Effect of calcium-oxide on the removal of calcium during industrial directional solidification of upgraded metallurgical-grade silicon

Author's Accepted Manuscript

Effect of calcium-oxide on the removal of calcium during industrial directional solidification of upgraded metallurgical-grade silicon C.H. Gan, X. Zeng, M. Fang, L. Zhang, S. Qiu, J.T. Li, D.C. Jiang, Y. Tan, X.T. Luo

www.elsevier.com/locate/jcrysgro

PII: DOI: Reference:

S0022-0248(15)00434-0 http://dx.doi.org/10.1016/j.jcrysgro.2015.06.004 CRYS22889

To appear in:

Journal of Crystal Growth

Received date: 18 April 2015 Revised date: 5 June 2015 Accepted date: 6 June 2015 Cite this article as: C.H. Gan, X. Zeng, M. Fang, L. Zhang, S. Qiu, J.T. Li, D.C. Jiang, Y. Tan, X.T. Luo, Effect of calcium-oxide on the removal of calcium during industrial directional solidification of upgraded metallurgical-grade silicon, Journal of Crystal Growth, http://dx.doi.org/10.1016/j.jcrysgro.2015.06.004 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

1

1

Effect of calcium-oxide on the removal of calcium during industrial directional

2

solidification of upgraded metallurgical-grade silicon

3

C.H. Gan a,c, X. Zeng a, M. Fang b,c, L. Zhang a,c, S. Qiu a,c, J.T. Li a, D.C. Jiang b,c, Y. Tan b,c,

4

X.T. Luo a*

5

a. Fujian Provincial Key Laboratory of Advanced Materials, College of Materials, Xiamen

6

University, Xiamen 361005, China;

7

b. School of Material Science and Engineering, Dalian University of Technology, Dalian

8

116023, China;

9

c. Qingdao Longsun Silicon Technology Company Ltd, Qingdao 266200, China

10 11 12 13 14 15 16 17 18 19 20 21 22 23

1

1

 Corresponding author: X.T Luo (E-mail: [email protected], Tel: +86-592-2188503)

2

24

ABSTRACT

25

Directional solidification is often used to remove metallic impurity in the photovoltaic

26

industry for the low equilibrium distribution coefficient between solid and melt. However, in

27

our present experiments, compared with other impurities, the removal of calcium is variable

28

at the low height of ingot, which is caused by the existence of insoluble CaO particle. CaO

29

exists as insoluble particle in the feedstock. During directional solidification stage, CaO

30

motions with the melt convection, and it is likely to envelop in solid. Consequently, the

31

content of calcium is relatively high if many CaO particles are just contained, which is

32

verified by the analysis of SEM-EDS. In a word, the removal efficiency depends upon the

33

chemical state of calcium. The reason why CaO exists is studied, and the envelopment of the

34

particle is mainly discussed by means of thermodynamics, especially on gravitational force,

35

repulsive force, and drag force.

36

Keywords: A1. Directional solidification A1. Upgraded metallurgical-grade silicon

37

A1. Envelopment A2. Removal

B1. CaO particle

38 39

1 Introduction

40

Use of coal, oil and natural gas induces pollution in the environment. It’s necessary to dig

41

new friendly resources. Solar cells are one of the promising energy [1,2]. The majority of

42

solar cells are made of crystalline silicon. Multicrystalline silicon is an important material

43

with the advantages of low production cost and relatively high conversion efficiency, while it

44

needs high purity [3,4]. There is still no general agreement about the maximum impurity

45

content in solar grade silicon. However, investigators agree that most of the metallic

46

impurities can form defects and enhance the formation of dislocations, which act as

47

recombination centers of photo-carriers and give rise to the decrease of conversion efficiency

48

of solar cells [5]. Thus, removal of impurity to acceptable levels for solar cells is of great

49

importance. Nowadays, chemical method can produce qualified products, however, it’s costly

50

and environment polluted. What’s worse, the manufacturing route is dangerous. Physical

51

method (metallurgy method) is considered relatively cost-effective, environment friendly and

52

relatively safe [3]. Among the metallurgy methods, directional solidification method is quite

3 53

effective to remove metallic impurity which has a small equilibrium distribution coefficient

54

between solid and melt silicon (far less than 1) [6,7]. After directional solidification, many

55

impurities, especially, the metallic impurities, are segregated to melt, and purer solid is

56

achieved. As a result, the former part of the ingot to solidify is much purer than the part that

57

solidified later. Moreover, if solidification velocity is much low, for example, less than 5μm/s,

58

the impurities removal ratio of the former part could be constant.

59

Many studies have shown the removal of transition metals by directional solidification

60

method or modified methods [8-11]. However, in terms of calcium, it is sparsely investigated

61

and has not been well understood yet. Some investigators claimed that the tolerance of

62

calcium in crystalline silicon solar cells is 40ppmw [12]. It may be reasonable for considering

63

the sole calcium element. But in multicrystalline silicon of solar cells, calcium impurity may

64

interact with other impurities, and even form precipitates or refractory compound, doing harm

65

to solar cells performance [13]. Therefore, it is better to remove calcium as possible as we

66

can. As we all know, calcium can be removed by oxygen blowing or slag refining for it easily

67

reacts with oxygen. Calcium is also removed by vacuum melting or electron beam melting

68

method for its high saturated vapor pressure [14]. But in our co-workers’ experiments,

69

calcium is not always removed efficiently by electron beam melting or slag refining method,

70

especially, removal ratio is not constant. So, calcium is a stubborn element in the silicon.

71

Calcium has small equilibrium distribution coefficient, about 1.6×10-3 [15], therefore, we can

72

try to remove it by directional solidification method. Here, one important thing should not be

73

neglected is the route of feedstock, that is to say, how the feedstock has been handled and

74

what condition the feedstock has experienced play an important role in the following

75

impurity removal stage. The chemical state of impurity in the feedstock has heavy influence

76

on its removal. If the impurity exists stably as insoluble compound, for example, CaO, it will

77

give rise to negative effect on the removal.

78

In the present paper, under the given experiment condition, we remove calcium by

79

industrial directional solidification, investigating the redistribution of calcium along the

80

crystal growth direction. Compared with other impurities, like aluminum, copper and boron,

81

calcium has significant feature, namely, its removal ratio is variable. We discussed the

82

causing, influence factors and mechanism in detail.

4

83

2 Experiments

84

The configuration of multi-heaters directional solidification furnace (Jingsheng-450) is

85

shown in Fig.1. Quartz crucible is supported by graphite susceptors, and the inner side of the

86

square crucible is 830×830×450mm3. Silicon nitride painted on the crucible is used as

87

anti-wetting layer and facilitate the ingot demoulding after solidification. In case of the

88

silicon is contaminated by the thermal decomposition of graphite resistance heaters, the

89

crucible is covered with a carbon fiber plate. A cutout is set in the plate center as passageway

90

of argon. Thermocouple 1 (TC1) is installed near the surface of top heaters to measure and

91

control the furnace chamber temperature, and thermocouple 2 (TC2) is installed through the

92

directional solidification block to measure the temperature at the crucible bottom. A quartz

93

rod is inserted from the furnace top into the crucible to detect the crystal growth rate. The

94

furnace chamber pressure is kept at 600mbar (6×104 Pa) by adjusting the argon flow. The

95

furnace wall is cooled by water. Thus, it is considered as a constant temperature boundary.

96

Thermal field is controlled by two ways: (1) controlling the power of graphite resistance

97

heaters; (2) closing or pulling the insulation cage upward.

98

Two

group

experiments

(grouped

1,

2)

were

conducted.

435kg

upgraded

99

metallurgical-grade silicon feedstock (about 4N purity) was used to study for the two groups,

100

respectively. Feedstock was loaded into the crucible and melted. During the melting step, the

101

insulation cage was closed entirely. Furnace chamber was about 1823K at 415 min, then, the

102

temperature decreased gradually to 1698K. After this step, the furnace chamber temperature

103

was controlled appropriately by the combination of resistance heaters and insulation cage,

104

and solidification began. Finally, the temperature at the furnace chamber top and bottom was

105

about 1690K, 1267K, respectively. At the same time, solidification came to an end. After

106

annealing, furnace cooling to ambient temperature directly.

107

The parameters in two groups were set in common, average crystallization rate was about

108

1.62μm/s for each group, and the ingots were both 830×830×270 mm3. 156×156×270 mm3

109

brick was cut from the central ingot for further investigation, as showed in Fig.2. Samples

110

were took brokenly along the sampling line, and the impurity content was detected by the

111

Inductively Couple Plasma Mass Spectrometry (ICP-MS, Thermo Fisher, ICAP QC).

5

112

Fig.1 Schematic configuration of directional solidification furnace

113 114

115 116

Fig.2 Sampling brick (156×156×270mm3) was cut from the central ingot, and the detected

117

samples were took brokenly along the sampling line.

118

3 Results

119

The content of impurities was analyzed by taking samples brokenly along the sampling

120

line shown in Fig.2. Boron (B), aluminum (Al), copper (Cu) and calcium (Ca), these four

121

elements are chose to discuss. Fig.3(a) shows the content of B, Al, Cu and Ca along the ingot

122

height in group 1. There is a clear trend that all content profiles show an accumulation of

123

impurities at the top of the ingot, which is the last part to solidify. What’s more, the contents

6 124

of all impurities except Ca are almost constant along approximately 70% (0.7) or more of the

125

ingot. In fact, at the lower solidification rate, the content of all impurities should have been

126

removed as constant at the lower height of ingot. Considering the content of Ca, it increases

127

along the height, but discontinuously. At the low height of ingot, some points are obviously

128

higher than that of the others, with the content differences of orders of magnitude.

129

In order to verify this phenomenon is not occasional, we conducted the other experiment

130

with setting the same parameters, namely, group 2. We took samples as group 1, but 12

131

samples. The result is shown in Fig.3(b). It shows similar content profiles of impurities as

132

Fig.3(a), and the Ca is still discontinuously, with the content differences of orders of

133

magnitude.

134 135

Fig.3(a) Content of impurities along the ingot height (crystal growth direction) in group 1, (b)

136

content of impurities along the ingot height (crystal growth direction) in group 2.

137

4 Discussions

138

4.1 Causing of the discontinuous content of calcium

139

The basic mechanisms of segregation has been reported clearly [9,11]. The pressure is

140

600mbar (6×104 Pa) in the present experiment, thus, not much lower than ambient pressure.

141

As a result, the effect of saturated vapor pressure can be ignored. We choose B, Al and Cu to

142

compare with Ca to discuss, because their equilibrium distribution coefficients are

143

representative (ranging from 4×10-4 to 0.8). The equilibrium distribution coefficient of B

144

closes to 1 [7], as we all know, its segregation is not obvious as shown in Fig.3 (a) and (b).

145

The equilibrium distribution coefficient of Cu is 4×10-4, lower than that of Al and Ca by one

146

order of magnitude. Removal efficiency of Cu is mainly dependent upon segregation, and the

147

result is as good as expected. The equilibrium distribution coefficient of Al and Ca is similar,

7 148

namely, 2×10-3, 1.6×10-3, respectively[7,15], and their saturated vapor pressure is similar if

149

considered [14,16]. For these sorts of consideration, the removal efficiency of Ca and Al

150

should be similar. However, according to Fig.3(a) and (b), Ca presents variable content at the

151

low height of the ingot as described in section 3, and Al remains relatively constant, as well

152

as B and Cu. It is not likely caused by experimental technique, such as solidification velocity

153

or undercooling. Average original content of B, Al, and Cu is 0.15 (true value ranges 0.07-0.2)

154

ppmw, 0.17 (true value ranges 0.08-0.44) ppmw, 0.39 (true value ranges 0.11-0.86) ppmw,

155

respectively. Table 1 shows the original content of Ca in the feedstock. According to Table 1,

156

the content of Ca in the feedstock is also drastically fluctuating. Maximum content is

157

6.86ppmw, and the minimum content is 0.22ppmw. The content differences of orders of

158

magnitude are consistent with the results shown in section 3. In fact, with respect to the

159

original content of B, Al and Cu, there also exists difference of an order of magnitude, but not

160

as scatter as Ca.

161

Table 1 Original content of calcium in the feedstock (ppmw) Sample number 1 2 3 4 5 6 Content 0.50 2.04 0.89 5.33 1.58 0.22 Sample number 8 9 10 11 12 13 Content 0.35 0.48 6.65 4.17 0.50 0.51

7 6.86 14 0.77

162 163

In order to further investigate the causing of discontinuous content, we chose several

164

samples to be polished and treated by ultrasonic cleaning, then detected by Scanning Electron

165

Microscope (HITACHI, SU 70). One of the samples, at about 40% (0.4) height, near the

166

sampling line, presents high content of Ca. Fig.4(a) is the image, and big linear defect (we

167

call it big grain boundary) is observed. The black cluster defect on the surface shown in the

168

image is caused by polishing. Fig.4(b) is the local part of Fig.4(a) which is marked by a white

169

square. The width of the big grain boundary is about 5μm. Single spherical particle (the

170

diameter is about 1μm) and agglomeration which is agglomerated by spherical particle

171

arrange in a row along the big grain boundary. We conducted line scanning by scanning

172

electron microscope equipped with energy dispersion spectrum analysis (SEM-EDS) across

173

the big grain boundary. Si, Ca, O, S, and C elements are detected, which is shown in Fig.5.

174

The variation of pulse number (CPS) of carbon is not obvious. Compared with the CPS, in

8 175

the single spherical particle or agglomeration region, CPS of Si decreases from 19000 to

176

16000, and CPS of Ca, O , S increases from about 50 to about 400, 50 to about 450, 50 to

177

about 500, respectively. It means the content of Ca, O, and S is relatively high in the region.

178

The line scanning (SEM-EDS) analysis is qualitative analysis, and its detection limit is about

179

0.1%. Consequently, the impurity, for example, aluminum, which content is less than 0.1% at

180

the grain boundary or around the particle can not be detected. However, it also means the

181

content of Ca more than 1000 ppmw at the big grain boundary, so do O and S, and the

182

content of Ca, O and S respectively larger than the other impurities at least one order of

183

magnitude. This content differences of orders of magnitude is in accordance with the analysis

184

in section 3. According to the analysis above, we think the particle consists of insoluble CaO,

185

and S is just attracted at the CaO particle surface. It is the Ca exists as agglomerated insoluble

186

CaO particle that makes the Ca is difficult to be removed somewhere during the directional

187

solidification. Consequently, compared with the other impurities, the content and removal

188

ratio are variable.

189

Because the melting temperature of CaO is very high, more than 2833K in the reductive

190

atmosphere and 3223K in the oxidative atmosphere [17], significantly above that of silicon.

191

Once produced, it exists as a particle and survives in the directional solidification process.

192

During the solidification process, Ca atom can easily diffuse upward from solidifying

193

interface, but CaO particle is difficult to diffuse and is easy to settle down at solid phase.

194

Single particle or agglomeration envelops in solidifying interface and stays at solid, then,

195

grain boundary forms. In return, more CaO particles easily settle down at the grain boundary

196

shown in Fig.4(b). If the sample we chose just the position where CaO particle had

197

agglomerated to form an agglomeration, the content of Ca not only include the isolated

198

substitutional Ca element, but also include the Ca existed as insoluble CaO particle. As a

199

result, the content of Ca is much higher than the sample which does not contain CaO particle.

200

The purity of feedstock is about 4N, especially, it was refined by electron beam melting to

201

decrease oxygen content to less than 0.05ppmw. What’s more, during directional

202

solidification, the furnace was full of melt silicon and argon atmosphere. Therefore, Ca and O

203

were not likely to react to produce CaO. In the present experiment, the particle radius is

204

1-2μm. Buonassisi et al hold the opinion that the refractory compound is not likely produced

9 205

during the directional solidification for the particle size is so large[13]. Before directional

206

solidification process, the feedstock had been refined with metallurgical method, such as

207

carbonthermal, oxygen blowing, which were full of oxygen. In these processes, the content of

208

Ca was relatively high, and the content of O was also adequate. According to Ellingham

209

diagram shown in Fig.6, compared with other elements, CaO is much easier to produce [18].

210

Thus, CaO was produced in the former steps and was brought in as trapped insoluble

211

inclusion of foreign material. Although the feedstock had been refined with electron beam

212

melting before directional solidification, the CaO particle was hardly removed as residual.

213

In a word, insoluble CaO particle or agglomeration produced and existed in the feedstock

214

before directional solidification process. During directional solidification process, it was easy

215

to envelop in the solidifying interface, giving rise to the discontinuous content.

216

217 218

Fig.4(a) Image of a sample taken from 40% (0.4) height near the sampling line,

219

(b) enlarged view of local part of Fig.4(a)

220 221 222 223 224 225

10

226

227

228

229

230

231 232 233 234

Fig.5 Image and qualitative constituent analysis of Fig.4(b) by line scanning of scanning electron microscope-energy dispersion spectrum (SEM-EDS)

11

235

Fig.6 Ellingham diagram for some representative elements

236 237

4.2 Envelopment of CaO particle

238

Based on the discussion in section 4.1, variable removal of Ca is caused by the

239

envelopment of insoluble CaO particle. Not all of the particles will be enveloped in

240

solidifying interface, only the particle or agglomeration which is in the appropriate condition

241

can be enveloped. According to thermodynamics, in the vicinity of the solidifying interface,

242

when the Gibbs free energy of particle-solid lower than that of particle-melt, particle can be

243

enveloped [19], but its kinetics procedure is very complicated. Density of CaO is 3.35g/cm3,

244

higher than the density of melt silicon [20]. Particle is assumed as sphere and gravitational

245

(buoyancy) force is expressed as [21]: 4 FG  R3g 3

246

,  ,

(1)

247

where

248

gravity acceleration, respectively. Before the starting of crystal growth of silicon, the larger

249

particle will directly settle down at the bottom of ingot because of significant gravity. This is

250

why the content of Ca at ingot bottom is relatively high, as shown in Fig.3(a) and (b).

251

However, some smaller particles will be pushed upward into the top, and some particles will

252

still motion with the melt convection, then be enveloped in the solidifying interface or pushed

253

upward when the crystal growth begins.

R

g

are particle radius, difference of density between particle and melt,

12 254

Whether the small particle will be enveloped in solid or pushed upward during the

255

solidification is mainly determined by convection, crystal growth velocity, interface shape,

256

and particle radius [21-24]. Solidifying interface shape plays an important role in the particle

257

envelopment. If the longitudinal temperature gradient is not high enough and lateral

258

temperature gradient are not stable, the solidifying interface is not planar and particle is

259

enveloped easily [23]. It was reported that the higher vertical temperature gradient at the

260

interface does good to push particles upward, but the particle will be easily enveloped in the

261

high crystal growth velocity [24]. Melt convection ahead of the solidifying interface alters

262

particle behavior in the vicinity of the interface. It was reported that the higher convection

263

will give rise to less particle settles down at the interface, and lower gravity acceleration leads

264

to more particles are enveloped in solid [21]. The viscosity of melt silicon is roughly equal to

265

water at room temperature [20], about 0.55 mPa.s. Therefore, melt flows easily. The thermal

266

field and crystal growth velocity in our experiment are given. We ignore the effect of particle

267

motion on the melt convection. Particle motions with the melt convection, and small particles

268

agglomerate together to become agglomeration. Consequently, agglomeration moves

269

downward because of significant gravity increasing. When the agglomeration in the vicinity

270

of interface, i.e. solute boundary layer, which is a mushy zone. In this zone, the viscosity is

271

higher than the upper bulk melt [25]. Therefore, particle becomes harder to motion. Here,

272

repulsive force which is due to the difference of interfacial energy and drag force which is

273

due to the viscosity in the melt these two fundamental forces that acting on the particle should

274

not be ignored. Repulsive force aids particle pushing and drag force aids particle

275

envelopment. They are expressed as equation (2), (3), respectively [21]:

276

FR  2R 0 (

277

FD  6U

a0 )k a0  d

R2 2 k d

(2) (3)

278

where FR , FD ,  0 , a0 , d , k ,  ,

279

interfacial energy between particle and solid, inter-atomic distance, separation between

280

particle and interface, ratio of particle and molten thermal conductivity, dynamic viscosity of

281

molten, velocity of particle perpendicular to the interface, respectively. The other parameters

282

have the same physical meanings with equation (1). It is difficult to acquire these parameters

U

are repulsive force, drag force, difference of

13 283

quantitatively, so, we just qualitatively analyze it. In terms of the particle or qgglomeration,

284

when the sum of gravitational force and drag force larger than the repulsive force in the

285

vicinity of solidifying interface, the particle will set down at the interface, and it will be

286

enveloped by the growth of solid latter. Because the longitudinal temperature gradient is not

287

so large, and the interface is variable, the particle envelopment may be enhanced. On the

288

other hand, the variable interface means disordering crystal growth, and disordered crystal

289

growth hinders the particle pushing forward, then defects are formed. More particles are

290

attracted here to form an agglomeration for the lower chemical potential. As a result, the

291

particle or agglomeration accumulates at the defects shown in Fig.4(a) and (b).

292

5 Conclusions

293

The removal of Ca is investigated by industrial directional solidification, however, CaO

294

exists as insoluble particle of foreign material in the feedstock. In the solidification step, CaO

295

particle directly settles down at solidifying interface, or agglomerates together to settle down

296

at solidifying interface. As a result, the content of Ca is relatively high if the sample we took

297

contains many CaO particles, but relative low if it just contains isolated substitutional Ca.

298

Therefore, CaO causes variable content at the low height of ingot. Whether the insoluble CaO

299

particle will be enveloped in solid or not, it’s mainly determined by gravitational force, drag

300

force and repulsive force in the vicinity of the solidifying interface.

301

According to the discussions above, we conclude that the route of feedstock is very

302

important. The chemical state of impurity, for example, existing as an isolated element or

303

high melting foreign insoluble inclusion, plays a critical role in the impurity removal.

304

Acknowledgments

305

The authors grateful acknowledge financial support from Qingdao Longsun Silicon

306

Technology Company Ltd, projects (51334004 and 51204143) supported by the National

307

Natural Science Foundation of China and project (2006L2003) supported by the Scientific

308

Technological Innovation Platform of Fujian Province. Especially, we thank very much Mr

309

Hou for the helpful discussions.

310

References

311

[1] A. Bahrami, S. Mohammadnejad, S. Soleimaninezhad, Photovoltaic cells technology:

14 312 313 314 315 316

principles and recent developments, Opt. Quant. Electron. 45 (2013) 161-197. [2] G. Masson, S. Orlandi, M. Rekinger, Global market outlook for photovoltaics 2014-2018, in: European photovoltaic industry association, 2015, 9-11. [3] S. Pizzini, Towards solar grade silicon: Challenges and benefits for low cost photovoltaics, Sol. Energ. Mat. Sol. C. 94 (2010) 1528-1533.

317

[4] J. Hofstetter, J.F. Leleievre, C. Canizo, A. Luque, Acceptable contamination levels in solar

318

grade silicon: From feedstock to solar cell, Mater. Sci. Eng B. 159-160 (2009) 299-304.

319

[5] A.A. Istratov, T. Buonassisi, M.D. Pickett, M, Heuer, E.R. Weber, Control of metal

320

impurities in “dirty” multicrystalline silicon for solar cells, Mater. Sci. Eng B. 134 (2006)

321

282-286.

322

[6] T. Shimpo, T. Yoshikawa, K. Morita, Thermodynamics study of the effect of calcium on

323

removal of phosphorus from silicon by acid leaching treatment, Metall. Mater. Trans. B.

324

35 (2004) 277-284.

325 326 327

[7] B.R. Bathey, M.C. Cretella, Review solar-grade silicon, J. Mater. Sci. 17 (1982) 3077-3096. [8] T. Luo, G.Q. Lv, W.H. Ma, K.X. Wei, X. Yang, S.Y. Li, Numerical and experimental study

328

of vacuum directional solidification purification process for SOG-Si in metallurgical

329

route, J. Cryst. Growth. 384 (2013) 122-128.

330 331 332 333 334 335 336

[9] M.A. Martorano, J.B. Ferreira Neto, T.S. Oliveira, T.O. Tsubaki, Refining of metallurgical silicon by directional solidification, Mater. Sci. Eng B. 176 (2011) 217-226. [10] N. Yuge, K. Hanazawa, Y. Kato, Removal of impurities in molten silicon by directional solidification with electron beam melting, Mater. Trans. 45 (2004) 850-857. [11] W.A. Tiller, K.A. Jackson, J.W. Rutter, B. Chalmers, The redistribution of solute atoms during the solidification of metals, Acta. Metall. 1 (1953) 428-437. [12] A.K. SØiland, M.G. Dolmen, J. Heide, U. Thisted, G. Halvorsen, G. Ausland,

337

K. Friestad, P. Preis, K. Peter, O. Graf, T. Bartel, R. Tronstad, First results from a

338

simplified Elkem solar route-input to tolerance limits, Sol. Energ. Mat. Sol.C. 130

339

(2014) 661-667.

340 341

[13] T. Buonassisi, A.A. Istratov, M.A. Marcus, B. Lai, Z.H. Cai, S.M. Heald, E.R.Weber, Engineering metal-impurity nanodefects for low-cost solar cells, Nat. Mater. 4 (2005)

15 342 343

676-679. [14] Y. Tan, S.Q. Ren, S. Shi, S.T. Wen, D.C. Jiang, W. Dong, M. Ji, S.H. Sun, Removal of

344

aluminum and calcium in multicrystalline silicon by vacuum induction melting and

345

directional solidification, Vacuum. 99 (2014) 272-276.

346

[15] B. Ryningen, T. Ervik, E. Olsen, E. Øvrelid, G. Stokkan, M. Kivambe, C. Modanese,

347

O. Lohne, On the role of calcium in multicrystalline silicon, in: 27th European

348

photovoltaics solar energy conference and exhibition, 2012, 1018-1022.

349

[16] K. Liu, J.J. Wu, K.X. Wei, W.H. Ma, K.Q. Xie, S.Y. Li, B. Yang, Y.N. Dai, Application

350

of molecular interaction volume model on removing aluminum from metallurgical grade

351

silicon by vacuum volatilization, Vacuum. 114 (2015) 6-12.

352

[17] D. Manara, R. BÖhler, L. Capriotti, A. Quaini, Z. Bao, K. Boboridis, L. Luzzi, A.

353

Janssen, P. PÖml, R. Eloirid, R.J.M. Konings, On the melting behaviour of calcium

354

monoxide under different atmospheres: A laser heating study, J. Eur. Ceram. Soc. 34

355

(2014) 1623-1636.

356 357 358 359

[18] K. Morita, T. Miki, Thermodynamics of solar-grade-silicon refining, Intermet. 11 (2003) 1111-1117. [19] D.R. Uhlmann, B. Chalmers, K.A. Jackson, Interaction between particles and a solid-interface, J. App. Phys. 35 (1964) 2986-2993.

360

[20] W.K. Rhim, K. Ohsaka, Thermophysical properties measurement of molten silicon by

361

high-temperature electrostatic levitator: density, volume expansion, specific heat

362

capacity, emissivity, surface tension and viscosity, J. Cryst. Growth. 208 (2000)

363

313-321.

364

[21] A.V. Bune, S. Sen, S. Mukherjee, A. Catalina, D.M. Stefanescu, Effect of melt

365

convection at various gravity levels and orientations on the forces acting on a large

366

spherical particle in the vicinity of a solidification interface, J. Cryst. Growth. 211

367

(2000) 446-451.

368

[22] S.Sen, B.K. Dhindaw, D.M. Stefanescu, A. Catalina, P.A. Curreri, Melt convection

369

effects on the critical velocity of particle engulfment, J. Cryst. Growth. 173 (1997)

370

574-584.

371

[23] E.M. Agaliotis, C.E. Sschvezov, M.R. Rosenberger, A.E. Ares, A numerical model study

16 372 373 374 375

of the effect of interface shape on particle pushing, J. Cryst. Growth. 354 (2012) 49-56. [24] A.W. Rempel, M.G. Worster, The interaction between a particle and an advancing solidification front, J. Cryst. Growth. 205 (1999) 427-440. [25] M.A. Martorano, J.B. Ferreira Neto, T.S. Oliveira, T.O. Tsubaki, Macrosegregation of

376

impurities in directionally solidified silicon, Metall. Mater. Trans.A. 42 (2011)

377

1870-1886.

378

17

379

Highlights

380

.Compared with other impurities (aluminum and copper), the content of calcium is

381

variable and discontinuous at low height part of the ingot, although it also has the

382

trend accumulating at the top.

383 384 385 386 387 388

.The variable removal is caused by the existence of insoluble calcium-oxide particle, whose melting temperature significantly above that of silicon. .Calcium-oxide exists as foreign insoluble material in the feedstock and it is enveloped in solid or pushed upward during the solidification step. .The mechanism of envelopment of the particle is discussed by thermodynamics, especially on gravitational (buoyancy) force, repulsive force and drag force.