layer seeks to extend these techniques. We have investigated a wet, thermally oxidised and a dry, fluorinated AlAs layer as pattern transfer masks and compared these masks with deposited dielectric layers. In particular, a steam oxidised layer is a robust intermediate pattern transfer mask, with selectivities greater than 100:1 over GaAs in SiC14.
0167-9317/99/$- see front matter © 1999 Elsevier Science BY. All rights reserved. PII: S0167-9317(99)00096-9
328
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
2. EXPERIMENTAL The MBE grown wafer consisted of a 20 nm GaAs cap, a 75 nm AlAs layer which was to be modified to act as the intermediate pattern transfer layer, a 919 nm thick GaAs layer and an AlAs etch stop layer 20 nm thick grown on a (100) GaAs substrate. A thickness of 75 nm was chosen for the top AlAs layer as the oxidation rate decreases rapidly for layers thinner than 50 nm[8]. The resist patterns used here were initially generated on a Leica EBPG-HR5 machine at an acceleration voltage of 50 kV. The nominal resolution is 5 nm and the spot size used was 15 nm. The PMMA resist used was Dupont Elvacite (average molecular weight 350,000) dissolved in the solvent o-xylene to give a PMMA weight percentage of 4% (nominal monolayer thickness was 100 nm, when spun-coated at 5,000 rpm and baked at 180 °C). Two layers of PMMA were used to avoid pin-holing effects in the resist and to provide a thick enough layer for high fidelity pattern transfer to both the intermediate dielectric layer (SiO 2 or SiNx) and the heterostructure. The exposed PMMA was then developed in 2:1 IPA:MIBK for 30 seconds; these being the optimum conditions for our laboratory at the time of the experiments. The silica (SiO2) and silicon nitride (SiNx) films used in these experiments were deposited using Plasma Enhanced Chemical Vapour Deposition (PECVD) in an Oxford Plasma Technology gP80 Plus machine. Two reactive-ion etching machines were used, the first machine is an Oxford Plasma Technology RIE80 machines and the other machine is an Oxford Plasma Technology System 100 RIE machine (S100). The wet thermal oxidation was camed out in a novel oxidation rig based on a commercial alloying furnace (Bio-Rad RC2400); Figure 1 shows the furnace schematic. The chamber stage incorporated a low-thermal-mass heater for rapid thermal cycling up to 750°C, with a control unit providing accurate timed heating of the stage and digital temperature control. The furnace temperature used in this work was 380 °C with a nitrogen carrier gas flow of 0.6 l/s flowing through a water bath at 90 °C; these
conditions being previously controllable oxidation.
optimised
for
Flow controller
Water bath t~ ~z
Alloying furnace Exhaust
!
¢:u
~a~
J
Heated stage Figure 1. Schematic of the oxidation rig used. 3.1
AlAs O X I D A T I O N One-dimensional gratings were written with periods varying from 150 to 250 nm with a 50% duty cycle. The resist features were then transferred into the top 100 nm of the epitaxial structure by SIC14 RIE in the S100 machine (15 sccm, 2mT, 200 W), which was sufficient to expose the sides of the AlAs layer. Examination of the etched patterns after cleaving indicated that there were problems of native oxide formation (Fig. 2); native oxide here means the oxide that results from the room temperature reaction of atmospheric oxygen with the A1 in the epitaxial layer. Spontaneous native oxidation is a problem with AlAs, but this problem could be overcome by using a high A1 molar fraction AlxGal_xAs layer (0.8
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
The AlAs layers exposed by the first RIE process were steam oxidised. The samples were oxidised for a maximum of 5 minutes, which was sufficient time to oxidise laterally between adjacent ridge edges; the lateral oxidation rate had been calibrated previously, but the colour change in the epitaxial material, upon oxidation, allows the front to be monitored in-situ via the optical microscope fitted above the furnace chamber. The sample was then returned to the S 100 machine where it was etched in SiC14/O 2 using conditions optimised for the vertical transfer of submicron patterns into GaAs (15/1 sccm, 3 mT, 250 W). Firstly, one sample with an oxidised AlAs layer was etched in order to determine the etch rate (2-5 nm/min). An etch rate of 200 nm/min for GaAs leads to a selectivity of 100:1 between the mask and semiconductor. This high selectivity, when combined with highly anisotropicaly RIE, allows the fabricatation of nanometre scale structures with high aspect-ratios (Fig. 3). 3.2
AlAs FLUORINATION Initial patterns were made using photolithography where stripe waveguides were defined in Shipley S1818 resist. The GaAs cap was then removed using a selective process in the RIE80 machine, which
Fig. 2. Native oxide formation after RIE due to previous thermal cycling
329
etches GaAs over A1GaAs or AlAs (2 sccm, 3 m T , 10 W)[9]. The chamber was then purged with nitrogen for 20 minutes, while still under vacuum, after which the exposed AlAs layer was exposed to a n S F 6 plasma (23 sccm, 50 mT, 50 W) for 2 minutes. This treatment modifies the exposed areas and leaves the resist covered areas untreated. The sample was then removed from the chamber, the photoresist was removed in acetone and thereafter the sample was etched in the S100 machine in SiC14/O 2. Initial inspection revealed that the process did produce a negative mask and reproduced the general outline of the resist features, but closer inspection showed that the mask was not uniform (Fig. 4). At present it is not clear which mechanism modifies the AlAs layer but a possible explanation would be the chemical reaction of the As with the F, dissociated from the plasma, to produce AsF 3, which is volatile. This reaction would leave behind A1 to react with the F and form A1F3, a dielectric, and then oxidise on exposure to the atmosphere after the sample is removed from the RIE chamber. This technique, if developed further, could provide an alternative form of negative lithography, which does not involve a lift-off process.
Fig. 3. Steam oxidised AlAs layer used a mask in SiC14/O2 to etch GaAs.
Fig 4. Fluorinated AlAs layer reproduces general features, but the mask is not uniform
330
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
3.3
C O M P A R I S O N OF M A S K S As the fluorination process does not produce a uniform mask, we shall only compare the oxidised AlAs layer with traditional deposited dielectric films. Table 1 shows the etch rates in a SIC14/O2 RIE plasma for all the intermediate masks being considered and the selectivities of these masks with respect to GaAs. As seen, the two dielectric masks have approximately the same etch rate and selectivity. However, the etch rate of the A1Ox layer is almost an order of magnitude smaller. No limitations of using the A1Ox approach for pattern transfer purposes have been seen in feature sizes down to 50 nm, so this technology has wide-spread appeal for the fabrication of GaAs-based device structures. Moreover, it does not require the potentially damaging post-growth deposition of a dielectric film.
Material Etch rate tnrn/minl Selectivit), SiO2 19 10.5 SiNx 21 9.5 A10~ 2-5 40-100 Table 1. Comparison of different masks in a SiCI4/O 2 plasma; selectivity is of the mask with respect to GaAs. 4. CONCLUSIONS We have demonstrated the ability to incorporate an intermediate layer into the original growth for use as a pattern transfer mask. Dry fluorination produces a non-uniform but robust mask, whereas wet thermal oxidation of an AlAs layer produces a robust mask which allows the subsequent fabrication of highaspect ratio, sub-micron features in GaAs due to selectivites greater than 100:1. Future devices may be fabricated using A1GaAs layers both for pattern transfer purposes and for optical and current confinement via selective oxidation.
5. A C K N O W L E D G E M E N T S We wish to thank colleagues in the Nanoelectronics Research Centre, the Optoelectronics Research Group, the MBE Research Group and the Dry Etch group at Glasgow for technical support. CJMS is supported by a UK EPSRC CASE award with B.T. Laboratories and TFK is supported by a Royal Society Research Fellowship. REFERENCES [1] K. D. Choquette, F. M. Geib, C. I. H. Ashby, R. D. Twesten, O. Blum, H. Q. Hou, D. M. Follstaedt, B. E. Hammons, D. Mathes, and R. Hull, IEEE Journal of Selected Topics in Quantum Electronics, 3, pp. 916-925, 1997. [2] N. Holonyak, Internation patent WO 92/12536, 1992 [3] R. D. Twesten, D. M. Follstaedt, K. D. Choquette, and R. P. Schneider, Applied Physics Letters, 69, pp. 19-21, 1996. [4] C. C. Cheng, A. Scherer, R.-C. Tyan, Y. Fainman, G. Witzgall, and E. Yablonovitch, Journal of Vacuum Science and Technology B, 15, pp. 2764-2767, 1997. [5] S. K. Murad, N. I. Cameron, S. P. Beaumont, and C. D. W. Wilkinson, Journal of Vacuum Science and Technology B, 14, pp. 3668-3673, 1996. [6] H. R. Borsje, H. M. Jaeger, and S. Radelaar, Microelectronic Engineering, 17, pp. 311-314, 1992. [7] S. F. Pellicori and E. Colton, Thins Solid Films, 209, pp. 109-115, 1992. [8] K. M. Geib, K. D. Choquette, H. Q. Hou, and B. E. Hammons, in Vertical-Cavity Surface Emitting Lasers, vol. 3003, K. D. Choquette and D. Deppe, Eds.: SPIE, 1997, pp. 69-74. [9] S. K. Murad, C. D. W. Wilkinson, and S. P. Beaumont, Microelectronic Engineering, 2 3, pp. 357-360, 1994.
Microelectronic Engineering 46 (1999) 327-330
M o d i f i e d AlAs epitaxial layers for use as pattern transfer m a s k s C.J.M. Smith ~, T.F. Krauss ~, S.K. Murad ~, C.D.W. Wilkinson", A. Boyd~, C.R. Stanley ", M. Dawson b and R. M. De La Rue" ~Department of Electronics and Electrical Engineering, University of Glasgow, Glasgow G12 8LT, Scotland blnstitute of Photonics, University of Strathclyde, 106 Rottenrow, Glasgow G4 0NW, Scotland The viability of using a thermally oxidised AlAs layer as a robust mask for the transfer of nanometre scale features is reported. Plasma fluorinated AlAs layers were investigated also, but despite selectivities for fluorinated masks of 100:1 between the mask and GaAs, this technique is currently not suitable for pattern transfer as it produces robust, but non-uniform, intermediate masks. In contrast, aspect ratios of 10:1 or greater can be obtained with a selectivity of 100:1 between the oxidised layer and GaAs for sub-micron features. 1. I N T R O D U C T I O N The ability to fabricate sub-wavelength optical structures has introduced exciting new device possibilities, such as diffractive optical elements and photonic bandgap structures. Fabrication technology has now reached the stage where strong light-matter interactions are being observed in epitaxial semiconductor structures, which is opening up a whole new area of device engineering. The fabrication of these new optical devices requires the use of high-resolution lithographic and pattern transfer processes, including electon-beam lithography (EBL) and reactive-ion etching (RIE). Unfortunately, these technologies have limiting factors, with one such limitation being the poor resilience of polymethylmethacrylate (PMMA), the high-resolution resist used in EBL, to most RIE plasmas. Although much research has been done to tackle this issue, some of the new resists are too sensitive and push current electron beam projection systems past their maximum working frequency. Therefore, PMMA is still commonly used today. The problem of its poor resilience has been overcome by the use of an intermediate layer, such as S i O 2, Si3N 4 or certain metals. However, both dielectric and metal masks have their disadvantages: the need to deposit these thin films, which can damage the underlying material; the need to etch the dielectric layer after initial patterning and the difficulty of reliable sub-micron lift-off. An alternative strategy, which addresses some of these
points is the inclusion of the intermediate mask layer at the growth stage and its modification after growth to give improved masking properties. The example of an "in-situ" epitaxial layer that we have investigated is AlAs and its modification by wet thermal oxidation or treatment in a fluorinebased RIE plasma. The wet, thermal oxidation of A1GaAs layers (x>0.8) has been used widely in recent years [1] after Holonyak et al.[2] first demonstrated the process. They showed that at temperatures of approximately 400°C, in the presence of steam, A1GaAs can be transformed into a durable oxide; this oxide has almost the same density as the original A1GaAs layer[3], but is significantly more robust when exposed to RIE plasmas[4]. The treatment of AlAs in fluorine-based plasmas has not been reported yet, to the authors knowledge, but fluorine addition to chlorine plasmas has been shown to provide selective etching of GaAs over A1GaAs[5], by the formation of an A10×Fy film at the A1GaAs surface. Sputtered metal halide films have also been used as an electron-beam resist[6] and are used for IR coatings[7]. The use of plasma treatment to create an amorphous A1Fx layer seeks to extend these techniques. We have investigated a wet, thermally oxidised and a dry, fluorinated AlAs layer as pattern transfer masks and compared these masks with deposited dielectric layers. In particular, a steam oxidised layer is a robust intermediate pattern transfer mask, with selectivities greater than 100:1 over GaAs in SiC14.
0167-9317/99/$- see front matter © 1999 Elsevier Science BY. All rights reserved. PII: S0167-9317(99)00096-9
328
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
2. EXPERIMENTAL The MBE grown wafer consisted of a 20 nm GaAs cap, a 75 nm AlAs layer which was to be modified to act as the intermediate pattern transfer layer, a 919 nm thick GaAs layer and an AlAs etch stop layer 20 nm thick grown on a (100) GaAs substrate. A thickness of 75 nm was chosen for the top AlAs layer as the oxidation rate decreases rapidly for layers thinner than 50 nm[8]. The resist patterns used here were initially generated on a Leica EBPG-HR5 machine at an acceleration voltage of 50 kV. The nominal resolution is 5 nm and the spot size used was 15 nm. The PMMA resist used was Dupont Elvacite (average molecular weight 350,000) dissolved in the solvent o-xylene to give a PMMA weight percentage of 4% (nominal monolayer thickness was 100 nm, when spun-coated at 5,000 rpm and baked at 180 °C). Two layers of PMMA were used to avoid pin-holing effects in the resist and to provide a thick enough layer for high fidelity pattern transfer to both the intermediate dielectric layer (SiO 2 or SiNx) and the heterostructure. The exposed PMMA was then developed in 2:1 IPA:MIBK for 30 seconds; these being the optimum conditions for our laboratory at the time of the experiments. The silica (SiO2) and silicon nitride (SiNx) films used in these experiments were deposited using Plasma Enhanced Chemical Vapour Deposition (PECVD) in an Oxford Plasma Technology gP80 Plus machine. Two reactive-ion etching machines were used, the first machine is an Oxford Plasma Technology RIE80 machines and the other machine is an Oxford Plasma Technology System 100 RIE machine (S100). The wet thermal oxidation was camed out in a novel oxidation rig based on a commercial alloying furnace (Bio-Rad RC2400); Figure 1 shows the furnace schematic. The chamber stage incorporated a low-thermal-mass heater for rapid thermal cycling up to 750°C, with a control unit providing accurate timed heating of the stage and digital temperature control. The furnace temperature used in this work was 380 °C with a nitrogen carrier gas flow of 0.6 l/s flowing through a water bath at 90 °C; these
conditions being previously controllable oxidation.
optimised
for
Flow controller
Water bath t~ ~z
Alloying furnace Exhaust
!
¢:u
~a~
J
Heated stage Figure 1. Schematic of the oxidation rig used. 3.1
AlAs O X I D A T I O N One-dimensional gratings were written with periods varying from 150 to 250 nm with a 50% duty cycle. The resist features were then transferred into the top 100 nm of the epitaxial structure by SIC14 RIE in the S100 machine (15 sccm, 2mT, 200 W), which was sufficient to expose the sides of the AlAs layer. Examination of the etched patterns after cleaving indicated that there were problems of native oxide formation (Fig. 2); native oxide here means the oxide that results from the room temperature reaction of atmospheric oxygen with the A1 in the epitaxial layer. Spontaneous native oxidation is a problem with AlAs, but this problem could be overcome by using a high A1 molar fraction AlxGal_xAs layer (0.8
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
The AlAs layers exposed by the first RIE process were steam oxidised. The samples were oxidised for a maximum of 5 minutes, which was sufficient time to oxidise laterally between adjacent ridge edges; the lateral oxidation rate had been calibrated previously, but the colour change in the epitaxial material, upon oxidation, allows the front to be monitored in-situ via the optical microscope fitted above the furnace chamber. The sample was then returned to the S 100 machine where it was etched in SiC14/O 2 using conditions optimised for the vertical transfer of submicron patterns into GaAs (15/1 sccm, 3 mT, 250 W). Firstly, one sample with an oxidised AlAs layer was etched in order to determine the etch rate (2-5 nm/min). An etch rate of 200 nm/min for GaAs leads to a selectivity of 100:1 between the mask and semiconductor. This high selectivity, when combined with highly anisotropicaly RIE, allows the fabricatation of nanometre scale structures with high aspect-ratios (Fig. 3). 3.2
AlAs FLUORINATION Initial patterns were made using photolithography where stripe waveguides were defined in Shipley S1818 resist. The GaAs cap was then removed using a selective process in the RIE80 machine, which
Fig. 2. Native oxide formation after RIE due to previous thermal cycling
329
etches GaAs over A1GaAs or AlAs (2 sccm, 3 m T , 10 W)[9]. The chamber was then purged with nitrogen for 20 minutes, while still under vacuum, after which the exposed AlAs layer was exposed to a n S F 6 plasma (23 sccm, 50 mT, 50 W) for 2 minutes. This treatment modifies the exposed areas and leaves the resist covered areas untreated. The sample was then removed from the chamber, the photoresist was removed in acetone and thereafter the sample was etched in the S100 machine in SiC14/O 2. Initial inspection revealed that the process did produce a negative mask and reproduced the general outline of the resist features, but closer inspection showed that the mask was not uniform (Fig. 4). At present it is not clear which mechanism modifies the AlAs layer but a possible explanation would be the chemical reaction of the As with the F, dissociated from the plasma, to produce AsF 3, which is volatile. This reaction would leave behind A1 to react with the F and form A1F3, a dielectric, and then oxidise on exposure to the atmosphere after the sample is removed from the RIE chamber. This technique, if developed further, could provide an alternative form of negative lithography, which does not involve a lift-off process.
Fig. 3. Steam oxidised AlAs layer used a mask in SiC14/O2 to etch GaAs.
Fig 4. Fluorinated AlAs layer reproduces general features, but the mask is not uniform
330
C.J.M. Smith et al. / Microelectronic Engineering 46 (1999) 327-330
3.3
C O M P A R I S O N OF M A S K S As the fluorination process does not produce a uniform mask, we shall only compare the oxidised AlAs layer with traditional deposited dielectric films. Table 1 shows the etch rates in a SIC14/O2 RIE plasma for all the intermediate masks being considered and the selectivities of these masks with respect to GaAs. As seen, the two dielectric masks have approximately the same etch rate and selectivity. However, the etch rate of the A1Ox layer is almost an order of magnitude smaller. No limitations of using the A1Ox approach for pattern transfer purposes have been seen in feature sizes down to 50 nm, so this technology has wide-spread appeal for the fabrication of GaAs-based device structures. Moreover, it does not require the potentially damaging post-growth deposition of a dielectric film.
Material Etch rate tnrn/minl Selectivit), SiO2 19 10.5 SiNx 21 9.5 A10~ 2-5 40-100 Table 1. Comparison of different masks in a SiCI4/O 2 plasma; selectivity is of the mask with respect to GaAs. 4. CONCLUSIONS We have demonstrated the ability to incorporate an intermediate layer into the original growth for use as a pattern transfer mask. Dry fluorination produces a non-uniform but robust mask, whereas wet thermal oxidation of an AlAs layer produces a robust mask which allows the subsequent fabrication of highaspect ratio, sub-micron features in GaAs due to selectivites greater than 100:1. Future devices may be fabricated using A1GaAs layers both for pattern transfer purposes and for optical and current confinement via selective oxidation.
5. A C K N O W L E D G E M E N T S We wish to thank colleagues in the Nanoelectronics Research Centre, the Optoelectronics Research Group, the MBE Research Group and the Dry Etch group at Glasgow for technical support. CJMS is supported by a UK EPSRC CASE award with B.T. Laboratories and TFK is supported by a Royal Society Research Fellowship. REFERENCES [1] K. D. Choquette, F. M. Geib, C. I. H. Ashby, R. D. Twesten, O. Blum, H. Q. Hou, D. M. Follstaedt, B. E. Hammons, D. Mathes, and R. Hull, IEEE Journal of Selected Topics in Quantum Electronics, 3, pp. 916-925, 1997. [2] N. Holonyak, Internation patent WO 92/12536, 1992 [3] R. D. Twesten, D. M. Follstaedt, K. D. Choquette, and R. P. Schneider, Applied Physics Letters, 69, pp. 19-21, 1996. [4] C. C. Cheng, A. Scherer, R.-C. Tyan, Y. Fainman, G. Witzgall, and E. Yablonovitch, Journal of Vacuum Science and Technology B, 15, pp. 2764-2767, 1997. [5] S. K. Murad, N. I. Cameron, S. P. Beaumont, and C. D. W. Wilkinson, Journal of Vacuum Science and Technology B, 14, pp. 3668-3673, 1996. [6] H. R. Borsje, H. M. Jaeger, and S. Radelaar, Microelectronic Engineering, 17, pp. 311-314, 1992. [7] S. F. Pellicori and E. Colton, Thins Solid Films, 209, pp. 109-115, 1992. [8] K. M. Geib, K. D. Choquette, H. Q. Hou, and B. E. Hammons, in Vertical-Cavity Surface Emitting Lasers, vol. 3003, K. D. Choquette and D. Deppe, Eds.: SPIE, 1997, pp. 69-74. [9] S. K. Murad, C. D. W. Wilkinson, and S. P. Beaumont, Microelectronic Engineering, 2 3, pp. 357-360, 1994.