Journal of Crystal Growth 6 (1970) 253-260 © North-Holland Publishing Co.

, Amsterdam

SOLUTE DISTRIBUTIONS AT NON-PLANAR, SOLID-LIQUID GROWTH FRONTS

I. STEADY-STATE CONDITIONS

R. M. SHARP and A. HELLA WELL

Oxford University, Department of Metallurgy, Oxford, England

Received 1 August 1969; revised manuscript received 25 August 1969

The shapes of steady-state growth fronts have been varied in
Al-Cu and Al-Ag alloys by independent control of concentra-
tions, freezing rates and temperature gradients. The information
reported concerns (a) the scale and shape of the interfaces, and
(b), the efficiency oflateral v longitudinal solute redistribution at
cellular or dendritic interfaces.
The observations are discussed with reference to the interfacial
temperature and stability, and it is suggested that the interfaces
adjust their shape in such a way as to eliminate any constitutional
supercooling.

1. Introduction produced in liquid which was mixed by diffusion

In a previous paper!) it was shown how long range
and short range solute profiles could be studied in
quenched alloy specimens using an electron micro-
probe analyser. Studies of this type have been con-
tinued with aluminium base alloys, solute distributions
being examined at a variety of cellular or dendritic
growth fronts.
Experimental details were as before!) and alloy speci-
mens 12 cm long and 2 mm diameter were quenched
into water when the growth front had advanced about
halfway along the rod. The shapes ofthe growth fronts
were controlled by the freezing rates, V, the alloy con-
centrations, C oo and the temperature gradients, G. The
freezing rates were assumed to be those of the rates of
traction. The concentrations were measured by micro-
probe analyses (see later) and confirmatory atomic ab-
sorption analyses. The temperature gradients were
varied by control of the temperature of the furnace
from which specimens were solidified and were the most
difficult quantities to measure - in addition to approxi-
mate values obtained with thermocouples, estimates
were made from the compositions and dimensions of
interfaces in alloys containing primary phase and eutec-
tic material.
Alloys examined, Al-Cu and Al-Ag, were such that
the solute was more dense than the solvent, so that
when solidified vertically upwards, with positive tem-
perature gradients, steady-state solute profiles were
only2,3).
Aluminium-copper alloys were studied in the range
2-20 wt % Cu. In this system the liquidus and solidus
are essentially linear and the equilibrium solid-liquid
distribution coefficient, k o , was determined from planar
growth fronts!) as 0.17. In the aluminium-silver system
the liquidus and solidus are not linear, and in the range
studied, from 7 to 24 wt % Ag, k o varied from 0.35 to
0.49. In the Al-Cu system the limit of solid solubility
is ~ 5.5 wt % Cu, in the Al-Ag system the solid-solubil-
ity limit is 55.6 wt % Ag (ref. 4).
2. Interface morphology
Allowing for uncertainty about precise values of the
temperature gradients, the conditions for the break-
down of the planar growth front coincided c10sely with
the expression for constitutional undercooling 3):
(1)
where G, V, C oo and k o are as stated, m = liquidus
gradient and D the liquid diffusion coefficient (which
was measured from previous work as 3.0 x 10- 5 cm 2
sec). This breakdown occurred in Al-Cu alloys when
GjVC oo = 6.3 X 10 3o C sec wt% mm- 2, and in Al-Ag
alloys over a range corresponding to the variable value
of k o.
The cellular morphology varied from slightly curved
projections of the type shown in fig. la to those of I b
and lc at which there occurred scalloped or branched

253
254 R. M. SHARP AND A. HELLA WELL

Fig. la

Fig. Ib

Fig. 1. Longitudinal sections of quenched growth fronts at different values of the product VC",; x 150. (a) VC", 13.um
wt %/sec; (b) VC., = 20.um wt %/sec, (c) VC", = 67 .um wt %/sec.
 SOLUTE DISTRIBUTIONS AT NON-PLANAR, SOLID-LIQUID GROWTH FRONTS. I 255

Fig. lc

Fig. 2. Quenched growth front illustrating a case in which the <100> growth axis did not coincide with the normal to the growth
front; x81.
256 R. M. SHARP AND A. HELLA WELL
shapes. The change in morphology with the product
VC oo was such that the curvature increased steadily
with a profile of the type shown in fig. la, but at some
point (VC oo ::::; 25 11m wt %/sec) the cell broke into more
complex shapes on which the curvature continued to
increase locally. However, while the morphology
changed in this way, the distances between the crowns
of cells or dendrites - i.e. between the tips of the co m-
pIex forms of figs. 1b or 1c - remained at approxima-
tely 75 ± 25 11m. Measurements were somewhat un-
certain because typica1 transverse sections included on1y
some 15-20 cells across the diameter of a specimen, and
the shapes became progressively more complex, - but
the impression was that this dimension was very in-
sensitive to the product VC oo over a range from 3.3 to
500 11m wt %/sec, in either system.
No discontinuous change in the morpho10gy could
be detected so that it is a matter of terminology whether
or not the curved projections at the interface should be
called cells or dendrites. Inasmuch as there is transition
in the morphology, however, it seemed possible to
distinguish as dendrites those forms which grew in crys-
tallographic directions, irrespective of the imposed di-
rection of heat flow. In the sections of figs. 1a-1c, the
heat flow and <100) dendrite axes coincide, but in a
case such as that of fig. 2 it may be seen how the den-
drite axes in the quenched liquid deviate from the inter-
face normal by '" 25 on that section. The morphology
0
of the steady-state form on the same section, however,
was such that the growth was taking place at '" 15 to 0
the direction of heat flow, so that this and other cases
were intermediate between the two extremes and thus
defied any system of rigorous classification.
To a first approximation, for any temperature gra-
dient, the changes in morphology progressed with the
product VC oo , but clearly the freedom for dendritic
side branches to grow (as distinct from the detailed tip
geometry and preferred growth direction) was partly
determined by where the composition fell on the phase
diagram. Typically, the regular cells of fig. 1a were
characteristic of compositions well within the solid
solubility ranges, while the dendritic structures, fig. 1c,
were produced with compositions well beyond the solid-
solubility limits. Within the range of conditions em-
ployed in the present work (up to VC oo = 500 11m
wt %jsec units, and with GL from 7.5 to 16.0°C/mm)
side branches did not ass urne perfectly crystallographic
orientations with respect to the primary sterns of den-
drites.
Some attempts were made to estimate the radii of
curvature, r, of cells, from longitudinal sections, the
larger radii being taken as approximately diametric
sections through individual cells. The measured values
fell from a maximum of ",75 11m close to the cellular-
planar transition, as fig. 1a, to an estimated tip radius
for the central spine, fig. 1c, of the order of 15-20 11m.
Measurements taken at different gradients also indicat-
ed that the radii were inversely proportional to some
function of G, so that a relationship of the form:
(2)
existed. Previous workers 5, 6) have concluded that
values of x and y = 1- were obeyed approximately, and
plotting r oc 1j(VC)t, for two gradients (7.5 and
16.0 °C/mm) appeared as in fig. 3. The scatter of points
80
70
"
10
OL---------~n~1--------~~72--~
V'I~ '"
(Vcr\
Fig. 3. Plot of the radius of curvature at the tips of cells, r,
against l/(VC)~for specimens grown under two different tem-
perature gradients.
is, however, very wide, as might be expected from
measurements on random longitudinal sections. On
this plot, fig. 3, the relative slopes of the radii are in
approximately the same ratio as the relevant temper-
ature gradients, and plotting these results against
k~jG(VC)t, fig. 4, shows fair agreement for a wide
range of conditions, and in the two systems with either
copper or silver as the solute.
3. Solute distribution
With the formation of curved interfaces, the solute
rejection becomes both lateral and normal to the ma-
 SOLUTE DISTRIBUTIONS AT NON-PLANAR, SOLID-LIQUID GROWTH FRONTS. I
80
60
x x
rl x specimens were made as an extra check against micro-
40
" tions agreed to ± 2 %jwt % Cu or Ag.
• x AI·2I.Cu G'75
•
x
0
c AI-ISZAg Gd60
20 o AI-27. Cu G,160
• AH07.Cu G'160
OL-----'2~----~4----~6~----n8----~1O~~
~_.
G-'CVCjk o)"' x 10"4
Fig. 4. Plot of r against ko'/G( vC)t for specimens ofboth Al-Cu
and Al-Ag alloys.
croscopic front and the result of the curvature can be
observed by measuring the solute concentration, cs,
immediately behind the tips of cell-dendrites - posi-
tion A in fig. 5a. Dendrite compositions, cs, were then
related to the bulk liquid concentrations, Ce,,, which
were measured from 300 .um li ne scans of the chilled
liquid, BC of fig. 5a taken beyond the solute enriched
layer. With the steady state conditions this was the
same as the bulk solid concentration. Solute profiles
were also plotted by driving the specimens under 50 .um
line scans to give profiles of the form shown in fig. 5b
from areas of the type OPQR in fig. 5a.
Micro-probe analyses were made using a Cameca
Mk.II, MS46, instrument, concentrations being meas-
ured from CuKet t and AgLet t radiations, with accelerat-
(a)
(b)
Fig. 5. Diagrammatic representation of the points and areas
when: electron micro-probe analyses were made, showing the
form of a solute profile obtained from the shaded area.
257
ing potentials of 30 kV and 20 kV respectively. Ab-
sorption corrections were made according to the proce-
dure described by Thomas 7) and bulk analyses (which
were made by atomic adsorption spectroscopy at IMI
laboratories, Witton, Birmingham) of homogeneous
probe results. The corrected and analysed composi-
The efficiency of solute rejection at a celljdendrite
could be expressed by an effective distribution coeffi-
cient, k; = CsjC oo ' and for alloys of different composi-
tions andjor different systems, the relative solute ac-
cumulation at the celljdendrite tip could be expressed
in a normalised form as (k e - ko)jk o. Estimates of k e
could be made with some accuracy (see above) from
X-ray intensity measurements counted over periods of
10 sec at different positions. The liquid interfacial con-
centrations, Ci> and the interfacial gradients, dcjdx,
could not be measured with the same accuracy because
line scans from the chilled liquid were only apparently
tangential to the curved fronts and the longitudinal sec-
tions of celljdendrites were not necessarily diametric.
Within the limits of accuracy of these latter measure-
ments the interfacial concentration was related to that
in the solid by the equilibrium distribution, coefficient,
Cs = koC j , as assumed above. To a first approxima-
tion, also, the interfacial solute gradients, normal to the
celljdendrite tips were compatible with the conserva-
tion of solute rejected at the tips, namely
(3)
In fig. 6, k e is plotted for Al-Cu alloys against
log VC oo , for two temperature gradients. It may be seen
kJ
o •
.. .0:
2 4
IOi:vc
Fig. 6. Plot of the effective solid-liquid distribution coefficient,
k e , for cell/dendrites grown under two different temperature gra-
dients in Al-Cu alloys, against log VC oo •
258 R. M. SHARP AND A. HELLA WELL

that the solute accumulation falls only gradually with
increasing rate and concentration, and tends to zero as
VCoo approaches very high values. The effect of in-
creasing the temperature gradient is the same as de-
creasing the product VC oo and, for a given value of
VC"" there is a correspondingly larger solute build up
at the cell/dendrite tip. At the same time, fig. 3,as there
is a larger solute accumulation, the radius of curvature
is somewhat smaller.
4. Temperature gradient
To obtain estimates of the temperature gradients
which were more reliable than those measured directly
with thermocouples, alloys of AI-20 wt % Cu were
of the interdendritic liquid it is considered that these
va lues relate primarily to the liquid phase.
5. Discussion
5 . 1. INTERFACE MORPHOLOGY
The present results are not strictly comparable with
previously reported measurements of cell/dendrite di-
mensions in metal systems 6) because the conditiom
described he re were steady state and those of other
workers involved stirred liquid. However, the preseot
results are remarkable in as much as the intercellular
spacings did not appear to be a sensitive function ot
freezing rate, but the more important dimension, the

-------l.06mms - - - - - - - - - - 1
Fig. 7. Diagram showing the method of estimating temperature gradients. The temperatures marked are obtained from the
phase diagram by measuring the composition of dendrites at the interface.

frozen at various rates (6 to 48 mm/sec) and examined
microscopically and by micro-probe analysis. From
such specimens, fig. 7, the distance between the cell/
dendrite tips and the eutectic front could be measured
and the corresponding temperature interval obtained
from the tip composition and the liquidus and eutectic
temperatures on the phase diagarn (assuming these to
be independent of curvature). The me an temperature
gradients so measured were 16.0 and 7.5 °C/mm for the
data plotted in figs. 3, 4 and 6, and since the volume
occupied by the primary crystals was sm aller than that
radius of curvature of the interface, followed a similar
relation to that observed in transp:lrent organic analo-
gues 5 ).
lt would be expected that as cell curvature increased
and lateral solute rejection became more pronounced,
the influence of one cell upon thc next would also
change or require adjustment ofthe intercellularsepara-
tion. It would appear from the present results that in-
crease in curvature becomes local beyond some limiting
value and that thereafter the relatively long range se-
parations between cells are insensitive to loeal changes
 SOLUTE DISTRIBUTIONS AT NON-PLANAR, SOLID-LIQUID GROWTH FRONTS. I
in geometry. These morphological changes may not be
typical of conditions where there is liquid stirring by
e.g. convection, but may be peculiar to the situation
where there is a long range solute profile normal to the
isotherms and where relatively steep temperature gra-
dients are employed - i.e. the measurements may not
be sufficiently sensitive in the range of conditions em-
ployed to detect any changes. In this work, changes in
the cell/dendrite spacings are only observed during
transient break down conditions - see later Part II (to
be published).
5 .2. INTERFACIAL TEMPERATURE
The principal observation from the previous section
is that reduction of solute build up at the advancing
front takes place only gradually with V, C oo or G, and
although there is a sharp drop in Ci when a planar
front is replaced by a curved one, the solute undercool-
ings are stilliarge (in the range 0 to 15°C) but decrease
with freezing rate in an alloy of given composition. By
comparison, the undercoolings associated with curva-
ture, which can be written as 2y/r LlS (where y is the
interfacial energy and LlS the entropy of fusion) are
very small ( < 10- 1 0c) for the radii measured, and can
certainly be neglected beside the solute undercoolings
which are larger by factors of 10 2 to 103 • In these
materials the thermal conductivities are such that the
undercooling associated with heat abstraction must
also be very small in the ranges studied and the results
must show that as the rate of freezing rises for a given
alloy, so does the actual freezing temperature. It fol-
lows, therefore, that the radii of curvature of cell/den-
drites and their consequent solute build-up do not ad-
just themselves in such a way as to minimise the under-
cooling - i.e. much sharper growth forms would reduce
the solute accumulation to a negligible barrier while
still producing a relatively small curvature term, but
this does not happen. In this respect a simiJar conc1u-
sion could be made from the measurements of Kramer,
Bolling, and Tiller 8 ) who measured interfacial temper-
atures at a variety of non-planar interfaces and deduced
aseries of k e values which were not dissimilar in trend
to those reported here.
5.3. INTERFACIAL STABILITY
Since the temperature at which the growth front ad-
yances is not a maximum it must be that so me other
259
criterion restricts the interfacial curvature and conse-
quent solute distribution. Thus, when liquid ahead of
a plan ar front is said to be constitutionally undercool-
ed, fig. 8, the curved interface which develops is re-
stricted in some way which is determined by the varia-
bles V, C oo and G. One arbitrary solution might be that
the curvature increases only until there is no further
AT
1
distance ~
Fig. 8. Diagrammatic illustration of the dependence of under-
cooling on: (a) temperature gradients for a given growth rate or
(b) growth rates for a given temperature gradient, assuming a
tangentialliquidus-gradient is maintained.
constitutional supercooling, i.e., until the liquidus gra-
dient and mean temperature gradient are equal at all
points on the interface, as is represented for various
gradients, fig. 8a, and velocities, fig. 8b. Such a situa-
tion might then be compared with the stability of a
planar front as has been discussed by Mullins and
Sekerka 9 ,10), the stable interface being one which is
stable to perturbations once the temperature and liqui-
dus temperature gradients are eq ual, irrespective of the
actual interfacial temperature. Such a condition could
be expressed by the solute conservation, eq. (3), in
which the solute gradient is equated with temperature
gradient:
and since Ci Cooke/k o, this can be expressed:
ke DG
-
k o- ke
Such a solution applies strictly only to a planar front
260 R. M. SHARP AND A. HELLA WELL
when steady state conditions do not apply, as with
partial liquid stirring 11 ), and in the present context
leads to an impossible conelusion 12 ) that k e --+ 0 as
VC oo --+ 00. A correct solution should also inelude in
the solute gradient a function describing the radial or
lateral redistribution, such that k e --+ k o with VC oo •
Qualitatively, however, it may be seen how the vari-
ables V, C", and G could control the interface.
It follows from the present results and interpreta-
tion that increase of temperature gradient, fig. 8a, for
a given liquidus curve, eventually causes the cellular
interface to freeze at a temperature elose to that of the
eutectic and then allows the front to grow in a coupled
manner over a range of compositions, as has been de-
monstrated in the Pb-Sn 13) and Al-Cu 14) systems.
5.4. COMPARISON BETWEEN SYSTEMS
Measurements of k e for the system Al-Ag were made
over a similar range of the product VC oo as those
plotted in fig. 6 for Al-Cu alloys. The temperature
gradient was 16.0 °Cfmm.
In order to compare the results for the two systems
the functions ke/k o -I were plotted against log VC oo
values, measured above the critical value log (VC oo )* at
which breakdown of the planar front occurred in either
l·S
t 5) K. A. Ross, B.Se. Thesis, Oxford University, 1967.
k,- k o
T 6) T. S. Plaskett and W. C. Winegard, Can. J. Phys. 34 (1956)
I,
." ""
0
o,s tI
o 21 (1953) 1987.
Fig, 9. Plot of the function (ke-ko)jk o against log VC"" as a
comparison of results for the AI-Cu and Al-Ag systems.
log (VCjVC*) = 0 corresponds to the planar-cellular breakdown
in either system (see text).
case, fig. 9, and the two curves have similar slopes over
the same ranges of ke/k o -1. This comparison shows
that the rate of decrease in the height of the solute
profile with VC oo is approximately the same in each
system, i.e. the range of the product VC oo over which
the interfacial concentration falls to C oo is roughly pro-
portional to I/k o for the system in question.
Work is in progress in which transient solute distri-
butions are being studied by fluctuations in the growth
rates and with partial liquid stirring.
Acknowledgements
The authors would like to thank Professor P. B.
Hirsch, F.R.S., for laboratory accommodation and
facilities essential to the present work. The research
was supported by the Science Research Council and
one of us (R.M.S.) was supported by the British Steel
Corporation (The English Steel Corporation Ud., Shef-
field).
References
1) R. M. Sharp and A. Hellawell, J. Crystal Growth 5 (1969)
155.
2) F. Weinberg, Trans. Met. Soc. AlME 227 (1963) 231.
3) W. A. Tiller, J. W. Rutter, K. A. Jackson and B. Chalmers,
Acta Met. 1 (1953) 428.
4) M. Hansen, The Constitution 01 Binary Alloys (McGraw-
Hili, New York, 1958).
96.
7) P. Thomas, A.E.R.E. (Harwell) Report No. R4593 (1964).
8) J. J. Kramer, G. F. Bolling and W. A. TiIler, Trans. Met.
Soc. AlME 227 (1963) 374 .
9) W. W. Mullins and R. F. Sekerka, J. Appl. Phys. 35 (1964)
444.
10) R. F. Sekerka, J. Crystal Growth 3, 4 (1968) 71.
11) J. A. Burton, R. C. Prim and W. P. Slichter, J. Chem. Phys.
12) W. A. TilIer, Can. J. Phys. 37 (1959) 1204.
13) F. R. Mollard and M. C. F1emings, Trans. Met. Soc. AlME
239 (1967) 1526, 1534.
14) R. M. Jordan and J. D. Hunt, Department of Metallurgy,
Oxford University.