Distribution of Solute in Crystals Grown from the Melt. Part II.

Experimental
J. A. Burton, E. D. Kolb, W. P. Slichter, and J. D. Struthers

Citation: J. Chem. Phys. 21, 1991 (1953); doi: 10.1063/1.1698729

View online: http://dx.doi.org/10.1063/1.1698729
View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v21/i11
Published by the AIP Publishing LLC.

Additional information on J. Chem. Phys.

Journal Homepage: http://jcp.aip.org/
Journal Information: http://jcp.aip.org/about/about_the_journal
Top downloads: http://jcp.aip.org/features/most_downloaded
Information for Authors: http://jcp.aip.org/authors

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 SOL UTE DIS T RIB UTI 0 N INC R Y S TAL S G ROW N FRO M THE MEL T. I 1991

Then the concentration of solute in the melt next to the ~.~ being the ith positive root of
interface is found to be (~5) cot(~5)= (!-ko)~.
More generally, let C, be the initial concentration at the
Co(t) 1 interface just before a step-function change is made in
C L at 1=0, and let C, be the final concentration at the
interface after a new steady state is reached. Then

~ ~,~ sin~,5· exp { - [ (

xL: - - - - - - - - - - -
: r + (;L,~)2 ]~: }
(31)
from Eq. (31) it follows that the concentration at the
interface, Co, varies with time as shown in Fig. 3, for
several values of ko and ~.
ACKNOWLEDGMENT
It is a pleasure to express appreciation to Dr. H.
sin2~,5)[( -~)2+ (;L,5) 2]
>==1
1- --( Reiss and Dr. J. A. Lewis for several valuable dis-
2~,~ . 2 cussions.

THE JOURNAL OF CHEMICAL PHYSICS VOLUME 21. NUMBER 11 NOVEMBER. 1953

Distribution of Solute in Crystals Grown from the Melt. Part II. Experimental
J. A. BURTON, E. D. KOLB, W. P. SLICHTER, AND J. D. STRUTHERS
Bell Telephone Laboratories, Murray Hill, Nw Jersey
(Received May 18, 1953)

Experiments have been performed on the distribution coefficients of a number of solute elements in
germanium crystals grown from the melt. The variation of distribution coefficient with conditions of crystal-
lization is examined in the light of the theory of Part I. The inc~rporation of solute elements into the crystal
is shown to depend critically upon the transport processes occurring in the melt.

T HE distribution of solute atoms in germanium

crystals grown from the melt was examined in
Part I, with particular attention to the effect of trans-
However, this work has been done for concentrations of
the order of one percent or greater, and affords little
information on the distribution coefficients in this low
port processes in the melt. The distribution coefficient, concentration region.
defined as the ratio of the solute concentration in the For germanium crystals containing impurities less
solid to that in the body of the liquid, was related by a than a part per million, the ordinary methods of direct
simple model to the growth parameters and to certain
chemical analysis are too insensitive. Radioactive
physical constants characteristic of the solvent-solute
tracer techniques are applicable, however, for any
system. In Part II, the distribution coefficient in
germanium is examined experimentally. element having a radioactive isotope with a suitably
long half-life and a sufficiently high specific activity.
1. EQUILmRIUM DISTRmUTION COEFFICIENT Such measurements have been made with the following
The equilibrium distribution coefficient was defined radioactive isotopes:* (60 day Sb 124), (26.8 hr AS76) ,
in Part I. For most elements in germanium the solid (14.3 hr Ga 72 ), (14.3 day P32), (50 day InIl4), (12.9 hr
solubility is small, so that solute atoms are ordinarily CU 64), (85 yr Ni 63 ) , (2.7 day AU198), (270 day AgllO) ,
present in germanium crystals to the extent of about (2.7 yr T1204), (250 day Zn 65), and (5.3 yr C0 60). The
1: 108 to 1: 10 6• At these small concentrations, the laws concentration of conventional donor or acceptor ele-
of dilute solutions should apply, and the liquidus and ments in solid germanium can also be determined from
solidus lines in the phase diagram should be virtually electrical conductivity measurements, combined with a
straight. Thus, the equilibrium distribution coefficient
knowledge of the electron and hole mobilities.t Both
is usually independent of concentration in this concen-
methods of analysis have been applied to single crystals
tration range.
Phase diagrams of many binary germanium a,lloys * The isotopes were obtained from Brookhaven and Oak Ridge
have been determined by Klemm and his co-workers. 1 National Laboratories on allocation from the Isotopes Division,
1 W. Klemm and H Stohr, Z. anorg. u. allgem. Chern. 241, 305 U S. Atomic Energy CommiSSIOn.
(1939); ibid. 244, 205 (1940). W. Klemm et al., ibid. 256, 239 t Distribution coefficients have been determined at the Bell
(1948). See also M Zumbush et al., ibid 242,237 (1939); also, M. Telephone Laboratories, from conductivity measurements, by
Hansen, Der Aufbau der ZweistojJlegierunger, (Springer, Berlin, H. C. Theuerer, M. Sparks, W. G. Pfann, G. L. Pearson, and W. P.
1936). Slichter.

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1992 BURTON, KOLB, SLICHTER, AND STRUTHERS

TABLE I. there is one donor or acceptor level per impurity atom,

n (or p) is identified with Cs, at room temperature.
Tetrahedral
covalent Conductivity is readily measured, but some uncertainty
radIUs Approx1mate exists concerning the value of 1-',6 for high conductivity
Element (Pauhng) CS/CL Notes
samples.
P 1.10 0.12 a
As 1.18 0.04 a Table I summarizes the distribution coefficients 6 for a
Sb 1.36 0.003 a number of solute elements in germanium, obtained
Bi 1.46 4XlO- s b from radioactive tracer and conductivity measurements.
B 0.88 >1 c
AI 1.26 010 d In Fig. 1, the measured values of CS/C L are plotted
Ga 1.26 0.10 a against the tetrahedral covalent radius 7 of the solute
In 1.44 0001 e
TI 1.47 4X10- 5 a atom. It can be seen that CS/CL covers a wide range for
Cu 1.3"5 15XlO- s a the different solutes, and that there is a fairly consistent
Ag 1.53 10-' f relation between the distribution coefficient and the
Au 1.50 3X10-s f
Ni 5X1O- 6 a atomic radius. As would be expected, the larger solute
Zn 1.31 0.01 a atoms are difficult to incorporate in the germanium
Co 10- 6 f
Ge 1.22 10
lattice, and tend to remain in the liquid phase. The
Group III-A and V-A elements are believed to enter the
Both radioactive tracer and conductivIty measurements.
a
germanium lattice substitutionally with tetrahedral
Conductivity measurements only.
b covalent bonds. Hence, the tetrahedral radius should be
'Conductiv1ty measurements only. Th,s value for boron 1S based on
work by R. N. Hall, Phys. Rev 78, 645 (1950) and by W. G Pfann (un- an approximate measure of atomic size for these ele-
pubhshed).
d Conductiv1ty measurements only. Th,s value for alummum is based ments. It is not known whether the other elements are
on work by H. C. Theurer.
e Both radIoactive tracer and conductivIty measurements. ThIS value incorporated substitutionally or interstitially. However,
has been confirmed by J. J. Dowd and R. L Rouse, Proc Phys. Soc
(London) 6, 60 (1953) the distribution coefficients seem to depend on tetra-
f RadIoactIve tracer measurements only.
hedral radius in much the same way, with the exception
of copper. Since copper has an unusually large diffusion
prepared by the method of Teal and Little,2 and to constant in germanium, 8 it is not surprising that copper
polycrystalline ingots. 3 should show singular behavior in other respects com-
The techniques used to measure the absolute concen- pared with other solute elements of comparable radius.
trations of tracer elements were essentially the same as Thus, it is seen that the distribution coefficient of copper
those used by Pearson, Struthers, and Theuerer. 4 A is also irregular, being about 1/200 as large as that for
known quantity of tracer element is added to the melt
antimony, an atom of like tetrahedral radius.
from which a crystal is grown. The analytical technique
It should be emphasized that the values for Cs/C L
gives the concentration of solute, C s in a given region of
the crystal. The concentration of solute in the melt, given in Table I are only approximate. These measure-
CL, at the time when that part of the crystal was formed ments were made on crystals grown at appreciable
is obtainable from the sum of the tracer content of the rates. Where possible, such measurements have been
subsequent portions of the solid; or it is obtainable by corrected to the zero growth rate values, using the
subtracting the tracer content of the previous portions theory of Part I. The uncertainties in this procedure and
from the original quantity of solute added to the melt the experimental errors are of the order of a factor of
(assuming none to have been lost by evaporation). The two. In most cases, the errors are probably in the direc-
ratio C siC L gives the distribution coefficient for the tion to cause the values quoted here to be larger than
given region of the crystal. the true equilibrium values.
The use of conductivity measurements to determine
concentration of donor and acceptor elements is con- 2. DISTRIBUTION COEFFICIENT AT FINITE
venient for most Group III-A and Group V-A elements, GROWTH RATE
and is particularly valuable for elements like boron and The steady-state distribution coefficient was found
aluminum which lack radioactive isotopes of long half- in Part I to depend upon growth conditions in the
life. Use is made of the well known relation
6 W. Shockley, Electrons and Holes in Semiconductors (D. Van
(1) Nostrand Company, New York, 1950), p 336. Mobility has been
studied recently by P. P. Debye and E. M Conwell, Phys. Rev.
where rT is the conductivity, e is the 'electronic charge, 87, 1131 (1952). See also J. R. Haynes and W. Shockley, Phys
Rev. 81, 835 (1951).
I-'is the mobility of electrons (or holes), and n (or p) is 6 Preliminary values of distribution coefficients have been re-
the density of electrons (holes). If only one donor or ported previously by Struthers, Buehler, Theuerer, and Burton, J.
Metals 4, 149 (1952); and by J. A. Burton at the lnst. RadiO
acceptor element is present in significant amount, and if Engrs and Am lnst. Elec. Engrs Conference on Semi-conductor
Device Research, June, 1952.
G. K. Teal and J. B. Little, Phys Rev. 78, 647 (1950).
2 7 L. Pauhng, Nature of the Chemical Bond (Cornell University
H. C. Theuerer and J. H. Scaff, Trans. Am. lnst. Mining Met.
8 Press, Ithaca, 1945), p. 179
Engrs., 189 (1951). J. Metals, 59 (1951). 8 C. S. Fuller and J. D. Struthers, Phys. Rev. 87, 526 (1952);
• Pearson, Struthers, and Theuerer, Phys. Rev. 77, 809 (1950). W. P. Slichter and E. D. Kolb, Phys Rev 87, 527 (1952).

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 SOLUTE DISTRIBUTION IN CRYSTALS GROWN FROM TH.E MELT. II 1993

following manner: t PULL

ROTATION
k* fo I ---
,e DONOR OR ACCEPTOR

k .:1= D' k*=k*(f). (2) ADDITION

k*+ (1- k*)e- A ' 1----"

Here again k is the effective distribution coefficient
and k* is the ratio C slC o, describing the solute distri-
I 'QUARTZ ENVELOPE

bution at the solid-liquid interface. The thickness, 0, I

SEED HOLDER ------- -------
of the diffusion-dominated region depends upon growth SINGLE CRYSTAL
rate, f, upon the diffusion coefficient in the liquid, D, SEED --------------

and upon the crystal rotation rate. The quantity k* de- SINGLE CRYSTAL ____ 1____ _ ,,_ --,-CARBON CRUOBLE

pends in general upon J. The dependence of k upon MELT ______ J __ ,_

--
COIL FOR RF
HEATING
growth conditions, for steady-state crystallization, has
been examined experimentally for the solutes gallium
and antimony. In these studies, single crystals were
pulled2 from a melt of germanium which had been
'"'-WM_j-I--
TEMPERATURE
CONTROL

purified by the molten zone technique. 9 The crystal- -"--11_~Ll---,

lization apparatus is shown schematically in Fig. 2.+

When the desired crystal diameter was attained, about
HZ OUTLET
2 cm, the heat input to the apparatus was regulated to
keep the diameter uniform for the rest of the crystal- FIG 2 Schematic diagram of the apparatus used to grow
single crystals of germanium for the melt.
lization. Constancy of diameter was considered to be
indicative of a steady state. When the proper diameter
control was in effect, a known quantity of the solute can markedly increase the apparent distribution
element was added to the melt during the course of coefficient of the antimony. In the present case, however,
growth. In the case of gallium, which has a fairly large spectrochemical analysis showed the antimony to be
value of ko C......,O.l), the solute was added in the form of free of arsenic. With both the gallium and the antimony,
a weighed pellet composed about 99.9 percent of the quantity of solute was limited so as to keep the
germanium. The gallium content of the admixture was resulting resistivity not less than about one ohm-em,
found by chemical analysis. In the case of antimony, in order to minimize the uncertainty concerning the
which has a very small value of ko("'-'0.003), the solute mobility of the current carriers.
was added in pure form. Arsenic is a common impurity The concentration of solute in the solid before and
in antimony, and since ko for arsenic is relatively large just after the impurity addition was found by conduc-
("'-'0.04), even a small contamination by this element tivity measurements. The volume of solid previous to
impurity addition was measured, and from this were
0 determined the melt volume and hence the value of CL
I
IB at the time of addition. The ratio of C s, the solid solute
I
I Ge concentration due to the added impurity, to C L gave
1
-- the effective distribution coefficient.
~ Crystals were grown under a variety of growth rates
. ~
I Al
At
1'1GaL......---
11 I
and rotation rates. Figure 3 displays the results for the
case of antimony in germanium. The curves were fitted
1 ~n to the experimental points by the method of trial and
I
2

I _y error. The curves were calculated on the basis of Eq.

I
I In
(15) of Part I, which involves the assumption that
3 : " I equilibrium exists at the solid-liquid interface (k*=ko),
r-- - i
I virtually independently of growth rate. An appropriate
I
I ' A,9 value for ko (0.003) was adopted, and by trial a value of
l-
I I I
i
I
B,Ntuf-
olD was chosen for one rotation rate. The values of olD
for the other rotation rates were then found from this
1
U 1f? 1 first value by means of Eq. (10) of the theory in Part 1.
•
10- L ~ +
08 09 10 11 12 13 14 15 16 Figure 4 shows similar data for the case of gallium, with
TETRAHEDRAL COVALENT RADIUS IN ANGSTROMS (PAULING)
ko taken to be 0.085, and with the curves computed in
FIG. 1. Experimental values of the distribution coeffiCients of like manner.
solute elements between liquid and solid germanium, plotted
against the tetrahedral covalent radii of the elements. With both solutes, it appears that reasonable agree-
ment is obtainable between experiment and the simple
W. G. Pfann and K. M Olsen, Phys. Rev. 89, 322 (1953).
9
theory, within the limits of experimental uncertainty.
t Theapparatus and methods used were those developed by
G. K. Teal and hIS co-workers (see reference 2). This uncertainty is seen to be not less than ± 10 percent.

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1994 BURTON, KOLB, SLICHTER, AND STRUTHERS

X 10- 3 is S.5XlO-5 cm2 secl. It is interesting to note that the

.
11.0.

10. o . f - - .
o
RPM
57
144
values thus obtained from the theory are of the same
magnitude as diffusion coefficients reported in the
literature for a wide variety of solutes in molten
'" 1440

elements.
V"
0.

/ 3. UNIFORM DISTRIBUTION OF SOLUTE

~ B.O
Z
w
i3
t(SEC/CMl / The validity of the theory of Part I may be tested in
iL
Hi 10
16Y
another way. It is well known that when freezing pro-
8
z
V • ceeds at a constant rate, the solute concentration gener-
o
t= 60
RP.y
/ /' ally increases markedly in the direction of freezing,u'
al
ii
57 ~O An extension of the theory permits one to derive the
7:
../
1;; growth conditions required to keep the concentration
(5 5.0
144/ uniform over a large part of the solid. Experiments
40. V l..)V following this approach reflect on the correctness of the
L-V o 14~O ~ ..-::- ~ basic theory.
~ i-'"'" For a given fraction of crystal grown, the greater the
3.0.
0
rate of freezing, the greater is the concentration of
20. solute brought into solid solution. For the early stages
o I<) 20. 30 40 50. 60 70 eo of crystallization, therefore, let a fairly high growth
GROWTH RATE IN MICRONS PER SECOND
speed be used, so that the rate at which impurity is
FIG. 3. Distrubution coefficient of antimony in germamum,
as a function of crystal growth rate and rotatIOn rate. frozen into the solid will be enhanced, compared with

0.22
Although it may be true that k* depends in some degree
~ 020
upon j, one is not forced to conclude that the effect is w RPM J I V
important with the solutes investigated. Indeed, if the
\)018 - • 57 ~(SEC/CMl............
[ 0 144

theory used to plot the curves is correct, the data do not 8 01

6- 0

'"
575
1440 R~ i-'"'"
134.........
.~
Z 014
entitle one to a conclusion which is any more detailed Q 0
~ ~ 144
o 0
~
than the simple relation k* = k o•
It is quite obvious that the effective distribution
~ 01 2
~ 010 .--
0

~~ ~ ~
1440' - 2 7
~'"
0

coefficient depends critically upon the intensity of '"0008

0.0. 6
stirring. This fact shows that diffusion in the liquid 0. m ~ m ~ ~
GROWTH RATE IN MICRONS PER SECOND
~ ro eo
contributes significantly to the over-all kinetics. This
finding is in conflict with Hall's beliepo that he had FIG. 4 Dlstrubltion coefficient of gallium in germanium,
as a function of cyrstal growth rate and rotatIOn rate
obtained virtually complete mixing in the melt, with a
crystal rotation rate of only 200 rpm. the amount obtainable at reduced speeds. Then, as the
The experiments on distribution coefficient lead to accumulation of solute in the melt tends to increase the
the choice of values of the ratio olD which for a given concentration in the solid still more, let the growth rate
solute depend upon the rotation rate and are independ- be progressively decreased. By the proper program for
ent of growth if j is small enough. The physical con- the decrease of growth speed, uniform solute concentra-
stants required for the exact calculation of 0/D from the tion is in principle obtainable over an extended region
theory evidently have not been investigated. But it is of the crystalP
proper to ask whether the values for olD found some- If the mean concentration of solute in the system is
what artificially from the present experiments lead to CM, the following relation exists between CL and the
reasonable figures for the diffusion coefficients. Equa- fraction of melt frozen, X, since C s is to be constant:
tion (11) of Part I may be rewritten to yield an expres-
sion for the diffusion coefficient, D, ~n terms of the (4)
angular velocity, w, and the corresponding value of 0/D.
(The small amount of solute in the boundary layer next
to the crystal is neglected.) Let the maximum extent of
(3) uniform solid obtained when the growth rate has been
progressively reduced to zero be termed X o• At this
Taking a reasonable value for the kinematic viscosity, point, Eq. (4) becomes
(0.0025 cgs unit), and the experimental values of olD, CL O(l-X o)+CSXO=CM. (5)
it then develops from (3))hat D for gallium in molten
germanium is 7.5XlO- 6 cm2 secl, and for antimony D 11See reference 5 in Part I.
12This technique was first investigated empirically by G. K.
111 R. N. Hall, Phys. Rev. 88, 139 (1952). Teal and E. Buehler (unpublished).

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 SOL UTE DIS T RIB UTI 0 N INC R Y S TAL S G ROW N FRO M THE MEL T. II 1995
10
Equating (4) and (5), and using (2), one can obtain a
o
relation which defines the required variation of growth S08
~" 06 If. ~
rate with fraction of material solidified, for a crystal '\V\; IA
grown under steady-state conditions. iii
",
~ 04
~ IV \ II V\ J\1\ I~ [\\ If V~
",
Z
l-k* l-ko l-Xo :
... 02
--·e-"-=--·--. (6)
o
k* ko I-X
Both A and k* depend upon j, but if there is virtual
equilibrium at the solid-liquid interface (k*=ko),
Eq. (6) assumes a simple form:
e-"-= (I-Xo)/(l-X). (7)
70
h h I I
It is evident that the fraction Xo cannot be reached Z",
_0
60
I f\ I: ", I I
50
in finite time. However one can choose a hypothetical ~8
-.., 40
:' I I I: I \1 ti , f\ I I \ ~
>-",
X o, and can in practice approach this value quite closely .!.~
30
I ~: I I \ 1 I 1'1 I \ I , I I: :

within a convenient time. For a chosen value of X o, one !!'u

-'2
20
10
I I, l! IV • ~ 1'. ./ \' "'-I ': \
has some freedom in the selection of crystal rotation Os
I I
w ~ ~ ~
rate and initial growth rate. Figure 5 shows the varia- DISTANCE IN MILLIMETERS

tion of resistivity along the length of a representative FIG. 6. Optical transmission of radioautograph, electrical
resistivity, and hole lifetime for a germanium crystal exhibiting a
1.4 straited distribution of antimony (data of J. D. Struthers, G. L.
\, Pearson, and J. R. Haynes).
12
,1 I .,
~
, 10
\ REG10N OF
tf- - PROGRAMMED GROWTH-- The concentration fluctuations are best seen with the
::I!
~ 08
I'J 0 I aid of radioautographs, which are made by cutting the
crystal longitudinally along the direction of growth, and
~ 06
0
\ placing this flat ground surface in contact with a photo-
:; graphic plate. The radiations from the tracer element
;::
'"
Vi 04
~ expose the emulsion to a degree which is roughly
't::
'0
UJ
a:
proportional to the concentration of tracer in the crystal.
02
J. R. Haynes and G. L. Pearson13 have shown that
o
o 02 04 06 08 10 12 14 16 1.8 20
accompanying irregularities occur in the lifetime of
DISTANCE FROM POINT OF FIRST GROWTH IN INCHES injected current carriers and in the resistivity. Figure 6
FIG. 5. Example of the variation of resistivity with shows the corresponding variations in solute concentra-
distance in a crystal grown with a programmed growth rate. tion, hole lifetime and resistivity in one of these irregular
crystals.
crystal, a portion of which was grown as specified by Figure 7 shows crystals grown from a poorly stirred
Eq. (7). In this case, arsenic was used as the solute. The melt, where the speed of crystal rotation was only four
crystal diameter was kept constant over the region of rpm, so that the fluid flow was probably largely the
programmed growth, as a criterion for steady-state result of uncontrolled convection currents. Growth rates
growth. It is seen that the solute concentration is quite were in the range 1 to 2 mils/sec. In both cases shown,
uniform over the programmed region. This result offers the radioautographs reveal bands or striations of high
further substantiation of the basic theory. and low solute concentration parallel to the solid-
liquid interface, and spaced about 0.5 to 2 mm apart.
4. STRIATED CONCENTRATION DISTRIBUTIONS In contrast, Fig. 8 shows radioautographs of two
The importance of transport processes in the melt to germanium crystals which were prepared with good
the over-all reaction of solute incorporation during stirring. Both crystals were grown at 2 mils/sec. For the
crystallization is further seen from a study of the effects crystal shown on the left, both the crystal and the
which prevail when the mechanical stirring of the melt crucible containing the melt were rotated at 120 rpm,
is deliberately made negligible. Then fluid motion comes in opposite directions. The crystal shown on the right
only from thermal convection currents which doubtless was rotated at 60 rpm, with a stationary crucible. It is
fluctuate, and so the contribution of diffusion to the noteworthy that the more intense stirring produced
transport process continually changes. Thus the crystal homogeneous concentrations in the crystals, except for
is grossly inhomogeneous in its solute composition. normal segregation in the direction of growth.
Variations in concentration can also result from fluctua- In Fig. 7 (left) and Fig. 8 (right), one can detect
tions in growth rate, caused by cyclic or erratic varia- crystallographic twins; these regions are quite distinct
tions in the temperature control. 13 Unpublished studies.

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions
 1996 BURTON, KOLB, SLICHTER, AND STRUTHERS

TWIN
\.

FIG. 8. Radioautographs of germanium crystals prepared from

well-stirred melts: (left) germanium single crystal rotated at 120
rpm and grown from a melt to which three antimony pellets were
added; (right) twinned germanium crystal containing radioactive
phosphorus and grown with a rotation of 60 rpm.

FIG. 7. Radioautographs of germanium crystals prepared from

poorly stirred melts (crystal rotated at 4 rpm): (left) radioactive growth rate, which is essentially Hall's viewpoint,to or to
antimony added; (right) radioactive gold added. a change in ko with crystallographic orientation.
Thus, Figs. 7 and 8 point out the importance of
in the original radioautographs. It is seen that a change adequate stirring to the growth of homogeneous crystals
in solute concentration, and hence a change in the and by inference show the importance of the transport
effective distribution coefficient, accompanies the shift process in the melt to the over-all reaction.
in crystal orientation. From measurements of the optical
density of the radioautographs, it is estimated that the ACKNOWLEDGMENT
distribution coefficient changed by not more than five The assistance of Miss A. D. Mills and Miss K. B.
percent. It is not clear whether this change is to be Wolfstirn in making a number of the measurements is
attributed to a change in the dependence of k*(f) upon gratefully acknowledged.

Downloaded 21 Sep 2013 to 128.123.35.41. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions