A vertical-cavity surface-emitting laser �VCSEL� con-sists of an active medium, often multiple quantumwells, placed between two high-reflecting ��99.9%�mirrors. Most of VCSEL development has beenmade with AlGaAs–GaAs material systems. TheVCSEL has a short cavity, normally a few wave-lengths, which makes the gain per pass quite low.Therefore high-reflecting mirrors are needed for laseroscillation. The short resonator also means that thelaser will operate in a single longitudinal mode.VCSELs are today used commercially in, for instance,short-distance-fiber-based optical data links. Thefact that VCSELs emit light from their surfacesopens the possibility of fabricating dense two-dimensional arrays of VCSELs, which makes themsuitable for laser printing and optical storage.
To obtain sufficient beam control of the emittedlaser light, one must use one or several optical ele-ments, for instance, to collimate or focus the laserbeam. In applications for optical processing and op-tical interconnects, more-advanced beam shaping is
The authors are with The Angstrom Laboratory, Uppsala Uni-versity, SE-751 21 Uppsala, Sweden. M. Karlsson’s e-mail ad-dress is [email protected].
Received 8 May 2001; revised manuscript received 1 October2001.
in many cases desired. Both rudimentary and more-complex optical operations can be resolved by use ofdiffractive optical elements �DOEs� fabricated in suit-able materials. The most attractive way to achievethese operations is to monolithically integrate DOEswith VCSELs, which can lead to better system per-formance.1 For instance, DOEs manufactured inGaAs and integrated with VCSELs that use GaAs asthe substrate material will not suffer from refractive-index or thermal-expansion discontinuities.
We report our studies of the transfer of DOEs orig-inally made in resist into GaAs for further integra-tion with VCSELs.
2. Fabrication of Diffractive Optical Elements
Although binary etched GaAs structures have al-ready been demonstrated, to our knowledge there isonly one publication on the transfer of continuous-relief DOEs into GaAs.2 DOEs can be fabricatedeither as continuous-relief elements or as binarystructures. Because of the stepped approximation,binary elements typically suffer from lower diffrac-tion efficiency. The aim of our study is to integratecontinuous-relief DOEs with GaAs-based VCSELs.The DOEs were fabricated with direct-write electron-beam �e-beam� lithography �system: JEOL JBX-5DII�, which permits continuous-relief elements to beexposed. The DOE pattern was written in electron-sensitive resist �PMGI SF15, MicrolithographyChemical Corporation� spun onto semi-insulating un-doped �100� GaAs substrates. Both blazed gratings
and diffractive lenses of the Fresnel type were ex-posed. Typical feature sizes of the blazed gratingwere a period of 10 �m and a grating depth of 1 �m.We also used test samples of GaAs structured withlamellar gratings in Shipley 1813 photoresist to beable to measure the etch rates in GaAs, e-beam resist,and photoresist. Knowing the etch rate in photore-sist is of interest because photoresist can also be acandidate for use in the structuring of micro-opticalstructures, by, e.g., photolithography.
All samples were examined with atomic-force mi-croscopy �AFM�. The transfer of the diffractivestructures from resist into GaAs was made in aninductively coupled plasma �ICP� etching system.Previous studies focused on using either reactive ionetching3,4 �RIE� or ion beam milling for the transferprocess.5 However, with ICP, which is a high-density plasma system, it is possible to generate iondensities 3 to 4 orders of magnitude higher than inRIE.6 Ion beam milling suffers from low etch rates,and it is difficult to control the selectivity betweendifferent materials. Another advantage of ICP etch-ing compared with RIE is that one is able to controlthe ion density and the bias independently, whichmakes it possible to decrease plasma damage on thewafer. A relatively high plasma density can thus bemaintained at low pressures ��1 mTorr�, which leadsto high etch rates and anisotropic etching.
3. Etching Experiment
For the transfer of DOEs from e-beam resist intoGaAs, we used a home-built ICP etching system.The ICP source operates at 13.56 MHz. The ICPsystem is also equipped with a water-cooled rf-powered wafer chuck for independent bias control�operating at 13.56 MHz�. All samples weremounted by use of vacuum grease on the rf-poweredaluminum chuck. In this study we used differentetch gases under various etch conditions. Etching ofGaAs is usually done in Cl-based plasma chemistry,most with using Cl2, BCl3, or both.6–11 BCl3 isknown to be an effective attractor of water vapor andwill immediately remove the native oxide layer fromthe GaAs surface, which makes BCl3 suitable for re-producible GaAs etching. However, BCl3 may causea problem because B and its compounds can be de-posited upon the wafer surface. By adding Ar it ispossible to sputter these nonvolatile products away;Ar also stabilizes the plasma and acts as a heat re-mover. Adding Cl2 will further increase the etchrate of GaAs. O2 is normally used to etch resist, andN2 can be added to BCl3 plasmas to degrade theresist.7 Electronic-grade BCl3, Cl2, Ar, N2, and O2 indifferent compositions, with a total gas flow of 15cubic centimeters per minute at STP �sccm�, wereused as etch gases. The ICP source power was var-ied from 450 to 600 W, the bias from �80 to �200 V,and the process pressure from 2 to 7 mTorr �etchtime, 5–10 min�. The etch depth of the test sample�lamellar grating, 50-�m period� was measured witha stylus profilometer over the binary structures. Wealso measured the remaining resist thickness on the
test sample with an optical interferometer to calcu-late the etch rates in GaAs and Shipley resist. Wethen examined the transferred DOE with AFM tomeasure the height of the structure, the surface mor-phology, and the accuracy of the transfer process.Also, a perspicuous inspection of the transfered DOEwith scanning electron microscopy �SEM� was made.
4. Results and Discussion
A. Transfer of the Diffractive Elements
It is of vital importance to keep the reaction chamberfree from any unwanted gases and materials thatcould react with the etch gases and contaminate thesubstrates. Problems appeared when efforts weremade to enhance the etch rate in e-beam resist byaddition of O2 to the plasma. The GaAs surface wasthen covered with a dark, grasslike corrugation, prob-ably as a result of micromasking of B2O3 when theamount of O2 exceeded 10% of the total gas amount.
The transfer of the DOEs for various etch gas com-positions, ICP powers, biases, and pressures wasstudied. At the start a mixture of BCl3 and Ar wasused for the transfer. Figure 1�a� shows a blazedgrating �period, 10 �m; depth, 635 nm� in e-beamresist, and Fig. 1�b� shows the same grating etchedinto GaAs. One can notice the low surface rough-ness of the etched element. The etch conditionswere as follows: gas flows of 5 sccm BCl3 and 10sccm Ar, chamber pressure of 2.5 mTorr, 600 W ofICP power, �200-V bias, and 5.5-min etch time,which gave etch rates of 435 and 500 nm�min inresist and GaAs, respectively �190 nm�min in Shipleyresist�. These conditions give a selectivity betweenGaAs and e-beam resist of 1.15, which means that theheight of the structure in this case will be slightlystretched. Lowering the bias to �150 V and retain-ing the other parameters as above yielded etch ratesof 400 nm�min in resist and 450 nm�min in GaAs�160 nm�min in Shipley resist�, for a selectivity of1.12. In the transfer of optical elements it is conve-nient to have selectivities both less than and largerthan 1. One can then compensate for the height ofthe structures to match the wavelength of the illumi-nating light. The height of a diffractive structure isdirectly related to the design wavelength; deviationsfrom this optimum relief height will lower the diffrac-tion efficiency. By adding Cl2 to the plasma weachieved a selectivity of 0.75. The decrease of selec-tivity was somewhat unexpected, because the etchrate should increase in GaAs when Cl2 is added, ow-ing to the higher density of reactive Cl2 species.This increase did occur, but the etch rate in thee-beam resist increased as well, by a higher factorthan in GaAs. Figure 2 shows a plot of selectivityrelative to the amount of Cl2. Similar selectivitybehavior occurred when N2, instead of Cl2, was addedas the etch gas. This behavior can be explained bythe fact that when N2 is added to a BCl3 plasma thelatter will have a higher density of reactive Cl2 spe-cies, owing to enhanced dissociation of BCl3.8
�150 V; pressure, 2.5 mTorr; 5-sccm BCl3, 10-sccmAr� we got a selectivity of 1.05. The etch rates werethen 290 nm�min in GaAs and 275 nm�min in e-beamresist �85 nm�min in Shipley�. The lowered etchrates are due to lowered density of reactive chlorineradicals. If instead the bias were lowered to �80 V,and for an ICP power of 600 W and other conditionsas above, the selectivity increased to 1.42, which gave
Fig. 1. �a� Top, AFM scan of a blazed grating �10-�m period;depth, 1 �m� in an e-beam resist; bottom, AFM picture of thegrating surface. �b� Top, AFM scan of the grating shown in �a�transferred into GaAs by ICP etching; bottom, AFM picture of theetched grating surface.
etch rates of 340 and 240 nm�min in GaAs ande-beam resist, respectively �Shipley, 75 nm�min�.This lowering of the etch rate in GaAs comes from adecrease in physical bombardment. The etch rate ine-beam resist, owing to a decrease in bias, drops evenmore, and the Shipley resist etch rate goes almost tozero. These results agree well with earlier reports ofthe etch rate dependence on Shipley resist.9 Whenthe pressure was increased to 7 mTorr �ICP power,600 W; bias, �150 V; 5-sccm BCl3 and 10-sccm Ar�,the etch rates increased: GaAs to 600 nm�min,e-beam resist to 530 nm�min, and Shipley resist to210 nm�min as a result of increased residence time ofthe reactive species upon the substrate surface. Allthese results were obtained with a 10-�m blazedgrating with a height of 1 �m.
Blazed gratings with a period of 3 �m �height, 600nm� were also transferred into the GaAs substrate.Figure 3�a� shows a blazed grating before etching,and Fig. 3�b� shows the grating after etching. In thiscase the distortion of the profile is greater than for the10-�m blazed grating. This distortion of the profileis fundamental for dry etching and comes from thefact that the etch rate of a structured material islocally dependent on the angle at which the materialis exposed to the reactive etch species. From ourexperiments we could see that it was the bias thatwas the main factor in the distortion, whereaschanges in ICP power, chamber pressure, and gascomposition had little or no effect on the profile dis-tortion. For higher bias ��200 V�, the least distor-tion occurred. During normal dry etching, when amask is used to open curtain areas on the substrateand then etch down to a certain depth this distortiondoes not do much harm. In our case, when we trans-fer an overlying structure into a substrate �GaAs� foroptical purposes, we need to have control of the sur-face distortion and of the effect that it will have on theoptical performance of the structures, as we discussin Subsection 4.B. Another difference between thetransferred 10-�m blazed grating and the 3-�mblazed grating lies in the selectivity. For the 10-�mgrating we had a selectivity of 0.75–1.42, but for the
Fig. 2. Selectivity between GaAs and e-beam resist versusamount of Cl2 �ICP power, 600 W; bias, �150 V; pressure, 2.5mTorr; 8-sccm Ar; total flow, 15 sccm�. The etch rates for GaAs,e-beam resist, and Shipley 1813 resist are shown �solid curves�.Based on etching experiments of a 10-�m blazed grating.
3-�m grating we had a selectivity of 0.35–0.77. Thisreduction in selectivity comes partially from the mi-croloading effect12; a smaller opening to the substratematerial leads to a smaller depth during dry etching,but it is also an effect of the angular dependence ofthe etch rate.
Fresnel lenses were also transferred into GaAswith good results. Figure 4 shows a surface scan ofa Fresnel lens before and after etching. As can been
Fig. 3. AFM scans of �a� a blazed grating �3-�m period; depth, 600nm� in e-beam resist and �b� the grating shown in �a� transferredinto GaAs by ICP etching.
Fig. 4. AFM scans of �a� a Fresnel lens in e-beam resist and �b� aFresnel lens transferred into GaAs.
seen, the Fresnel lens exposed to the e beam alreadyshows different relief heights over the variousFresnel zones as a result of scattering of electronswhen the resist is exposed �proximity effect�.13 Be-cause the period becomes smaller �seen from the cen-ter� in a Fresnel lens, the microloading and theangular effects are viewed well in this sort of struc-ture. The height difference in the transferredFresnel lens can of course be compensated for in thee-beam writing step. Figure 5�a� shows an AFMscan of a Fresnel lens transferred into GaAs; Fig. 5�b�shows a SEM picture of the same lens.
The etch process was highly stable for all the pro-cess parameters investigated and produced smoothetched surfaces with low surface roughness. Theroot-mean-square roughness for the etched surfaceswas �1 nm, measured with AFM.
Fig. 5. AFM picture of a Fresnel lens in GaAs; �b� SEM picture ofa Fresnel lens in GaAs �the faint lines are a result of Cartesiane-beam exposure�.
By using a two-dimensional topography etch-simulating computer program �2Dinese �Ref. 14��,one can predict the shape distortion of the micro-optical surface during the etch process. The mainfactor that is responsible for the distortion is thedependence of the etch rate on the surface relief an-gle.15,16 We do not know the exact dependence, butby trying various likely dependencies we were able tofind good agreement between etched results and sim-ulated etched profiles. Figure 6 shows a blazed grat-ing profile before and after etching, simulated in2Dinese by use of the best-fitted etch rate relative tothe angle of incidence �Fig. 7�. One can see that thesimulated etched profile �Fig. 6�b�� has a profile sim-ilar to the actually etched structure �Fig. 1�b��.
Using this software, one can also see that the dis-tortion of the diffractive structures will increase asthe structure is further etched into the substrate. Ifthe etching continues, the diffractive structure willfinally disappear and the surface will be almost flat�Fig. 6�c��. This fact means that the distortion in our
Fig. 6. �a� Profile of a blazed grating before etch simulation in 2Dintransferred into GaAs by use of data from Fig. 7 �etch simulationthickness of e-beam resist, 2 �m for the same angle dependenciessimulation in 2Dinise for the blazed grating, with all parameters�d� AFM scan over the same grating as in Fig. 1�b� but after furth
case could be less if the optical structures, exposedfirst in e-beam resist, were developed precisely downto the GaAs substrate. In our experiment all opticalstructures had a layer of 1-�m e-beam resist betweenthe GaAs substrate and the bottom of the structures.In Fig. 6�d� the blazed grating from Fig. 1�a� is shownafter continued etching for 5.5 min; one can see that
�b� Etch profile, after simulation in 2Dinise, for the blazed gratinge, 5.5 min; etch rate in GaAs and e-beam resist, 400 nm�s; totalth GaAs and e-beam resist�. Cf. Fig. 1�b�. �c� Etch profile after�b� except that the simulated etch time is 11 min. Cf. Fig. 6�d�.
5-min etching.
Fig. 7. Best fitted etch rate versus angle of incidence for etchingresist upon GaAs.
ise.tim
for boas forer 5.
height is reduced and the distortion is increased.We also performed an experiment in which westopped the etch process while the diffractive struc-ture had not yet reached the GaAs substrate. Wethen examined the optical structures by AFM andcould verify that most of the final distortion had al-ready occurred.
C. Optical Testing and Evaluation
A transferred blazed grating with a period of 10 �mwas used for optical evaluation. We used a 940-nmsingle-mode VCSEL �Avalon Photonics, Zurich� to-gether with a lens collimating the light to a diameterof �1 mm. The size of the blazed grating area was5 mm � 5 mm. In this test a blazed grating with afive-step approximation was used. The ideal etchdepth for this grating is �5�6���n � 1� and has atheoretical diffraction efficiency in the first order of1 � sinc2�1�N�. Using relevant numbers resultedin an optimal etch depth of 332 nm and a theoreticaldiffraction efficiency of 87.5%. Figure 8 shows themeasured intensity in some diffraction orders. Thereason that we could not use the grating structureshown in Fig. 1 was that the blazed grating area�200 �m � 200 �m� was too small. The measureddiffraction efficiency in the first order was 67%. Noantireflection coating was provided. The diffractionefficiency does not reach the theoretical maximumchiefly because of the shape distortion mentionedabove.
We also used data from a two-dimensional AFMscan of the diffractive structures �i.e., a blazed grat-ing� to calculate the diffraction pattern. We madethis calculation by taking the Fourier transform ofthe surface structure. Figure 9�a� shows the calcu-lated diffraction in some orders for a 10-�m blazedgrating after e-beam exposure; Fig. 9�b� shows thesame grating transferred into GaAs �we used theAFM data from the scans shown in Figs. 1�a� and1�b��. The calculated diffraction efficiencies in thefirst order were 83% and 60%. As these simulationsare based on true continuous profiles, the theoreticalmaximum of the first-order diffraction efficiency is100%. In the same way we calculated the diffractionefficiencies for the 3-�m blazed grating as 60% and34% for unetched and etched gratings, respectively�cf. Figs. 3�a� and 3�b��. This method is useful for a
Fig. 8. Measured intensity in some diffraction orders for a 10-�mGaAs blazed grating.
quick evaluation of the optical dependence of theshape distortion on the diffractive structures. How-ever, a direct comparison of calculated and measureddiffraction efficiencies is difficult because the calcu-lation is based on a surface scan only, whereas themeasured data integrate over the whole surface thatis illuminated.
5. Conclusion and Directions for Future Research
We have studied the transfer of continuous-relief dif-fractive structures into GaAs by ICP etching. ABCl3–Ar chemistry was found to give satisfying re-sults with possibilities of facilitating selectivity be-tween GaAs and e-beam resist of 1.0–1.42. Theselectivity between GaAs and e-beam resist also de-pends on the period and the angle of the diffractivestructures, the height of the structures, the thicknessof the e-beam resist between the GaAs substrate andthe bottom of the structures, and finally on the timeof further etching after the structure is etched intoGaAs. By adding Cl2 or N2 it is possible to lower theselectivity to 0.75. The transfer process producedsmooth surfaces with low surface roughness �rms, �1nm�.
The distortion that comes from the angular depen-dence of the etch rate was reduced by use of a higher-bias voltage of the ICP. In our etch simulations wefound that it is of great importance to reduce thedistortion in the e-beam exposure �proximity effect�because that distortion will be magnified in the trans-fer step. The distortion has a vital influence on thediffraction efficiency of the optical structure. Wecan also see that the distortion sets a limit on howsmall the periods that we can make be, because
Fig. 9. �a� Calculated diffraction patterns �a� for a 10-�m blazedgrating after e-beam exposure �with the AFM scan from Fig. 1�a� asinput� and �b� for the blazed grating �as in �a�� transferred intoGaAs �with the AFM scan from Fig. 1�b� as input�.
smaller features will suffer more distortion and hencegive lower diffraction efficiency. The diffraction ef-ficiency for our 10-�m transferred blazed grating was67% in the first order, with a theoretical value of 87%.We now intend to fabricate the diffractive structureupon the same substrate that was used for theVCSEL.
We thank Jorgen Bengtsson of Chalmers Univer-sity of Technology for help with e-beam exposure andIlia Katardjiev of Uppsala University for help withthe 2Dinise program. This research was supportedby SUMMIT, The Swedish Center of Surface andMicrostructure Technology, which is financed by theSwedish Agency for Innovation Systems.
References1. H. Martinsson, J. Bengtsson, M. Ghisoni, and A. Larsson,
“Monolithic integration of vertical-cavity surface-emitting la-ser and diffractive optical element for advanced beam shap-ing,” IEEE Photon. Technol. Lett. 11, 503–505 �1999�.
2. F. S. Pool, D. W. Wilson, P. D. Maker, R. E. Muller, J. J. Gill,D. K. Sengupta, J. K. Liu, S. V. Bandara, and S. D. Gunapala,“Fabrication and performance of diffractive optics for quantumwell infrared photodetectors,” in Infrared Detectors and FocalPlane Arrays V, E. L. Dereniak and R. E. Sampson, eds., Proc.SPIE 3379, 402–409 �1998�.
3. F. Nikolajeff, T. A. Ballen, J. R. Leger, A. Gopinath, L. TzuChen, and R. C. Williams, “Spatial-mode control of vertical-cavity lasers with micromirrors fabricated and replicated insemiconductor materials,” Appl. Opt. 38, 3030–3038 �1999�.
4. E. M. Strzelecka, G. D. Robinson, L. A. Coldren, and E. L. Hu,“Fabrication of refractive microlenses in semiconductors bymask shape transfer in reactive ion etching,” Microelectr. Eng.35, 385–388 �1997�.
5. H. Sankur, R. Hall, E. Motamedi, W. Gunning, and W. Ten-nant, “Fabrication of microlens arrays by reactive ion milling,”in Miniaturized Systems with Micro-Optics and Micromechan-ics, M. Motamedi, ed., Proc. SPIE 2687, 150–155 �1996�.
6. R. J. Shul, G. B. McClellan, R. D. Briggs, D. J. Rieger, S. J.Pearton, C. R. Abernathy, J. W. Lee, C. Constantine, and C.
Barratt, “High-density plasma etching of compound semicon-ductors,” J. Vac. Sci. Technol. A 15, 633–637 �1997�.
7. C. Constantine, R. J. Shul, C. T. Sullivan, M. B. Snipes, G. B.McClellan, M. Hafich, C. T. Fuller, J. F. Mileham, and S. J.Pearton, “Etching of GaAs�AlGaAs rib waveguide structuresusing BCl3�Cl2�N2�Ar electron cyclotron resonance,” J. Vac.Sci. Technol. B 13, 2025–2030 �1995�.
8. T. Maeda, J. W. Lee, R. J. Shul, J. Han, J. Hong, E. S. Lambers,S. J. Pearton, C. R. Abernathy, and W. S. Hobson, “Inductivelycoupled plasma etching of III–V semiconductors in BCl3-basedchemistries. I. GaAs, GaN, GaP, GaSb and AlGaAs,” Appl.Surf. Sci. 143, 174–182 �1999�.
9. S. Agarwala, O. King, S. Horst, R. Wilson, D. Stone, M. Da-genais, and Y. J. Chen, “Response surface study of inductivelycoupled plasma etching of GaAs�AlGaAs in BCl3�Cl2,” J. Vac.Sci. Technol. A 17, 52–55 �1999�.
10. J. W. Lee, J. Hong, E. S. Lambers, C. R. Abernathy, S. J.Pearton, W. S. Hobson, and F. Ren, “Plasma etching of III–Vsemiconductors in BCl3 chemistries. I. GaAs and relatedcompounds,” Plasma Chem. Plasma Process. 17, 155–167�1997�.
11. Y. Z. Juang, Y. K. Su, S. C. Shei, and B. C. Fang, “Comparingreactive ion etching of III–V compounds in Cl2�BCl3�Ar andCCl2F2�BCl3�Ar discharges,” J. Vac. Sci. Technol. A 12, 75–82�1994�.
12. C. Hedlund, H. O. Blom, and S. Berg, “Microloading effect inreactive ion etching,” J. Vac. Sci. Technol. A 12, 1962–1965�1994�.
13. F. Nikolajeff, J. Bengtsson, M. Larsson, M. Ekberg, and S.Hard, “Measuring and modeling the proximity effect in direct-write electron-beam lithography kinoforms,” Appl. Opt. 34,897–903 �1995�.
14. I. V. Katardjiev, G. Carter, M. J. Nobes, S. Berg, and H. O.Blom, “Three-dimensional simulation of surface evolution dur-ing growth and erosion,” J. Vac. Sci. Technol. A 12, 61–68�1994�.
15. C. Hedlund, L. B. Jonsson, I. V. Katardjiev, S. Berg, and H. O.Blom, “Angular dependence of the polysilicon etch rate duringdry etching in SF6 and Cl2,” J. Vac. Sci. Technol. A 15, 686–691 �1997�.
16. W. Daschner, M. Larsson, and S. H. Lee, “Fabrication of mono-lithic diffractive optical elements by the use of e-beam directwrite on an analog resist and a single chemically assistedion-beam-etching step,” Appl. Opt. 34, 2534–2539 �1995�.
