Journal o f Crystal Growth 27 (1974) I 18-125 9 North-llolland Publishsin# Co.

DEFECTS IN EPITAXIAL MULTILAYERS*

I. MISFIT DISLOCATIONS

J. W. MATTHEWS and A. E. BLAKESLEE

I B M Tlwmas J. lVatson Research Center, Yorktown Heiohts, New York 10598, U.S.A.
Received 6 March 1974; revised manuscript received I0 July 1974

Multilayers composed of many thin films of GaAs and GaAso.sPo.5 were grown epitaxially on GaAs surfaces
inclined at a few degrees to (001). Examination of the multilayers by transmission and scanning electron
microscopy has revealed that the interfaces between layers were made up of large coherent areas separated
by long straight misfit dislocations. The Burgers vectors of the dislocations were inclined at 45 ~ to (001) and
were of type ]a (110~. Dislocations in adjacent interfaces were usually not independent of one another. They
often lay on the same slip plane and when this was so they were clearly products ofthe same source. The layer
thickness at which misfit dislocations were formed was in satisfactory agreement with the predicted thickness.
However, the fraction of the total misfit accommodated by dislocations (once the critical thickness for disloca-
tion generation was passed) was much smaller than predicted. This large discrepancy seems to arise from
difficulties associated with the creation of misfit dislocations. Although there are many processes which can
impede dislocation generation, the most important one in GaAs]GaAso.sPo.5 multilayers appears to be the
impaction of dislocations on one glide plane against dislocations in another.

1. Introduction within a few percent. They were prepared for the 'semi-
conducting superlattice' device proposed by Esaki and
The a c c o m m o d a t i o n o f misfit across the interface TsuT).
between an epitaxial film and its substrate has been con-
sidered by F r a n k and van der Merwel'Z). They show 2. Experimental details
that a misfit smaller than about 7 percent will be accom- Alternating layers o f G a A s and GaAs0.sPo.5 were
m o d a t e d by uniform elastic strain until a critical film grown on G a A s substrates using the G a - A s H 3 - P H 3-
thickness is reached. Thereafter, it is energetically H C I - H 2 v a p o r system, in this system GaC! is formed by
favorable for misfit to be shared between dislocations reaction o f G a with HCI and transported by the H2
and strain. Experimental tests o f these predictions have carrier gas, with either AsH3 or A s H 3 + P H 3, into the
been made on m a n y bicrystals. In the majority o f these deposition zone. Reactive deposition produces an
tests the agreement between predictions and experiment epitaxial layer o f G a A s or Ga(As,P) on the substrate
has been satisfactory. However, in some the fraction o f surface. The details o f the apparatus used to g r o w the
misfit a c c o m m o d a t e d by elastic strain has been larger alternating layers have been described elsewhereS). A n
than predicted3-5). The discrepancies between pre- important feature o f it is the ability to inject PH3 into
diction and experiment in these systems are believed to the A s H 3 v a p o r stream so that there is little mixing o f
result from difficulties associated with the generation o f the AsH3 and A s H 3 + P H 3 pulses as they move to the
misfit dislocations 6). deposition zone. However, some mixing is inevitable,
The aim o f this paper is to describe a study o f misfit and as a result o f it, the interfaces between layers are
a c c o m o d a t i o n in epitaxial multilayers. The multilayers not perfectly sharp.
were made by depositing a succession o f gallium The duration o f the injection o f P H 3 was controlled
arsenide and gallium arsenide-phosphide films on a by a solenoid valve activated by an electronic timer.
gallium arsenide substrate. All layers were single The relative thicknesses o f the G a A s and Ga(As,P)
crystals in the orientation o f their substrate. They were layers could be adjusted by varying the ratio o f the off
accurately planar and their thickness was uniform to to on times. However, the on and off times were equal
* A summary of this work was presented at the Conference on in all specimens described here. This ensured that the
Vapor Growth and Epitaxy, Jerusalem, May 1973. thicknesses o f the G a A s and Ga(As,P) layers were

118
 DEFECTS IN EPITAXIAL MLILTILAYERS 119

approximately equal. The multilayers contained either and 700 A layers. However, they were not present in
60 or 120 layers and the layer thickness ranged from specimens composed of thinner layers. This means that
75 to 700 A. Layer thickness was determined from scan- the critical thickness for the generation of misfit dis-
ningelectron micrographs ofthemultilayersviewed from locations lay somewhere between 160 and 350 A.
the side or from the positions of satellite peaks in X-ray Four misfit dislocations between 700 A layers are
diffraction patternsg). The resolution limit of the scan- seen in fig. 1. A feature of these and many of the other
ning microscope was'smaller than 150 A. This resolution dislocations present to accommodate misfit between
enables.one to determine the wavelength, or repeat individual layers is that they are paired. The Bragg
distance, in a multilayer made of GaAs and Ga(As,P) reflections responsible for image contrast in fig. 1.
layers that are each about 75 A in thickness. All layers were 220 in (a), 040 in (b), ~20 in (c) and 400 in (d). Two
with the exception of the thickest were grown in 1 sec. pairs of dislocations are visible in (a) and (c), the upper
The injection time used for the thickest (700 A) layers pair is visible in (b), and the lower one in (d). The in-
was 5 sec. visibility of the lower pair of dislocations in (b) shows
The GaAs substrate surfaces were chemically polish- that the Burgers vectors of this pair lay in (010). If we
ed and were inclined at between 2 ~ and 3 ~ to (001). The assume 1o) that stable, complete dislocations in GaAs
rotation away from (001) was about a (110) axis in or Ga(As,P) have Burgers vectors of type 89a (110)
(001). The density of dislocations in the substrates was then the Burgers vectors of the lower pair were either
always less than 5 x 104/cm 2 and usually less than _ 89a [101] or 4- 89a [10T]. The Burgers vectors of the
1 x 104/cm 2. The wafer temperature during the deposi- upper pair were either ___ 89a [011] or 4. 89a [011]. All
tion of the multilayers was 750 ~ An epitaxial GaAs these possible Burgers vectors are inclined at 45 ~ to
layer 10 to 20 lam in thickness was grown on the sub- (001) and at 60 ~ to the dislocation lines. The misfit
strate surface before the deposition of the alternating accommodated by dislocations with this geometry is
layers began. This layer was doped with between 1017 only half that accommodated by edge dislocations with
and 101 s sulphur atoms per cm 3. The multilayers were Burgers vectors in (001). The dislocations in fig. I are
sometimes doped similarly. thus inefficient misfit dislocations. However, they are
Samples were prepared for transmission electron the most efficient complete misfit dislocations that can
microscopy in the following way. Wafers were lapped be made by glide to the interface on {111} slip planes.
on the substrate side to a total thickness of ,-~ 250 lain The dislocations are straight because the lines of inter-
and cleaved into small squares so as to fit into the section of {111} slip planes and the interface are
sample holder of the microscope. The multilayer side straight.
of each square was then attached to a thin glass cover Dislocations similar to those in fig. I have been found
slide with a methanol-insoluble grease. The slide was in silicon doped by diffusion11), in GaAs-Ga(As,P)
mounted vertically and a fine jet of a solution contain- samples12), and in deposits of one fcc metal on an-
ing 15 drops of Br 2 in 100 ml of CH3OH was directed otherla'14). The fact that the dislocations in fig. 1
against the center of the sample. This removed GaAs at go out of contrast in pairs rather than singly is con-
approximately 12 lam/min and produced a polished sistent with the hypothesis (proved below) that paired
surface. Etching was stopped as soon as a small hole dislocations have antiparallel Burgers vectors.
appeared in the sample. Specimens were removed from The separation, S, of paired dislocations was found,
the slide by dissolving the grease in trichlorethylene. in the majority of cases, to obey the following rela-
tion:
3. Observations
S = h cot 55 ~ (1)
3.1. THE GEOMETRYOF MISFIT DISLOCATIONS where h is the thickness of individual GaAs or Ga(As,
Specimens composed of layers 75, 160, 350, 380, 440 P) layers, and 55 ~ is the angle between { 111 } slip planes
and 700 A in thickness were examined. Dislocations and the (almost) (001) interface between layers. Paired
that accommodated part of the misfit between layers misfit dislocations that did not obey eq. (1) had spacings
were found in specimens composed of 350, 380, 440, that were either three or five times the spacing given by
120 J. W. MATTtlE~VS AND A. E. BLAKESLEE

Fig. 1. Four micrographs of a specimen composed of 700 A layers. The layers were almost perpendicular to the incidentelec-
tron beam. The reflections responsible for the image contrast were ]~0 in (a), 040 in (b), 220 in (c) and 400 in (d).

eq. (l). These results suggest that paired dislocations specimen made up of 700 A layers are seen in fig. 3
lay on the same {l I l} plane and in interfaces that were (a, b and c). Some of the properties of the dislocation
either one, three or five layers apart. arrays are listed below.
Paired dislocations separated by three 350 A layers (i) The dislocation lines were parallel to the lines of
are seen in the micrographs in fig. 2. An important intersection of {l 1l} slip planes and the interface. As
feature revealed by this figure is that paired dislocations the interface plane was close to (001) this means that
are joined to one another. They are portions of the the dislocation lines were approximately parallel to the
same dislocation line. This means that paired disloca- [110] and []'10] directions in the (001) plane.
tions have antiparallel Burgers vectors. (ii) The Burgers vectors of the dislocations were of type
Although many of the misfit dislocations in samples 89a ( I f 0 ) and were inclined at 45 ~ to (001). This
composed of layers 350 A or more in thickness were result follows from fig. 3 (b) if one assumes that the
paired, there were many others arranged in parallel and Burgers vectors of stable, complete dislocations in fcc
uniformly spaced arrays. Portions of two arrays in the crystals are of type 89a ( l I0).
 D E F E C T S 1N E P I T A X I A L M U L T I L A Y E R S 121

" !

Fig. 2. Successive micrographs which show the removal of paired misfit dislocations. The dislocations were separated by three
350 A layers and they lay on the same (I I I) glide plane. The layers were approximately perpendicular to the incident electron beam.

(iii) The average separation of adjacent parallel dis- the Burgers vectors of the parallel portions of arrays
locations obeyed S = h cot 55 ~ alternate in sign as shown in fig. 5.
(iv) Members of an array lay in the same or in nearby (vii) The area occupied by individual arrays was often
{I11} planes. This result is suggested by (iii) and con- hundreds of square microns. The arrays in fig. 3
firmed by fig. 3 (c) which was recorded after the sample extended beyond the borders of the figure and,
had been tilted to bring a set o f { l 11} planes parallel to although we do not have direct evidence for this, there
the incident beam. It can be seen that this tilt brought is little doubt that they extended to interfaces above and
the members of an array vertically above one another. below those present in the thinned sample. Evidence
(v) Most arrays terminated against other arrays. An that arrays often involved almost all the interfaces
example of this is present in fig. 3. present in a multilayer is provided by micrographs like
(vi) The end of the visible array in fig. 3 (b) shows that the one in fig. 6. This figure is a scanning electron image
the dislocations in the array were joined to one another of an etched {110} cleavage surface perpendicular to
in pairs. Examination of both ends of arrays has re- the multilayer plane. The layer thickness was 440 A.
vealed that they consist of single dislocations that bend The horizontal dark and light lines are images o f
back and forth as illustrated in fig. 4. Tfiis means that individual layers. The rows of pits that pass obliquely

Fig. 3. Two arrays of misfit dislocations in the specimen composed of 700 A layers. The layers were approximately perpendicular
to the incident electron beam in (a) and (b). The specimen was tilted in (c) so as to view (I I 1) planes edge on.
122 J. W. MATTIIE~,VS A N D A. E. B L A K E S L E E

k,,, C
B)
( c
B)
T
Fig. 4. A n array o f misfit dislocations.

"z
," C
,y",x. B

"2",,y-" B Fig. 7. Dislocation loops left after paired dislocations h a d an-

9' C nihilated o n e a n o t h e r at several points along their length. Layer
"& B thickness = 350 A.

Fig. 5. A section t h r o u g h an array o f misfit dislocations.

the dislocations lie on the same {111} plane this motion
is by glide. If they lie on adjacent or nearby {111} planes
their motion involves climb as well as glide. A train of
dislocation loops left after cancellation had occurred at
various points along the length of a pair of dislocations

: :::::/7/_z ::::::::
,: : : 9 .'..'.'" ?'..::" : " . "A
(A) with antiparallel Burgers vectors are present in fig.
7. Two of the loops are arrowed. The thickness of
V.-iL: ; . : : . . " ~._':~i: :.-' : . " ~ individual layers in fig. 7 was 350 A.
The elimination of misfit dislocations from the
..: ....... ,. "_-.. ::-.::t..-:_--i specimen composed of 350 A layers suggests that
diffusion reduced the misfit between adjacent layers
Fig. 6. S c a n n i n g e l e c t r o n m i c r o g r a p h o f a multilayer seen f r o m
the side. T h e thickness o f the layers was 440 A. Pits m a r k the
sufficiently for misfit dislocations to be unstable. (By
emergence points o f dislocations in arrays o n (111). this we mean that diffusion raised he from below 350 A
to above.) This is discussed further in 4.2 and 4.3.
across the figure mark the emergence points of disloca-
3.3. MISFIT ACCOMMODATED BY DISLOCATIONS
tions arranged in arrays on two parallel {111} planes.
In addition to showing the extent of arrays this figure The misfit accommodated by dislocations with the
reveals that dislocations are sometimes absent from a geometry described in 3.1 is
few of the interfaces present.
6Be = b/2d, (2)
3.2. ELIMINATION OF MISFIT DISLOCATIONS
where b is the strength of the dislocations, and d the
Creation of misfit dislocations has not been observed, average distance between dislocation lines in the same
but the reverse process has. Figs. 2 (a) and 2 (b) are interface. Measurements o f d i n the specimen composed
successive electron micrographs of the sample com- of 700 A layers showed that 6Be ~ 10 -4. This is less
posed of 350 A layers. The parallel lines near the center than one percent of the misfit between GaAs and
of each micrograph are misfit dislocations that have GaAso.sPo.s; it is compared with the predicted value
antiparallel Burgers vectors and Iines that are separated for 6Be in 4.4.
by three layers. The micrographs reveal the removal of
short lengths of these dislocations by the motion of the 4. Discussion
threading dislocation that connects them.
4.1. DISLOCATION FORMATION
If paired dislocations are separated by a single layer
they can be destroyed by a process which.is slightly 4.1.1. Dislocation arrays
different from that in fig. 2. Dislocations in adjacent One mechanism for formation of arrays is clear from
interfaces can move towards one another and cancel. If fig. 4. A substrate dislocation with suitable Burgers
 D E F E C T S IN E P I T A X I A L M U L T I L A Y E R S 123

\ B J .\ 8 J
A C
B
(o)
A

p O
C ) k P Q
o (" ) ) (; k
c ) ( ( B )
a ( )
A
(bl
Fig: 8. Mechanism for creation of dislocation arrays.

vector is replicated in the deposit. This dislocation C

glides back and forth under the influence of coherency C 8 >
strain and generates interfacial dislocations as it does so. c C
B
Although this mechanism gives arrays with the
A
geometry of those observed it is unable to explain the
Fig. 9. Mechanism for formation of paired dislocations.
number of arrays present. The dislocations that
terminate on the substrate surface can account for less has been discussed elsewhere t 6). It will not begin until
than one percent of the observed arrays. From this we the thickness of the first layer is large enough for it to
may conclude that many threading dislocations are contain a half-loop stable under the misfit strain.
created during layer growth. Comparison of the number The number of threading dislocations made as in fig.
of dislocations that terminate on the substrate surface 8 will only be significant if the misfit strain is large
with the number that terminate on the surface of the enough for dislocation nucleation to be probable at the
multilayer confirms this conclusion. The density of temperature of film growth. The calculations o f
dislocations in the substrate surface was < 5 x 104/cm 2. Frank 17) and Hirth 18) suggest that the 1.8yo misfit
The density in the multilayer surface was N [0*/cm 2. between GaAs and GaAso.sPo.5 is sufficient for this
A mechanism for the creation of pairs of threading purpose.
dislocations, and for conversion of these threading
dislocations into dislocation arrays, is illustrated in 4.1.2. Paired dislocations
fig. 8. The misfit strain in the first B layer results in the Although paired dislocations are made when one of
nucleation of a dislocation half-loop on a {111} plane. the loops in an array (see fig. 4) travels further than the
This loop grows by glide to make a pair of threading remainder, this may not be the only mechanism for
dislocations and a length of misfit dislocation line in their generation. An alternative process is shown in fig.
the AB interface [fig. 8(a)]. The growth of additional 9. A half-loop similar to that in fig. 8 (a) is created in a
B and C layers is accompanied by back and forth layer other than the first. During the growth of the next
motion of the threading dislocations to make misfit layer portions of the loop labelled P and Q move to-
dislocations in BC and CB interfaces [fig. 8(b)]. The wards one another and annihilate. Annihilation is
process is termihated if dislocations like P and Q in probable because, as has been shown elsewhere I s), it is
fig. 8(b) meet during the growth of the layer in which energetically favorable for the dislocation content o f
they lie and annihilate one another. This is an im- the final interface to exceed that of the penultimate or
probable event, however. This is because is) it is semifinal one.
energetically favorable for 6AB to exceed the dislocation
4.2. EFFECTS OF DIFFUSION
content of the final BC interface when the stress-free
lattice parameters of the substrate (A) and C layers are A realistic diffusion coefficient for GaAs-Ga(As,P)
equal. at 750 ~ is believed to be 10 -15 cm2/sec. If we assume
The nucleation of half-loops like the one in fig. 8(a) this value then for all specimens except the one com-
 124 J. x,V. M A T T t l E W S A N D A. E. B L A K E S L E E

posed o f 700 A layers the diffusion distance [2 \ ! (Dt)]

was -,- 6 A at the end of the growth of a single layer,
0
was < 70 A after the growth of a multilayer, and was C
,-~ 130/'x by the time specimens were examined.
YB
4.3. CRITICAL-THICKNESS FOR THE FORMATION OF
MISFIT DISLOCATIONS Fig. 10. Threading dislocations in coherent (a), critical (b), and
incoherent (c) multilayers.
Interface structure in multilayers can be predicted
from the forces on dislocation linest9). Two of the It is perhaps worth emphasizing that the calculations
important forces are F,, the force exerted by the misfit made above hold for all
threading dislocations. It does
strain, and F~, the tension in the dislocation line. If the not matter whether they are made by replication of dis-
elastic constants of B and C are equal and isotropic, locations present in the substrate or result from nuclea-
and ha = h e then tion processes like the one illustrated in fig. 8.
The values o f h c predicted by eq. (5) is four times as
F~= 2G(l + v)
(l - v )
bilecos 2. (3) great as the critical thickness expected in a system
composed of a single epitaxial film on a substrate of
G is the shear modulus of B and C, v is the Poisson ratio infinite thickness. Half of this factor of four arises from
and 2 is the angle between the slip direction and that the fact that the elastic misfit strain in a multilayer is
direction in the film plane which is perpendicular to the shared between all layers. Half the layers are compress-
line o f intersection of the slip plane and the interface. ed and the other half are stretched. The remainder of
The tension in the dislocation line is approximately the factor of four arises because the motion of a disloca-
tion like that in fig. l0 makes two misfit dislocations.
h)
Ft -- 4~(1-v)Gb2(l - v cos 2 r (ln~ + 1 , (4) Migration of a threading dislocation in a specimen
composed of a thin film on a thick substrate makes only
where c~ is the angle between the dislocation line and one.
its Burgers vector.
The maximum value of the strain is era,, = 89f. If 4.4. RELAXATION OF MISFIT STRAIN

F, . . . . the value of F, at e. . . . is less than" 2F t then In ideal circumstances, the motion of a dislocation like
threading dislocations will have geometry similar to (c) in fig. l0 reduces e so as to keep F~ = 2F v The misfit
(a) in fig. l0 and the interfaces between layers will be accommodated by dislocations when balance is main-
coherent. If F,m,. = 2 Fi, threading dislocations will tained is
have the geometry shown by (b). If F,r,," > 2 Ft,
b(l-vcos2~) ( h )
dislocations will move and assume the geometry of (c).
This motionreduces e and destroys the coherence of
6ac=f-nh(l+v) cos). ln~+l . (6)

the interfaces between layers. The layer thickness at The value o f 6Bc predicted by this equation for h =
which F~.... = 2 Fl is 700 A is --, l 0 - 2 . This is about one hundred times the
observed value (~ee 3.3). Discrepancies of this magnitu-
tic -~ 2r~b(l-vc~176
) ( l + v ) c oInsThc~ + i . (5) de have been found in other systems and are believed to
result from processes that impede relaxation of misfit
hc for GaAs-GaAso.sPo.s (where v = ~, b = 4 A , strain6). Examples of these processes are the Peierls-
cosct = 89 cos 2 = 89 a n d f = 0.018)is ,-- 250 A. Thus, Nabarro stress, the barrier to the nucleation of disloca-
the presence of misfit dislocations in interfaces between tions, and the interaction between dislocations.
350, 380, 440 and 700 A layers (see 3.1) is expected. The effect of the Peierls-Nabarro stress on the relax-
Escape of misfit dislocations during examination of ation of misfit strain in semiconductors has been dis-
350 A layers (see 3.2) is also not surprising. Diffusion cussed elsewhere t 6). It is important when the substrate
after layer growth (see 4.2) would be expected to in- temperature and density of threading dislocations are
crease h c from ,-- 250 A to roughly 350 A. low. The substrate temperature (750 ~ and the density
 DEFECTS IN E P I T A X I A L M U L T I L A Y E R S 125

of threading dislocations (108/cm z) in the GaAs-Ga order to relieve elastic stresses present as a result of the
(As,P) multilayers suggest that the influence of the misfit between the multilayer taken as a whole and its
Peierls-Nabarro stress in these samples was small. substrate. Part II[ will describe how multilayers free
Indeed, if it were the only impediment to dislocation of misfit dislocations, threading dislocations, pile-ups,
formation, 6nc would be expected to approach its slip lines and cracks can be prepared.
optimum value [eq. (6)] in the time taken to grow a
single GaAs or Ga(As,P) layer t 6). Acknmvledgements
The observation that most arrays of misfit disloca- We would like to thank B. K. Bischoff for his help
tions were terminated by impaction against arrays on with the growth ofmultilayers and with the preparation
intersecting planes (see fig. 3) suggest that interaction of samples for microscopy. Scanning electron micro-
between existing misfit dislocations, coupled with the graphs were taken by C. G. Bremer, and the measure-
barrier to the nucleation of new ones ~7.18), was largely merit of layer spacing by the X-ray satellite technique
responsible for the low value of 6ac. w a s performed by J. Angilello.

5. Final remarks References

I) F. C. Frank and J. H. van der Merwe, Proc. Roy. Soc.
The multilayers described in this paper were pre- (London) A 198 (1949) 216.
pared by CVD. It would be imprudent to suggest that 2) J. H. van der Merwe, J. Appl. Phys. 34 (1963) 117;
all features of CVD multilayers would be present in J. H. van der Merwe, in: Single Crystal Fihns (Pergamon,
New York, 1964) p. 139.
specimens made by LPE or other techniques. However, 3) B. Borie, C. J. Sparks and J. V. Cathcart, Acta Met. 10 (1962)
there are some properties of CVD multilayers that one 691.
might expect to be quite general. The layer thickness at 4) U. Gradmann, Ann. Physik 17 (1966)91.
5) J. W. Matthews and E. Klokholm, Mater. Res. Bull. 7 (1972)
which the formation of misfit dislocations with the 213.
geometry described above is expected to begin is in- 6) N. Cabrera, Mem. Sci. Rev. Met. 62 (1965) 205.
dependent of preparation technique. Nucleation of 7) L. Esaki and R. Tsu, IBM J. Res. Develop. 14 (1970) 61.
8) A. E. Blakeslee, J. Electrochem. Soc. 118 (1971) 1459.
dislocation half-loops is expected in all multilayers 9) A. Segm~iller and A. E. Blakeslee, J. Appl. Cryst. 6 (1973) 19.
where growth temperature and misfit strain are suitably 10) F. C. Frank and J. F. Nicholas, Phil. Mag. 44 (1953) 1213.
large. However, the misfit strain required for disloca- i 1) J. Washburn, G. Thomas and H. J. Queisser, J. Appl. Phys.
35 (1964) 1090.
tion nucleation is influenced by surface steps, inclu- 12) M. S. Abrahams, J. R. Weisberg, C. J. Buiocchi and J. Blanc,
sions, and other defects that cause localized high J. Mater. Sci. 4 (1969) 223.
stresses. The concentration and effectiveness of these 13) J. W. Matthews, Phil. Mag. 13 (1966) 1207.
14) J. W. Matthews and W. A. Jesser, Acta Met. 15 (1967) 595.
stress raisers may depend on preparation technique. 15) J. W. Manhews and A. E. Blakeslee, IBM Research Report
Processes that hinder the elongation of misfit disloca- RC 3854, May, 1972.
tions are found in all systems. However, the magnitude 16) J. W. Matthews, S. Mader and T. B. Light, J. Appl. Phys. 41
(1970) 3800.
of the effects they produce ate known to vary from one 17) F. C. Frank, in: Proc. Syrup. on the Plastic Deformation o f
epitaxial system to another 16). Crystalline Solids (Carnegie Inst. of Technology, Pittsburgh,
As its title implies, this paper is the first of a series of 1950) p. 89.
18) J. P. Hirth, Relation between Structure and Strength in ~Ietals
articles on defects in multilayers. Part I[ will be con- and Alloys (H. M. Stationery Office, London, 1963) p. 218.
cerned with dislocation pile-ups and cracks formed in 19) W. A. Jesser and J. W. Matthews, Phil. Mag. 15 (1967) 1097.