You are on page 1of 12

Dynamical Calculations for RHEED from MBE Growing Surfaces. II.

Growth Interruption and


Surface Recovery
Author(s): L.-M. Peng and M. J. Whelan
Source: Proceedings: Mathematical and Physical Sciences, Vol. 435, No. 1894 (Nov. 8, 1991), pp.
257-267
Published by: The Royal Society
Stable URL: http://www.jstor.org/stable/51964 .
Accessed: 10/05/2013 01:37
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Proceedings:
Mathematical and Physical Sciences.

http://www.jstor.org

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

Dynamical calculations for


MBE

RHEED

from

growing surfaces. II. Growth interruption


and surface recovery

BY L.-M. PENG AND M. J. WHELAN


Department of Materials, University of Oxford, Parks Road, Oxford OX] 3PH, U.K.
Detailed

dynamical calculations for reflection high-energy electron diffraction


from surfaces growing by molecular beam epitaxy have been made to
investigate the technique of growth interruption and surface recovery kinetics. A
birth-death growth model and a systematic reflection approximation to RHEEDhave
been used. It is found that whilst the RHEEDintensity oscillation behaviour is very
sensitive to the incident glancing angle, the shape of the intensity recovery curve is
insensitive to the diffraction condition. It is further shown that the RHEEDintensity
recovery curves bear a resemblance to the corresponding surface-roughness recovery
curves. Sensible quantitative studies of recovery can therefore be made by analysing
the RHEEDintensity recovery curves. A similarity between the surface recovery and
RHEED intensity recovery has been established.
(RHEED)

1. Introduction
Novel devices grown by molecular beam epitaxy (MBE),such as modulation-doped
field effect transistors and quantum well lasers, consist of complex multilayer
structures. Because in electronic devices a two-dimensional carrier gas is localized
at an interface, and in optoelectronic devices a two-dimensional electron-hole gas is
strongly localized in a potential well formed by materials with a smaller band gap
bounded by material with a larger band gap, the device performance is critically
dependent on the electrical and morphological quality of epitaxial layers and
interfaces (see, for example, Herman & Sitter 1989).
The reflection high-energy electron diffraction (RHEED) intensity oscillation
technique (Harris et al. 1981; Wood 1981) has been extensively used to study
interface formation and interrupted growth effects in the preparation of heterojunctions, quantum wells and superlattices (see articles in Larsen & Dobson
1988). In the case of GaAs-AlAs quantum wells, it has been shown that although the
interfaces grown by conventional MBE are in general abrupt, there exist asperities of
one atomic layer of height 2.83 At, leading to the well-width fluctuation 6Lz of + I
atomic layer within one quantum well (Sakaki et al. 1985). On the other hand, by
utilizing the RHEEDintensity oscillation technique and the phase-locked epitaxy
(PLE) technique (Sakamoto et al. 1984) considerable improvement has been made to
the growth of very thin films and superlattice structures.
The PLE technique utilizes the 'recovery' effect (Neave et al. 1983), whereby on
cessation of growth the specular beam was reported to return to its pre-growth value
at a rate dependent on the precise point in the layer sequence at which the growth
1 A = 10-10 m = 10-1 nm.
Proc. R. Soc. Lond. A (1991) 435, 257-267
Printed in GreatBritain

257

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

L.-M. Peng and M. J. Whelan


Figure 2

258
Figure I
t)

00
l 0.8 - /
0.4 -8

-'

b04 0.3

Cr

l--F

* r

ITI

'

. I

. -l

...

monolayers deposited (t/T)


monolayers deposited (t/v)
1.
1
to
for
curves
diffusive
Surface
(from
8)
homoepitaxial growth on a
Figure
layer coverage
GaAs(001) substrate. The growth is interrupted at 6 monolayers deposition, and a diffusion
parameter of k = 15 has been used.
Figure 2. Variations of the surface roughness parameter R with the deposition of Ga and As species
and with time after the molecular beam shutters are closed at 6 monolayers deposition.

is interrupted. This recovery process has been shown to result from the recovery of
surface smoothness, and has profound effects on the optical properties of the grown
multilayer structure (Sakaki et al. 1985; Bimberg et al. 1986).
In this paper we will first examine the technique of growth interruption and
corresponding surface recovery kinetics. A birth-death model (Cohen et al. 1989) will
be used to describe MBE growth as well as surface recovery, and a systematic
reflection approximation will be made to the general matrix formulation of the
dynamical theory of electron diffraction (Peng & Whelan 1990a), to calculate RHEED
from MBEgrowing surfaces. In the following paper the simple birth-death growth
model for epitaxial growth of a single type of atom species will be generalized to
include the deposition and diffusion of multiple species, and thus enable the
heteroepitaxial growth as well as quantum well formation to be described. The
recovery effect will be applied to the growth of heterostructures, and the technique
of PLE will be examined.

2. Growth models
In this paper we will use the diffusive growth model as proposed by Cohen et al.
(1989) to simulate the epitaxial growth on a low-index surface. For simplicity of
presentation, the growth model used will be limited to epitaxial growth of a single
type of atom species. The growth model can easily be extended to include the
deposition and diffusion of multiple atom species and thus to give a more realistic
description of the growth of compound semiconductors or semiconductor alloys. This
generalization is presented in the following paper concerning the growth of
heterostructures and quantum wells.
Following Cohen and co-workers (Cohen et al. 1989) we use a solid-on-solid (sos)
approximation, excluding vacancies and overhangs, and assume that the net
diffusion of an atom species from an upper layer to the layer immediately below is
proportional to a product of the number of mobile (i.e. uncovered) atom species in
the upper layer and the total number of uncovered sites in the second layer below.
The coverage 0n of the nth surface layer then satisfies a set of nonlinear differential
equations
dOn/dt = (n,_,-

-n)/T+

k(O(n+l -On+2)

(0_,-l-

n)-

k(O n-

n+l) (n,_2-,,1)

Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

(1)

MBE RHEED intensity

(a)

259

oscillations.
(b)

l 0.5

0.5

Figure 3. Growthfront profilesfor ten growingsurfacelayers. In (a) the profilecorrespondsto 6


monolayersdeposition,and in (b)the profilecorrespondsto a recoveredsurfaceafter the growth
is interruptedfor a time 6T.
during the growth, and
dOn/dt = k(Onl

-On+2)

(On - On)- k(On-

n+l) (On-2- 0n-l),

(2)

after the growth is interrupted. T is the time to deposit a monolayer of atoms, and
k is a diffusion parameter that measures the mobility of surface atoms. In general,
the higher the substrate temperature the higher the atom mobility or the diffusion
parameter k. Shown in figure 1 are the variations of eight surface-layer coverages for
GaAs homoepitaxial growth on a GaAs(001) substrate and with a diffusion parameter
of k = 15 for Ga atoms (the growth of GaAs single crystal is mainly determined by
the deposition and diffusion of Ga, while the As species is plentiful on the surface).
The growth has been interrupted at t = 6r. Throughout this paper time t is given in
units of r and k is given in units of -1.
At this stage it is convenient to introduce a roughness parameter R to describe the
roughness or smoothness of the growing surface. For a continuous growth process, we
can define the roughness parameter for the surface as
oo

no

R(t)

n + -n) (1 -n(t))
(n0

=
n=l

in which

(n-no) On(t),

(3)

n=nO+l

_ NINT(t/T), t < to,


n
=?

NINT(to/r),

t> t

(4)

where NINT(x) is defined as the nearest integer to the argument x, and the growth
is interrupted at time to.
The variations of the surface roughness R during MBEgrowth and after the growth
interruption, which corresponds to the growth conditions of figure 1, are plotted in
figure 2. In this figure the surface roughness is seen initially to increase very quickly
for less than half a monolayer deposition. After the initial stage the roughness
parameter oscillates with an approximate periodicity equal to T, and the average
roughness increases slowly with increasing deposition. After the growth is interrupted
at 6 monolayers deposition, the roughness decreases towards zero and the growth
front profile approaches a perfect-step function as shown in figure 3. The significance
of this effect will be considered further in the following section.

3. Growth interruption and recovery effect


In this section we will examine the growth interruption technique (Lewis et al.
1985; Madhukar et al. 1985; Sakaki et al. 1985), utilizing the 'recovery' effect first
observed by Neave et al. (1983). It was found that on cessation of growth the specular
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

L.-M. Peng and M. J. Whelan

260
(a)

0.07 (/0.06-

y_

^0.3800.4.38) f

0.36

0.05 0.04 -

t/r

0.41 : (c)

) 0.39 l(-

:0.0

0.06
06

10

t/
4. intensity
RHEEDam
recovery aorresponding
Figure
a time equivalenttothat fordeposcillation

curves, growth to
interruptionat
olayers.RHEDcalculationshave been made for

10 keV primary beam energy and glancing incident angles of (a) 75.6 mrad, and (b) 64.8 mrad. The
recovery curves of (a) and (b) are plotted together in (c), showing that they are in fact identical.
In (c) the full line is the recovery part of (a) and the circles are the recovery part of (b). The incident
intensity is unity.
unity.
beam intensity

beam intensity recovers almost to its pre-growth value in a manner dependent on the
precise point in the intensity oscillation at which the Group III element shutter is
closed. There is always an initial rapid response, followed by a slower stage, and in
general the 'recovery' obeys an expression of the form
I = A -A1 exp (- t/1)

-A2 exp (- t/r2),

(5)

where A1 and A2 are usually positive constants, an ndIand rT are temperaturedependent time constants of the fast and slow stages respectively.
Much effort has been made to investigate the 'recovery' effect and the surface
recovery kinetics. For GaAs(001) 2 x 4 reconstructured surfaces, Neave et al. (1985)
reported an activation energy of about 0.2 eV associated with the rapid initial stage
of 'recovery', and attributed the energy to Ga-As bond dissociation at step edges.
In a more detailed investigation of growth kinetics and mechanisms, Lewis et al.
(1985) suggested that the fast process can be understood as a rapid smoothing of the
growth front profile,a andthe slow process can be regarded as recovery of long-range
order, i.e. rearrangement of terrace geometries and/or reduction of one-dimensional
disorder. More recently, from their Monte Carlo simulation, Clarke et al. (1989) and
Vvedensky & Clarke (1990) proposed that during the fast stage the surface loses any
'dendritic' structure (single-bonded atoms) resulting from the non-equilibrium
growth phase, and the initial rapid stage is followed by a slow equilibration of the
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

MBE RHEED intensity

oscillations.

II

261

surface, as features such as step edges or islands undergo gross changes of shape,
resulting from the adatom clusters evolving to form the maximum number of
nearest-neighbour bonds.
All the previous studies of recovery have suffered from the inability to treat
dynamical effects in RHEED, and therefore a direct comparison with measured
RHEED recovery curves has not proved possible. We have made detailed calculations
of the dynamical RHEED intensity evolution curves during MBE growth and after the
growth is interrupted, using the diffusive growth model previously discussed. The
calculations are based on the general matrix formulation of the dynamical theory of
electron diffraction (Peng & Whelan 1990a), and detailed procedures for its
application to calculating RHEED intensities from MBE growing surfaces have been
published earlier (Peng & Whelan 1990b, 1991). Shown in figure 4a, b are two
intensity oscillation curves for 10 keV fast electrons incident at a GaAs(001) growing
surface, with a glancing incident angle of 75.6 mrad and 64.8 mrad for figures 4a and
4b respectively. A diffusion parameter of k = 20 has been used. In these figures it is
clear that although the incident electrons are reflected from an identical growing
surface, the intensity oscillation behaviour is quite different depending on the
incident conditions, as has been observed experimentally (Zhang et al. 1987). It was
also demonstrated by Joyce et al. (1988) that the time constants are substantially
different for recovery curves obtained for different diffraction conditions, but under
fixed growth conditions and from the same termination point in the oscillation
waveform. This fact throws doubt on all the previous quantitative studies of
recovery, since the characteristic time constants as obtained from fitting the
recovery curves to the form of (5) are not unique.
For our present model of epitaxial growth and diffraction, we found, however, that
whereas the intensity oscillation behaviour is very sensitive to the diffraction
conditions, the recovery curves are much less sensitive to the incident condition.
Plotted in figure 4c are intensity recovery curves, taken from figure 4a, b, showing
clearly that the two intensity recovery curves are practically identical in shape, and
therefore give identical time constants T1 and r2 for the initial fast and subsequent
slow recovery stages respectively. RHEED calculations have also been made for some
other diffraction conditions and primary beam energies; the conclusion that the
time constants as obtained from RHEED intensity recovery curves are not sensitive
to the incident conditions is found to be valid for all cases. An example is shown in
figure 5, for 20 keV primary beam energy, which shows four RHEED intensity
evolution curves during growth and after the growth is interrupted at 6 monolayers
deposition. All the curves have been calculated under the identical growth condition,
i.e. k = 20, but with changing glancing incident angle. It is obvious that both the
intensity oscillation amplitudes and phases are substantially different for different
incidence. Surprisingly, the recovery time constants are found to be identical for all
the four curves, and as an illustration the recovery curves for figure 5a, d are plotted
together in figure 5e, showing that apart from the sense of recovery and a scaling
factor the two curves are in fact identical. The uncertainty in time constants as
observed by Joyce et al. (1988) is caused by the scattering events parallel to the
epilayers, i.e. intralayer scattering. This effect will be minimized however, if a
systematic diffraction condition is chosen as in the present study.
Shown in figure 6a are calculated RHEED intensity oscillations during the growth
of 6, 6.2, 6.4, 6.6, 6.8 and 7 monolayers, and the corresponding intensity recovery
curves up to 10r. Both the growth and diffraction conditions are identical to those
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

L.-M. Peng and M. J. Whelan

262
0.07 ^~~-D

4^

0.04

W
0.046,-

'0.04 -

W1

0R;~~~~~~~~~
(c)
F

0.40

'~~0.08

~~~~(d)

----0.170.16-

0.38-

0.15

0.36-0.1514

t/t
0.075 4D

(e)

mi

uui

ili l

l-ll

0.145

^~

1~0.^~075-

d?

0.150 I

0.07 -

..
;

0.065-,155
6

?0.155 -

10

t/t
Figure 5. RHEED intensity oscillation and recovery curves for 20 keV fast electrons incident at the
MBE growing surface with a glancing incident angle of (a) 46 mrad, (b) 50 mrad, (c) 53 mrad and
(d) 57 mrad. Shown in (e) are recovery curves for (a) full line, and (d) circles; it is seen that apart
from the sense of recovery and a scaling factor, the two recovery curves are identical.

used in figure 4a. The recovery curves for growth interruptions at various points
between 6 and 7 monolayers have been fitted to the form (5), and an example is
shown in figure 6b for growth interruption at 6.2 monolayers. We found that the
results are consistent and reproducible with high accuracy for T1, but the time
constants thus obtained are not consistent and accurate for T2, i.e. the initial fast
stage can always be adequately represented by a single exponential but cannot be
adequately so represented for the subsequent slow stage, and this agrees with the
observations of Neave et al. (1983) and Lewis et al. (1985). A plot of T1 against layer
completion at which the growth is interrupted is shown in figure 7 for two different
surface diffusion parameters k = 5 and k = 20, the latter case corresponding to
figure 6. This plot shows clearly a maximum near-growth interruption at layer
completion and a minimum near-growth interruption at half-layer completion.
The surface roughness curves and the roughness recovery curves corresponding to
figure 6a are shown in figure 8. It is found that the roughness recovery curves can
also be fitted with the form (5), and that the curves bear a resemblance to the RHEED
intensity recovery curves of figure 6a. Shown in figure 9 is an example of the
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

MBE RHEEDintensity

oscillations.

II

263

0.380

0.41
01
a)

*z 0.372

0.38

0.364

0.35
0 -

8.0

6.4

9.6

Figure 6. (a) RHEED intensity oscillations and recovery curves corresponding to growth
interruption at 6.0r, 6.27, 6.47, 6.67, 6.8r and 7.07. (b) The recovery curve resulting from growth
interruption at 6.2 monolayers deposition is fitted with the form (5) with Ao = 0.418, A1 = 0.018,
A2 = 0.037, T1 = 0.24- and T2 = 3.5 x 106. The figure shows that the fitting is almost perfect.
, fitted; o, calculated.

-Figure7

Figure8

1.0
T1

0.5

0
7.0
6.5
interruption time (t/T)
Figure 7. Plot of time constant 71 for the initial fast recovery stage against time at which the
growth is interrupted. T1 in units of 7.
Figure 8. Surface roughness and roughness recovery curves for growth interruption at 6.07, 6.27,
6.47, 6.67 and 7.07. The growth condition is identical to that of figure 6a.
6.0

0.41 -

Figure9

Figure 10
- 0.1

40

/2~ 1.0

0L
0
-

Fl

C)
-4-

0.2

0.39-

bc~

I.-1

zc

Ti

0.5 -

0.3 ;,

02t

0.37 -I ....

_,

-20

............

III....I....I....I...
.I

10

6.()

6.5
interruption time (t/T)

7.0

Figure 9. Plot of RHEED intensity recovery (thick line) and surface-roughness recovery (thin line)
curves, corresponding to growth interruption at 6 monolayers deposition. A diffusive growth model
has been used, with a diffusion parameter of k = 20.
Figure 10. Plot of the time constant 71 for the fast recovery stage of the surface-roughness recovery
curves against time at which the growth is interrupted. 71 in units of 7.
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

L.-M. Peng and M. J. Whelan

264

resemblance. A RHEED intensity and a surface roughness recovery curve are plotted
together, both of which correspond to growth interruption at 6 monolayers
deposition. The time constant r1 for the fast recovery stage, as obtained by fitting the
surface roughness curves with (5), is plotted in figure 10 against the layer
completion at which the growth is interrupted. This plot also bears a resemblance to
that derived from intensity recovery curves (figure 7). A plot of time constant for
both the intensity recovery curves and surface roughness recovery curves against
diffusion parameter k is given in figure 11, showing that for a wide range of k there
exists a similarity between RHEED intensity recovery and surface roughness recovery,
i.e. the RHEED intensity recovery curve is a faithful representation of surface
recovery.
Because, in our calculations, the diffraction conditions have been so chosen that a
systematic reflection case is realized and reflected beam intensity is insensitive to the
lateral distribution of atoms or the shape of surface clusters, the two recovery stages
as manifested in the RHEED intensity recovery curves must have origins differing
from those previously proposed (Lewis et al. 1985; Vvedensky et al. 1990). In our
present model the reflected beam intensity is sensitive only to the surface layer
coverages, and we will therefore propose an explanation depending on the variation
of the layer coverages after interruption of growth.
Roughly speaking, the appearance of two recovery stages simply reflects the fact
that in our growth model the diffusion of an atom species from an upper layer to a
largely unfilled layer is much easier than to a nearly completed layer. To illustrate
this point we will consider an ideal two-level model. We assume that the number
of the mobile atoms on the upper level is 01, that the number of uncovered sites
in the lower level is 1-00, and that the growth is interrupted at layer completion
f0(0 < f < 1). According to the growth-rate equations (2), and noticing that 06 = 0
for all n > 1 and 08 = I for all n < O, we have
dO/dt

= k01(1-00),

d0j/dt = - k0(1-

).

The solutions are given by


01(t)=

(1 -fo) 01(0) exp [-(1 -f)


(1 -fo)+ 01(0){l1-exp[-(1--f)

kt]
kt]}

and

0o=fo-01.
Initially, when t < 1l/(1-f0) k, the above solution for 01 reduces approximately to
01 = 01(0) exp {-[(I -fo) + 01()] kt},
(6)
namely the initial recovery stage is dominated by an exponential term, and the time
constant for this initial stage is given by
(7)
1/[(1-fo) +01(0)] k.
For large t and a finite value of the diffusion parameter k such that t > 1/(1 -f0) k,
have
(8)
we have
= (_ff
-f) kt].
()p
(t)we
(1-f0) + 01(0)
Ti =

This solution clearly shows that again the process is governed by an exponential
term, but with a larger time constant
2 = 1/(1--fo)

k.

(9)

It should be mentioned that while the initial fast recovery process can always be
Proc. R. Soc. Lond. A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

oscillations.

MBE RHEED intensity

100

0.40 -Figurel
0.3

7.5

0.3

1I

265

Figure12
-

5.0 0.2-

2.5 -/
0.1

Il

ll

ll

1l

(l1

20

1.

1l1

60
100
20
60
100
k
k
Figure 11. Plot of the time constantT1for the fast recoverystage of both RHEED
intensityrecovery
and surface-roughness
recoveryagainst diffusionparameterk. The growthwas interruptedat 3T.

in units of r. o, intensity; o, roughness.


Figure 12. Plot of T11against diffusion parameter k for RHEED intensity recovery and growth
interruption at 3r.

T1

described by using equation (6) and a single value of the time constant T1, the validity
of (8) to describe the later recovery process depends on the exact time of growth
interruption. In particular, for growth interruption near f0 = 1 (for deposition of a
nearly complete monolayer), (1-f) kt < 1 for moderate times, and the general
expression given above for 01(t) has a limiting form
01(t) = 01(0) [1 + 01(0) kt]-1,

which clearly cannot be fitted to the expression (5) at large times. It is therefore not
surprising that we found difficulties in fitting the whole recovery curves to the form
(5) for growth interruption at layer completion (curves corresponding to growth
interruption at times equivalent to 6.0 and 7.0 monolayers deposition in figure 6a),
and this agrees with the observations of Lewis et al. (1985).
The variation of the time constant rI against layer completion can also be
understood, by using the relation T1 = 1/[(1-fo)+ 0(0)]k. If the variation of 0 (0)
with f0 is small, for example for the case of fast surface diffusion, T1 will then have
a maximum at layer completion I-f0 = 0, and a minimum at half-layer completion
1-f0 = 0.5 (see figures 7 and 10 for the curve with k = 20). On the other hand, if the
variation of 01(0) with f0 is not negligible in comparison with the term 1-f0, the
variation of T1 with layer completion may then be different. The situation could be
further complicated by the fact that the ideal two-level model used here is not always
valid. A somewhat irregular curve could be expected for growth at lower temperature
(see the curves associated with k = 5 in figures 7 and 10)
It is a corollary of the ideal two-level argument that in the high-temperature or
high-surface diffusion regime, where typically only two growing layers are involved
and the two-level model applies, from equation (7) a linear relation is expected
between 1/T1 and k. Shown in figure 12 is a plot of 1/r1 against the diffusion
parameter k. It is seen that the plot approaches a straight line as the surface diffusion
parameter k becomes large. If we go a step further and assume an Arrhenius form for
the diffusion parameter k:
/BT
k = koexp (- U/kB T),
(10)
where k0 is a constant and kB is Boltzman's constant, the activation energy U can
then be related to the time constant Tr through the relation
rT = rOexp (U/kB T),

(11)

Proc. R. Soc. Lond. A (1991)


Vol. 435.

o10

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

L.-M. Peng and M. J. Whelan

266
with

(12)
To= 1/[(1-fo)+0o(0)]ko,
and the activation energy U can be easily obtained from an Arrhenius analysis of the
fast recovery stage, i.e. by plotting In rT against l/kB T within the high-temperature
regime.

4. Conclusion
Detailed dynamical calculations for RHEEDfrom MBEgrowing surfaces have been
made to investigate the technique of growth interruption and surface recovery
kinetics. We found that, while both the RHEED intensity oscillation amplitude and
phase are very sensitive to the incident condition, the shape of the intensity recovery
curves and thus the characteristic time constants are insensitive to the diffraction
condition. This enables sensible quantitative studies of surface recovery to be made.
It has been further demonstrated that RHEEDintensity recovery bears a resemblance
to the corresponding surface roughness recovery, and a similarity between RHEED
intensity recovery and surface recovery has been established.
In our present model the diffraction condition has been so chosen that the specular
reflected beam intensity is sensitive only to the surface layer coverage, rather than
to the detailed lateral distribution of atom species on the surface. All the previously
proposed recovery mechanisms for interpreting the two characteristic features of
measured RHEED recovery curves are inapplicable to the results as presented in this
paper. For the present growth and diffraction model we have shown that the
appearance of two recovery stages in RHEEDrecovery curves simply reflects the fact
that the diffusion of atom species from an upper layer to an almost unfilled layer
below is much easier than to a nearly completed layer.
The authors thank Professor Sir Peter Hirsch for the provision of laboratory facilities. One of us
(L.-M.P.) is supported by the U.K. Science and Engineering Research Council (SERC), through a
Research Assistantship, and by Wolfson College, Oxford, through a Junior Research Fellowship,
both of which are gratefully acknowledged.

References
Bimberg, D., Mars, D., Miller, J. N., Bauer, R. & Oertel, D. 1986 J. Vacuum Sci. Technol. B 4,
1014.
Clarke, S., Vvedensky, D. D. & Ricketts, M. W. 1989 J. Cryst. Growth95, 28.
Cohen, P. I., Petrich, G. S., Pukite, P. R., Whaley, G. J. & Arrott, A. S. 1989 Surface Sci. 216,
222.
Harris, J. J., Joyce, B. A. & Dobson, P. J. 1981 Surface Sci. 103, L90
Herman, M. A. & Sitter, H. 1989 Molecular beam epitaxy-fundamentals and current status.
London: Springer-Verlag.
Joyce, B. A., Zhang, J., Neave, J. H. & Dobson, P. J. 1988 Appl. Phys. A 45, 255.
Larsen, P. K. & Dobson, P. J. Ed. 1988 RHEEDand reflectionelectronimaging of surfaces. New York:
Plenum Press.
Lewis, B. F., Grunthaner, F. J., Madhukar, A., Lee, T. C. & Fernandez, R. 1985 J. Vacuum Sci.
Technol. B 3, 1317.
Madhukar, A., Lee, T. C., Yen, P., Kim, J. K., Ghaisas, S. V. & Newman, P. G. 1985 Appl. Phys.
Lett. 46, 1148.
Neave, J. H., Joyce, B. A., Dobson, P. J. & Norton, N. 1983 Appl. Phys. A 31, 1.
Neave, J. H., Dobson, P. J., Joyce, B. A. & Zhang, J. 1985 Appl. Phys. Lett. 47 (2), 100.
Peng, L.-M. & Whelan, M. J. 1990a Proc. R. Soc. Lond. A 431, 111, 125.
Peng, L.-M. & Whelan, M. J. 1990b Surface Sci. Lett. 238, 446.
Proc.R. Soc. Lond.A (1991)

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

MBE RHEED

intensity oscillations. II

267

Peng, L.-M. & Whelan, M. J. 1991 Proc. R. Soc. Lond. A 432, 195. (Part I.)
Sakaki, H., Tanaka, M. & Yoshino, J. 1985 Jpn J. appl. Phys. 24, L 417.
Sakamoto, T., Funabashi, H., Ohta, K., Nakagona, T., Kawai, J. & Kojima, T. 1984 Jpn J. appl.
Phys. 23, L 657.
Vvedensky, D. & Clarke, S. 1990 Surface Sci. 225, 373.
Wood, C. E. C. 1981 Surface Sci. 108, L 441.
Zhang, J., Neave, J. H., Dobson, P. J. & Joyce, B. A. 1987 Appl. Phys. A 42, 317.
Received 18 December 1990; revised 20 March 1991; accepted 17 April 1991

Proc. R. Soc. Lond. A (1991)


10-2

This content downloaded from 140.120.31.217 on Fri, 10 May 2013 01:37:32 AM


All use subject to JSTOR Terms and Conditions

You might also like