Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

ARTICLE IN PRESS

Journal of Crystal Growth 311 (2009) 684–687

Contents lists available at ScienceDirect

Journal of Crystal Growth


journal homepage: www.elsevier.com/locate/jcrysgro

Segregation control of vertical Bridgman growth of Ga-doped germanium


crystals by accelerated crucible rotation: ACRT versus angular vibration
L.C. Wang a, Y.C. Liu a, W.C. Yu b, B. Roux c, T.P. Lyubimova d, C.W. Lan a,e,
a
Department of Chemical Engineering, National Taiwan University, Taipei, Taiwan 10617, ROC
b
Department of Molecular Science and Engineering, National Taipei University of Technology, Taipei, Taiwan, ROC
c
Laboratoire Modélisation et simulation numérique en mécanique, L3M/FRE 2405 MSNM: CNRS—Universités d’Aix-Marseille, France
d
Institute of Continuous Media Mechanics UB RAS, Perm, Russia
e
Photovoltaics Technology Center, Industrial Technology Research Institute, Hsinchu, Taiwan

a r t i c l e in f o a b s t r a c t

Available online 27 September 2008 The control of dopant segregation by accelerated crucible rotation was investigated for the vertical
Keywords: Bridgman growth of Ga-doped germanium crystals. The accelerated crucible rotation technique (ACRT)
A1. Accelerated crucible rotation and the angular rotation technique (AVT) were considered. Both techniques were found effective in
A1. Angular vibration radial segregation control; however, ACRT generated more global mixing and thus slightly larger axial
A1. Convection segregation. Computer simulation was also carried out for the experiments. The results were also
A1. Segregation consistent with previous simulation predictions and low-temperature transparent experiments.
B2. Bridgman method & 2008 Elsevier B.V. All rights reserved.

1. Introduction demonstrated for a low-temperature model material, and com-


puter simulation has also been carried out for illustration [7].
The control of convection and segregation is important in Although numerical simulations and low-temperature trans-
vertical Bridgman (VB) crystal growth. Because of a lack of control parent have been carried out for both ACRT and AVT [7–9], they
over stirring conditions, the use of external forces is often have not yet implemented and compared directly on semicon-
adopted. The accelerated crucible rotation technique (ACRT) ductor growth. In this paper, we present experimental results for
proposed by Scheel and Schulz-DuBois [1] is a promising way to using both techniques on the segregation control of a VB growth
introduce a forced convection, i.e., the Ekman flow, and thus the of gallium (Ga)-doped germanium (Ge) crystals. In the next
melt mixing. The control of rotation scheme, especially the cycle section, the experimental setup and procedures are briefly
period, is crucial to the flow structures and thus the segregation described. Section 3 is devoted to results and discussion followed
during crystal growth [2]. As a general rule, in order to develop the by short conclusions in Section 4.
Ekman boundary layer, the time for affi spin-up or spin-down cycle
pffiffiffiffiffiffiffiffiffiffiffiffi
is usually in the order of 2RC = OR nm [3,4], where RC is the crystal
radius, OR the rotation amplitude, and nm the kinematic viscosity 2. Experimental setup and procedures
of the melt. This time constant, as a quarter of the cycle period, is
usually several seconds for most of the materials having a The growth of Ga-doped Ge crystals in a quartz ampoule
diameter of several centimeters. Although ACRT is an effective coated with boron nitride was carried out in a directional
way in flow control, the control of radial segregation often solidification furnace. The schematic sketch of the experimental
requires an intensive flow, which leads to significant growth setup is shown in Fig. 1, where the measured temperature profile
striations. An alternative approach to ACRT is to use a cycle time is attached. The boron nitride coating was used to avoid the
that is much shorter than the Ekman time. This method is known reaction between the germanium and the quartz ampoule. Before
as the angular vibration technique (AVT) [5,6]. In this technique, the growth, a (111) seed of 18 mm in diameter and 40 mm in
the ampoule is vibrated in the angular direction at high frequency length was placed at the bottom of the ampoule. Then, Ga-doped
(typically greater than 5 Hz). The control of solute segregation and Ge pellets were placed in the ampoule, and vacuum sealed under
the morphological stability using this technique has been an argon pressure of 0.08 Torr.
The crystal growth experiments were carried out in a furnace
having a thermal gradient at the seed/feed interface of about
 Corresponding author at: Department of Chemical Engineering, National 20 K/cm, which was controlled by a hot zone and a water-cooled
Taiwan University, Taipei, Taiwan 10617, ROC. Tel./fax: +886 2 2363 3917. cold zone temperatures. Before crystal growth, the ampoule was
E-mail address: cwlan@ntu.edu.tw (C.W. Lan). heated up at a constant rate of 10 1C/min and then was kept at

0022-0248/$ - see front matter & 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jcrysgro.2008.09.087
ARTICLE IN PRESS

L.C. Wang et al. / Journal of Crystal Growth 311 (2009) 684–687 685

Fig. 1. A sketch of experimental setup, thermal profile, and the appearance of grown crystals: (a) no crucible rotation; (b) 10 Hz AVT; and (c) 60 rpm ACRT. The arrow in the
photographs indicates the interface after back melting.

1200 1C in the hot zone for 12 h to ensure a homogenous 1.0


dopant mixing in the melt. To start crystal growth, the ampoule No crucible rotation
was pulled down at 5.8 cm/h from for 30 mm at the beginning 0.9
10 Hz AVT
and then at 17.4 cm/h for the rest of the growth. Finally, the
furnace was cooled down slowly to room temperature in 14 h. 0.8 60 rpm ACRT
The same growth condition was applied to all the cases with and
without accelerated crucible rotation. The ACRT cycle scheme and 0.7
AVT amplitude are similar to the previous low-temperature
0.6
experiments [7–9]. Each ACRT cycle consisted of 5 s acceleration,
C/C0 (-)

5 s deceleration, and 2 s rest. For AVT, the amplitude of vibration


0.5
was set to be 0.15.
After crystal growth, the ingots were taken out and sliced into 0.4
a few wafers along the growing direction. These wafers were
polished and then etched for 10–30 s by an acid solution 0.3
(H2O2:CH3COOH:HF ¼ 1:1:1). With this etching procedure, the
seed/crystal interface could be revealed [10]. The axial and radial 0.2
dopant concentration profiles were determined by the resistivity
r measured by the four-point probe technique using the following 0.1
equation [11]:
0.0
r ¼ N a =B, (1) -1 -0.5 0 0.5 1
where N stands for the actual impurity concentration in r (cm)
atoms/cm3. The parameters a and B vary when the impurity Fig. 2. Comparison of radial dopant concentration profiles at the fraction 0.5 for
concentration switches into different orders. In our case, for the the three crystals.
impurity concentration of 1018oNo1019, the values of a and B are
0.653 and 1.61 1010, respectively.

case of no crucible rotation. However, in the case of 60 rpm ACRT,


3. Results and discussion due to the longer acceleration and deceleration time in each cycle
(24 s), striation-like appearance was observed from the rougher
To examine the effect of accelerated crucible rotation on radial surface.
and axial segregations during crystal growth, Ga-doped Ge single Because of the difference in the growth lengths from run-to-
crystals grown with and without crucible rotation were compared run, the radial dopant concentration profiles were compared on
in terms of their appearance, and the radial and axial dopant the basis of its solidification fraction of the entire growth.
profiles. Three photos of the grown crystals with different rotation Fractions of 0.5 and 0.8 were chosen to represent the radial
conditions are provided at the right corner of Fig. 1. The segregation of slow and fast growth stages for each case.
demarcation of the back-melting interface and the growth area The radial dopant profiles at fraction 0.5 for the three cases are
can be clearly identified from the appearance of each grown compared in Fig. 2. The values of the concentration in Fig. 2 were
crystal. Because the thermal gradient was only 20 K/cm, the normalized by their doping concentration C0. It was supposed that
amount of back melting, and so as the starting point of the growth without crucible rotation yielded the highest con-
the growth, was difficult to control. In the case of 10 Hz AVT, the centration at the center because the toroidal thermal convection
appearance of the grown crystal has no much difference from the near the concave interface induces solutal accumulation there
ARTICLE IN PRESS

686 L.C. Wang et al. / Journal of Crystal Growth 311 (2009) 684–687

[12]. However, as shown in Fig. 2, due to the asymmetric growth, 3


the highest dopant concentration moved to the right. In the case No crucible rotation
of 10 Hz AVT, the uniformity of radial concentration was slightly 10 Hz AVT
improved, and the values of C/C0 distributed from 0.3 to 0.75, but 2.5
not effective. In the case of 60 rpm ACRT, the overall concentration 60 rpm ACRT
was dramatically reduced to about 0.35, and the profile was the
most uniform among the three cases. It should be noted that the 2
radial dopant profiles in the case of no crucible rotation and 10 Hz
AVT were both asymmetric. Indeed, with a long melt height,

C/C0 (-)
crucible tilting and asymmetric heating were difficult to avoid, 1.5
and they led to asymmetric flow and dopant fields [13]. Such an
asymmetry often enhanced dopant mixing, and thus reduced the
radial segregation. The effect of Schlitching flow generated by AVT 1
could be affected by the asymmetric flow as well, and this might
explain why AVT was not as effective as we anticipated in the
axisymmetric ones [6]. 0.5
Fig. 3 shows the radial dopant profile at fraction 0.8 of each
grown crystal. Without crucible rotation, the dopant accumulated
more severely at the center in the grown crystal. Such a dramatic 0
distribution indicated that the radial segregation was severely 0 0.2 0.4 0.6 0.8
developed in the fast growth. Due to the shorter melt height, the fraction (-)
dopant profile became more symmetric as a result of the more
axisymmetric flow. Also, in the case of 10 Hz AVT, C/C0 distributed Fig. 4. Axial dopant concentration profiles along the centerline of the grown
crystals.
from 0.9 to 2.2 indicating a much smaller radial segregation; the
10 Hz AVT yielded the most uniform profile among the three
cases. In the case of 60 rpm ACRT, C/C0 distributed from 1.0 to 3.0
and the improvement on radial segregation was also obvious. acceleration and deceleration times that developed larger Ekman
In Fig. 4, the axial dopant profiles along the centerline of the flow and generated stronger global mixing. Therefore, a larger
grown crystals are plotted for comparison. The values were also axial segregation was found in the case of ACRT. These axial
normalized by their doping concentration and compared on the dopant profiles should be compared with care because the
basis of solidification fraction. In the case of no crucible rotation, measurement along the centerline might miss the highest values
the concentration profile was very similar to that shown in when the radial profile was not axisymmetric.
Ref. [10], which laid between the theoretical curves of no Computer simulation using an axisymmetric model [7–9] was
convection and complete mixing. With 10 Hz AVT, the axial also conducted for the previous growths, and the flow and dopant
segregation was not changed much, which was consistent with fields are shown in Fig. 5 for comparison; the gallium diffusion in
the previous simulation predictions [6,7]. In the case of 60 rpm the solid was neglected. Because of the asymmetric growth
ACRT, the curve tended to be closer to the complete mixing conditions in reality, the one-to-one comparison is not possible.
and thus increased the axial segregation. It can also be seen from Nevertheless, from the simulated flow structures, it is clear that
Fig. 2 that the significant reduction of overall radial concentration both AVT and ACRT generate a small flow cell near the growth
at the beginning for the case of ACRT was caused by the longer front. Because their flow direction is opposite to the thermal
convection, as a result, the dopant accumulation at the crystal
center caused by the thermal convection is significantly reduced.
7.0 As a result, the radial segregation with AVT and ACRT is improved,
No crucible rotation which can be seen from the dopant fields in the crystal. Also, from
10 Hz AVT the maximum value of the normalized dopant concentration in
6.0
60 rpm ACRT the crystal, the one with ACRT is also significantly lower (0.3081)
than the others. This indicates that ACRT generates a stronger bulk
5.0 dopant mixing than AVT, which is consistent with the measured
axial dopant profile in Fig. 4. The maximum dopant concentration
of AVT lies in the middle, which is consistent with the measured
4.0 one as well. Interestingly, from the dopant fields of ACRT shown in
C/C0 (-)

Fig. 5c, significant dopant fluctuations are observed, which


indicates that growth striations may be generated by ACRT if
3.0
the solid state diffusion is not fast enough to smooth out the
inhomogeneity. Indeed, some striations were found in the crystal
2.0 grown with 60 rpm ACRT.

1.0 4. Conclusions

In this paper, we have compared the vertical Bridgman growth


0.0 of Ga-doped Ge crystals with different conditions of accelerated
-1 -0.5 0 0.5 1
crucible rotations: no crucible rotation, 10 Hz AVT, and 60 rpm
r (cm) ACRT. Both AVT and ACRT were found effective in improving radial
Fig. 3. Comparison of radial dopant concentration profiles at the fraction 0.8 for segregation for both low and fast growths. In our observation, AVT
the three crystals. reduced radial dopant segregation and meanwhile generated less
ARTICLE IN PRESS

L.C. Wang et al. / Journal of Crystal Growth 311 (2009) 684–687 687

No Rotation 10 Hz AVT 60 rpm ACRT

Ψmax = 0.000000 Ψmax = -0.02474 Ψmax = -0.01147


6.5 Ψmin = 0.230267 Ψmin = 0.22036 Ψmin = 0.05925

k(C/C0) (-) k(C/C0) (-) k(C/C0) (-)


6 0.58765 0.38804 0.30811
0.55428 0.36797 0.29336
0.52090 0.34790 0.27861
0.48752 0.32783 0.26386
0.45415 0.30776 0.24911
5.5
z (cm)

0.42077 0.28769 0.23435


0.38739 0.26762 0.21960
0.35401 0.24755 0.20485
0.32064 0.22748 0.19010
5 0.28726 0.20741 0.17535
0.25388 0.18734 0.16060
0.22051 0.16727 0.14585
0.18713 0.14720 0.13109
4.5 0.15375 0.12713 0.11634
0.12038 0.10706 0.10159

3.5
0 1
r (cm)

Fig. 5. Simulated melt flow fields (streamlines) and dopant fields in the crystal: (a) no rotation; (b) 10 Hz AVT; and (c) 60 rpm ACRT.

global mixing. Due to the longer acceleration and deceleration References


times in the rotation cycle, ACRT generated more global dopant
mixing and thus higher axial segregation as compared with AVT. [1] H.J. Scheel, E.O. Schulz-Dubois, J. Crystal Growth 8 (1971) 304.
Computer simulation was also conducted and the results were [2] P. Capper, J.J.G. Gosney, C.L. Jones, J. Crystal Growth 70 (1984) 356.
[3] J.C. Brice, Crystal Growth Process, Wiley, New York, 1986.
consistent with the measured ones, as well as the previous [4] A. Yeckel, J.J. Derby, J. Crystal Growth 209 (2000) 734.
predictions from the low-temperature experiments and simula- [5] W.C. Yu, Z.B. Chen, W.T. Hue, B. Roux, T.P. Lyubimova, C.W. Lan, J. Crystal
tion. Growth 271 (2004) 474.
[6] C.W. Lan, J. Crystal Growth 274 (2005) 379.
[7] W.C. Yu, Z.B. Chen, W.T. Hue, B. Roux, T.P. Lyubimova, C.W. Lan, J. Heat Mass
Trans. 50 (2007) 58.
Acknowledgements
[8] Y.C. Liu, B. Roux, C.W. Lan, J. Crystal Growth 304 (2007) 236.
[9] Y.C. Liu, W.C. Yu, B. Roux, T.P. Lyubimova, C.W. Lan, Chem. Eng. Sci. 61 (2006)
This work was sponsored by the National Science Council 7766.
(NSC) of Taiwan and National Taiwan University. Partial support [10] P. Dold, F.R. Szofran, K.W. Benz, J. Crystal Growth 234 (2002) 91.
[11] S.M. Sze, J.C. Irvin, Solid-State Electron. 11 (1968) 599.
from the Orchid program between NSC and PIA of France and the [12] P.M. Adornato, R.A. Brown, J. Crystal Growth 80 (1987) 155.
Russian Foundation for Basic Research was appreciated. [13] M.C. Liang, C.W. Lan, J. Crystal Growth 167 (1996) 320.

You might also like