The effect of silicon doping on the lattice parameter of gallium arsenide grown by liquid-phase epitaxy, vapour-phase epitaxy and gradient-freeze techniques

Journal of Crysral Growth 50(1980) 648—653 © North-Holland Publishing Company

THE EFFECT OF SILICON DOPING ON THE LATTICE PARAMETER OF GALLIUM ARSENIDE GROWN BY LIQUID-PHASE EPITAXY, VAPOUR-PHASE EPITAXY AND GRADIENT-FREEZE TECHNIQUES P.F. FEWSTER and A.F.W. WILLOUGHBY Engineering Materials Laboratories, The University, Southampton 509 SN!!, IlK Received 7 December 1979;manuscript received in final form 16 February 1980

An extensive study of lattice parameters of silicon doped gallium arsenide grown by liquid-phase epitaxy, vapour-phase epitaxy and gradient-freeze techniques has been undertaken. Lattice parameters were measured by the Bond techniques to a precision of 1 ppm, and the material studied was also cbaracterised independently for carrier concentration, carrier mobility, and, in some cases, for silicon site distribution by infra-red LVM measurements. It is concluded that the p-type LPF layers show a very large lattice contraction compared with undoped material, while VPL and gradient-freeze material have parameters much closer 4 A at hole concentration levels of 6 x I o’ 8 to the undoped value. The lattice contraction in Si-doped LPE layers is about 6 X I o cm3, and the contraction is roughly proportional to the free-hole concentration from ito 6 X 1018 cm3. It is concluded that the p-type Si-doped LPE layers studied contain a high concentration of a contracting defect associated with silicon whose concentration is related to the free-hole concentration. The most likely defect responsible is Slca, present in high concentrations as a compensating donor, although other defects cannot be ruled out.

1. Introduction

2. Experimental

An extensive study has been undertaken to examine the change in lattice parameter of GaAs caused by the presence of commonly used dopants. Large deviations in lattice parameter have been reported previously for Sn- and Te-doped GaAs [1,21 with free-electron concentrations > 1018 cm3. Baker, Hart, Halliwell and Heckingbottom [3} have measured the lattice parameter on a melt-grown Si-doped GaAs slice (n ~ (3.5—4.5) X iO’8 cm3) and concluded that the lattice was dilated, In this present work gradient-freeze, VPE and LPE material, in the carrier concentration range (0.3—6) X 1018 crnt, has been examined; the former two growth methods produced n-type and the latter p-type material. Since Si is an amphoteric impurity in GaAs one might expect a certain amount of electrical compensation leading to difficulty in assessing the total Si concentration, and hence in establishing its effect on the lattice.

Most of the gradient-freeze material was grown at Standard Telecommunication Laboratories (1—larlow), (STL), and the method is described by Greene [41. The growth temperature was 1238°C with an arsenic overpressure of 1 atm. The free-electron concentrations of the ingots were measured either by the Van der Pauw technique or by infra-red plasma resonance. as indicated in table 1. In the cases indicated in the table the mobilities have been estimated either from measurements taken at the front and rear of each ingot, or by the general trend in mobility values of crystals, grown under the same conditions, with similar free-electron concentrations. The remaining melt-grown material originated from Mining and Chemical Products (MCP), were the growth rate was 60—70% faster than ingots grown at STL. The arsenic pressure was maintained close to 1 atm. The free-deetron concentrations and mobiities were measured by the Van der Pauw technique at MCP. 648

P.F. Fewsrer, A.F. W. Willoughby

/ Si doping and lattice parameter of GaAs

649

Table 1 Silicon doped gallium arsenide Slice

Thickness (urn)

Melt-grown (STL) NF39/55

Type

Cone. 18 cni3) (X10

I.LRT (cm2 V1

1.3 a) (LUS) 2.0 2.3 2.5 2.8 a)

2250 2000 2000 h) 1900 1750

5.65356 5.65378 5.65360 5.65383 5.65354

29a) 3.2 3.0—3.6 a) (RHS) ~ a) 3.0—3.7 a) (LHS) 3.6 a) 4.0 a) 4.0 a) (RH5) 4.2 a) 4.2 ~ a) 44a) 4.7 ~ a) (RHS)

1750 1700 b) 1700 1650 1700 1650 1450

5.65361 5.65364 5.65362 5.65361 5.65359 5.65357 5.65367 5.65366 5.65362 5.65367 5.65362 5.65360 5.65364 5.65 350

NF6/52 793/21 781/I1C(B) 781/1IC(A)

n n n n n n

793/11 NF39/ll 781/11C(D) NF39/5(A) 781/I1C(C) 781/1IR/(A) NF39/5(B) 781/11R(B) 793/1 781/I1R(C) 781/I1R(D) NF6/1 NF39/5(C)

n mm n n n n n n n n n n n

Melt-grown (MCP) XK41O/2 XK3650 XK367/2 MCP(D2)F5 MCP(D2)B2

n n n n n

1.3 2.1 3.5 —2.0 —3.0

2200 1800 1600

9 14 —20 8 —20 12

p p p p p p

1.4 3.1 3.2 4.4 5.7 5.9

47 28 29 21 20 18

26 28 27

n n n

0.36 1.1 3.7

77617

LPE (STL) HPI3 l-1P14 SS216 l-lPl6 SS217 1-1P17 VPE (RSRE) S23R10 S23R11 S23R18

~1)

Tgrowth IC)

—-

1400 1400 1350 1350 1300 b)

± ~

±

3 5 2

±

3

±

4

± ±

1 4

±

5

± ±

5.65359± 7 5.6535 1 ± 2 5.65366 ± 5 5.65348 ± 5 5.65354 ±11

— —

3520 2800 1670

a27oC

755—732 755—702 763—730 755—732 769—725 751—705

5.65362 ± 2 5.65339±11 5.65330 5.65332 ± 4 5.65310 ± 5 5.65296± 1

750 750 750

5.65359 5.65353 5.65358

±

2

± 7 ± 10

a) n Measured by infra-red plasma resonance technique. b) Estimated values.

The p-type LPE material was grown at STL, where their layer thicknesses were measured as well as their free-hole concentrations and mobilities by the Van der Pauw technique. Three slices of VPE Si—GaAs grown and electrically characterised at RSRE (Malvern) have been examined. The free-carrier concentrations, mobiities,

layer thickness and temperature, where applicable. are given in table 1. The lattice parameters were measured by the method of Bond [5] with an APEX goniometer. The 800 reflection of GaAs was used (0 80.07°) with CuK~3 radiation (X = 1.392218 A), -to measure the lattice parameter to an accuracy of I ppm. All

P.I~Fewster, A.F. W. Willoughhj’ / Si doping and lattice parameter of GaAs

650

samples had surface orientations close to {10O~.The results were all corrected for refraction, polarisation and Lorentz factors. The temperature was maintained stable within values <±0.2°Cduring any single measurement and for all results within the range 25 ± 3°C.The temperature was monitored throughout to obtain the interpolated average value from measurements at the beginning and end of each lattice parameter determination. All the lattice parameters were adjusted to correspond to a temperature of 27°C by the coefficient of linear expansion (a = 4.54 X 6 °C’, derived from measurements on a nomil0 nally undoped GaAs crystal).

3. Results and discussion Where ever possible each sample was measured in several regions to indicate the degree of inhomogeneity and the ranges are represented by error-bars in fig. 1. The lattice parameter difference from that of undoped material is plotted against the free-carrier concentrations in fig. 1. The lattice parameters are given in table 1. The undoped value for the lattice parameter at 27°Cis 5.65362(3) A. The results are presented graphically in fig. 1 and the most striking feature is the large contraction observed for p-type LPE Si—GaAs, for free-hole concentrations >1018 cin3. This strain is approximately

20

I I

0

1

-20

I

4

‘

5

6

~

-40

linearly related to the free-hole concentration and is larger than the “superdilation effect” in Te-doped GaAs [2]. The mobility in this LPE material is very low, partly because it is p-type and presumably because it is highly compensated. This large contraction is in direct contrast with the measurements on n~melt-grown and VPE material which gave lattice parameters much closer to the undoped value, even at the highest doping levels. In considering possible explanations of this large contraction in p-type LPE layers, we first note that it is generally accepted [12,15] that silicon can be incorporated into the GaAs lattice on a gallium site as a donor, i.e. Si(;a, or on an arsenic site as an acceptor i.e. SiM and numerous other complex centres involving silicon have also been proposed. When GaAs is grown from a Ga-rich solution containing silicon, p-type material is obtained at low temperatures and n-type at high temperatures - the transition temperature depending on the arsenic vapour pressure [6], the substrate orientation and the silicon concentration in the solution [7]. The simplest model of silicon incorporation is to assume that the net acceptor concentration is equal to the difference between the SiAS acceptor concentration and the SiGa donor concentration. Again taking a simplistic model of misfit of silicon in the lattice, we note that both the Pauling [8] and Phillips [9] radii of silicon are smaller than both Ga and As and hence would predict a lattice contraction (see table 2), both in n~and p-type material, whereas a significant contraction was only measured in p-type material. The compensation in p-type material might, however, account for the difference since this implies a much higher total concentration of Si~. 1+ SiAS. The 2 mobility values measured for this p-type (47—l8 cm V~s~)are very low compared with uncompensated

Fable 2 Pauling 181

-60 o o

r

(511)

gradient lreeze(MCP) VPE{RSRE)

————-

•

• Fig. 1. Lattice parameter difference versus free-carrier concentratlons.

—-—

(;a

t.26

As SiGa Si~

1.18 1.17 1.17

Phillips ~r

—

-—-

—

—0.09 —0.01

r —-—-

191 ~r

—

1.225 1.225 1.173 1.173

—0.052 --0.052

P.F, Fewster, A.F W. Willoughby / Si doping and lattice parameter of GaAs

p-type GaAs (>100 cm2V1s’) [19]. An additional factor of potential importance is that silicon will in general be associated with a charged defect,and so an extra contribution should be made to the change in lattice parameter. In the case of Group IV dopants, this charge effect, according to the calculations of Greene [101, would dilate the lattice whether ionised as donors or acceptors. The additional contribution to the lattice strain is calculated to be +2.3, +4.6 and +8.9 ppm for dopant concentrations of 2, 4 and 6 X 1018 cm3 respectively. Consequently we would expect highly compensated material to exhibit a very large dilation. If, on the other hand, the charge effects of Sj~aand Si~ tend to counteract each other then the large drop in lattice parameter with free-hole concentration might be explained by simple misfit theory. Ignoring these charge effects, however, and assuming that S~Gaand SIAS are the only contributors to the lattice strain, we may estimate the concentrations necessary to account for the large lattice strain on the basis of a simple misfit model and applying Vegard’s law, while acknowledging that such a simple model is rarely accurate, particularly in covalent crystals [18]. However, this model is probably adequate for an order of magnitude estimate. Assuming also that p = [Si~] [SiGa] the concentrations in table 3 were obtained, giving values of [SlGal/[S~As]of ‘—0.9 and total concentrations of [S~Ga] + [SiAs] of frorr~ —3.5 X l0’~ cm3 to -—1.1 X 1020 cm3. Unfortunately, however, other evidence indicates that the situation may not be so straightforward as this, Infra-red local vibrational mode studies (LVM) have provided valuable independent data on the role

651

of silicon in gallium arsenide, but, at present, there is a measure of disagreement about the detailed defect structure of LPE layers. Spitzer and Panish [lii coneluded that SiAS is not the dominant acceptor defect and that ESi~a]> [SiAS] in p-type material. They therefore postulated that an additional acceptor defect (perhaps SIGa—~S~ASnearest neighbour pairs) must be considered, although annealing studies [12] indicate that this also may not be the dominant acceptor defect. Kachare et al. [12] later concluded that, although the defect structure of LPE layers is basically different from that of the melt-grown material, each requires the presence of a different unknown acceptor species. Laithwaite and Newman [15], however, have made a thorough LVM study of LPE layers and Bridgman Si-doped GaAs, and their calibration of the strength of the infra-red absorption from SlGa and SiM, while not being very different to that used by Spitzer and Panish [11], does lend to a radical change in the conclusions for the LPE layers, i.e. it was not necessary to invoke the presence of unknown acceptors. The Laithwaite-Newman calibration has, in fact, now been used successfully in characterising a wide variety of other material [16,17]. We therefore cannot, on LVM evidence, entirely reject our very simple model at this stage, but we cannot rule out the possible role, in p-type LPE layers, of other defects in the large lattice strain, since other defects (e.g. S~Ga~S~As pairs) have been postulated by some authors, although it seems likely that the main contracting defect is charged since the contraction is proportional to the free-hole concentration. In the case of the melt-grown Si doped GaAs, the

Table 3 p (X1D’8 cm3)

Pauling radii /9] 6.0 4.6 2.8 Phillips radii [8/ 6.0

4.6 2.8

Z~a(X10° A)

Conc. (X1018 cm3) [SiAsi

[SiGal

[SiGal

6 4 2

62.7 42.3 21.6

56.7 37.7 18.8

119.4 80.1 40.4

0.90 0.89 0.87

6 4 2

58.3 39.2 19.8

52.3 34.6 17.0

110.7 73.8 36.9

0.90 0.88 0.86

+

lSjAsl

[SiGal/[SiAsl

LI

652

P. F Ecwster, Al. W. R’illouglihr / Si doping and lattice parameter of GaAs

[S~M]![S~Ga] ratio has been measured by LVM studies on samples of crystal NF/39, adjacent to those measured in this study as being about 0.2 [13] and mass-spectrometric measurements of total Si were not in disagreement with this estimate. It can be seen from fig. I that most of these samples, in contrast to the LPE layers, show little difference in lattice parameter from the undoped value. Some of the samples investigated do deviate from the undoped value by as much as 2.1 X i0~ A. although the variation in parameter within any one slice was not greater than ‘-1.1 X I o~ A about its mean value. Samples on regions of a slice with a lattice parameter lower than the undoped value could be a consequence of a higher concentration of contracting defects (possibly material that is more compensated, cf. LPE Si—GaAs). For example the two samples MCP(D2)F5 and B2 both have lattice parameters less than the undoped value and could be caused by a high concentration of Si-related contracting defects resulting in greater compensation of the donor population, and hence the free-electron concentration is limited, as experienced from this ingot, by the upper limit of [Si] that can be incorporated in GaAs during melt growth (1238°C),before faceting occurs, The two slices measured at the lower carrier concentration range which gave large lattice dilations are difficult to explain. These two slices were from the front ends of ingots and so they may l1ave been susceptible to contamination, although the major onwanted impurity was thought to be Al, but not in concentrations to account for the observed dilation, It is also difficult to explain these deviations in terms of differing growth conditions; since slices 793/1 and 11 are unstrained whilst 793/21 is one of the heavily strained slices although from the same crystal. The conditions are unlikely to have changed during growth, but possibly the level of contamination was sufficient at the front end of the ingot to disturb the point defect equilibrium and create this dilation, The lattice parameter of the n-type VPE grown Si—GaAs is close to the undoped value in agreement with the melt-grown material. The mobility of this VPE material is comparable to the melt-grown Si—GaAs and presumably the compensation ratio is similar, The variation in lattice parameters for the various slices of n-type Si—GaAs (melt-grown and VPE)

might thus be explained partly by differences in concentration of some contracting defect associated with silicon, and by impurity levels disturbing the defect equilibrium, giving rise to •a dilation of the lattice. There is also the possibility that a dilating defect exists in melt-grown Si—GaAs as postulated in Te—GaAs [2], and the two samples exhibiting an increase in lattice parameter are associated with a high concentration of this defect at low Si concentrations, but this dilation is compensated by contracting defects in other slices. We note, also, that Baker et a!. [3] measured a dilation in one sample at similar Si 3). levels (n ‘-—1 X 1018 cm It is interesting to note that there is no detectable difference in lattice parameter between n-type meltgrown (1238°C) and n-type VPE (750°C)Si—GaAs, particularly as they are grown at very different ternperatures. The p-type LPE material exhibits a large contraction in the lattice which is linearly related to the freehole concentration above 1018 cm3, and could be caused by very high concentrations of some contracting defect, which could be SIGa providing it is electrically compensated. The calculated compensation ratio (ISiGaJ/[SiAsl , table 3) is constant over this range at ‘-‘-0.9, If Si~ is not the major acceptor in Si—GaAs then still a high compensation ratio requires a high [SiGa] level and according to the covalent radii of Pauling [8] S~Ga is the important contracting defect. Whatever the contracting defect might be, it must be a species whose concentration is related to the free-hole concentration. Rode, Brown and Afromowitz [14] observed a linear dependence in lattice dilation with free-hole concentration in LPE Ge—GaAs. They assumed that the bond-breaking contributed by free-holes was much larger than size effects. This dilation was small (~aIa‘-‘-l0~ for p ~~.1019cm’-3) compared with that observed in the present work, but it is noteworthy that a defect species whose concentration is related to the free-hole concentration may also be important in material with this Group IV dopant. The influence of contaminants must also be considered to explain the difference in lattice parameter between Si-doped p-type LPE and n-type VPE and melt-grown materials. The clear linear dependence would require any unwanted impurity to be strongly related to the free-hole concentration and be present

LI

P. F Fest’ster, A. F. W. Willoughby / Si doping and lattice parameter of GaAs

in high concentrations (>10’~cm’-3 at p ‘—6 X cm°)which is unlikely.

1018

653

Acknowledgements

The concentration of the defect causing this con-

The authors acknowledge the following for supply

traction clearly does not differ significantly from VPE to melt-grown material which are grown at different temperatures; this could either mean that the concentration of the defect is essentially temperature independent in n-type Si-GaAs or that the defect responsible is specific to p-type material. All the p-type LPE Si-GaAs samples examined were grown at similar temperatures (‘-‘-750°C),remote from the p—n transition temperature, and therefore any relation between contracting-defect concentration and growth temperature cannot be observed. Clearly further work is required to ascertain the defect that contracts p-type LPE Si-GaAs, for example; full characterisation before and after heattreatment, and examination of material grown at different temperatures (perhaps above and below the p—n transition temperature). Such experiments could

of GaAs material and useful discussions: P.D. Greene, K. Shaw, D. Ashen and J.B. Mullin. Discussion with R.C. Newman, D.C. Newton and DJ. Stirland are also gratefully acknowledged. This work has been carried out with the support of the Procurement Executive, Ministry of Defence and sponsored by DCVD. Financial support by SRC is also acknowledged.

help to establish the relative role of charged and neutral defects in the equilibrium situation in Si—doped LPE material. From the present study Si~a appears to be the most probable defect for contracting the lattice, since it has a smaller covalent radius than GaGa, but it is not possible, at this stage to rule out a wide variety of other possible defects.

4. Conclusions It is concluded, from an extensive study of the lattice parameters of silicon doped gallium arsenide grown by liquid-phase epitaxy, vapour-phase epitaxy and gradient freeze techniques, that p-type LPE layers show a very large lattice contraction compared with undoped material, while VPE and gradientfreeze material have parameters much closer to the undoped value. The lattice contraction in Si-doped LPE layers is about 6 X io~A at hole concentration levels of 6 X 1018 cm3 and the contraction is roughly proportional to the doping level from (1—6) X l0’~cm3. It is concluded that the p-type Si-doped LPE layers studied contain a high concentration of a contracting defect associated with silicon whose concentration is related to the free-hole concentration. The most likely defect responsible is S1Ga present in high concentrations as a compensating donor, although, other defects cannot be ruled out.

References Ill

J .B. Mullin, B.W. Straughan, C.M.H. Driscoll and A.F.W. Willoughby, J. Appi. Phys 47 (1976) 2584. [21 P.S. Dobson, P.F. Fewster, D.T.J. Hurle, P.W. Hutchin-

son, 131

J.B. Mullin, B.W. Straughan and A.F.W. Willoughby, Inst. Phys. Conf. Ser. No. 45 (1979) ch. 2, p. 163. J.F.C. Baker, M. Hart, M.A.G. Halliwell and R. Heckingbottom, Solid State Electron. 19 (1976) 331.

[41 PD. Greene, unpublished data. [51 W.L. Bond, Acta Cryst. 13(1960)

814. [61 J.O. McCaldin and R. Harada, J. Appi. Phys. 31(1960) 2065. [7J B.H. Ahn, R.R. Shurtz and C.W. Trussell, J. App!. Phys. 42 (1971) 4512. [8] L. Pauling, The Nature of the Chemical Bond (Cornell Univ. Press, Ithaca, NY, 1960) ch. 4. [9] J.C. Phillips, Bonds and Bands in Semiconductors (Academic Press, New York, 1973) ch. 1. 1101 P.D. Greene, Solid State Commun. 21(1977) 827. 1111 W.G. Spitzer and M.B. Panish, J. Appi. Phys. 40(1969) 4200. [121A.H. Kachare, W.G. Spitzer, J.M. Whelan and G.H. Narayanan, J. Appi. Phys. 47 (1976) 5022. [13] R.C. Newman, unpublished data. 1141 D.L. Rode, R.L. Brown and M.A. Afromowitz, J. Crystal Growth 30 (t975) 299. [151 K. Laithwaite and R. C. Newman, J. Phys. C (Solid State Phys.) 9 (1976) 4503. [16] K. Laithwaite, R.C. Newman, J.F. Angress and G.A. Gledhill, GaAs and Related Compounds (Inst. Phys. No.J.33a, 1977)R.C. p. 133. [17] Conf. MR. Set. Brozel, Butler, Newman, A. Ritson, D.J. Stirland and C. Whitehead, J. Phys. C (Solid State Phys.) 11(1978)1857. [181 C.M.H. Driscoll, A.F.W. Willoughby, J.B. Mullin and B.W. Straughan, Inst. Phys. Conf. Set. No. 24 (1975) 275. [19] S.M. Sze and J.C. Irvin, Solid State Electron. 11(1968) 599