Surface processes controlling the growth of GaxIn1−xAs and GaxIn1−xP alloy films by MBE

Journal of Crystal Growth 44 (1978) 75—83 © North-Holland Publishing Company

SURFACE PROCESSES CONTROLLING THE GROWTH OF Ga~Ini_~As and Gaxlni_xP ALLOY FILMS BY MilE C.T. FOXON and B.A. JOYCE Philips Research Laboratories, Redhill, Surrey, England Received 14 November 1977; manuscript received in final form 24 January 1978

The kinetics of surface processes involved in the growth of certain Ill—Ill—V alloy films from beams oftheir constituent elements has been investigated by modulated beam techniques in combination with RHEED and AES. It has been shown that the principle limitation to the growth of these alloy films by MBE is the thermal stability of the lesser stable of the two Ill—V compounds of which the alloy may be considered to be composed. This leads to the preferential desorption of the Gp V element (as a dimer), leaving an excess surface population of the Gp III elements. At somewhat higher temperatures the situation is further complicated by the preferential desorption of the more volatile Gp III element. Surface segregation effects as such appear to be minimal.

1. Introduction

cation in a range of optoelectronic devices has been demonstrated. It appears, however, that the Ga~Ini~As~P1._~ quaternary system is less easy to

Molecular beam epitaxy (MBE) is now a widely used technique for the preparation of thin epitaxial films of Ill—V and Il—VI semiconductor compounds and alloys. It is based on the condensation of directed beams of the appropriate elements on a heated substrate surface under UHV conditions to obtain highly controlled film growth. Several reviews [1—31have appeared recently dealing with various aspects of the subject. For the growth of Ill—V compounds and alloys, the group III element flux is always atomic, but the group V element beam can be either diatomic or tetratomic, depending on the source material. The kinetics of the interaction of Ga with both As2 and As4 beams on GaAs surfaces has been extensively investigated by Arthur [4—6] and Foxon and Joyce [7—9]. The only work reported on kinetic processes in Ill—V alloy film growth however, has been the determination of the relative sticking coefficients of As4 and P4 in the growth of GaAs1~ alloys [10.11], but in practical terms, an empiric approach has proved quite successful in the growth of Al~Ga1~As alloys (see e.g. ref. [1]). Within this system films can be grown having good crystallographic perfection and compositional control, and their appli-

control, at least as determined by the photoluminescent properties of the films [12]. In this paper, we report the results of a study of surface adsorption, desorption and segregation processes occurring in the growth of the ternary alloys Ga~In1_~Asand Ga~In1_~P on {l0O} GaAs substrates. To establish greater reliance on our data we have used a wide range of techniques, including modulated beam mass spectrometry [13] AES and RHEED. We are here primarily concerned with surface processes as such; their influence on film properties (structure, composition, electrical and optical behaviour) will be the subject of a separate communication.

2. Experimental The basic experimental arrangement has been described before [7,9,13], but for the work reported here a fluorescent screen was installed immediately opposite the CMA, so that the coaxial gun could also be used to generate RHEED patterns (at 3 keV). The beam sources, containing Ga, In, As and P, were 75

76

C. T. Foxon, BA. Joyce

/ Surface processes controlling growth

of 111—111—V alloy films

independently shuttered and fluxes could be measured with a calibrated ion gauge mounted in the beam path. The calibration can be made absolute for Ga and In (the ion current is directly related to the mass of each element condensed on a plate in the

3. Results

beam path) but only relative calibration is possible for As4 and P4 because of non-unity sticking coefficients. In general, however, the jon gauge was only used to ensure reproducibility of fluxes, not to measure absolute magnitudes. Substrates were {100} oriented GaAs slices, -~l5 mm square and 0.25 mm thick. They were polished in a 7 1 : 1; H2S04 : H202 H2O etch before being mounted on an internally heated Mo block in the UHV system, using a thin layer of In to provide good thermal contact between block and substrate, in a simple modification of the method Choa and Arthur [1]. Temperatures were described measured by with butt-welded chromel—alumel thermocouple passing through the Mo block, but only 0.25 mm below the

plete picture of the surface processes involved in alloy growth we need to begin by considering the cases where x = 1 and x = 0, i.e. GaAs and InAs. We have previously investigated the Ga—As4—GaAs{100} systern in considerable detail [7], but the relevant measurements were repeated for this work and confirmation obtained of their reproducibility. To study the In—As4—InAs{lOO} system, parallel orientation epitaxial films of InAs, —~i000A thick, were grown in-situ on the GaAs substrate prepared as described in section 2, using 2s~, In andand As4 fluxes oftemperature ~~,1013atoms a substrate of (molecules) cm 600 K.

We will describe in detail results for the Ga~Ini~As system, but qualitatively identical behavjour was observed for Ga~In1~P. To obtain a com-

3.1. Surface interactions involving As 4 and As2

surface to which substrates were attached. Using this arrangement we believe that measured values will be correct to ±5K. The base pressure after bakeout was —P5 X 10_li Torr, and the9 operating with beams increase on, was and 10_bpressure, Torr, the pressure between i0 being attributable almost entirely to As 4 (or P4). Substrate surfaces were prepared in-situ by heating to between 850 K and 890 K for 20 mm RHEED in an As4 flux 2s1. The pattern of ~~~10l3molecules cm from a surface so prepared showed a streaked, weakly reconstructed As-stable surface, i.e. GaAs(100) (2 X 1). Before any kinetic measurements were made, a GaAs film, ~1000 A thick, was always grown in-situ on the substrate, using Ga and As 4 beam fluxes of 2s~). equal intensity (~~.l0l3 a toms/molecule cm The film surfaces were strongly reconstructed, Asstable, c(2 X 8) or (2 X 4), the RHEED patterns also being highly streaked [14,15] *~ AES examination showed all surface contaminants to be below the detection limit (0.01—0.001 monolayer). This was the type of surface used as the basis for all kinetic measurements.

desorption We first determined the sticking coefficient of As4 (SM4) as a function of temperature on InAs and GaAs substrates, constant (JM4) but with no incidentusing Ga ora In fluxes.As4 The flux starting surface was

As stable RHEED pattern

o 6r

I

molcm

/1

I

o.4 SAS, o

S

I

InAs

T

4

o.4

1

I / ~

o4 I

I II 0~—1~—~

500 It is not possible to distinguish between these two possibil-

C(8x2)or(4x2)

o

I

*

In Stable RHEED pattern

—

C(2x8)or(2x4)

11

~ II

I

I’

J__-,-’

a

x

13

~ /i~0~AS

~ As Stable RHEED pattern C(2x8)or(2x4)over

1—T—-~1

I

whole temperature range

600

700

600

T 5 (K)

ities from the RHEED patterns. For a fuller discussion on this point see refs. [211 and [22).

Fig. 1. Sticking coefficient of As4 as a function of substrate temperature on {lOO}GaAs and {100}InAs surfaces.

CT. Foxon, B.A. Joyce

/ Surface processes controlling growth

in each case As-stable, i.e. produced a c(2 X 8) or (2 X 4) RHEED pattern. The measurements were made by examining the response of the mass spectrometer while modulating the desorption flux, with an unmodulated As incident flux ofin~..10b3mole2s~. The 4results are shown fig. 1, from cules cm which it is clear that at temperatures a little over 550 K SAS becomes measurable on both InAs and 4

77

1.5

I -

13mo1 cm2S

~As

1

4 r15x10

~

..~

1.0

As

.

GaAs {iOO} surfaces. However, its temperature dependence is much stronger on InAs, although in neither case does it exceed 0.5 [7]. RHEED patterns obtamed while SAS4 was being determined showed that for GaAs the surface remained As-stabilized (c(2 X 8) or (2 X 4)) over the whole temperature range investigated, but for InAs a transition to the In-stable reconstruction (c(8 X 2) or (4 X 2)) occurred at ~720 K (for 2s~).The an incidentnature As4 flux 1.5 X 1013beam molecules of theofmodulated expericm ment did not allow the sharpness of the transition to be determined with any degree of accuracy. The addition of the RHEED facility made lt possible to ascertain that the limitmg value of 5As 4 = 0.5 was only reached for an In-stable surface; for Asstable it was <0.5. It has in fact been generally confirmed during the course of this work 5Ai that for materials in the Gaxlni_xAsyPiy system, 4 (and Sp4) only reaches its maximum value of 0.5 when the surface has reconstructed to the group III element-stable form. It is always <0.5 on surfaces which are group V element-stable. It has previously been suggested [6,7,9] that the 5As temperature dependence of 4 results from the formation of a temperature dependent Ga (or In) surface population produced by loss of As2 from the substrate. On the basis of this model, the sticking coefficient results imply that As2 is lost much more rapidly from InAs than from GaAs as the temperature is increased from 500 K. This has been confirmed by direct observation, as illustrated in fig. 2, which shows the temperature dependence of the As~/As~ ion intensity ratio as measured in the mass spectrometer. The measurement was performed by modulating the flux desorbing from the surface when an unmodulated As4 flux was incident on the surface. The ratio is sensibly constant up to the temperature at which 5As 4 becomes measurable, and then the proportion of As~increases significantly, but much more rapildy from an InAs surface than from a GaAs one. As~is

of Ill—Ill— V alloy films

Stable

—..in stable

.

.2

RHEED pattern

/

RHEED pattern

0

C(2x8)or(2x4)

/

C(8x2)ar(4x2)

~ 0.5

.~-InAs

a

•~

,2

O__o

Ga As

As stable RHEED pattern C(2x8)or(2x4) aver whole temperature range 0 600 700 800 T5 (K) Fig. 2. Ion intensity ratio (A4/As~) as a function of sub000

strate temperature. The ratio corresponds to the flux desorbing from IlOo}GaAs and InAs surfaces with an incident As 2s~, but no 1.5 X 1013 molecules cm incident 4Gaflux or Inoffluxes.

.

always present in the fragmentation pattern of As4, so it is the increasing ratio which is important, and confirms that As2 as such is leaving the surface, in addition to the fraction of incident As4 which desorbs directly. It has been well established [16,17] that it is only the dimeric form of the Gp. V element which is lost from the surface of Ill—V compounds during thermal dissociation, in agreement with the above observation. The final measurement for the individual compounds was of the (apparent) sticking coefficient of As4 on InAs and GaAs as a function of temperature in the presence of In and Ga beams respectively. These flux conditions obviously correspond to cornpound growth, and the results are shown in fig. 3. Below ‘--600 K, SAS4 is temperature independent on both InAs and GaAs, confirming the previous5M results for GaAs [7]. The apparent rapid decrease of 4 to zero between 600 and 700 K on InAs (and the much slower decrease on GaAs) are,however, only artefacts of the measurement technique. The reference level for the amplitude of the As4 desorption signal in the absence of a Gp III element flux to the surface is assumed to correspond to zero sticking coefficient, but this is only valid if there is no significant Gp III

CT. Foxon, BA. Joyce / Surface processes controlling growth of 111—111— Valloy films

7g

03

13mol cm2s~

0.6 0.5

JJj~, r8x1012 ri 4x10 atoms cm~s~ JGa~8x1O1 atoms cm ~

(2x4)over whole

0~ SAS 4

JGO~1x10’3atomscm2s~

02

I ri—. In

—x

temperature range

0 3 02

~

0O—0~o 0 G~As As stable *— RHEED pattern C(2x8)or(2x4)

01

500

600 T(K)

temperature.

01

In stable RHEED pattern ~ C(8x2)or(4x2)

_____

400

irreproducible above this

I

-~-~—.

I

in flux only

x Ga~Influxes

and In ~Ga results are

SAS4

.~\

~

.l&~s~ 4xlOt3mol cm2s1

As stable RHEED pattern C(2x8)or

700

ô.

0

800

Apparent temperature dependence of SAS4 on 2s1In {100}GaAs and InAs surfaces in the presence of Ga and beams respectively. = 1.4 x 1013 molecules cm for ~th = 8 X 1012 ~As4 atoms cm2s1. ~As 2s_i for~Ga = 8 X 1012 atoms4cm2s~. 1.4 x 1013 molecules cm

500

1

6001

Fig. 3.

element surface population produced by thermal dissociation. The results presented in fig. 3 simply confirm, therefore, the formation of In, and to a lesser extent Ga surface populations, lndependent of the external flux. With the additional In flux however, In-stable RHEED patterns are found at a rather lower temperature (650, cf. 720 K), although for GaAs the surface remained As-stable over the complete temperature range investigated, Having established conditions for the loss of As2 from the two compounds representing each extreme of the alloy range, both during growth and in the presence of an As4 flux only, we are in a position to examine the effects occurring during alloy growth on a substrate of one of the compounds. In particular, we need to determine to what extent the growth of an alloy relates to the growth of the separate 5As compounds. The measurement made, was of 4 as a function of temperature, with the following beam fluxes incident on an As-stable {lOO} substrate: 2s1, GaAs ~in =~Ga= 1 X 1013 ~4 atoms cm2s1. The As ~As4 X 1013 molecules cm 4 desorption flux amplitude was determined for incident fluxes of As4 alone, As4 + Ga, As4 + In and As4 + Ga + In. Each source was independently shuttered so the sequence could be repeated as many times as required. The results are shown in fig. 4, from which it may be5M seen that up to a substrate temperature of ‘-620 K, 4 is temperature independent and proportional to the flux of the Gp III element, the values being completely reproducible. More importantly, In and Ga are completely

RHEED pattern As stable C(2x8)or(2x4) far GaAs,In As I

I

and GaIn As.

I

I

700 T~(K)

800r

____

900

I

RHEED pattern As stable

RHEED Ga/In pattern stable

for GaAs In stable RHEED pattern

I for andGaGaAsIn In AS. As

I

C(8x2)or(4x2) I for In As and I Ga In As. Fig. 4. Apparent temperature dependence of SAS growth in the system Gaxlni_xAs.

4 for alloy

equivalent in this respect, one Ga atom having exactly the same influence on SA54 as one In atom, so that if J1~= JGa, and SAS4 = n in the presence of the Ga flux alone, it equals 2n when both Ga and In fluxes are incident. In this temperature range the surface is always reconstructed to the As-stable form. Above ‘--620 K, however, SAs4 becomes completely nonreproducible for all cases when an In flux is, or has just previously been incident on the surface, i.e. there is a finite In surface population. RHEED shows the surface to be always Gp III element stable whenever In is present at the growth temperature (>620 K), while Auger spectra obtained after cooling the substrate down to ‘--500 K show surface enrichment of In and Ga, with1),a corresponding reduction in the As signal ((table compared with films grown at <620 K). .For the interactions of Ga, In an P4 beams on GaAs substrates, leading to the formation of Ga~In1~P films, almost identical sticking coefficient and RHEED pattern behaviour to that for Ga~In1~Asis observed. There are, however, two additional features in the Auger spectra which are important in considering all of the processes occurring during alloy film growth. Firstly it is apparent from fig. 5a that for the phosphorus alloy on a GaAs sub-

C. T. Foxon, B.A. Joyce / Surface processes controlling growth of 111—Ill— Valloy films Table 1 Normalized Auger peak heights for GaInAs alloy films grown at 570 and 670 K; incident flux ratios correspond to Ga 0~5In05Asfor the film composition Auger transition

Auger peak heights normalized to the ASL3M4M4 (1228 eV) peak

228eV) A5L3M4M4(1 ASM 4VV (31 eV)

Growth at 570K 1

Growth at 670K 1

2.52

0.50

GaL3M4M4 (1069 eV)

0.68

1.09

GaM3M4M4 1~MNN(404 (51 eV)eV)

0.37 1.42

0.56 1.74

79

features to be related to the low congruent evapora-. tion temperature of InP, which is ‘--635-K [18]. Growth above this temperature can result in surface decomposition and the formation of free In. Similar AES results could have been obtained if alloy growth above 620 K occurred as a discontinuous 3D film. Nomarski interference micrographs of the film surfaces do not support this interpretation, however. 3.2. Annealing experiments on Ga05In05As films The measurements so far described clearly indicate that during alloy growth at temperatures >620 K there is an increase Ga and In be surface atom populations, which incantheapparently attributed mainly to the desorption of As

strate, even after growth of 250 A of material, there is still a strong As signal in the Auger spectrum for substrate temperatures >620 K. For growth below this temperature there is no detectable As signal at this thickness (fig. Sb). The second effect is the strongly enhanced In signal obtained from the film growth at the higher temperature. We believe both

2. However, experimental times are quite long, and it is possible that surface segregation processes could provide some contribution to the accumulation of In and Ga. To assess its relative importance an epitaxial film of In0 5Ga0 5As, approximately 1000 A thick was grown on a Ii00} GaAs substrate at —~560K (i.e. below the temperature at which any excess surface population would be formed), and its relative surface composition checked

Thickness = 250A, 1.5 ~Ga~P

~pi

.2

1:1 :2 (a)T5=670K

\s1228eV In 404ev

~

~I1~- _ LA5~570K

I

25

ao0.5

4-. L 40)) 0) 4,

I

In

‘—~--—--—‘

Ga

II

400 900

‘—.-------.

As

4-~

a_

~

.L._ 0~____________________ x

\

5 ~As1228eV / ~A 0GalO69eV As Ga 1228eV 51 eV

x

—

X_)

As 31 eV

I

1300

E (eV)

Fig. 5. Auger spectra obtained from surfaces of 250 A thick epitaxial alloy films grown with flux ratios of ~In: ~Ga Jp4 = 1: 1: 2 on 1100 jGaAs substrates. (a) GaInP at 670 K, (b) GaInP at 570 K. Note the large In enrichment and the large As signal at the higher temperature.

c

_______________________________

700 750 800 T5 (K) Fig. 6. Normalized Auger peak heights as a function of annealing temperature after 30 min anneals for a 1000 A spitaxial Ga0 5In0 5As film on a {100}GaAs substrate. 550

600

650

80

C. T. Foxon, BA. Joyce / Surface processes controlling growth of 111—111— V alloy films

by AES. It was then heated for periods of 30 mm without removal from the UHV system in a flux of As4 of 1.5 X 1013 molecules cm2s~ at a series of increasing temperatures. After each anneal the relative surface composition was again checked by AES, the substrate being cooled back to the growth ternperature (560 K) while the spectra were being recorded. In addition, for one temperature (640 K) the anneal was repeated for a second period of 30 mm, but no differences were observed in the Auger spectra. The results are presented in fig. 6 as Auger peak heights normalized with respect to the As L 3M4M4 (1228 eV) peak height. It is apparent that any surface accumulation of In directly attributable to surface segregation is small, although the slight increase in relative In concentration up to ‘-‘700 K is probably a real effect, outside experimental error. That a larger segregation effect is being masked by the occurrence of In as surface droplets, thereby decreasing its effective Auger signal, is very improbable since for the temperatures concerned and with the large incident As4 fluxes, surface dissociation does not occur on any significant scale. This interpretation is reinforced by the concomitant decrease in the relative Ga surface concentration as determined from the normalized Ga 51 eV (M3M4M4) peak height, whereas the normalized Ga 1069 eV (L3M4M4) peak height stays virtually constant to 700 K. The low energy peak is effectively providing information only about the first atomic layer, while the escape depth of 1069 eV electrons corresponds to several atomic layers. It is important to note also that the relative As surface population stayed constant over the whole temperature range, as determined from the normalized 31 eV (M4VV) As peak height. Of much greater significance, however, is the rapid decrease in the relative In surfaceand concentration at temperatures >700 K. Goldstein Szostak [191 have reported the preferential evaporation of In from Ga~In 1~Asalloys at temperatures ~770 K, so it appears that at higher temperatures, at least in the static situation where no In is being supplied externally, In surface atoms are desorbed at a rate faster than that at which they can be replaced from the bulk by diffusion or segregation. This process could in principle influence growth, but in that situation we are much more concerned with dynamic desorption effects, i.e. the desorption of incident In atoms, or

desorption from an In adatom population, which we have established can form during growth. Direct measurements of In desorption are described in the next section. Desorption rates of incident Ga atoms in this temperature range (<875 K) are more than a decade slower than In, so do not provide a limiting effect on the growth process. 3.3. Indium desorption measurements We describe briefly in this section the results of a series of experiments aimed at determining directly In—GaAs {lOO} surface interactions, particularly In desorption behaviour with and without a coincident As4 flux. In practice, this system has proved quite complex, the details of which will be reported sepa-

1~ (K 910

830

770

710

10

102

o

o 14atoms

.~

10

E

i.5~0.1eV

~ J1~~2.8x1O ‘ cr~~2 o

1

.j~,,

~ lx i014atoms cm2s~

112

113

114

1000/Ts(K)

Fig. 7. In desorption rate from {100}GaAs as a function of temperature, ED = 1.50 ±0.1 eV.

C. 1’. Foxon, BA. Joyce / Surface processes controlling growth of 111—Ill— V alloy films

rately; we are here only concerned with those aspects directly relevant to film growth. The first measurement was to determine the desorption energy of In from an As-stable {l00} GaAs surface, using a large, constant (~~l014 atoms cm2s’) incident flux of In. The method used was to modulate the In flux desorbing and measure the desorption signal amplitude as a function of substrate temperature. The results are shown in fig. 7, and assuming a rate equation of the form of eq. (1): —dn/dt

=

k exp(—ED/kl),

(I)

~

2’

2

10

results are shown in fig. 8. At low incident rates the desorption flux is proportional to the incident flux, as expected, but with increasing flux the desorption rate becomes constant for a fixed substrate temperature. This is clearly consistent with the build-up of a steady state surface population of In atoms, and cornbined with the results of fig. 7 it can be deduced that the activation energy for desorption from this population is “-1 .5 eV, which is 1 eV lower than the (equilibrium) enthalpy of sublimation of In. No differences were observed in the desorption flux over any part of the incident flux—substrate temperature range 2s’ when an As4 influx of ‘--1toX the 1013Inmolecules cm was incident, addition flux. Finally, by modulation of the incident flux of In, we established that its interaction with GaAs Il00} surfaces was very non-linear. The details are not relevant to this article, but it may be noted that surface lifetimes are very long (—1 s) even at temperatures where the sticking coefficient is low. With respect to the growth of GaInAs alloys, however, the important

- -

~7x

0/

~

~10

/

a

the activation energy for desorption, ED, is 1.50 ±0.1 eV, where —dn/dt is the desorption rate and k an overall rate constant. This value of ED may be compared with the enthalpy of sublimation of In, 298this K~oftemperature 2.50 eV [20]. That dependence could be observed by modulating the desorption flux implied that a steady state population of In surface atoms was being formed, such that the desorption rate was not controlled by the arrival rate of In atoms, i.e. the situation equivalent to alloy growth at temperatures >620 K. To confirm this, the desorption flux was measured as the incident flux was varied over almost three decades (2 X 1012_lOis atoms cm2 s1) at three substrate temperatures (825, 850, 875 K). The

81

~.2 I

T

o

5=875K T585O K

o

T5=825K I

10’s 2s~) Inc:dentrate In from flux 1100)-GaAs (atoms çmas a function of Fig. 8. In desorption incident In flux at three substrate temperatures (875, 850 and 825 K). ~O12

1013

x

1014

results from the In—GaAs interaction experiments is that In can desorb from an In adatom population with an activation energy which is low compared to the enthalpy of vapourisation. It has already been established that such a population can form by preferential loss of As 2 during growth at temperatures >620 K.

4. Discussion and conclusions Perhaps the most critical factor in alloy film growth by MBE is to produce material whose cornposition is homogeneous and related in a comparatively simple way to the incident fluxes. We have established that for the growth of Ill—Ill—V ternary alloys two surface processes can occur which seriously impair compositional uniformity and control. These are respectively preferential desorption of the Gp V element, and, at some higher temperature, preferential desorption of one of the Gp III elements. If we consider first the desorption of the Gp V element, the results indicate that in this respect the alloy may simply be considered as a mixture of the two binary compounds of which it is composed, i.e. InAs and GaAs in the case of Ga~In 1_~As. As2 then desorbs to leave a surface with an enriched In and Ga

82

C. T. Foxon, BA. Joyce / Surface processes controlling growth of 111—111— V alloy films

adatom population. In principle this loss by desorption could be balanced by increasing the incident As4 flux, but in practice, as the temperature is increased the required flux becomes impracticably high. This is especially true for GalnAs and GaInP. AES results conclusively demonstrate that provided growth is carned out at temperatures below which InAs (or InP in the case of Ga~Ini_~P) are thermally stable, the surface composition of the alloy reflects the relative flux intensities of Ga and In. The preferential desorption of As2, which increases with increasing temperature, clearly leads to the temperature dependent sticking coefficient of As4, but it is important to note that in agreement with our previous model for GaAs [71,SA54 never exceeds 0.5, and in fact this limiting value is only reached when the surface has reconstructed to the Gp III element stable form. These conditions presumably correspond to the maximum number of surface atoms possible before the free metal begins to form on the surface, When alloy growth on a compound substrate is carried out above the temperature at which significant desorption of the Group V element begins from the alloy, the Gp V element in the substrate can be transported to the surface of quite thick films (see fig. 5). The corresponding surface enrichment of one of the Gp III elements (In in this case) suggests that the effect results from growing above the congruent evaporation temperature of one of the binary end members (InP). The annealing experiments, which were performed primarily to investigate possible segregation effects, provide a convenient link between preferential desorption of the Gp V element and the Gp III element. It appears that a very limited amount of surface segregation can occur, in that there is a small enrichment of In with respect to Ga in Ga0 5In0 5As up to —700 K. This is in accordance with the simplest thermodynamic

explanation of surface segregation, that the component with the lower

which requires heat of vapourisation should be enriched in the surface to minimise the surface free energy. By far the greater effect, however, is the loss of In from the surface by evaporation at temperatures 700 K, a result confirmed by direct determination of InGaAs desorption behaviour, using modulated beam techrnques, which indicates the overall activation energy for

desorption of In from an In adatom population to be —1.5 eV, or 1 eV lower than the heat of vapourisation. Since this value is maintained for quite high In coverages, it may be suggested that the rate limiting step is not desorption itself, but surface diffusion to a preferred desorption site. This is to some extent supported by the non-linearity of the system response when the incident In flux is modulated. Simple atom desorption as such can only follow first order behavior, so clearly some additional process is occurring. We can conlude, therefore, that the principal limitation to the growth of Ill—Ill—V alloy films by MBE is simply the thermal stability of the lesser stable of the two Ill—V compounds of which the alloy may be considered to be composed. In addition, preferential desorption of the more volatile Gp III element can occur at higher temperatures. It is perhaps fortunate that GaxAli_xAs was the ternary alloy chosen for a range of devices which have been successfully fabricated using MBE [1]. It happens that AlAs is more thermally stable than GaAs so by using growth parameters previously determined for GaAs, it is to be expected that films of GaxAli_xAs, with good compositional uniformity and control could be prepared by MBE.

Acknowledgements The authors are grateful to Mr. J.H. Neave for advice and help in installing the RHEED facility, and for valuable discussions on the results, and to Dr. G. Laurence for assistance with the surface segregation experiments.

References [1] A.Y. Cho and J.R. Arthur, Progr. Solid State Chem. 10 (1975) 157. 121 BA. Joyce and CT. Foxon, ESSDERC 1976, Inst. Phys. Conf. Ser. 32 (1977) 17. [3] BA. Joyce and CT. Foxon, Japan. J. Appl. Phys. 16 (Suppi. 16-1) (1977) 17. [4] JR. Arthur, in: Proc. Conf. on Structure and Chemistry of Solid Surfaces, Ed. GA. Somorjai (Wiley, New York, 1969) P. 46. [5] JR. Arthur, J. Appl. Phys. 39 (1968) 4032. [6] J.R. Arthur, Surface Sci. 43 (1974) 449. [7]CT. Foxon and B.A. Joyce, Surface Sci 50 (1975) 434.

C. T. Foxon, L.A. Joyce / Surface processes controlling growth of 111—111— V alloy films [8] B.A. Joyce and C.T. Foxon, J. Crystal Growth 31 (1975) 122. [9] C.T.Foxon and B.A. Joyce, Surface Sci. 64 (1977) 293. [10] JR. Arthur and J.J. Le Pore, J. Vacuum Sci. Technol. 6 (1969) 545. [11] Y. Matsushima and S. Gonda, Japan. J, Appl. Phys. 15 (1976) 2093. [12] A.Y. Cho, H.C. Casey and P.W. Foy, AppI. Phys. Letters 30 (1977) 397. [13] C.T. Foxon, M.R. Boudry and B.A. Joyce, Surface Sci. 44 (1974) 69. [14]A.Y. Cho, J. Appl. Phys. 42 (1971) 2074. [15]A.Y. Cho, J. AppI. Phys. 47 (1976) 2841.

83

[16]CT. Foxon, J.A. Harvey and BA. Joyce, J. Phys. Chem. Solids 34 (1973) 1693. [17] C.T. Foxon, BA. Joyce, R.F.C. Farrow and R.M. Griffiths, J. Phys. D7 (1974) 2422. [18] R.F.C. Farrow, J. Phys. D7 (1974) 2436. [19] B. Goldstein and D. Szostak, AppI. Phys. Letters 26 (1975) 685. [20] G.J. Macur, R.K. Edwards and P.G. Wahlbeck, J. Phys. Chem. 70 (1966) 2956. [21] J.H. Neave and BA. Joyce, J. Crystal Growth 43 (1978) 204. [22]J.H. Neave and BA. Joyce, J. Crystal Growth, to be submitted.