Journal of Crystal Growth 208 (2000) 219}230
In situ studies of Cd
Zn Te nucleation and crystal growth
1~x x
B.W. Choi*,1, H.N.G. Wadley
Intelligent Processing of Materials Laboratory, School of Engineering and Applied Science,
University of Virginia, Charlottesville, VA 22903, USA
Received 27 April 1999; accepted 30 September 1999
Communicated by R.S. Feigelson
Abstract
The nucleation and growth of Cd
Zn Te crystals in a multi-zone vertical Bridgman growth furnace have been
1~x x
observed and measured using in situ eddy current sensor techniques. A two-coil eddy current sensor measured coil
impedance changes for multifrequency which were then interpreted using an electromagnetic "nite element analysis. The
sensor was used to characterize the initial melting of a charge and the subsequent nucleation of solid during solidi"cation.
Fully remelted in situ compounded charges were exposed to signi"cant melt superheating and were found to undergo
large melt undercoolings (of up to 203C), spontaneous crystal nucleation and rapid solidi"cation (velocities approaching
60 mm/h which was more than 10 times the furnace translation rate). Post-growth metallography revealed that about
20 mm of polycrystalline solid was formed in this way before recalescence arrested the solidi"cation interface. In partially
remelted charges neither undercooling nor unstable growth were observed. These results indicate that eddy current
sensors can be used to monitor critical aspects of the vertical Bridgman crystal growth of semiconducting materials and
may simplify the implementation of seeded crystal growth concepts in this, and other, semiconductor crystal growth
processes. ( 2000 Elsevier Science B.V. All rights reserved.
PACS: 81.70.Ex; 81.05.Dz; 81.10.!h; 81.30.Fb
Keywords: II}VI semiconductor crystals growth; Multifrequency eddy current sensor; Finite element analysis; Vertical Bridgman
growth; CdZnTe; Liquid}solid interface location
1. Introduction
Infrared transparent Cd
Zn Te alloys are
1~x x
used as substrates for infrared focal plane arrays
and as solid state c-ray detectors [1,2]. These
* Corresponding author.
1 Present address. Chemistry and Materials Science Directorate, Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, CA 94550, USA.
semiconductor alloys are typically grown by an
unseeded vertical Bridgman method [3,4] using
shallow thermal gradients ((l03C/cm) to minimize
thermal stresses during and after ingot growth [5].
It has been di$cult to perfect methods for reliably
growing large single-grained high transparency
Cd
Zn Te alloys. This has stimulated numerous
1~x x
experimental and modeling e!orts to correlate the
controllable growth parameters of a vertical Bridgman process with resulting characteristics of the
solidi"ed ingot such as degree of single crystallinity,
0022-0248/00/$ - see front matter ( 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 2 2 - 0 2 4 8 ( 9 9 ) 0 0 5 3 6 - 9
220
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
dislocation density and solute segregation [5}8].
The initiation of solidi"cation appears to be critical
for controlling subsequent bulk ingot quality. If
one dominant grain can be grown preferentially
from the "rst solidi"ed region at a crucible tip,
a large single crystal can sometimes evolve. In the
vertical Bridgman growth of Cd
Zn Te, at1~x x
tempts to induce growth of a dominant grain have
met with mixed success [4,9,10]. This has prompted
an interest in developing a better understanding of
thermal conditions at the crucible tip and the mechanisms of solidi"cation in the thermal environments of large-scale commercial growth furnace.
Eddy current sensor approaches to the monitoring of melting and solidi"cation have attracted
signi"cant interest because all semiconductors exhibit signi"cant changes in their electrical conductivity at their melting point [11}17]. In the CdTe
system, the electrical conductivity of the solid is 4}5
times less than that of the liquid at the melting
point [12]. Thus, an eddy current sensor positioned
in the "rst to freeze region of a vertical Bridgman
furnace may be able to detect the formation of solid
providing new insights about nucleation and early
growth characteristics of vertical Bridgman grown
Cd
Zn Te alloys.
1~x x
Here, a multi-frequency eddy current sensor has
been noninvasively integrated into a commercial
scale vertical Bridgman furnace and used to characterize melting and the initial solidi"cation of
Cd
Zn
Te ingots. Results from an electro0.955 0.045
magnetic "nite element method (FEM) analysis of
the sensor}ingot interaction have been used to analyze the response of the eddy current sensor. The
study reveals the existence of a very large undercooling and high initial solidi"cation velocity during growth from fully melted ingots. This result is
compared with growth from an incomplete remelted charge where no undercooling was detected,
and the solidi"cation velocity was close to that of
the thermal gradient translation rate.
2. Simulation of sensor response
Fig. 1 shows the axisymmetric geometry of the
sensor and sample in the r}z plane (where r and
z are the radial and the axial coordinates). An
Fig. 1. The eddy current sensor and sample geometry. A six turn
primary coil was used to excite an electromagnetic "eld. A four
turn pickup coil then detected the perturbation of the primary
coil's "eld resulting from the presence of the conducting sample.
electromagnetic "nite element model can be used to
relate the multifrequency response of the eddy current sensor to the position of a liquid}solid interface located near the tip of an ampoule. The
modeled problem consisted of a Cd
Zn
Te
0.955 0.045
sample contained within a conically shaped nonconducting crucible. A cylindrical 80 mm diameter
sample that was either liquid, solid, or containing
a solid/liquid #at interface was analyzed. We "rst
solved for the magnetic vector potential, A(r, z),
using an electromagnetic FEM code [18]. The
model's inputs were the sensor's geometry and test
frequency, the sample's diameter, the liquid}solid
interface's location, and the electrical conductivities
of the solid (1400 s/m) and the melt (6550 s/m). The
deduced magnetic vector potential was then used to
obtain the sensor's electrical impedance as a function of test frequency. The detailed method and
procedures can be found in Refs. [11,12].
Fig. 2 shows the calculated normalized impedance response for both an entirely liquid (open
circles) and an entirely solid (solid circles) sample as
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
Fig. 2. Normalized impedance curves for an all liquid, all solid
and partially solidi"ed sample (solid/liquid interface located at
19.1 mm from cone tip). Note the existence of a frequencydependent `interface shifta in the imaginary impedance component. Its magnitude depends on the interfaces location.
a function of test frequency. Normalizing the response by the empty sensors impedance, resulted in
data for the completely liquid and solid cases falling
on the same characteristic `comma shapeda curve.
The shape of this curve depends only upon the
sample's diameter, the cone angle and placement of
the sensor's coils relative to the sample [11,12]. The
only di!erence between signal for the `all solida
and `all liquida cases is a shifting of frequency
points around the impedance curve. This is consistent with the well-known result that a "xed frequency impedance point moves counter clockwise
around the impedance curve as the sample conductivity decreases [12].
To simulate the response of the sensor during the
onset of nucleation and subsequent propagation of
the solid/liquid interface, a series of FEM calculations were performed with eleven di!erent locations of a #at solid/liquid interface in the conical
region of the ampoule. Fig. 2 shows an example of
the calculated normalized impedance curve for
a #at liquid}solid interface located 19.1 mm above
221
the cone tip. Above a frequency of about 50 kHz
samples that contained two distinct electrical conductivity regions resulted in impedance curves that
no longer fell on the characteristic curve of homogeneous (solid or liquid) samples. Examination of
the vector potential "eld revealed that as the solid
volume increased (i.e. as the interface moved upwards) more of the electromagnetic #ux linked solid
material and the sensors response converged towards that of the homogeneous solid. This resulted
in an `interface shifta (Fig. 2), whose magnitude was
a function of the test frequency and interface height,
h, measured from the bottom of the crucible.
The result above indicates that prior to the onset
of nucleation, a sensors imaginary impedance component will have a value determined by the liquid's
electrical conductivity. After a solidi"cation front
has passed through and beyond the range of the
sensor's electromagnetic "eld, the imaginary impedance component converges to a value (here near
unity on the imaginary axis) determined by the
solid's electrical conductivity. Thus, the impedance
increases towards unity as the interface height
changed from zero to in"nity (Fig. 3). The largest
variations in response to an interface position
change occurred at test frequencies between
500 kHz and 1.2 MHz. However, the skin depth is
relatively shallow at high frequencies and the eddy
current density at a solid nucleation in the interior
of an ampoule is more reliably detected at a lower
frequency. Our experiments indicated that the frequency range best suited for sensing the nucleation
and growth of solid Cd
Zn Te was between 200
1~x x
and 500 kHz.
3. Experimental procedures
3.1. Sensor construction/furnace installation
A two-coil `encirclinga eddy current sensor with
a design similar to that analyzed above, was placed
near the crucible tip region of a commercial scale
multi-zone vertical Bridgman furnace (Fig. 4). To
minimize thermal disturbances to the crystal
growth environment, the sensor coils were wound
on pre-existing concentric alumina tubes that are
normally used to support the conically shaped
222
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
ampoule inside the furnace. In order to constrain
the movement of the wires during heating and
cooling, grooves with the same depth and width as
the winding wires were machined on the surfaces of
the two tubes. A 6 turn, 1.02 mm diameter platinum
wire was used to wind the 38.1 mm long primary
coil while a 4 turn, 0.25 mm diameter platinum wire
was used to wind a 12.7 mm long secondary coil.
3.2. Measurement methodology
Fig. 3. The sensors calculated imaginary impedance component
variation with interface position at four frequencies.
The detailed operating principle and measurement methodology of two-coil multifrequency eddy
current sensors has been described in Refs. [12,13].
Brie#y, a continuous signal was supplied to the
primary coil and the secondary coil voltage was
measured. Frequency-dependent gain (g) measurements for the two-coil system were obtained by
recording the ratio of the voltage induced across
the secondary coil with the voltage drop across
a 1 ) low-inductance precision resistor in the
Fig. 4. Schematic diagrams of multi-zone vertical Bridgman furnace and eddy current sensor located at the ampoule tip.
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
primary circuit. The phase di!erence (/) between
these two voltage signals was also monitored. The
input currents to the primary were kept below
100 mA so that negligible eddy current heating of
the sample occurred. The sensor's gain/phase response was normalized by that of the empty sensor.
These empty sensor measurements were made at
the growth temperature and gave reference empty
coil gain (g ) and phase (/ ) measurements at each
0
0
test frequency. The real and imaginary components
of the normalized impedance (Z) in the presence of
a sample were then obtained at each test frequency
by computing:
A B
A B
Re(Z)"
g
sin(/!/ ),
0
g
0
(1)
Im(Z)"
g
cos(/!/ ).
0
g
0
(2)
During a growth experiment, gain/phase data
were collected at 101 logarithmically spaced test
frequencies between 50 kHz and 5 MHz. Since the
translation rate of the furnace was slow (less than
2 mm/h), the data collected and downloaded to
a personal computer once every 5}10 min throughout each growth experiment.
223
3.3. Growth experiments
Three sequential growth runs (1, 2, and 3) using
the same charge composition were monitored to
observe melting, nucleation and initial growth with
the sensor. A 3.3254 kg ingot was synthesized from
IRFPA substrate grade precompounded CdTe and
pieces of Te, Cd and Zn (all supplied by Johnson
Matthey Electronics, Spokane, WA). The charge
was placed in a 80 mm inner diameter conical
pyrolitic boron nitride crucible and then sealed
in an 85 mm diameter quartz ampoule under
10~6 Torr pressure. Run 1 compounded the constituents (and thus allowed the monitoring of
precompounding and melting) and was followed by
solidi"cation of a fully melted charge. The position
of the furnace with respect to the stationary ampoule was conveniently referenced by a pointer
attached to the furnace (Fig. 4). The pointer position for the start of this run was 7.4 cm. The same
charge was used in the second run, but with a raised
furnace (at a pointer position of 9.3 cm to start).
Run 3 was a repeat of the run 2 with the same
charge and start position. The starting position for
the run 1 is typical of that used for precompounding where higher temperatures are needed to ensure
complete mixing/in situ-compounding. Normally,
Table 1
Growth conditions for run 1
Segment d
Activity
Duration (h)
1&2
3
4
5
6
Ramp up to 7003C with furnace pause at 7.4 cm
Ramp up to 11533C with furnace pause at 7.4 cm
Stationary furnace at 7.4 cm
Move furnace up at 1.87 mm/h to 9.3 cm
Move furnace up at 1.49 mm/h to 11.2 cm
1
2
10
10
13
Table 2
Growth conditions for runs 2 and 3
Segment d
Activity
Duration (h)
1&2
3
4
5
6
7
Ramp up to 7003C with furnace pause at 9.3 cm
Ramp up to 11533C with furnace pause at 9.3 cm
Stationary furnace at 9.3 cm
Move furnace up at 1.49 mm/h
Stationary furnace at 11.24 cm
Move furnace up at 1.49 mm/h to 13.17 cm
1
7.5
10
13
10
13
224
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
growth of previously compounded material is accomplished from a cooler region of the furnace
[9,10] and the follow-up runs (runs 2 and 3)
explored the consequence of this using the remelted
charge. Each run consisted of a series of segments
for furnace heating, translation and holding
(Tables 1 and 2 ).
3.4. Furnace temperature proxle
To characterize the thermal environment during
growth, the axial temperature pro"le was measured
from the tip of an empty quartz ampoule (of identical diameter to that used later to contain CdZnTe
sample) to the top of the furnace. In Fig. 5, temperature pro"le data as a function of distance from the
ampoule tip was obtained from every 1 cm apart
at the best approximation to the pro"le during
real growth runs. The maximum axial temperature
variation of the region monitored by the sensor
was $13C. The empty ampoule measurements
indicated the temperature gradient where solidi"cation initiated was about 8.63C/cm. Fig. 5
Fig. 5. Temperature pro"les in the cone area for run 1 (pointer
start position of 7.4 cm) and runs 2 and 3 (pointer start position
of 9.3 cm).
also shows the crucible tip locations for the two
furnace starting positions used during growth.
Note that the melting point for an alloy of the
Cd
Zn
Te composition was 1098$13C
0.955 0.045
[12]. For run 1, the temperature at the cone tip was
10903C while at the cone shoulder it was 11083C.
For the second and the third runs, the temperature
near the cone tip varied from 1080 to 11003C. The
ampoule tip temperature was about 103C below the
melting point for the "rst run and about 203C
below for the second and the third runs. This resulted in signi"cantly di!erent melting behaviors in the
tip region. In run 1, complete melting of the charge
occurred before the start of the growth run, whereas in the other runs melting was incomplete.
4. Results
4.1. Compounding (growth run 1)
Fig. 6 shows the normalized imaginary impedance as a function of time during growth run
1 beginning at the time when furnace heating was
"rst commenced. The shaded areas in Fig. 6 represent periods during which the furnace was held
stationary. A sharp drop in impedance occurred at
1.1 h as the temperature of the furnace approached
7003C. This was correlated with melting of elemental Zn and Cd added to precompounded
equiatomic CdTe to bring the target composition
to Cd
Zn
Te. When viewed with a higher
0.955 0.045
time resolution, the imaginary impedance component returned back towards unity (the empty sensor's value). This was a consequence of each
(metallic) elements dissolution to form a solid with
an overall Cd
Zn
Te composition which
0.955 0.045
has a very low conductivity at temperatures below
7003C [12]. This solid state mixing process proceeded as the furnace temperature was increased
towards a set point of 11503C (at the end of segment d3). As the end of segment d3 was approached, the impedance began to drop sharply,
consistent with the beginning of melting (and an
associated rise in conductivity) in the region
monitored by the sensor.
From thermal measurements with the empty
furnace, upon entering segment d4, the cone tip
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
225
Fig. 6. Estimated cone tip temperature and the measured impedance response at three frequencies as functions of time during the runs.
Melting/compounding occurred during the "rst 5 h. Spontaneous nucleation of solid occurred at about 28 h into the process.
region of the charge had reached a temperature of
10903C, while the cone shoulder reached 11053C.
Both are above the melting temperature 10983C for
a Cd
Zn
Te composition ingot. The con0.955 0.045
tinued drop in impedance is therefore consistent
with the need for approximately 2 h (during segment d4) for the ingot material to melt completely
at the cone tip, presumably as a consequence of the
low thermal conductivity of CdTe alloys. After this
melting transient, a small increase in the imaginary
impedance component was observed during the
remainder of segment d4. This may have been due
to further mixing/precompounding during this 10 h
hold. It is interesting to note that in the past, it has
usually been assumed that a melting and mixing of
the charge was completed by the end of segment
d3. The eddy current results indicate this not to be
the case. The normalized impedance curve mea-
sured at the end of segment d4 compared well
with the FEM results for an entirely liquid sample.
The measured impedance data were consistent
with a sample whose conductivity was about
6400 mho/m, which is close to previously measured
values for the fully liquid state [12].
4.2. Spontaneous nucleation (growth run 1)
During segment d5, the furnace was translated
upwards for 10 h (from a starting pointer position
of 7.4 cm) at a rate of 1.87 mm/h. The temperature
of the cone tip decreased during this segment while
the eddy current data showed a very slight rise
consistent with the retention of a supercooled
liquid whose conductivity was slightly decreasing
(see Fig. 2). At the end of segment d5, the supercooling was estimated to be about 183C while the
226
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
melt conductivity was approximately 6100 mho/m.
Continued cooling in the cone area occurred during segment d6 as the furnace was translated upwards at a slower rate of 1.49 mm/h. At a process
time of 27.8 h, an abrupt return of the imaginary
impedance towards its null value was seen at all test
frequencies, Fig. 6. The furnace position at this
moment was 9.9 cm and the estimated supercooling
at the tip of the crucible was about 203C. The
interface height was determined using results from
electromagnetic FEM calculations of the sensors
interactions with a #at solid/liquid interface in the
conical region of the ampoule. This is justi"ed by
observations of reasonably uniform radial temperature pro"les in the conical region of similar
ampoules analyzed under equivalent growth conditions. For example, model predictions for a very
similar system were reported by Derby et al. [19],
and are consistent with experimental evidence (by
interface demarcation using a radioactive dopant)
during vertical Bridgman CaTe growth [20].
Examination of the normalized impedance curve
immediately after the sharp rise in impedance indicated an interface shift that increased with test
frequency. This abrupt impedance change was
therefore consistent with nucleation of solid at the
cone tip. It is possible to estimate the height of the
liquid}solid interface from the imaginary impedance data using the FEM result shown in Fig. 3.
Fig. 7 clearly shows that nucleation from a homogeneous liquid abruptly occurred at a process time
of 27.8 h. The unstable solidi"cation event caused
the interface position to move upwards about
19 mm from the cone tip over a period of 20 min.
The solidi"cation rate during this nucleation event
was estimated to be 57 mm/h, compared to a temperature gradient (i.e. furnace) translation rate of
only 1.49 mm/h. The nucleation event terminated
with a small, short duration, melt back at about
29.5 h into the run consistent with recalescence in
this low thermal conductivity system. After the
recalescence event, solidi"cation continued at
a rate close to that of the furnace's translation.
4.3. Remelting (growth runs 2 and 3)
The second and third experiments (runs 2 and 3)
were conducted using the sample from run 1 but
Fig. 7. The solid}liquid interface position, h, as a function of
time during initial nucleation and growth of solid for run 1.
with a di!erent furnace starting position (refer to
Fig. 6) closer to that normally used for precompounded growth runs. Runs 2 and 3 were nominally identical and exhibited very similar eddy current
sensor behaviors. Detailed experimental results are
therefore only presented for the run 3 experiment.
Fig. 8 shows the imaginary component of the sensors impedance normalized by that of the empty
sensor for three test frequencies as a function of
time.
The variation of the imaginary impedance component with process time is seen to be quite di!erent to that of the run 1 (Fig. 6). During the "rst 20 h
of the process period, no abrupt variations in imaginary component were observed. Instead, the imaginary impedance component gradually decreased
with process time as the samples electrical conductivity gradually increased. Note that the imaginary
impedance component was still decreasing at the
end of segment d4, indicating that the sample had
not reached a quiescent state prior to the start of
furnace translation, even though the furnace temperature had been "xed at 11533C for 10 h.
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
227
Fig. 8. Measured imaginary component impedance response for run 3. The estimated cone tip temperature is also shown for
comparison.
4.4. Resolidixcation (growth runs 2 and 3)
Fig. 8 shows that during the "rst 19 h of the
growth run, the 500 kHz imaginary impedance
component dropped to 0.970 whereas for run 1 the
drop was to 0.962. This is consistent with incomplete melting of the charge material during the
run 3. Evidence of this incomplete melting was also
seen in the impedance diagrams which were characteristic of a sample containing a liquid}solid interface. Using the FEM result of Fig. 3, the interface
position, h, was determined from the measured
imaginary impedance data, Fig. 9. Melt back was
seen to have occurred progressively with process
time during segment d4. However, it appeared to
have arrested 9.6 mm above the cone tip at the
beginning of the segment d5.
As the furnace was raised (in segment d5), the
imaginary impedance was seen to immediately begin to increase consistent with a liquid}solid inter-
face that propagated upwards away from the cone
tip. The initial solidi"cation velocity was found to
be 1.1 mm/h which was a little less than the furnace
translation rate of 1.49 mm/h. As the furnace translation process continued beyond 25 h, the interface
velocity was observed to accelerate slightly and
reached a constant speed similar to that of the
furnace. It is interesting to note that in industry,
this remelting process sequence was thought to
have caused complete melting and to then have
induced nucleation from an undercooled melt. The
results shown in Fig. 9 provide clear evidence that
this does not occur.
5. Discussion
The in situ eddy current sensor results for the
growth runs described above indicate that the approach used to ensure complete melting of ingots in
228
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
Fig. 9. Interface position area during the early remelting/solidi"cation stage of run 3. Complete melting was not achieved. The interface
descended to within 9.6 mm of the cone tip before commencement of solidi"cation at an initial velocity of about 1.13 mm/h.
a vertical Bridgman furnace is incurred at the expense of signi"cant melt superheating. This appears
to be a result of the low thermal conductivity of the
CdTe system, and the resulting axial temperature
gradient in the region of melting. Recent work has
shown that as superheating increases, the extent of
subsequent melt supercooling is also increased [5].
The eddy current sensor data has revealed that
a quite large melt undercooling of &203C occurred
following ingot precompounding prior to directional solidi"cation. The sensor revealed that unstable solid nucleation then occurred during
solidi"cation. The liquid}solid interface velocity
was measured to be &57 mm/h. During this spontaneous nucleation event, the interface advanced
approximately 19 mm from the cone tip.
Fig. 10a shows a sample grown in the vertical
Bridgman furnace just before the cone sensor was
installed. This ingot was produced under the same
thermal gradient and furnace starting position conditions as for run 1. Multiple equaixed grains of
roughly the same small size are seen to extend
20 mm upwards from the cone tip. This structure is
consistent with a completely liquid charge in which
solid was nucleated and rapidly propagated about
20 mm along the ampoule. It is also interesting to
note that after the interface had stabilized, a large
grained ingot was subsequently grown from what
now amounted to a polycrystalline seed. Fig. 10b
shows the grain structure of the tip region after
run 3. In this case, eddy current monitoring indicated that melt back terminated about 10 mm
above the ampoule tip. In this case, no "ne grained
polycrystalline region is observed. Instead, a few
large grains in the conical region were observed.
A high angle boundary was located at about 10 mm
from the cone tip coincidently corresponding to the
eddy current determined amount of melt back.
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
229
6. Summary
The initial melting and growth of solid
Cd
Zn
Te has been successfully monitored
0.955 0.045
by a two-coil eddy current sensor installed in a vertical Bridgman growth furnace. A "nite element
model was used to convert sensor impedance data
to a liquid}solid interface's position so that solid
nucleation and the growth velocity could both be
deduced from the sensed signal. The sensor was
used to determine the degree of melting, the extent
of liquid supercooling, the moment of solid nucleation and the initial growth velocity. It revealed
large supercoolings and unstable nucleation/
growth from in situ compounded melts. The normal
growth process conditions used for already precompounded charges were found to result in incomplete
melt back, no supercooling and a growth velocity
slightly less than that of the furnace translation rate.
The sensor approach could be useful for perfecting
the growth conditions of current processes. It also
creates the possibility of using in situ sensing to
control seeded solidi"cation processing.
Acknowledgements
Fig. 10. Photographs of the cone area of two Cd
Zn
Te
0.96 0.04
ingots: (a) corresponds to growth from a completely melting
state, (b) was obtained after the incomplete melting run 3.
(Photographs of the ingot provided by courtesy of Johnson and
Matthey Electronics.)
The results above indicate that to avoid unstable
solidi"cation it is necessary to ensure a charge is
completely melted at its tip without incurring large
superheats elsewhere in the melt and therefore large
undercoolings during subsequent solidi"cation.
Either a lower thermal gradient to prevent superheating [21] or the use of a seed crystal to prevent
undercooling during solidi"cation could be used to
improve the growth process. In the latter case, the
use of in situ sensing of the remelting melting phase
during heating might simplify control of seed tip
remelting and then seeded regrowth. Careful redesign of thermal environment might then allow
a large grained ingot to grow by a seeded growth
method [21].
This work has been performed as a part of the
research of the Infrared Materials Producibility
Program conducted by a consortium that includes
Johnson Matthey Electronics, Texas Instruments,
II}VI Inc., Loral, the University of Minnesota and
the University of Virginia. We are grateful for the
many helpful discussions with our colleagues in
these organizations and inparticular to the sta! of
JME for their assistance in preparing the samples.
The consortium work has been supported by
ARPA/CMO under contract MDA972-91-C-0046
monitored by Raymond Balcerak.
References
[1] S. Sen, W.H. Konkel, S.J. Tighe, L.G. Bland, S.R. Sharma,
R.E. Taylor, J. Cryst. Growth 86 (1988) 111.
[2] F. Butler, F.P. Doty, B. Apotovsky, S.J. Friesenhahn,
C. Lingren, in: R.B. James, T.E. Schlesinger, P. Si!ert,
F. Franks (Eds.), Semiconductors for Room Temperature
Radiation Detector Applications, Materials Research Society, 1993, p. 502.
230
B.W. Choi, H.N.G. Wadley / Journal of Crystal Growth 208 (2000) 219}230
[3] R. Triboulet, Y. Marfaing, J. Electrochem. Soc. 120 (9)
(1973) 1260.
[4] P. Rudolph, M. Muhlberg, Mater. Sci. Eng. B 16 (1993) 8.
[5] M. Muhlberg, P. Rudolph, M. Laasch, E. Treser, J. Cryst.
Growth 128 (1993) 571.
[6] R.J. Naumann, S.L. Lehoczky, J. Cryst. Growth 61 (1983)
707.
[7] T. Wu, W.R. Wilcox, J. Cryst. Growth 48 (1980) 416.
[8] S. Brandon, J.J. Derby, J. Cryst. Growth 110 (1991) 481.
[9] Art Socha, Private Communication at Johnson Matthey
Electronics in Spokane, WA, 1995.
[10] Pok-kai Lio, Private Communication at Texas Instruments, Texas, TX, 1994.
[11] H.N.G. Wadley, K.P. Dharmasena, J. Cryst. Growth 130
(1993) 553.
[12] H.N.G. Wadley, B.W. Choi, J. Cryst. Growth 172 (1997)
323.
[13] K.P. Dharmasena, H.N.G. Wadley, J. Cryst. Growth 172
(1997) 337.
[14] G. Rosen, Ph.D. Dissertation, Clarkson University, 1994.
[15] H.N.G. Wadley, A.H. Kahn, W. Johnson, Mater. Res. Soc.
Symp. Proc. 117 (1988).
[16] J.P. Wallace, J.K. Tien, J.A. Stefani, K.S. Choe, J. Appl.
Phys. 69 (1991) 550.
[17] V.M. Glazov, S.N. Chizhevskaya, N.N. Glagoleva, Liquid
Semiconductors, Plenum Press, New York, 1969.
[18] Ed. B.E. MacNeal, MacNeal-Schwendler Corporation,
MSC/EMAS Modelling Guide, 1991.
[19] S. Kuppurao, S. Brandon, J.J. Derby, J. Cryst. Growth 155
(1995) 93.
[20] R. Route, M. Wolf, R.S. Feigelson, J. Cryst. Growth 70
(1984) 379.
[21] S. Kuppurao, S. Brandon, J.J. Derby, J. Cryst. Growth 158
(1996) 459.