JournalofELECTRONICMATERIALS,Vol.36,No.2,2007 RegularIssuePaper DOI:10.1007/s11664-006-0057-5 (cid:1)2007TMS InAlSb/InAs/AlGaSb Quantum Well Heterostructures for High-Electron-Mobility Transistors BRIAN R. BENNETT,1,3 J. BRAD BOOS,1 MARIO G. ANCONA,1 N. A. PAPANICOLAOU,1 GRAHAM A. COOKE,2 and H. KHEYRANDISH2 1.—NavalResearchLaboratory,Washington, DC20375,USA.2.—CSMA-MATS,Stoke-on-Trent, StaffordshireST47LQ,UK.3.—E-mail:[email protected] HeterostructuresforInAs-channelhigh-electron-mobilitytransistors(HEMTs) were investigated. Reactive AlSb buffer and barrier layers were replaced by more stable Al Ga Sb and In Al Sb alloys. The distance between the 0.7 0.3 0.2 0.8 gate and the channel was reduced to 7–13 nm to allow good aspect ratios for veryshortgatelengths.Inaddition,n+-InAscaps weresuccessfully deposited on the In Al Sb upper barrier allowing for low sheet resistance with rela- 0.2 0.8 tivelylowsheetcarrierdensityinthechannel.Theseadvancesareexpectedto result in InAs-channel HEMTs with enhanced microwave performance and better reliability. Key words: InAs, high-electron-mobility transistors (HEMTs), molecular beam epitaxy (MBE), FET, InAlSb, AlGaSb have been reported in the S-band,8 X-band,9,10 INTRODUCTION Ka-band,11andW-band.5,12,13Forexample,athree- High-electron-mobility transistors (HEMTs) with stage W-band low-noise amplifier (LNA) was InAschannelsandAlSbbarrierswerefirstreported demonstrated with 11 dB gain at a total chip dissi- over 15 years ago, as discussed in a recent review.1 pation of only 1.8 mW at 94 GHz.12 This is a factor Advantagesofthismaterialsystemincludethehigh of 3 lower power than comparable InP-based LNAs electron mobility (30,000 cm2/V s at 300 K) and at the same frequency. velocity (4 · 107 cm/s) of InAs and a large conduc- Significant increases in f and f have been T max tion band offset between InAs and AlSb (1.35 eV). achievedin InP-basedHEMTs byreducing thegate The large offset results in good carrier confinement length to less than 100 nm.14–16 A goal of this work and enhanced radiation tolerance.2 There has been is to develop heterostructures for sub-100-nm InAs- renewed interest in recent years. For example, two channel HEMTs. In order to maintain good charge groups have achieved 100-nm-gate-length InAs- control (e.g., minimal short-channel effects) in the channel HEMTs with unity-current-gain cutoff fre- scaled devices, it is essential that the gate-to-chan- quency, f , and unity-power-gain cutoff frequency, nel separation also be reduced. An additional T f , values of 200–300 GHz.3,4 Compared to state- advantage of reduced gate-to-channel separation is max of-the-art InP-based HEMTs with the same gate thatthethresholdvoltage,V ,willbecloserto0 V. th length, the InAs HEMTs provide equivalent high- To quantify this, we performed modeling of V as a th speed performance at 5–10 times lower power dis- function of sheet density and vertical spacing and sipation.4 These transistors exhibit low microwave compared the results to experimental data.1 At a noise, with noise figures of 0.6–0.8 dB at 10 GHz.1 density of 2 · 1012/cm2, |V | decreases from 1.0 V th High-frequencyperformancehasalsobeenachieved ataseparationof22 nmto0.5 Vfor10 nm,withall in antimonide-free InAs-channel HEMTs with In- in depletion mode. Low |V | is important because th AlAs barriers,5,6 as well as InSb-channel HEMTs.7 it allows operation at low drain voltage and hence Inthelast3 years,circuitsbaseduponInAsHEMTs low power dissipation. In addition, low |V | th reduces the gate-to-drain voltage and resulting breakdown phenomena.17 Additional goals of this (ReceivedSeptember20,2006;acceptedOctober30,2006) study include the elimination of highly reactive 99 Report Documentation Page Form Approved OMB No. 0704-0188 Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. 1. REPORT DATE 3. DATES COVERED SEP 2006 2. REPORT TYPE 00-00-2006 to 00-00-2006 4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER InAlSb/InAs/AlGaSb Quantum Well Heterostructures for 5b. GRANT NUMBER High-Electron-Mobility Transistors 5c. PROGRAM ELEMENT NUMBER 6. AUTHOR(S) 5d. PROJECT NUMBER 5e. TASK NUMBER 5f. WORK UNIT NUMBER 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION Naval Research Laboratory,4555 Overlook Avenue REPORT NUMBER SW,Washington,DC,20375 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S) 11. SPONSOR/MONITOR’S REPORT NUMBER(S) 12. DISTRIBUTION/AVAILABILITY STATEMENT Approved for public release; distribution unlimited 13. SUPPLEMENTARY NOTES 14. ABSTRACT 15. SUBJECT TERMS 16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF ABSTRACT OF PAGES RESPONSIBLE PERSON a. REPORT b. ABSTRACT c. THIS PAGE Same as 6 unclassified unclassified unclassified Report (SAR) Standard Form 298 (Rev. 8-98) Prescribed by ANSI Std Z39-18 100 Bennett,Boos, Ancona,Papanicolaou, Cooke,and Kheyrandish AlSb from the heterostructures and the addition of transistor structures in the past; more details are n+ cap layers for lower access and contact resis- given elsewhere.18 Table I includes a summary of tances. all the samples grown for this study. Hall/van der Pauw transport measurements were made on EXPERIMENTAL 5 · 5 mm2squaresatroomtemperatureand0.37 T. Results were confirmed at 1.0 T for selected sam- A typical heterostructure is shown in Fig. 1a. ples. Some samples were grown for whole-wafer Samples were grown by solid-source molecular circuit processing. Hence, destructive transport beam epitaxy (MBE) on a Riber 21T system. The measurements were not possible. For these sam- temperature was held constant near 500(cid:2)C for the ples, we report the mobility and sheet density AlGaSb buffer, InAs channel, and InAlSb spacer measuredbyaLehightonmobilitymappingsystem. layers. The temperature was decreased (usually to For all samples, we also report the resistivity mea- 450(cid:2)C) for the Te d-doping, the InAlSb barrier, and sured by a different Lehighton resistivity mapping the InAs cap layers. The temperature of the GaTe system. Atomic force microscopy (AFM) measure- cell was 600(cid:2)C for all samples reported here. This ments were performed on selected samples to yield corresponds to a density of about 2 · 1018/cm3 at a root-mean-square (rms) roughness. X-ray diffrac- 1 ML/s growth rate. The Te dose was varied by tion measurements were made on a double-crystal changing the length of time the stagnant surface system using Cu K radiation. Ultra-low-energy wasexposedtoaGaTeflux(soaktime).Thegrowth a secondary ion mass spectroscopy (SIMS) measure- process was similar to that used for InAs/AlSb ments were conducted using 300 eV Cs+, 500 eV Cs+, and 1,000 eV Cs+ ions. (a) RESULTS AND DISCUSSION A band diagram of the InAlSb/InAs HEMT InAs 2 nm cap structure as obtained from a calibrated density- gradient simulation is shown in Fig. 1b. Both the In Al Sb 2.5 or 6.5 nm barrier 0.2 0.8 AlGaSb and InAlSb layers provide a large conduc- tion band offset with respect to the InAs, allowing good confinement of electrons. A significant differ- In Al Sb 4.5 or 6.5 nm spacer ence between this structure and our earlier 0.2 0.8 HEMTs18,19 is that the InAlSb upper barrier re- places InAlAs and AlSb.20 One reason for this InAs 12 nm channel change was to avoid the use of pure AlSb, because Al Ga Sb 1500 nm buffer Miya et al. showed that replacing 20–40% of the Al 0.7 0.3 with Ga in AlGaAsSb greatly reduces oxidation.21 The x-ray diffraction data for sample Q is shown GaAs 0 or 230 nm buffer in Fig. 2. Peaks are visible for the GaAs substrate, AlGaSb buffer, InAs channel, and InAlSb barrier SI GaAs (001) substrate layers. Assuming the nominal layer thicknesses showninFig. 1a(13-nm InAlSb), simulationsusing (b) 1.5 RADS software from Bede Scientific yielded a good match to the experimental data for layers of Al 0.68- Ga Sb, R = 97.9%; InAs, R = 15%; and In 0.32 0.207- 1 Al Sb, R = 0%, where R is the degree of 0.793 Al Ga Sb relaxation. Such fits are not unique, because, for 0.7 0.3 example, other combinations of composition and ) V 0.5 relaxation of the Al Ga Sb layer can yield the e x 1-x ( sameBraggangle.Thevalueofx = 0.68waschosen y InAs g because it is in reasonable agreement with our r e 0 nominal value of 0.70 from MBE reflection high- n E energy electron diffraction (RHEED) oscillation calibrations and because the corresponding relaxa- -0.5 tionvalueisclosetowhatwehavefoundinthepast In Al Sb 0.2 0.8 for pure AlSb layers of similar thicknesses on GaAs.18IfR > 0isassumedfortheIn Al Sblayer, y 1-y -1 then larger values of y are required and are not 0 5 10 15 20 25 30 35 40 consistent with our nominal value. For the InAs Position (nm) layer, the experimental peak position can also be matchedforR = 0%if2%Sbcross-contaminationis Fig.1. (a)CrosssectionoftheInAlSb/InAs/AlGaSbHEMTstructure and(b)calculatedbandstructure. assumed. Our results suggest that the InAlSb and InAlSb/InAs/AlGaSb Quantum WellHeterostructures 101 Table I. GrowthParameters andRoom-Temperature Transport ResultsforAllSamples: R isthe Sheet SH1 ResistancefromRoom-TemperatureHall/VanDerPauwMeasurements(SamplesA-FandI-K)oraLehighton MobilityMapper(Samples G,H, andL-Q);and R isthe SheetResistance fromaLehighton Resistivity SH2 Mapper. The n+Cap Layer is20-nm InAsDoped with (cid:1)1 ·1019/cm3 Te Upper GaTe Soak SbFlux Spacer Bar Soak Temp. During GaAs n+ n l R R s 300K SH1 SH2 Sample (nm) (nm) (sec) ((cid:2)C) d Dop. Buffer Cap (·1012 cm)2) (cm2/V s) (W/h) (W/h) A 4.5 2.5 0 — — Y N 0.3 12,800 1600 650 B 4.5 2.5 33 450 Y Y N 1.6 20,000 200 176 C 6.5 2.5 33 450 Y Y N 1.4 23,500 197 184 D 6.5 2.5 30 450 Y N N 1.4 21,300 212 174 E 6.5 2.5 30 500 Y N N 1.2 20,000 256 330 F 6.5 2.5 30 450 N N N 1.2 18,100 289 218 G 6.5 2.5 30 450 Y N N 1.4 25,000 178 232 H 6.5 2.5 30 450 Y N N 1.4 24,000 186 228 I 6.5 2.5 30 450 Y N Y — — — 78 I 6.5 2.5 30 450 Y N Y 2.0 23,200 135 — J 6.5 2.5 50 470 Y N N 2.4 21,100 125 171 K 6.5 2.5 50 430 Y N N 2.4 18,400 144 195 L 6.5 6.5 30 450 Y N N 1.6 20,100 188 177 M 6.5 6.5 30 450 Y N Y — — — 74 N 6.5 6.5 30 450 Y N Y — — — 78 O 6.5 6.5 30 450 Y N N 1.5 19,700 211 160 P 6.5 6.5 30 450 Y N N 1.1 23,100 249 243 Q 6.5 6.5 30 450 Y N N 0.9 22,200 327 326 InAs layers are nearly coherent with respect to the widthto3.3 nm.Onthatbasis,weconcludethatthe relaxed AlGaSb buffer layer, with the InAs in ten- 300 eV profile should be a good approximation to sion (1.1% mismatch) and the InAlSb in compres- therealdistribution.ThisprofileisshowninFig. 3. sion (1.3% mismatch). The y-axis calibration is based upon our expected ThedepthresolutionofconventionalSIMScanbe total dose of Te, which was derived from transport poor due tointermixing induced bythe high-energy measurements on thick layers of GaAs(Te). The ions. Using lower ion energies can reduce this profile is relatively symmetric suggesting that problem. Exploiting this strategy, we analyzed diffusion and not segregation of Te in InAlSb dom- sample I with SIMS and found the d-doped Te pro- inates. The Te concentration drops by over two file full-width at half-maximum (FWHM) decreased orders of magnitude from the peak to the top from 4.7 nm to 3.5 nm when the ion energy was interface with the InAs quantum well where the reduced from 1,000 eV to 500 eV. A further reduc- value is near 2 · 1016/cm3. Throughout most of tion to 300 eV resulted in only a small change in the quantum well, background concentrations in the mid-1015/cm3 range are observed. Although Te may suffer fromsignificant desorption,segregation, and diffusion for some materials and growth condi- tions,22,23 the results of this study suggest that it can be a suitable dopant in InAlSb. These results are comparable to the Si d-doping of InAlAs, often used for InP HEMTs, where SIMS using 1,000-eV ions yielded a FWHM near 5 nm.24 In Fig. 4, we plot the room-temperature mobility versus the sheet carrier density for 12 samples (Table I). The intentionally doped samples have mobilities between 18,000 and 25,000 cm2/V s and densitiesbetween 1.2 and 2.4 · 1012/cm2. The sheet densitycanbecontrolledbyvaryingthesoaktimeof the Te d doping. Several other trends can also be observed from Table I. Three samples were grown with a 230-nm GaAs buffer layer. This should pro- vide a smoother surface for the subsequent deposi- tion of the AlGaSb buffer layers. The data show no apparent difference in the mobility and density for Fig.2. Double-crystal x-ray diffraction of sample Q near the (004) reflection.Thelowertraceisasimulation(refertotext). sampleswithandwithouttheGaAsbufferlayer.As 102 Bennett,Boos, Ancona,Papanicolaou, Cooke,and Kheyrandish the Sb shutter but observed no significant differ- ences. Our standard spacer thickness was 6.5 nm, but good mobility (20,000 cm2/V s) was achieved for 4.5 nm (sample B). The upper barrier thickness was either 2.5 or 6.5 nm with the lower value providing a better aspect ratio for very short gates, but the higher value probably yielding lower gate leakage currents (as has been demonstrated for 70- nm InGaAs-channel HEMTs).27 Werecentlyreportedtransistorcharacteristicsfor adevicewiththestructureofFig. 1a(6.5-nmspacer and 2.5-nm upper barrier) and a 350-nm gate length. The devices have a DC transconductance of 1 S/mm and an intrinsic f of 95 GHz,28 a result T consistent with other InAs HEMTs at comparable gate lengths.1 The significance of this result is that the structure contains no AlSb,29 and the distance Fig.3. Low-energySIMSshowingTeddopinginInAlSb. from the gate to the channel is only 9 nm. For comparison, the gate-to-channel separations for 100-nmInAsHEMTswere14–19 nm.3,10,30Wenote that a study of InP HEMTs demonstrated good microwave performance if the gate length was greater than the sum of the gate-to-channel sepa- ration and the channel thickness.16 For our struc- tures, the sum is 19–25 nm. The scaling properties may be different for InAs HEMTs. Experimental work to achieve sub-100-nm gates is in progress in our laboratory. Low sheet resistance, R , is desirable for good SH RFperformance.TheR valuesaslowas100W/sq SH can be achieved by high channel doping (e.g., n (cid:1) s 3 · 1012/cm2, l (cid:1) 20,000 cm2/V s).18 However, high valuesofn canresultinlargeV .1Toaddressthis s th dilemma, we are investigating a recessed-gate approach with a thick n+ cap layer. We initially grew n+ InAs caps on top of our standard InAlAs barrier layers. There is a large lattice mismatch Fig.4. Room-temperature mobility as a function of sheet density. betweentheInAlAs(a = 5.9A˚)andtheotherlayers TheGaTecelltemperaturewasfixedandthesoaktimeswerevaried o as indicated. All samples included here were capped with 2-nm (ao(cid:1)6.1A˚).Asexpected,duringInAlAsgrowth,the undopedInAs.NotethatsamplesPandQweregrownafteravent RHEED pattern becomes spotty, indicating three- andreloadingoftheGaTesource,resultinginachangeintheGaTe dimensional growth. Although we were able to flux;theyarenotincludedinthisplot. successfully cap the InAlAs with 2-nm InAs, we could not do so with 20-nm n+ InAs. These n+ caps were very rough and did not result in good device mentioned above, the growth temperature was properties. By changing the upper barrier to In 0.20- typically reduced from 500(cid:2)C to 450(cid:2)C for the Te Al Sb, which has a lattice constant of 6.20 A˚ and 0.80 d-doping. The goal was to reduce Te segregation is coherently strained, we were able to routinely and diffusion. Similar changes in growth temper- grow 20-nm n+ InAs caps with good surface mor- ature are sometimes used for Si-doped InP-based phology and etching properties. Our AFM rms HEMTs.25,26 To test the importance of the growth roughnessvalues(5 · 5 lm2area)arenear1 nmfor temperature, d-doping substrate temperatures of as-grown samples with and without the n+ caps as 430(cid:2)C, 470(cid:2)C, and 500(cid:2)C were also used. As shown well as for samples after removal of the n+ cap. As in Table I, there is no significant effect on the shown in Table I, we obtained as-grown sheet mobility or density. This is encouraging in that resistances of 74–78 W/sq. For sample I, we per- precise control of the growth temperature is not formedtransportmeasurements after removing the necessary. [In previous work on InAs/AlSb quan- n+ cap with a chemical etch. The mobility was good tum wells, we found that high mobilities could be (23,200 cm2/V s); the sheet density (2.0 · 1012/cm2) achieved over a relatively large range of growth was higher than usual for this GaTe soak time, temperatures for the InAs ((cid:1)470–520(cid:2)C).] Our probably because of surface effects.31 The device normal procedure was to leave the Sb shutter open properties are under investigation and will be during deposition of GaTe. For sample F, we closed reported elsewhere. InAlSb/InAs/AlGaSb Quantum WellHeterostructures 103 Another group recently fabricated high-perfor- 3. J.Bergman,G.Nagy,G.Sullivan,A.Ikhlassi,B.Brar,C. mancetransistorswithn+InAscapsonInAlAs/AlSb Kadowetal.,DeviceResearchConf.(2004),p.243. barriers.17 The InAlAs passivates the highly reac- 4. R. Tsai, R. Grundbacker, M.D. Lange, J.B. Boos, B.R. Bennett,P.Nametal.,GaAsMantechConf.(2004),p.4.4. tive AlSb, which enables the use of a gate recess 5. A. Leuther, R. Weber, M. Dammann, M. Schlechtweg, M. process.AnotherreasonforusingInAlAsbarriersis Mikulla,M.Walther,andG.Weimann,Proc.InPandRe- that, based upon the band structure, they should latedMaterialsConf.(Piscataway,NJ:IEEE,2005). provideabarrierforholesgeneratedinthechannel 6. Y. Royter, K.R. Elliott, P.W. Deelman, R.D. Rajavel, D.H. Chow,I.Milosavljevic,andC.H.Fields,TechDigestIEDM by impact ionization, hence reducing gate leakage (Piscataway,NJ:IEEE,2003), p30.7.1. currents. Our preliminary measurements on de- 7. S.Datta,T.Ashley,J.Brask,L.Buckle,M.Doczy,andM. vices with InAlSb/InAs/AlGaSb heterostructures Emeny, et al., Tech Digest IEDM (Piscataway, NJ: IEEE, suggest relatively low leakage currents may be 2005). 8. W.Kruppa,J.B.Boos,B.R.Bennett,N.A.Papanicolaou,D. possible without the InAlAs barrier. Park,andR.Bass,Electron.Lett.42,688(2006). The work described here involved Te-doped 9. B.R.Buhrow,V.Sokolov,P.J.Riemer,N.E.Harff,R.Tsai, structures. We have also grown structures using a A.Gutierrez-Aitkenetal.,Proc.InPandRelatedMaterials 1.2-nm InAs(Si) doping layer32,33 as well as InAlSb Conf.(Piscataway,NJ:IEEE,2005). upper barriers, AlGaSb buffer layers, and small 10. R. Tsai, M. Barsky, J.B. Boos, B.R. Bennett, J. Lee, N.A. Papanicolaou et al., Proc. GaAs IC Symposium (Piscata- gate-to-channel separations (12 nm). Some samples way,NJ:IEEE,2003),p.294. also included a 20-nm n+ InAs cap. The values of 11. J.B.Hacker,J.Bergman,G.Nagy,G.Sullivan,C.Kadow, mobility, density, and rms roughness were compa- and H.K. Lin, et al., IEEE Microwave Wireless Compon. rabletowhatwereporthereforTe-dopedstructures. Lett.14,156(2004). 12. W.R. Deal, R. Tsai, M.D. Lange, J.B. Boos, B.R. Bennett, All of the Si-doped structures contained a total of andA.Gutierrez,IEEEMicrowaveWirelessCompon.Lett. 57 nmofAlSb above andbelowthechannel,but we 15,208(2005). see no reason the AlSb could not be totally elimi- 13. P.J. Riemer, B.R. Buhrow, J.B. Hacker, J. Bergman, B. nated as wehave done for theTe-doped structures. Brar, B.K. Gilbert, and E.S. Daniel, IEEE Microwave WirelessCompon.Lett.16,40(2006). 14. K. Elgaid, H. McLelland, M. Holland, D.A.J. Moran, C.R. SUMMARY Stanley,andI.G.Thayne,IEEEElectron.Dev.Lett.26,784 (2005). We investigated heterostructures for InAs-chan- 15. T.Suemitsu,T.Ishii,H.Yokoyama,T.Enoki,Y.Ishii,and nel HEMTs containing Al Ga Sb buffer layers 0.7 0.3 T. Tamamura, Jpn. J. Appl. Phys. Part 2—Lett. 38, L154 and In0.2Al0.8Sb upper barriers with Te d-doping. (1999). TheSIMSmeasurementsindicatethatverylittleTe 16. Y. Yamashita, A. Endoh, K. Shinohara, K. Hikosaka, T. isdiffusingintothechannel.Thecarrierdensitycan Matsui,S.Hiyamizu,andT.Mimura,IEEEElectron.Dev. Lett.23,573(2002). be controlled by the GaTe soak time. Room-tem- 17. C.Kadow,M.Dahlstrom,J.U.Bae,H.K.Lin,A.C.Gossard, perature mobilities range from 18,000 to andM.J.W.Rodwell,etal., IEEETrans.Electron.Dev.52, 25,000 cm2/Vs.Thechannelthicknessis12 nmand 151(2005). the gate-to-channel separation ranges from 7 to 18. B.R.Bennett,B.P.Tinkham,J.B.Boos,M.D.Lange,andR. Tsai,J.Vac.Sci.Technol.B22,688(2004). 13 nm. These values should allow scaling to sub- 19. J.B. Boos, W. Kruppa, B.R. Bennett, D. Park, S.W. Kirc- 100-nm gate lengths. X-ray diffraction measure- hoefer,R.Bass,andH.B.Dietrich,IEEETrans.Electron. ments show that the In0.2Al0.8Sb is coherently Dev.45,1869(1998). strained. The AFM measurements indicate rela- 20. B.P.Tinkham,B.R.Bennett,R.Magno,B.V.Shanabrook, tivelysmoothsurfacesofthestandardstructuresas andJ.B.Boos,J.Vac.Sci.Technol.B23,1441(2005). well as samples with a 20-nm n+-InAs cap. The use 21. S.Miya,S.Muramatsu,N.Kuze,K.Nagase,T.Iwabuchi, andA.Ichii,etal.,J.Electron.Mater.25,415(1996). of n+ caps should result in HEMTs that simulta- 22. B.R. Bennett, R. Magno, and N. Papanicolaou, J. Cryst. neously have low sheet resistance (<100 W/sq) and Growth251,532(2003). moderatesheetcarrierdensity((cid:1)1 · 1012/cm2).The 23. S. Cohen, C. Cytermann, and D. Ritter, Proc. InP and RelatedMaterialsConf.(Piscataway,NJ:IEEE,2006). absence of reactive AlSb is expected to improve the 24. H.Sugiyama,H.Yokoyama,andT.Kobayashi,Jpn.Appl. reliability and manufacturability of the HEMTs. Phys.Part1—RegularPapersShortNotes&ReviewPapers 43,534(2004). ACKNOWLEDGEMENTS 25. A.S.Brown,R.A.Metzger,J.A.Henige,L.Nguyen,M.Lui, The Office of Naval Research and the Defense andR.G.Wilson,Appl.Phys.Lett.59,3610(1991). 26. L.D. Nguyen, A.S. Brown, M.A. Thompson, and L.M. Advanced Research Projects Agency supported this Jelloian,IEEETrans.Electron.Dev.39,2007(1992). work. The authors thank M.D. Lange, R. Tsai, and 27. M. Borg, J. Grahn, S. Wang, A. Mellberg, and H. Zirath, Y.-C. Chou, Northrop Grumman Corporation, for Proc. InP and Related Materials Conf. (Piscataway, NJ: technical discussions, and Vladimir Kuznetsov and IEEE,2005). 28. N.A.Papanicolaou,B.R.Bennett,J.B.Boos,D.Park,andR. Corwyn Canedy, NRL, for sample characterization. Bass,Electron.Lett.41,1088(2005). REFERENCES 29. A potential disadvantage of AlGaSb buffer layers is that they are less resistive than pure AlSb. Our preliminary 1. B.R. Bennett, R. Magno, J.B. Boos, W. Kruppa, and M.G. Hall measurements for 1.5-lm layers of Al Ga Sb yiel- 0.7 0.3 AnconaSolid-StateElectron49,1875(2005). ded resistivities of 7·105 W/sq, at least an order of mag- 2. B.D.Weaver,J.B.Boos,N.A.Papanicolaou,B.R. Bennett, nitudelowerthanforpureAlSb.Thiscouldimpactdevice D.Park,andR.Bass,Appl.Phys.Lett.87,173501(2005). isolationandpossiblymicrowaveperformance. 104 Bennett,Boos, Ancona,Papanicolaou, Cooke,and Kheyrandish 30. J.B. Boos, M.J. Yang, B.R. Bennett, D. Park, W. Kruppa, 32. B.R.Bennett,M.J.Yang,B.V.Shanabrook,J.B.Boos,and C.H.Yang,andR.Bass,Electron.Lett.34,1525(1998). D.Park,Appl.Phys.Lett72,1193(1998). 31. C.Nguyen,B.Brar,andH.Kroemer,J.Vac.Sci.Technol. 33. M.D.Lange,R.S.Tsai,W.R.Deal,P.S.Nam,N.J.Lee,R.S. B11,1706(1993). Sandhuetal.,J.Vac.Sci.Technol.B24,2581(2006).