Photoreflectance characterization of OMVPE GaAs on Si

Journal of Crystal Growth 93 (1988) 481-486 North-Holland, Amsterdam

481

PHOTOREFLECTANCE CtIARACTERIZATION OF OMVPE GaAs ON Si N. BOTTKA and D.K. G A S K I L L Naval Research Laborator); Washington, DC 20375, USA

R.J.M. G R I F F I T H S , R.R. BRADLEY and T.B. JOYCE Plessey Research and Technology, Caswed, Towcester, Northans, NN12 8EQ, UK

and

C. ITO and D. M c I N T Y R E Ford Microelectronics, Inc., Colorad;, Springs, Colorado 80908, USA

Photorefiectance, a contactless, non-destructive optical characterization tool, can be of great utility to the material scientist in identifying crystal growth problems and ascertaining material quality in a very short time. It provides a precise measurement of the spectral energy of the fundamental absorption edge and higher lying critical point transitions. Shifts in these energies are a measure of the built-in strain in strain-layered heterostructures. Moreover, the spectral energy of the so-called Franz-Keldysh oscillation extrema is directly related to the net carrier concentration of the semiconductor material under study. We have ased the Photoreflectance technique to determiae the crystal quality, the carrier concentration, and the built-in strain of epitaxial GaAs gro~,n by OMPVE on Si substrates. The study includes determination of these parameters as a function of various growth conditions, post-growth anneal cycle, and spot position on the wafer. Results are correlated with X-ray diffraction, and C - I / measurements.

1. Introduction Photoreflectance (PR), a contactless, non-destructive form of electroreflectance (ER) [1-5], has recently been shown to offer material diagnostic information in a manner practical to the material scientist. In this work we present details on using P R as a convenient tool to determine the net carrier concentration and the built-in strain in GaAs films grown on Si by OMVPE. The objective is to compare the GaAs on Si films grown at two different laboratories under completely different growth conditions, to correlated the PR results with other accepted measurements, and to provide assistance to the crystal grower in assessing the overall quality and uniformity of the GaAs on Si films. The analysis of our PR data is based on the so-called high-field limit of ER. In this limit, the PR spectra near the fundamental absorption edge exhilzits the so-called Franz--Keldysh oscillations (FKO) above the band gap energy, Ee, the ex-

trema of which can be used to determine the magnitude of the surface electric field by using the asymptotic exeression of the spectral lineshape function [6]: R

cos

h-O

- ¼

(d- 1) ,

(1)

where h~0 is the energy of the probe beam, h8 is the characteristic energy of a quantum mechanical particle of interband reduced mass tz being accelerated by the electric field F,

hO = ( e2F2hZ/8t~ ) t/3:

(2)

d is the dimensionality of the critical point, and the proportionality factor is an exponential damping term. For direct gap materials such a,,: GaAs, at the fundamental absorption edge d = 3, the cxtrema in eq. (1). [h~]¢. will occur for

[ho~],=hOFj+E~. F=[~rr(j+$)]

0022-0248/88/$03.50 © Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

2/3.

j=O. 1 . 2 . 3 . . . . .

(3a) (3b)

482

N. Bonka et al. / PR characterization of OMVPE GaAs on Si t .500 . . . . . . . . . .

....

.

o

2.o

4.0

6.o

F~ Fig. 1. Plot of the spectral position of the Franz-Keldysh oscillation extrema versus Fj = [~-rr(j + ½)]2/3 for the 3-inch GaAs on Si sample F16. The slope of the line is a measure of the built-in surface electric field (and thus carrier concentration, No ~-NA). The cross hatched area "A" comprises 70% of the central area measured.For the central area, N ~ = 1.6 × 1016 =m - 3 and varies by +5%. The inset shows a plot of the measured A R / R versus h ~.

The inset of fig. 1 shows a typical room temperature A R / R photoreflectance trace of GaAs on Si. As indicated by eq. (3), a plot of [h~0]j versus [~w(j + ½)]2/3 yields a straight line whose slope is proportional to the built-in surface electric field of the material u m e r study. The ordinate intercept is the bandgap energy of the material, which in the case of GaAs on Si, can be used as a measure of the built-in strain in the layer. In the F K O spectral region only the topmost layer is being probed. The feature " I " shown in fig. 1 will be discussed below. We have shown recently [3,4,7] that for n-type GaAs the electric field determined from PR experiments can be related to the carrier concentration, N = Nt~ + NA, and the built-in potential, V~, through the generalized Schottky equation: F,; ::

2 e ( U D +/~k)(VBxc0 -- Vp - k r / e )

],/2,

(4)

..I

where Fs is the mag,aitude of the electric field at

the surface, N D and NA are the density of ionized donors and acceptors, respectively, Vp is the quasi-equilibrium photovoltage present during PR measurements, k T / e is a thermal term, and go0 is the dielectric constant times the permittivity of free space. Details of the experimental arrangement for PR can be f o u n d in ref. [4]. In our experiments, the procedure for calculating N D + NA is as follows. Depending u p o n the ~,',ze of the sample, at least three spots were measured by PR. The F K O extrema was t h e n plotted versus Fj using eq. (3). The built-in electric field was obtained from the slope of this line assuming ~ = 0.057 m (the value for GaAs), and the intercept is Eg. N was *hen calculated from eq. (4), assuming • = 13.18, lip = 0.13, and VB = 0.73 eV, respectively [4,7,8]. The accuracy in measuring ~ was limited by the resolution of the F K O e x t r e m u m energy or + 0.002 eV. The estimated error in determining F s from the [ h ~ ] j versus ~ plot is +0.5 k V / c m . This yields for N = 1.0 × 1016 cm -3 an estimated + 3% resolution in determining N o + N A across the wafer. A well characterized homoepitaxial n-GaAs wafer was used as a standard for comparison. This Si-doped G a A s layer was grown simultaneously both on n + and semi-insulating G a A s substrates using the N R L vertical low pressure OMVPE reactor with rotating susceptor. The GaAs layer grown was 3.85 # m thick. The Van der P a u w - H a U mobility was 6,450 and 17,460 c m 2 / V • s at 293 and 77 K, respectively, The corresponding values of N o - N A (after correcting for depletion thickness) were 1.3 × 1016 and 1.1 × 1016 cm -3 at 293 and 77 K, respectively. The layer grown on the n + substrate was used to determine lip in a separate ER experiment [4].

2. C ~ s t a l growth Two sets of samples were characterized by PR: a set from F o r d Microelectronics, Inc. (designated by the sample identification letter F) and a set from Plessey Research and Technology (designated by the letter P). The Ford G a A s epitaxial layers were grown on 3-inch, 10--30 ,q cm p(100) Si substrates with

N. Bottka et aL / P R characterization o f O M V P E GaAs on Si

483

Table 1 S u m m a r y of the room t e m p e r a t u r e data for the Ford Microelectronic G a A s o n Si samples; growth temperature: 630 ° C Sample No.

Layer thickness (/~ m)

XRD (arc sec)

Es (eV)

(ND + NA ) × 1 0 1 6 ( c m - 3)

Si-orientation

Anneal cycle

F1 F2 F3 F4 F5

3.4 3.4 3.0 3.5

336 180 150

1.408 1.406 1.407 1.404

1.5 0.92 4.1 0.20

(100) (100) (100) 2 ° --* (110)

1x 2x

3.4

324

1.406

1.0

1 o __, ( 1115

-

F6

3.4

324

-

3.3 3.3 3.4 3.5 3.5 3.5 3.3 3.3 3.5 3.5

186 138 354 198 174 336 222 168 138 -

1.3 0.18 0.33

2 ° ~ (111)

F7

1.405 1.403 1.403 1.404 1.404 1.405 1.404 1.405 1.402 1.415 1.404

1 ° ~ (111) 1 ° --* (111)

1X 3x

0.94

4 ° ~ (111 )

-

0.94 0.39 0.98 0.60 0.55 0.1 1.6

4° 4° 6o 6° 6° 2° 1°

1x 3× 1× 3x 3x -

F8 F9 FIO

Fll F12 F13 F14 F15 F16 XRD Es N O + NA 1X

= = = =

~ (111 ) --* (111) --, (111 ) ~ (111) --* (111) --* (1105 ~ (111)

X-ray data ( d o u b l e crystal diffraction peak width F W H M ) gap energy as measured by the line intercept method in eq. (3) average value of t h e carrier concentration as measured from the line slope method in eq. (3). a single thermal a n n e a l cycle; 850 o C for 30 min under As ~ . erpressure

off-orientation tilt towards the (011) direction of 0, 1.0, 2.0, 4.0, and 6.0 __+0.5°. The wafers were degreased, etched in 1 0 : 1 DI w a t e r / H F , rinsed, and loaded into the reactor.

The Ford atmospheric OMVPE reactor is a customized, double chamber system with a maxim u m capacity of 39. 3-inch diameter (or two. 8-inch diameter) wafers per chamber [9]. The re-

Table 2 S u m m a r y of the room t e m p e r a t u r e data for the Plcssey G a A s on Si samples; growth temperature: 750 o C; "'initialization" layer thickness in parenthesis Layer thickness (,am)

XRD (arc sec)

Eg (eV)

( N D + NA)X 1016 (cm - 3 ) PR

C-V

3.0 1.0 6.0

4.3 (1.0) 4.3 (1.0) 4.3 (3.0)

190 190 1000

1.414 1.407 1.415

P4

5.6

3.2 ( 1.05

200

1.408

6.9 4.6 11 2.4 (0.2)

7.0 8.0 -

Single VPSLS Single VPSLS (RTA) l ×VPSLS+In 2 × SLS 300 ,~ sp (selected area)

P5

14.0

3.2 (0.75

244

1.413

5.7 (0.2)

6.0

2 x SLS 300 ,~ sp

P6

16.0

3.2 (0.7)

200

1.412

P7

9.3

4.2 (1.2)

210

1.413

~.~

~

.~..~

¢ C dta Jl ~. X

~, J~rl%

1 .~ 1 ¢.. "1 1 ...,e

P9 P10

3.6 1.0

4.8 (2.3) 4.8 (2.3)

280 -

1.410 1.415

2.0 4.0 1.5 -

2 x SLS 300 ,~ sp 2 x V P S L S 1000 ,~ sp

P8

1.9 (0.1) 4,2 (2.2) ,-I -=) , , I , ~ z.., q ~.~-r 3.0 (0.2) 14 (2.0)

Sample No.

Wafer size (cm 2)

P1 P2 P3

XRD PR Eg SLS VPSLS SSI, sp RTA

= = = = = = = =

I

X-ray data average N as m e a s u r e d by photoreflectance (standard deviation in parenthesis) gap energy as m e a s u r e d by the line intercept method in eq. (35 standard strained layer superlattice 10 periods G a A s / I n G a A s (60 , ~ / 6 0 ,~) variable period SLS single strained layer of l n G a A s (40 ,~5 GaAs spacer layer rapid thermal a n n e a l

Remarks

• ..c.T

~A~L~

o

_:_J V~ItCU

Sp

11 × SSL varied sp As-implanted Si

484

N. Bottka et aL / PR characterization of O M V P E GaAs on Si

actant gases are introduced vertically over a horizontally rotating, RF induction-heated susceptor. The growth procedure was as follows: 5 min bake at 9 0 0 ° C in 270 AsH3/H2; 20 n m GaAs nucleation layer grown at 425°C; and 3.5 /.tin GaAs grown at 630°C. The growth rate was 2.5 # m / h . Flows were 40 SLM H2, 90 SCCM AsH 3, and 150 SCCM TMG. Table 1 gives the film thickness and off-orientation for the wafers studied. The Plessey GaAs layers were grown on 3-inch, 0.02 ,g cm n(100) Si substrates. The Si wafer was chemically cleaned in 5 : 1 : 1 H20:NH 4 OH : H 2 0 2 and 6 : 1 : 1 H 2 0 : HCI : H202, etched in HF, rinsed in DI water. The Pies, ey horizontal reactor has a rectangular cross-section, is RF heated, and can handle two 3-inch wafers. The growth procedure was as follows: bake at 1150°C in H2; 30 period, 3 n m / 3 nm A I A s / G a A s superlattice at 400°C; 20 rain anneal at 750°C in A s H a / H 2 ; 700 nm GaAs buffer layer followed by one to 10 periods of 6 n m / 6 nm GaAs/In0.1Ga0.gAs strain layer superlattices (SLS) separated by 30 nm GaAs, all at 750°C; 3 /.tm GaAs at 7 5 0 ° C . The growth rate was 3 p.m/h. The TMG and AsH 3 mole fractions in the reactor were 5 × 10 -5 and 2.2 × 10 -3, respectively. Table 2 indicates any variation in the above procedure.

electric field homogeneity (and thus, the homogeneity in N ) across the 3-inch wafer. The cross-hatched region designated by " A " represents about 70% of the central area of the wafer. In this region N varies + 5% from the average value of 1.6 × 1016 cm -3. In the remaining 3070 (near the wafer perimeter) N increased by as much as a factor of two with respect to the center. For F16, the intercept corresponded to Eg = 1.404 + 0.002 eV for all the spots measured. There was an observed reduction in N after two or three anneal cycles, but no change in E~. We observed no direct correlation between N, Eg, and the Si wafer orientation, or the F W H M of the X-ray data. There was, however, an observed order of magnitude reduction in etch pit count after annealing the samples. The Plessey PR results for N and E~ are summarized in table 2. The original 3-inch wafers were diced and only selected pieces were measured by PR. Unlike the Ford layers grown in a vertical rotating susceptor reactor, the Plessey samples showed a monotonically changing variation in N perpendicular to the horizontal reactor flow direction. Fig. 2 gives a summary of the homogeneity in N and Eg for two of the Plessey samples. The

"5° T

3. Results and discussion The results for N and Eg as determined from the PR measurement for the Ford samples are summarized in table 1. Table 1 also shows the crystallographic orientation, the double crystal Xray diffraction peak width FWHM, and the postgrowth anneal condition for the wafer. Except for sample F16, all PR was done on a spot near the center of the 3-inch wafer. Sample F16 was measured on multiple spots across the wafer in two orthogonal directions. The same spots were remeasured after thermal annealing. A thermal annealing cycle (1 x ) corresponded to a 850°C bake for 30 min under AsH 3 overpressure. Fig. 1 summarizes the multiple spot PR results on sample F16. The variation in slope of [h~]~ versus Fj is a measure of the built-in surface

_J

=u

P6

/" ~_

1.450

294K

~X,~_)

1.400 . . . . . . . . . . . . . 0 2.0 4.0 F]

6.0

,

Fig. 2. The variation of N D + N A for samples P6 and P7 grown in a horizontal O M V P E reactor. The cross hatched area indicate the region measured in each sample. The arrows indicate the direction of flow within the reactor. N,,,~ = 2 . 0 x 1016 cm 3 and variations -i- 5% for P6. The N variation for sample P7 is given in table 2.

N. Bottka et al. / PR characterization of OMVPE GaAs on Si 1.600/ . . . . . . . . . . . . . .

t

I

g

,

/

2 ~

1.550

. 150K

200

250 294

B

ua

,~

.a

294K

1.450

1.400

0

2.0

4.0

- ~ G,0

Fj Fig. 3. Plot of the spectral position or the Franz-Keldysh oscillation extrema for: (a) the GaAs on Si sample P5 (solid line) as a function of temperature; (b) a 3.85 Fm thick homoepitaxial, Si-doped G a A s sample (dashed line) used as a s t a n d a r d (STD). The STD sample has a room temperature carrier conc e n t r a t i o n of 1.3 × 10 t6 c m - "~ as determined by Hall measurement. The slope of the line is proportional to ND + NA; its intercept is the gap energy, Eg, of the material.

cross hatched regions in fig. 2 comprise about 70% of the wafer area. Sample P6 represents the least, P7 the most N variation observed in the Plessey wafers studied. Selected area GaAs deposition through openings in silicon nitride on Si (P4) showed similar PR spectra. There was no significant change in Eg within the area studied. Photoreflectance was also done at 82 K on selected wafers. For both the Ford and Plessey samples studied, we observed a substantial decrease in the built-in electric field with temperature. From 294 to 82 K, samples F16 and P5 showed a reduction in ~ by a factor of 1.6 and 2.0, respectively. The value of Eg (E 0 transition) at 82 K was 1.482 eV for F16 and 1.490 eV for P5, re~neetivelv The spin-orbit split transition f E,, + /t o) of GaAs for sample P5 was 1.833 eV at 82 K, giving A o of 0.343 eV. Fig. 3 shows [hw]j versus Fj as a function of temperature for P5. The average rate of change of the measured gap energy with temperature, dEg/dT, was 3.7 x 10 -4 e V / K . For comparison, fig. 3 also shows the 294K [hoa]i versus F, plot for the homoepitaxial GaAs layer used as a standard. -

~r

.

.

.

.

.

.

J

.

.

.

.

.

,

v

485

The PR results presented above lead to the following observations. Thick GaAs layers grown directly on Si substrates show a larger shift in gap energy than those grown on "initialization layers" composed of strained layer superlattices. And once grown, thermal annealing has little effect on this shift. The PR measured shift in Eg, A Eg, implies a built-in strain in the plane probed. The values of A Eg can be related to the strain, c, produced by a bi-axial stress parallel to the crystal direction [100] and [010] using the theory of Asai and Oe [10]. X-ray diffraction studies have shown that the GaAS layers grown on Si substrates are under bi-axial strain [11]. The in-plane expansion is a result of the difference in thermal expansion coefficient between the two materials (6.8 x 1 0 - 6 / K for GaAs, and 3.26 × 1 0 - 6 / K for Si at 575K). Our measurements confirm this result. We used our measured bandgap shift to predict the epitaxial growth temperature from known values of the elastic stiffness constants and deformation potentials [12] and the thermal expansion coefficients [13]. For the Ford samples in table 1, AEg = :4 meV results in c = 2.3 ><, 10 -~. Using the expansion coefficients, the growth temperature is calculated to be 6 3 0 ° C , in excellent agreement with the actual growth temperature. In contrast, the Plessey samples, even though they were grown at higher temperature, exhibit a much smaller shift, A Eg = 7 meV. This implies c = 1.1 x 10 -3 and a growth temperature of 3 1 0 ° C if the GaAs thermal expansion coefficient is assumed or 535°C if AlAs expansion coefficient is assumed. Since neither of these values are equal to the actual growth temperature, we must conclude that a portion of the strain produced t;y cooling is being relieved. The relief of strain, we believe, is due to the reduction of the strain gradient at each of the S L S / G a A s interfaces when strain fields of the closely spaced SLS bands overlap. As expected, the PR measured shift in Eg at 82 K from bulk values is even larger than that at room temperature. For the Ford sample F16, AEg = 28 meV at 82 K. This result is in good agreement with the 30 meV red shift measured by Menna and coworkers [14] using photoluminescence at 77 K on GaAs grown on Si under similar conditions.

486

N. Bottka et al. / PR characterization of O M V P E GaAs on Si

The unintentionally doped GaAs on Si layers grown by Plessey showed a carrier concentration higher than those grown by Ford Microelectronics. This may be due to the higher bakeout temperature (1150°C) which may lead to Si contamination from the quartz reactor. The reduced value in N at lower temperatures seems to be characteristic of the OMVPE grown GaAs on Si. A plot of N versus 1/T for sample P5 in fig. 3 gives an activation energy of 18 meV. This activation energy could be associated either with an acceptor (most likely carbon) or some unknown deeper-lying defect related donor complex. The broad peak "I" at 15-20 meV below the gap-energy observed in PR (see fig. 1) could also be related to these unknown "impurity" centers, but its presence in PR is due to a mechanism that differs from electric field modulation of the bands.

4. Summary In summary, photoreflectance enabled us to determine that strain layer superlattice (SLS) initialization layers help to reduce built-in strain due to thermal expansion differences; that thermal annealing reduces the carrier concentration N D + NA, but not the built-in strain; that wafer inhomogeneities in N D + NA are not correlated to built-in strain; the Nt~ + NA is significantly reduced below 300 K indicating the presence of an unidentified level within the band-gap having an 18 meV activation energy; that GaAs layers grown at 750 °C on SLS initialization layers give a value of 3.7 x 10 -4 e V / K and 0.343 eV for dEg/dT and spin-orbit splitting, respectively; and that mea-

surement of ND+NA and Eg is possible on selected area epitaxy.

Acknowledgements The authors thank the following colleagues for their help: Drs. P.D. Augustus, P. Kightley, J.A. Beswick, R. Glosser, R.S. SiUmon, H. Lessoff, D. Law and K.V. Vaidynathan.

References [1] E.Y. Wang and W.A. AIbers, Phys. Letters 27A (1968) 347. [2] R.N. Bhattacharya, H. Shen, P. Parayanthal, F.H. Pollak, T. Courts and H. Aharoni, SPIE 794 (1987) 81. [3] R. Glosser and N. Bottka, SPIE 794 (1987) 88. [4] N. Bottka, D.K. Gaskill, R.S. Sillmon, R. Henry and R. Glosser, J. Electron. Mater. 17 (1988) 161. [5] D.E. Aspnes, in: Handbook on Semiconductors, Vol. 2, Optical Properties of Solids, Ed. M. Balkanski (North. Holland, Amsterdam, 1980). [6] D.E. Aspnes and A.A. Studna, Phys. Rev. B7 (1973) 4605. [7] D.K. Gaskill, N. Bottka and R.S. Sillmon, J. Vacuum Sci. Technol. B, in press. [8] W.E. Spicer, I. Lindau, P. Skeath, C.Y, Su and P. Chye, Phys. Rev. Letters 44 (1980) 420. [9] C. Ito, M. Feng, V. Eu and H.B. Kim, in: Heteroepitaxy on Silicon, Mater. Res. Soc. Syrup. Proc. 63 (MRS, Pittsburgh, PA, 1986)p. 197. [i0] H. Asai and K. Oe, J. Appl. Phys. 54 (1983) 2052. [11] H. Zabel, N. Lucas, R. Feidenhans, J. Als-Nielsen and H. Morkoc, Superlattices and Microstructures 3 (1987) 515. [12] S. Adachi, J. Appl. Phys. 58 (1985) R1. [13] M. Neuberger, Handbook of Electronic Materials, Vols. 5 and 7 (Plenum, New York, 1971). [14] R.J. Menna, S.Y. Narayan, R.T. Smith, M.S. Abrahams and C.W. Magee, RCA Rev. 47 (1986) 578.