Research on new method of electron beam
candle melting used for removal of P from
molten Si
D. C. Jiang1,2, Y. Tan1,2, S. Shi1,2, Q. Xu3, W. Dong*1,2, Z. Gu1,2 and R. X. Zou1,2
A new method, electron beam candle melting (EBCM), is proposed for the removal of P in molten
Si, to produce high quality material such as solar grade silicon for photovoltaic applications.
EBCM is designed to overcome the shortcomings of electron beam melting while utilising the high
saturated pressure of P in molten Si to effect refining. The experimental result showed that it could
remove P from Si effectively; in addition, the energy utilisation ratio was experimentally proved to
be high. The evaporation coefficient of P removal is in a reasonable region and comparable with
the theoretical value, which indicates that EBCM is a feasible method for the removal of P in
molten Si in low power.
Keywords: Electron beam, Phosphorus, Molten silicon, Evaporation coefficient, Solar grade silicon
This paper is part of a special issue on Energy Materials
Introduction
With the rapid development of the photovoltaic industry,
the requirements for solar grade silicon (SoG-Si) have
been increased dramatically in recent years. Metallurgical
methods to obtain SoG-Si have been given more
attention because of their low cost, low energy consumption and minimal environmental impact. In order to
avoid electron–hole recombination in silicon solar cells,
phosphorus needs to be removed from metallurgical
grade silicon (MG-Si). However, P can be removed
effectively at high temperature and in high vacuum due to
its relatively high saturated vapour pressure. Pires et al.
confirmed the technical feasibility of purifying Si by
electron beam melting (EBM) method, and the P removal
rate was reported to reach 98?2%.1,2 The EBM was
applied in refining 150 kg of MG-Si, and the P content was successfully reduced from 2?561023 to
,161025 wt-%.3
However, wide application of EBM in industrial
production is limited by some unresolved problems,
such as low energy utilisation ratio caused during
refining by the heat exchange between the water cooled
copper crucible and the molten Si.
Yuge et al.4 thought that the P removal rate was
controlled by free evaporation from the molten Si surface.
P can only leave from the surface of the molten pool,
and thus, P in molten Si should be diffused to
the surface of the molten pool for removal. Therefore,
the removal of P from molten Si mainly depends on the
surface area and the depth of the molten pool at a constant
temperature and a certain vacuum condition. Increasing
the surface area of the molten pool and decreasing the
depth of the molten pool are considered to increase the
removal rate of P and the energy utilisation ratio.
In this paper, a new method is proposed according to
the above consideration. The P removal rate is
confirmed by experimental results, and its feasibility is
explained theoretically.
Process of electron beam candle melting
(EBCM)
As shown in Fig. 1, EBCM, which has a large surface
area and a shallow molten pool, comprises several steps
in one cycle. Initially, a low energy electron beam is used
to melt the surface of the Si ingot. Irradiation is
maintained for several minutes until a desired amount
of P is removed. Afterwards, the electron beam energy is
increased to melt the top edge of the Si ingot. Liquid Si
with low P content flows down the sides of the ingot
sidewall into the receptacle (copper crucible).
The cycle is repeated until the P content in the whole
ingot is reduced to the targeted value. This process
resembles the burning of a candle, which explains why
the process is named EBCM.
Experimental
1
School of Materials Science and Engineering, Dalian University of
Technology, No. 2 Linggon Road, Ganjingzi District, Dalian 116023, China
2
Key Laboratory for Solar Energy Photovoltaic System of Liaoning
Province, Dalian 116023, China
3
Center of Analytical Chemistry, Faculty of Chemical, Environmental and
Science and Technology, Dalian University of Technology, No. 2 Linggon
Road, Ganjingzi District, Dalian 116023, China
*Corresponding author, email [email protected]
ß W. S. Maney & Son Ltd. 2011
Received 23 April 2011; accepted 11 October 2011
DOI 10.1179/1433075X11Y.0000000026
The Si ingot prepared by directional solidification was
used in the experiment. The remaining solid part of the
ingot (70658673 mm, 729 g) was used for EBCM
studies. The P content of the ingot used in EBCM was
1?4461022 wt-%, which is obtained by inductively
coupled plasma mass spectrometry.
Materials Research Innovations
2011
VOL
15
NO
6
406
Jiang et al.Jiang et al.
Electron beam candle melting for removal of P from molten Si
1 Process of EBCM
In the pretreatment, the Si ingot was washed with
ethyl alcohol to remove the oil and organic compounds
adhering to its surface, whereas 5 mol.-% hydrofluoric
acid was used to remove the metal impurities on its
surface. Then, the ingot was washed in deionised water
and exposed to supersonic waves until the deionised
water became neutral.
Figure 2 shows the schematic diagram of the EBCM
equipment used in the present study. The diagram
consists primarily of three parts: an electron beam gun, a
vacuum system and a water cooled copper crucible. The
nominal power of the electron beam gun is 30 kW when
the accelerating voltage is set to 20 kV. Various beam
trajectories or beam patterns can be obtained by
controlling the strength of the magnetic field in the x
and y directions. Two independent vacuum systems are
employed in the furnace: one for the gun chamber and
the other for the melt chamber.
The ingot was placed in the water cooled copper
crucible, and the chamber was evacuated to 261023 Pa.
The electron beam pattern was a circle with a diameter
of ,25 mm, and the centre of the beam pattern was
adjusted to coincide with the centre of the top surface of
the ingot. The electron beam voltage was set to 30 kV.
Compared with EBCM, the beam density was set to
200 mA and was required to be over 500 mA for melting
the same size of Si ingot during EBM.5,6 Irradiation with
the electron beam caused the top surface of the ingot to
melt gradually and form a molten pool, which can be
seen through the observation window. The upper part of
the ingot was orange red, whereas the lower part was
darker. The electron beam was sustained until the top
surface of the ingot was melted completely. However,
the molten Si remained on the top surface due to surface
tension. With an extended irradiation, the surface
tension cannot maintain the molten Si on the top
surface, which began to flow down the sides and
solidified around the bottom of the ingot in the crucible.
At this point, the unmelted part of the ingot emerged,
which is regarded as the melting cycle. Electron beam
irradiation was continued until the ingot melted
completely in the crucible.
Results and discussion
Figure 3a and b shows the Si ingot morphology before
and after EBCM respectively. During EBCM, no
splashing was observed through the observation window. After solidification, the ingot had a button shape
and a high surface luster, and only a thin layer of the
melted part can be observed on the crucible inner wall.
However, part of the Si in the edge did not melt
completely because the ingot was a cube, and the
electron beam pattern was a circle. This problem can be
solved using a cylindrical ingot.
During EBCM melting, the temperature of cooling
water in the inlet and outlet hardly changed because the
molten Si did not get in touch with the water cooled
copper crucible directly, compared with the EBCM, the
temperature of the cooling water increases with the
increase in power and the temperature difference
between the inlet and outlet of cooling water was
increased by 1 K when the power increased by 3 kW,7 so
the energy loss was less and the energy utilisation ratio
became high during EBCM.
As is known, the surface temperature of molten Si
during EBCM can be roughly estimated by the Hertz–
Knudsen–Langmuir equation (equation (1)) as follows
VSi ~(MSi =2pRT)1=2 PSi
(1)
where PSi is the vapour pressure (Pa) of Si, MSi is the
atomic weight of Si (g mol21), T is the free surface
2 Schematic diagram of EBCM equipment
a before EBCM; b after EBCM
3 Silicon ingot morphology
Materials Research Innovations
2011
VOL
15
NO
6
407
Electron beam candle melting for removal of P from molten Si
Jiang et al.Jiang et al.
ln
½%At
S
~{k ½%A
V
½%A0
(5)
where [%A]0 is the initial mass percentage of the solute.
In the experiment, [%A]0 and [%A]t can be obtained so a
straight line can be constructed by plotting ln [%A]t/
[%A]0 against time. The k value can be obtained by
calculating the slope of equation (5), and the result is
shown in Fig. 4.
In theoretical calculation, the temperature of the
molten surface is non-uniform in EBCM, so the
evaporation flux J (mol cm22 S21) should be obtained
first. The evaporation flux can be expressed by the
Langmuir equation as follows
J~ 4 Relationship between [P]/[P]0 and melting time
temperature (K) of the molten Si and R is gas
constant(J mol21 K21).
VSi can be expressed by equation (2)
VSi ~DmSi =At
(2)
where DmSi is the loss of Si during melting (g), A is the
surface area of the molten pool (m2) and t is melting
time (s).
PSi can be expressed by equation (3)
log PSi ~{20900=T{0:565| log Tz12:9
(3)
Equation (3) is supported by a database collected by
Kubachewski and Alock.8 The surface temperature of
molten Si during EBM at different powers was also
listed in the table from our previous work.7
As shown in Table 1, the average surface temperature
at the power of 6 kW during EBCM is comparable to
that at the power of 9 kW during EBM, as for this case,
nearly 30% energy estimated to be saved.
The content of P in Si after each cycle is shown in
Table 2. P is removed by ,60% at each cycle. With an
increase in melting time, the content of P is decreased,
which is the same as the observation of Ikeda and
Maeda.9 As 200 mA during EBM is not high enough to
melt the Si ingot completely,7 the energy utilisation ratio
during EBCM increased obviously.
The evaporation coefficient k (m s21) is used to
discuss the removal ratio of P from molten Si during
EBCM.
k was obtained from the theoretical calculations and
experiment performed. In the experiment, [%A] is the
mass percentage of the solute, in which the evaporation
of monatomic P is assumed. Therefore, the relationship
between the evaporation rate and [%A] can be written in
differential form as follows3
d½%A
S
~k ½%A
(4)
{
dt
A
where V is the volume of the molten pool (m3), and S is
the evaporation area (m2). Equation (4) can then be
integrated from 0 to t
pp
2pMp RT
(6)
1=2
where Pp is the equilibrium vapour pressure of P above
the molten Si surface (Pa), Mp is the molar mass of P, R
is the gas constant and T is the temperature of the
molten surface (K).
The temperature distribution of EBM in the region
approximates a normal distribution.10 Based on the
normal distribution, the relationship between the
temperature and radius of the region can be expressed
as follows
"
#
1
(r{u)2
T~
zb
(7)
exp {
2s2
ð2pÞ1=2 s
where u is the coordinate of the top surface temperature,
s is the variance in temperature and b is the temperature
at r50 and r5R, which are the melting points of Si.
The relationship between the evaporation flux and the
radius can be obtained from equations (6) and (7) as
follows
Pp ðsÞ1=2
J~ 1=2
:
Mp R
ð2pÞ0 25
(
"
#
){0:05
ðr{R=2Þ2
1=2
(8)
zTm ð2pÞ s
exp {
2s2
where Tm is 1687 K, which is the melting point of Si.
The equilibrium vapour pressure of P in the molten Si
surface can be determined using
Pp ~Xp cP Pop (T)
(9)
where Pop is the partial pressure of monatomic phosphorus (Pa), which is in equilibrium with pure liquid P,
cP is the activity coefficient of P, Xp is the mole fraction
of P on the surface of molten Si and Pop is a function of
temperature, which can be obtained from the equation
Table 2 Content of P in Si after each melting cycle
Melting time/s
Content/wt-%
First
cycle
Second
cycle
Third
cycle
Fourth
cycle
350
0.0034
300
0.0057
250
0.0065
270
0.0063
Table 1 Average surface temperature of molten Si during EBCM and EBM
Average temperature/K
EBCM (6 kW)
EBM (9 kW)
EBM (15 kW)
EBM (21 kW)
1936
1940
1978
2035
Materials Research Innovations
2011
VOL
15
NO
6
408
Jiang et al.Jiang et al.
Electron beam candle melting for removal of P from molten Si
considered, whereas P2 is neglected, and the deviation in
surface area between the actual and calculated values
has been observed.
As k is close is to the theoretical value of P
evaporation from Si, P can be removed effectively
during EBCM. However, the melting power and depth
of the molten pool during EBCM are less than those
during EBM, so the energy utilisation ratio is relatively
high. To increase the removal rate further, the melting
time should be increased.
Conclusions
5 Relationship
values of k
between
theoretical
and
experimental
1=2
of Takahiro et al.11 ðsÞ1=2 = Mp R
ð2pÞ1=4 is defined as
c, so equation (8) can be expressed as follows
#
( "
){1=2
ðr{R=2Þ2
1=2
zTm ð2pÞ s
(10)
J~cPp (r) exp {
2s2
The amount of P that evaporated from molten Si in unit
time M (mol s21) can be obtained by integration from 0
to the whole surface area
(
ð pR2
M~
"
cPp (r) exp {
0
){1=2
#
ðr{R=2Þ2
1=2
ð
2p
Þ
s
ds
zT
m
2s2
(11)
(11)
Equation (11) can be transformed into equation (12)
with u5(r2R/2)/s
ð R=2s
M~
0
{1=2
u2
zTm ð2pÞ1=2 s
2psRcPp (r) exp {
du
2
(12)
(12)
The temperature distribution of the molten surface was
obtained from a previous work.7 The M obtained by
Taylor expansion is 461024 mol s21, and the evaporation flux in EBCM is defined as
J~M=S
(13)
In the evaporation process, J can be described by
equation (11), from which k (m s21) can be obtained, i.e.
J~kXP
(14)
The values of k obtained by the experiment and
theoretical calculation are shown in Fig. 5. Compared
with previous work, the relationship between the
temperature and k of P and P2, which was investigated
by Takahiro et al., is also shown in this figure.11
As shown in Fig. 5, the k of P2 is larger than that of P
at a certain temperature. The results of Takahiro et al.
show that if the P concentration in Si is ,561023 wt-%,
then the monatomic P vapour dominates in the gas
phase. In this experiment, the P concentration in Si was
1?4461022 wt-%, and both P and P2 are suggested to
exist in the gas mixture, which resulted in a k value
between P and P2 k values.
The theoretical value is less than the experiment value,
which may only be the volatilised P. A monatomic P is
Electron beam candle melting for the removal of P in
molten Si is proposed and discussed in detail. The
method combines the characteristics of EBM and the
highly saturated pressure of P in molten Si. The
evaporation coefficient of P during EBCM obtained
from theory and experiment is close to the theoretical
value. The energy utilisation ratio was enhanced in
EBCM due to the shallow depth of the molten pool and
the low melting power. The removal ratio should be
increased by extending the melting time further.
Acknowledgements
The authors gratefully acknowledge the financial support from the National Natural Science Foundation of
China (grant no. 51074032) and the Liaoning Province
Key Technologies R&D Program of China (grant
no. 2006222007).
References
1. J. C. S. Pires, A. F. B. Braga and P. R. Mei: ‘Profile of impurities in
polycrystalline Si samples purified in an electron beam melting
furnace’, Sol. Energy Mater. Sol. Cells., 2003, 79, 347–355.
2. J. C. S. Pires, J. Otubo, A. F. B. Braga and P. R. Mei: ‘The
purification of metallurgical grade Si by electron beam melting’,
J. Mater. Process. Technol., 2005, 169, 16–20.
3. K. Hanazawa, N. Yuge and Y. Kato: ‘Evaporation of phosphorus
in molten Si by an electron beam irradiation method’, Mater.
Trans., JIM, 2004, 45, 844–849.
4. N. Yuge, K. Hanazawa and Y. Kato: ‘Removal of metal impurities
in molten Si by directional solidification with electron beam
heating’, Mater. Trans., JIM, 2004, 45, 850–857.
5. D.-C. Jiang, W. Dong, Y. Tan, Q. Wang, X. Peng and G.-B. Li:
‘Investigation on removal of aluminum impurity in metallurgical
grade silicon by electron beam melting’, J. Mater. Eng., 2010, 8, 8–
11.
6. X. Peng, W. Dong, Y. Tan, D.-C. Jiang, Q. Wang and G.-B. Li:
‘Evaporation behavior of calcium in metallurgical silicon by an
electron beam melting method’, J. Funct. Mater., 2010, 41, 117–
120.
7. D.-C. Jiang, Y. Tan, W. Dong, H. Shi, Z. Gu and R. X. Zou: ‘The
removal of phosphorus in molten silicon by electron beam candle
melting’, Mater. Lett., to be published.
8. O. Kubachewski and C. B. Alock: ‘Metallurgical thermochemistry’,
5th edn, 372; 1979, New York, Pergamon Press Ltd.
9. T. Ikeda and M. Maeda: ‘Purification of metallurgical Si for solar
grade Si by electron beam button melting’, ISIJ Int., 1992, 32, 635–
642.
10. D. M. Maijer, T. Ikeda, S. L. Cockcroft, M. Maeda and R. B.
Rogge: ‘Mathematical modelling of residual stress formation in
electron beam remelting and refining of scrap Si for the production
of solar-grade Si’, Mater. Sci. Eng. B, 2005, B390, 188–201.
11. M. Takahiro, M. Kazuki and S. Nobuo: ‘Thermodynamics of
phosphorus in molten Si’, Metall. Mater. Trans. B, 1996, 27B, 937–
941.
Materials Research Innovations
2011
VOL
15
NO
6
409