KAPL-P-000173 (K97057)

0A. l F-q 7 a s/ I 4- - GROWTH AND CHARACTERIZATION OF In,,.&,-,Sb DEVICE STRUCTURES

USING METALORGANIC VAPOR PHASE EPITAXY

G. Charache, H. Ehsani, et. al.

May 1997

ER

NOTICE

This report was prepared as an account of work sponsored by the United States Government. Neither the United States, nor the United States Department of Energy, nor any of their employees, nor any of their contractors, subcontractors, or their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness or usefulness of any information, apparatus, product or process disclosed, or represents that its use would not infringe privately owned rights.

KAPL ATOMIC POWER LABOFUTORY SCHEIWCTADY, NEW YORK 10701

Operated for the U. S. Department of Energy by KAPL, Inc. a Lockheed Martin company

DISCLAIMER

This npon was prepared as an account of work sponsored by an agency of the United Statu Governmew Neither the United States Government nor imy agency thereof, nor any of their anployas, makes any wammty. u p r u s or implied. or assumu any legal liability or responsibility for the aauracy, mmpicscneu, or use- fulness of MY information, apparatus, product, or ptoccu disclosai or rcprcscnts that its use would not infringe privately owned rights. R c f u c n a h a m to any rpc- afic CommerCiai product, proass, or service by trade name. rndcmuk. manufac- turer. or otherwise does not necessarily constitute or impiy tu cndonanent. m m - mend;rtion, or favoring by the United Statu Governmat or any rgtocy thenof. The views and opinions of authors e x p d h m m do not e-iy m t c or reflect those of tbe United States Government or any agency thermf.

DISCLAIMER

Portions of this document may be illegible in electronic image products. Images are produced from the best available original document.

C L 3 C

Growth and Characterization of In0,2Gag,@b Device Structum Using Metalorganic Vapor Phase Epitaxy

H. Ehsani, 1. Bhat, C. Hitchcock and R Gutmann Department of Electrical, Computer and Systems Engineering

Center far Integrated Electronics and Electronics Manufacturing Rcnsselaer Polytechnic Institute, Troy, NY 12180-3590

and G. Charache and M. Fmmm

Lockheed Martin Inc., Schenectady, New Yo& 12301-1072

Abstract: m.2GaO8Sb epitaxial layers and thennophotovoltaic (TPV) device structures have been grown on GaSb and OaAs substrates by metalorganic vapor phase epitaxy (MOVPE). Control of the n-type doping up to 1 x 10 cm was achieved using diethyltellurium (DETe) as

the dopant some. A Hall mobility of greater than 8000 cm /Vs at 77K was obtained for a 3 x

1017~m*3 doped In().2(&io,gSb layer grown on high-resistivity OaSb substrate. The Iq).2Ga0.8Sb epilayers directly grown on GaSb substrates were tilted with respect to the substrates, with the m u n t of t2t increashg with the layer thickness. Transmission electron microscopy (TEM) studies of the layen showed the presence of dislocation networks across the epiIayers parallel to the interface at different distances from the htaface, but the layers above this dislocation network were virtually frec of dislwadons. A strong carrelation between epi1;byer tilt and TPV device properties was found, with layers having mofe tilt providing better devices. The results suggest that the dislocations mving pardld to the interface cause lattice tilt, and control of this lam tilt may enable the fabrication of better quality device structures.

18 -3

2

INTRODUCTION

In,Gal.,Sb material system is atttactive for application in TPV cells since the energy bandgap can be varied h m 0.17 eV to 0.72 eV by wrying the composition (1). By adding arsenic into the system, lattice-matched InxGal-xAsySb1-y structures am be grown on GaSb substrates. Even though antimunides may have some advantages compared to an InXGal_Jls system for TPV applications, the growth technology for LnxGal-xSb is not as advanced. MOVPE growth of antbnides has emphasized lattice-matched InxGal-xAsySbl-y or the bjnary compounds OaSb and InSb. Doping studies and &location reduction techniques for tartice-mismatched epitaxy have not been investigated in detail fbr bGal-xSb lityen grown on GaSb substrates (2). Studies on doping characteristics and on methods to *duct dislocation den~ses in the active laps are required for better antimonide devices.

Pmliminary results on the growth and doping of Xag.~Gag.@b by M W P E (3) and p-type doping using Si have been presented before (4). In this paper, recent results on the n-type doping of hQ.2G@,8Sb by Te and structural characterization of ln(),2Gag,gSb layers am presented. These epilayers are tilted with respect to the substrates, with the amount of tiit increasing with

layer thickness. This res& has imponant implications for the design of step grading techniques for epitaxial growth of device quality I R ( ) . ~ G ~ o . ~ S ~ .

EXPERIMENTAL PROCEDURES

The ternary ep&yers were grown on (100) oriented OaSb and semi-insulating (100) GaAs substrates in 8 low pnssm, rf heated, horizontal MOVPE reactor. Trimethylgaltiurn, trimethylindium, rrlnrethylanbony and dicthyltellurium (DETe) were used as the Ga, In, Sb and Te sources, mespeaively. After degrcasing the wsfets in organic solvents, the GaAs substrates were etched in Caro’s etch (solution containing H~SO~:H~@:HZO, 5:l:l by volume) for two minutes and the GaSb wafiers were etched in 1% bromine-mcthanol solution for 30 scconds. The GaSb substrates were of low resistivity p-type at room temperature, but were of high resistivity at 77K (sheet resistance of 600 Q per square) so that Hall measu+errpents could be made on these at 77K.

Double crystal x-ray dfihctfon WM used to determine layer tilt With respect to the substrate, crystalline quality of the layers, CoIllpOsidon of the layers, and lattice relaxation. Variation of the peak separation between the epilayers and the substrates as a function of rotational angle (azimuth) was used to determine the l a t h misnatch and the epilayx tilt. The residual strain in so= f&ns was measured by x-ray diffhction specua of non-symnetrk planes such as (115) and symmaic plans such as (004). Details of the measurement technique can be found elsewhere (5).

RESULTS AND DISCUSSIONS

Earlier work (3) reported n-type doping of b(),2Ga().gSb with Te for DETe mole

fractions higher t h ~ 1 x d , where the carrier concentration actually decreased as the DETe mole Eraction was Increased. This decroase WBS attributed to the h t i o n of Te precipitates or Ga-Te and Sb-Te reactions at the growth surface. These reactions pmbably resulted in inclusions of undesired precipitates in thc layer, causing degradation of crystal quality. Secondary ion mss specvomeay (SIMS) measurtments showed that only 2 - 3% of the incorporated Te was active

when the total TC concenmtion was about 2 x 1019 mv3. he mobility of tht layers a l~o decreased as the DETe mole &tion was inneased.

In this work, this study was extended to Iowa concentrations of Te, by modifying the reactor with the addition of a “double dilution” scheme to deliver DETe. In this method, hydrogen through an additional line is used to dilute the DETc/Hz gas mixtures. Only a small portion of the DETe/HZ is delivered to the reactor, and the rest is passed through the exhaust lints. The mole

hction of DETe in the reactor can be conmUcd over a wide range (lo-* to IO-’ ) using this method.

Figure 1 shows the measured carrier concentration at 77K as a hction of the DETe mole h & n fw layers grown on GaAs and GaSb substrates. The canier concentration increases

linearly with the DETe flow i&ia&, but beyond lo1* ane3 the carrier concentration actually

decreases with the DETe Mw. The Continuaus Iinc in the graph corresponds to a a @ETe)’-*

suggesting a hrgh degree of Te activation at doping concentrations as high as 1 x 1018cm-3. TEM of lightly Te-doped Ino.2GaO.gSb (e 1 x 10 cm ) does not show presence of any precipitates. 13 -3

Figure 2 shows the Hall mobility at 77K versus DETe mole fraction for layers grown on GaAs and GaSb substrates. It was consistently found that the mobility of layers grown on GaSb were higher than those grown on GaAs substrates. For DETe mole &tion less than 5 x p- type layer with low mobility was obtained. The Hall mobility was fbmd to increase with the doping concentration as shown in figure 3, which i s oppuske to tht usually observed behavior in other ID-V compounds. Similar behavior (Le., increase of Hall mobility with the doping concentration) was observed by others as wen in both molecular beam epitaxially (MBE) grown GaSb (6) and MOVPE grown OaSb (7) doped with Te. Sirnilar behavior was also observed in bulk grown 3nSb material (8). Turner et al. (6) and Zittez et al. (8) attributed this to a reduced screening effect of the charged native acceptors by the donors as the donor concentration is reduced. Pascal et al.(7) believed that this is due to an inhomogeneous distriiution of Te in the layer and the limiting case ofrhis homogeneity being the presence of p and n regions in the same layers. The physical reason behind thls phenomenon is not clear at present. However, our experimental nsults an consistenr with the results reported by other groups.

After doping studies, several TPV device suucturcs were grown in fng.2Gag.gSb with various types of stcp-graded b e layers. Figure 4 shows two device sbuctures grown in this study. Figure 4(a) shows the technique commonly used, in which s e d step laycrs are grown with equal distribution of thickness and composition between the active layer and the substrate, This technique has been found to be success€ul in G a I n A W TPV structures (9). This method is particularly useful when the lattice rnismatch between the active layer and the substrate is small, and for materials that mhx very quickly beyond the critical layer thickness. Figure 4(b) shows another step-grading technique in which a thick l a p with a large lattice mismatch is grown as the fust step layer, foUowed by the growtb of a few thin sap layers that are closely lattice-matched.

The ideas behind the second method are as fullows: (1) Since there is a large lattice mismatch between the first step layx axxi the submte, many dislocatiOns are generated; therefbre, the probabiluy of dislocation interaction and annihilatiOn by forming dislocation loops increases. This process can be extremely effkctive at high threading diskation densities, but as the defm concentration decreases, the probability of further dislocation interactions decreases. (2) A thicker step layer allows for a complete relaxation of the layer at the growth temperature, which is especially importam for xxtterials which relax slowly. (3) The lattice mismatch between the adjacent top sup layers is smaU enough to prevent the nucleation of m y moft misfit dislocations at individual interfaces. (4) The top step layers will effiectively bend the residual thrtading dislocations that propagate from the initial interface, SO that the active layer will be relatiwly dislocatim-fite.

P d i m h r y studies have shown that the device performance of IqzGa().aSb layem grown by the method as shown in figure 4(b) is better than layers grown by the method shuwn in figure %a). (10) In order to understand the nhixation phenomena in these two structures, x-ray

I . J I A A IYV. J l U L l U L t J J

diffraction and TEM mEasuremnts were carried out on selected samples. X-ray wasurexnents show thai the epitaxiaI layers a~ actuaUy tilted with respect to the subsnates.

Figure 5 shows the VBSiatiOn of the relatiw 3ragg angle as a function of the azimuth rotation angle for a (004) refltction measured on a 6pm thick In0.2GaO.8Sb epilayer grown on (150) G a b substrate. The amplitude of the variation in thc Bragg angle for the mbstrate gives the substrate dscut, and the amplitude of the variation in the Bragg angle fbr the ephyer gives the substrate m k u t plus the relative e p i l a ~ tilt with respect to the substrate. As can be seen, the tilt angle between the Ing.2Ga().SSb epilayer and the OaSb substrate is much higher than the substrate misorientasion, which indicates that formation of the ephyer tilt in hGal-,Sb/GaSb system is not initiated by the substratc misorientation alone. To deterrmine whether the total tilt angle is f o d in the vicinity of rhe interface (where the density of misfit dislocation is large) or whether it gradually increases, a systematic study was conducted. In thfs set of txpe&nents, epilayers of e e n t thickness were grown kttpiug all other parameters constant. FigUte 6 shows that the tilt mgle increast s as the layer thiclooess increases, indicating that tht layex tilt is formed continuously throughout the epilayer.

The full width at halfmax;lmum (FWHM) of the kym which is related to the dislocation density was m a s d and the data are shown in figure 7. Several points shodd be noted in this figure: (1) The FwfIM of the 0.2pm Iry).2(fq.$Sb layer is lower than that of thicker layers, which c o d bc attributed to the layer not being M y relaxtd even though the layer is rhicker than the critical layer thickness for htticc relaxation. (2) Thc FWHM increases as the thickness increases until the thickness reaches 0.6pm, since a large population of misfit and threading dislocations form due to the 1.396 lattice mismatch- (3) Tht FWHM decreases between 0,Sp.m and 2 . 5 ~ owing to the 8Mlihilation of the dislocations. (4) The FWHM of the layers do not decrease significantly beyond 2 . 5 ~ Also, note that the tilt of the epitayer continuously varies with the thickness (see figure 6). This variation of tilt with thiclrntss conuibutes to the broadening of the x-ray peaks. Hence, it was concluded that the actual FWHM of the layers should be lower than that shown in fim 7.

A long-standing problem in the arm of epitaxial growth is the lack of undwstanding of the relaxation rate of the epitaxial layers and the tilt formation between the layers and the substrates. In lattice-mismatched hetero-epitaxy, misf3 dislocations arc formed at the interface, and in many cases, a crystallographic tilt of the layer with respect to the substrate is obtained. The tilt of the epilayers depends OA the subsuatc tilt as well as on the giowth conciitions.

Nagai (11) was the first to propose a model to explain the formation of tilt in a GaInAdGaAs system This thearetied mxfcl shows that the tilt of the epilayer is a function of the lattice mismatch and the substrate misorientation. Based on &is model, the amumt of tilt is always less than the substrate misorimtation, with tht tilt iwnreasing with the substrate misorientation and vanishing if tbe substrate is aorninally (100) Oriented. An increase in the tilt angle with respect to the substrate misorientation was also observed m many hetcro-epitaxy systems such as CdZnTe/GaAs(l2), ZnSeKiaAs(l3), and ZnSeKfe(14). Olsen and Smith(l5) proposed that tilt i s related tu the fiormation of misfit dislocations and that the component of the Burgers vector ptrpendicular to the growth plane is respoasible for the tik formtion. They showed that the tilt angle of epitiurial layers is W l y propodond to the misfit strain, except

when misfit is relieved by pure edge dislocations. Their -1 can only p&ct an upper limit for the magnitude of tilt.

To date, the experimental results showed that the amount of tilt angb between the

epitaxial layers and the substrates with eact (100) orientation (0.1-0.2O off) is less than a few hundred arc-stc. To the authors knolwledge, this is &st dm that a Iarp amount of tilt was observed for growth on nomidly (100) orknted substrates, and also that thc epilayer tilt increases with the thickness.

Tu investQate the propagation behavior of dislocations, "EM microscopy was carried out on several samples. Figures 8(a) and 8(b) show the E M bright field images for two a r e n t layers of nominally 2 p thkk lnxGal-xSb epibyers grown on GaSb substratc with a 0 . 3 ~ thick GaSb buff'er layer. Tht figures ckarly show a zone of dislocations buried at a depth of 1.5 - 1.7pm below the surface. On these and other samples that w m studied, long dislocations patallcl to the misfit networks were observed just above the misfit zone; and in several part of the samples these &locations sefm to bend paralf;el to the surf8ce at different depths ftom the surface. Most of the heading dislocations which originate at the mist3 networks are deflected parallel to the interface (converted to misfit disiocatim) and then terminated. Beyond approximately lp thick layers, few dislocations were observed in the m a studied by TEM. These dislocations which are parallel to the interface my be responsible for the lattice tilt. Since these dklwcations are bund to move parallel to the interface at different distances from the interface, the tilt is ais0 a function of the layer thickness.

The inclination angle of the layers grown with the two step grading schemes described before was compared It was found that when the step graded layer consists of 15 steps with qual thickness and 0.09% SatticC mkmtch between adjacent layers, the value of tilt angle was vwy low (about 35 arc-a). On the other hand, a large tilt angle was observed (about 2100 arc- s ~ ) when a thick initial layer of larger composition and lattice mismatch was grown fobwed by a few thinuer s q s . Since the lattice tilt docs not inhsically have any effect on the device performaace, the presence of tilt m y indicate that the layem are relaxed through dislocation motion along the growth plane so that active layers arc dislocation-free. ThereEore, layers with larger tiit yielded better quality TPV devices.

CONCLUSIONS

N-type doping studies carried out on Ia0.2Ga0.gSb grown on GaAs and GaSb show that 18 -3 the doping concentration increases with the DETe =le fiacbns as expected, up to 1x10 cm .

When the DETe mkfraction is increased further, the doping concentration is actually decreased. TPV device structures using two di&rtnt stepgrading techniques have been grown and characterized. It was found that the epitsxial layers are tilted with respect to the substrates and the mount of tilt increases with layer thiciaress. The layer tilt and the device quality are found to be correlated. TEM studies of some selected layers show that the dislocations Originating at the misfit networks move parallel to the interface at various distances fiom the interface, which rnay expiain the layer tilt.

1. 2.

3.

4. 5. 6. 7.

8. 9. 10.

11. 12.

13. 14. 1s.

REFERENCES

Milnes, A. 0. and Polyakov A. Y., Solfd State Electronics, 36,803 (1993). Bougnot J., Delannoy F., Pascal E, Grosse P., G h i A., Kaoulrab J., Bougnot 0.. Fourcade R., Walker P. J., Mason N. J., and Lambert B., J . Crysz. Growth, 107,502 (1991). Ehsani H., Bhat I., Hitchcock C., Barrego J., and Gutmann R., AIP Conference Proceedings, 358,423 (1996)- Ehsani H., Bhat I., Gutmann R., and Charache G., Appl. Phys. Len* 69,3867 (1996). Leiberich A. and Levkoff J., J. Crystal Growth, 100,330 (1990). Turner 0. W., Eglas S. J., and Strauss A. J., J. Vac. Set. Technol., B11,864 (1993). Pascal F., Delannoy P., Bougnot J., Gouskov L., Bougnot G., Grosse P., and Kaoukab J., J . Electron. Muter., 19, 187 (1990). Zimr R. N., Strauss A., and Attard A,, Phys. Rev 115,266 (1959). Sharps P. R. and Tinrmons M. L, MP Coderence Proceedings, 358,458 ( I 996). rxitchcock C., EhsaniH., Fxeeman M., Bhat I., Charache G., Barego J., and Gutmann R., “G&Sb and G W s S b TPV Device Fabrication and Characterization’*, this conference. Nagai H., J. Appl. Phys., OS, 3789 (1974). Ahlgren W. L., Johnson S. M., Smith E. J., Ruth R. P., Johnson B. C., Kdisher M. H., Cocknun C. A., James T. W., Arney D. L., Zieger C. K., and Lick W., J. Vac, Sci. Tech., A7,331 (1989). Ohki A., Shibata N., and Zcmbutsa S., J . AppI. Phys, 64,694 (1988). Kleiman J., Park R. M., and Mar= A,, J . Appl. Phys, 64, 1201 (1988). OIsen G. H., and Smith R, T., Phys. Stat. Sol, A31,?39 (1975).

r .

101gl j

0 Electron eonc. in GalnSb layer grown on GaSb subst. Electron conc. in GalnSb layer grown on GaAs subsl.

9- 6 v

c 0 .- c 2 c c a, 0 c 0 0 c

0 a, w

2 c -

DETe mole fraction

Figure 1 Carrier concentration at 77K versus DETe mole fraction in Ino.zGao.gSb grown on GaAs and GaSb substrates.

10000

1000

0

0

V

0 -

0 0

0

0 0 0

0

cp>

0

GalnSb on GaAs at 300 K (ptype) GalnSb on GaSb at 77 K (n-type) I 0 GalnSb on GaAs at 77 K (n-type)

100 1 o - ~ 10-8 10.’ 10-6

DETe mole fraction

I 0”

Figure 2 Hall mobility versus DETe mole fraction for Ino.zGao.sSb layers grown on GaAs and GaSb substrates.

1 ooo(

1 oa

101

0 .

0

0 0

Q

0 0

0 0

DETe mole fraction >1 0-6 DETe mole fraction -=1 OB

1017 1 o18

Carrier Concentration

Figure 3 Hall mobility at 77K versus carrier concentration for Ino.2Gao.sSb layers grown on GaAs substrate. Data for high flow of DETe (mole fraction > are also plotted.

I Emitter I In02Ga~.$b Active

layer

Step graded layer 15 steps, 0.31,~m

1 GaSb Substrate

Emitter

layer Ino.zGao.8Sb Active

Step graded layer 5 steps, 0 . 3 ~

GaSb Substrate

Figure 4 Two types of step grading methods used in this study

6000

4000

moa

0 90 180 270 360 450

Rotation Angle (degree)

Figure 5 Relative Bragg angle versus azimuth rotation angle. The amplitude difference between the epi and the substrate indicates the layer tilt with respect to the substrate.

30001 . . .

0

0

0 0 1 2 3 4 5 6 7

Film thickness (pm)

Figure 6 Epilayer tilt with respect to the substrate for different values of layer thickness.

7001 L

600. .. 500 *

400 *

300 * .

0 .

0

0

0 0

Film thickness (pm)

Figure 7 The FWHM of InGaSb layers grown on GaSb substrates as a function of layer thickness.

Figure 8 Bright field TEM micrographs of a Si doped (a) and Te doped (b) Ino.zGal-,Sb layer grown on GaSb with a 0.3pm thick buffer layer. Note the deflection of dislocations parallel to the interface, and at different distances from the interface.