硅在氮化硅涂层上的形核SiO2

Journal of Crystal Growth 331(2011)64–67

Contents lists available at ScienceDirect

Journal of Crystal Growth

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

Nucleation of silicon on Si 3N 4coated SiO 2

I. Brynjulfsen Ã, L. Arnberg

Department of Materials Science and Engineering, Norwegian University of Science and Technology, 7491Trondheim, Norway

a r t i c l e i n f o

Article history:

Received 30May 2011Received in revised form 6July 2011

Accepted 8July 2011

Communicated by P. Rudolph Available online 19July 2011Keywords:

A1. Nucleation A1. SolidificationA2. Undercooling B1. Silicon

a b s t r a c t

Control of the nucleation during directional solidificationof solar cell silicon is important in order to be able to control the growth and number of grains formed. A certain amount of undercooling is required to obtain dendritic growth with faceted twins (whichhas shown promising results for structure control), but a too high undercooling will lead to extensive nucleation which will oppose the positive effect of a small number of large grains with controlled growth directions. In the present experiments, the nucleation undercooling of silicon on Si 3N 4coated SiO 2with variation in coating parameters has been investigated. Experiments were performed with the sessile drop method, and with differential thermal analysis, with a cooling rate of 20K/min.There were no significantdifferences in nucleation undercooling between the different variations in coating. The undercooling does not seem to be dependent on coating thickness, oxygen concentration, wetting angle or roughness at the given cooling rate.

&2011Elsevier B.V. All rights reserved.

1. Introduction

The solar cell industry is developing fast in several directions, and in order for the multicrystalline solar cell to be able to compete with monocrystalline cells and other new alternatives, the efficiencyhas to be improved. The solidificationprocess of multicrystalline silicon is important for the finalefficiencyof the solar cell. Grain size, grain orientation, and impurity distribution/concentrationare all proper-ties dependent on solidificationparameters. Some of these char-acteristics like the number of, the size, and orientation of grains are again dependent on the nucleation of silicon, and it is therefore important to able to control this mechanism.

Recently several solidificationexperiments have been per-formed by Fujiwara et al. [1–4].They studied grain growth, and were able to increase the crystals size by an initial faceted dendritic growth followed by traditional planar front directional solidification.The dendritic growth results in fewer and larger grains, which again lead to less grain boundaries were recombi-nation can take place. Fujiwara et al. [1]investigated how different cooling rates influencedthe size of undercooling needed to obtain faceted dendritic growth. The present work has been performed in order to study how/ifthe substrate on which the silicon grows will influencethe undercooling and nucleation of silicon. Si 3N 4coated SiO 2has been chosen as a substrate since multicrystalline silicon ingots are normally cast in Si 3N 4coated SiO 2crucibles.

ÃCorresponding author. Tel.:þ4773594903; fax:þ4773550203.

E-mail address:[email protected](I.Brynjulfsen). 0022-0248/$-see front matter &2011Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2011.07.003

Nucleation is the dominant process in the beginning of solidifica-tion and leads to the establishment of the finalgrain number. Heterogeneous nucleation undercooling depends strongly on the wetting angle between the nucleus and the nucleating substrate. This implies that the nucleation is dependent on the substrate roughness, composition, thickness, etc. [5]. Another aspect is impu-rities. Impurities in the bulk have been studied by several authors, and it has been shown that silicon often nucleate from Si 3N 4-or SiC-particles [6]. This nucleation can cause the formation of an equiaxed zone instead of the desired columnar zone [7, 8].It has been documented that particles like this are present in the bottom of the ingot [9].

The substrate’sinfluenceon undercooling for solidificationof silicon has not been studied thoroughly, but some investigations in the area has been done. Appapillai et al. [10]investigated nucleation undercooling for silicon samples coated with different materials among others Si 3N 4. They found finelyspaced nuclea-tion sites near the edge of the samples coated with dry oxides (highundercooling), which indicated that the nucleation started in this region. For the silicon nitride coated samples the nuclea-tion sites were further apart. This resulted in the conclusion that a lower undercooling gave fewer grains, which is consistent with classical nucleation theory. This shows the importance of the to control more precisely. They also showed that the chemical composition played an important role in nucleation. The oxides had a higher interfacial stability resulting in a higher undercooling than the Si 3N 4.

The present work has been performed to investigate which coating parameters influencethe nucleation undercooling of silicon on Si 3N 4coated SiO 2. This is done in order to be able to

I. Brynjulfsen, L. Arnberg /Journal of Crystal Growth 331(2011)64–6765

control the nucleation and undercooling needed for a dendritic growth more accurately, since this growth has shown promising results for structure control of silicon ingots.

2. Experimental procedure

Three different parameters have been studied in this work; the coating thickness; the amount of oxygen in the coating; and the roughness of the coating. Two types of experiments have been used to study the nucleation undercooling. The sessile drop method performed in a wettability furnace and differential thermal analysis (DTA).

The samples for the wettability furnace, SiO 2pieces with the dimensions 0.9Â0.9Â0.3cm 3, were coated both manually with a spray gun and with a coating robot. Before coating the samples were preheated on a heater to a temperature of 1501C, but this dropped rapidly when the samples were coated. The samples were coated with a variation in number of layers. The samples were dried between each layer of coating to ensure a better adherence to the substrate and a smooth coating. The amount of coating was weighed and the coating thickness was both estimated and measured after the nucleation experiments, with correlating numbers. The samples were then firedat 11001C for 4h as a standard firingroutine. Firing temperatures of 900and 12001C, and holding time of 6h were used to obtain different oxygen levels in the coating. The oxygen concentration was measured by LECO. Similar samples were also produced by a commercial crucible producer, Vesuvius.

For the nucleation study in the wettability furnace a small piece of silicon (approximately20mg) was placed on the Si 3N 4coated substrate and mounted on the sample holder. The proce-dure and specificationsfor using the wetting furnace are explained elsewhere [11]. The sample was heated at a rate of 5K/minwhen it was close to melting and to the preset holding temperature. When the sample reached the holding temperature, which was always less than 50K above the melting point, it was cooled at a rate of 20K/min.The temperature was measured with

a thermocouple placed directly below the sample holder, and the melting point of silicon was used as a reference point for the amount of undercooling. When the droplet started solidifying it expanded in most cases, as can be seen in Fig. 1. The solidificationtemperature was set to be the temperature when the firstvisible changes in size and/orreflectionin the middle of the droplet took place. The high cooling rate was chosen in order to be able to visually see the point of solidification.A high cooling rate will give higher values of undercooling [12], but since the cooling rate was the same for all experiments it should not have influencedthe parameters under investigation in this study.

Some experiments were performed in a differential thermal analysis (DTA)equipment to confirmthe undercooling measured by the sessile drop method. The same heating and cooling routine as for the wetting experiments were used in these experiments. The crucibles were Al 2O 3coated with Si 3N 4, and therefore an Al 2O 3substrate was also tested in the wetting furnace. Because of the small diameter of the DTA crucibles the coating was not easy to spray uniformly. Some crucibles were spray coated and some painted with coating to investigate if this led to differences.

3. Results and discussion 3.1. Wetting experiments

As explained in Introduction a good wetting between solid and substrate leads to a lower energy barrier for nucleation, thus a lower nucleation undercooling. The surface energy balance for a solid droplet in a liquid is shown in Fig. 2(a).

For good wetting between the solid and the substrate, g Liquid ÀSubstrate must be higher than the sum of the other interface energies:

g Liquid ÀSubstrate Z g Solid ÀSubstrate þg Liquid ÀSolid cos y

ð1Þ

A large g Liquid ÀSubstrate will give wetting of the solid nucleus in Fig. 2(a).On the other hand, if g Liquid ÀSubstrate is large it can be

seen

Fig. 1. Melted droplet of silicon (a)before cooling and (b)solidified.

Fig. 2. The figureillustrates the difference between contributions affecting the wetting angle for the two cases:(a)Solid nucleus in liquid silicon. (b)Liquid drop on substrate in vapour atmosphere, sessile drop experiment.

66I. Brynjulfsen, L. Arnberg /Journal of Crystal Growth 331(2011)64–67

from Fig. 2(b)that this can lead to non-wetting in the liquid on substrate case. Wetting conditions in this second case is also dependent on the liquid–vapourtension and vapour–substratetension. Both of these interface tensions are dependent on the coating, because a high oxygen concentration can lead to the formation of an oxide filmand/ora high production of SiO gas. This will affect the wetting and nucleation conditions on the edges of the droplet, but not the nucleation conditions inside the droplet. Therefore it is difficultto predict whether or not non-wetting/wettingin sessile drop experiments indicate wetting/non-wetting of the solid nucleus forming inside the droplet.

Another factor from theory that will contribute to a good wetting is a rough coating which will lead to more nucleation points [13]. In addition, in the sessile drop experiments a low oxygen concentration will lead to a higher degree of wetting [11]. Since the dendritic growth requires a certain amount of under-cooling to take place, non-wetting in Fig. 2(a)is wanted. The aim of this study is therefore not only to investigate what factors influencethe nucleation undercooling, but also see which gives the highest undercooling at the given cooling rate. Since several factors are affecting the wetting conditions, the only conclusion that can be drawn is that wetting in Fig. 2(b)gives a higher probability for non-wetting in Fig. 2(a).A consideration that must be taken into account, and cannot be left out of experiments with solar cell silicon, is that non-wetting in Fig. 2(b)is important to prevent the silicon from sticking to the crucible.

Large variations in the undercooling were measured, from the lowest of 12K to the highest of 37K. The undercooling did not seem to be dependent on the oxygen concentration in the coating, see Fig. 3(a),or the measured wetting angle, see Fig. 3(b).

A thicker coating will limit the diffusion of oxygen from the silica substrate and in this way affect the reactions taking place between silicon and substrate. It was therefore believed that the thickness would influencethe undercooling, but variations in coating thickness

Undercooling as a function of oxygen concentration in coating

40)

35

K ( 30g n 251100°C, 4hi l o 1100°C, 6ho 20c 900°C, 2hr e 151200°C, 4h

d n 10U 50Oxygen concentration (wt.%)Undercooling as a function

of wetting angle

4035

)

K ( 30g n 25Author i l o Vesuvius

o 20Coated aluminac r e 15Pure silicaPure alumina

d n U 1050Wetting angle (°)

Fig. 3. Variation in undercooling with:(a)oxygen concentration. The tempera-tures and time displayed are the firingparameters for the coating. (b)Liquid-substrate wetting

angle.

Undercooling as a function of

coating thickness

4035g

n 30i l o 25o c 20

r e d 15Vesuvius

n U 1050Coating thickness (µm) Undercooling as a function of roughness

- samples from Vesuvius

4035)

K ( 30g n i 25l o o 20c r e 15d n U 1050Roughness (µm)

Fig. 4. Display of variation in undercooling with coating thickness (a)and roughness (b).

did neither show a trend; see Fig. 4(a).These samples were produced with the same firingroutine and hence had the same oxygen concentrations. Samples produced by Vesuvius with an up to 20times thicker coating also gave undercooling in the same range. The variations in coating thickness were applied due to earlier investiga-tions of wetting and oxygen concentration [11].

The comparison of samples with different roughness from Vesuvius did neither show any trend; see Fig. 4(b).The average roughness given in the figureis smaller than 5and 10m m Ra. The last value marked as 20m m Ra is for all the experiments with variation in coating thickness. The two finestcoatings are some-what thinner than the rough coatings.

The coating firedat 12001C for 6h displayed a different behaviour then the rest. It looked like the droplet reacted with the substrate after melting and formed an oxide layer around the droplet. It lost its round shape and imploded. This implies an effect of the oxygen concentration on the surface tension and hence the nucleation. Because of this behaviour the data for the coating firedat 12001C were only plotted in Fig. 4(a).

The present results are not directly comparable with the experiments in the study by Appapillai et al. [10], since in their case they coated the silicon samples completely with the sub-strate. But, as for Appapillai, the coated Al 2O 3substrates in the present project gave a somewhat lower undercooling than the coated silica substrates. Silicon wet these substrates most, and the spreading was faster and more significant.This implies that the chemical composition of the silicon nitride coating alone does not determine the undercooling.

All the samples in the present experiments were covered by a black layer, indicating a low interfacial stability, which can contribute to decrease the undercooling. This behaviour together with the clear reaction for the samples firedat 12001C showed, in accordance with Appapillai et al. [10], Koh et al. [14]and Vallat-Sauvain et al. [15], the importance of chemical composition.

I. Brynjulfsen, L. Arnberg /Journal of Crystal Growth 331(2011)64–6767

Table 1

DTA experiments performed. M is for mixed, S is for sprayed and P is for painted crucible. DTA-results Coating

M M S S P P P Undercooling (K)

17

15

24

29

22

18

20

Undercooling all substrates

40

35)

K 30( g n 25i l o o 20c r e d 15n U 1050

Fig. 5. Variation in undercooling for all the different substrates tested in this study. Experiments with the same undercooling on the same type of substrate are placed next to each other.

As mentioned above there is also the possibility for the nucleation to take place at inclusions in the melt. Si 3N 4particles can be formed due to dissolution and re-precipitation of the coating, and carbon is also available in the furnace atmosphere giving the possibility for formation of SiC. The substrates without coating were not exposed to nitrogen. (Thecoating has been identifiedas the main source of nitrogen in silicon.) If Si 3N 4particles were the main cause of nucleation of silicon, the substrates without coating should show a higher undercooling. This is not the case in this work, but SiC can not be ruled out as a compound influencingnucleation.

3.2. DTA-analysis

Two series of differential thermal analysis experiments were performed on Si 3N 4coated Al 2O 3crucibles, with the same heating and cooling cycle as in the wetting furnace. Because of the small diameter of the crucible an even coating was very difficultto obtain. The firstseries of coating were therefore still on the experimental level and not very complete or uniform. Two crucibles were coated in this way and are marked M in Fig. 1. In the second series, two crucibles were spray coated, S, and three painted with coating, P. The average undercooling was ca. 21K. The shapes of two of the DTA curves were varied from the others. These were the spray coated and most evenly coated crucibles. They have a somewhat higher undercooling than the others, but not significantlyenough to conclude with a real difference, see Table 1.

The undercooling of the samples in the DTA were in the same range as the sessile drop experiments. This means that even if the DTA is a more accurate measuring method, since the sessile drop experiments rely on a visual observation of start of nucleation, sessile drop experiments can be used to measure undercooling. Two of the DTA-values were somewhat lower, but as commented above these were not coated as good as they should have. This leads to an uncertainty in their accuracy. All the experiments are summarized in Fig. 5.

4. Conclusion

The results from the wetting experiments do not indicate that the coating alone plays an important role in the nucleation under-cooling. Different oxygen concentration, thickness, and roughness gave undercooling in the same range. There where on the other hand some variations in undercooling with different substrates, as the coated alumina substrates showed a lower undercooling, see Fig. 5. A very high oxygen concentration did affect the undercooling in the way that a clear reaction and deformation affected the silicon leading to a lower degree of undercooling. Even if the DTA results and the results from the sessile drop experiments were in the same range, the DTA-results showed variations, and a further study of the coating roughness should be done. Nucleation in the melt caused by inclusion should not be ruled out either. Further work is required to study if the nucleation undercooling experiments in the present investigation can be used to predict nucleation conditions during solidificationof large silicon ingots. Acknowledgement

This work was performed within The Norwegian Research Centre for Solar Cell Technology project number 193829, a Centre for Environment-friendly Energy Research co-sponsored by the Norwegian Research Council and research and industry partners in Norway. References

[1]K. Fujiwara, W. Pan, K. Sawada, M. Tokairin, N. Usami, Y. Nose, A. Nomura,

T. Shishido, K. Nakajima, Directional growth method to obtain high quality polycrystalline silicon from its melt, Journal of Crystal Growth 292(2)(2006)282–285.

[2]K. Fujiwara, W. Pan, N. Usami, K. Sawada, M. Tokairin, Y. Nose, A. Nomura,

T. Shishido, K. Nakajima, Growth of structure-controlled polycrystalline silicon ingots for solar cells by casting, Acta Materialia 54(12)(2006)3191–3197.[3]K. Fujiwara, K. Maeda, N. Usami, G. Sazaki, Y. Nose, A. Nomura, T. Shishido,

K. Nakajima, In situ observation of Si faceted dendrite growth from low-degree-of-undercooling melts, Acta Materialia 56(11)(2008)2663–2668.[4]K. Fujiwara, K. Maeda, N. Usami, G. Sazaki, Y. Nose, K. Nakajima, Formation

mechanism of parallel twins related to si-facetted dendrite growth, Scripta Materialia 57(2)(2007)81–84.

[5]W. Kurz, D.J. Fisher, Fundamentals of Solidification,Trans Tech Publications,

1998.

[6]A.K. Soiland, E.J. Ovrelid, T.A. Engh, O. Lohne, J.K. Tuset, O. Gjerstad, Sic and

Si 3N 4inclusions in multicrystalline silicon ingots, Materials Science in Semiconductor Processing 7(1–2)(2004)39–43.

[7]M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti, Impurities influenceon multi-crystalline photovoltaic silicon, Transactions of the Indian Institute of Metals 62(2009)505–509.[8]N. Mangelinck-No¨el, T. Duffar, Modelling of the transition from a planar

faceted front to equiaxed growth:application to photovoltaic polycrystalline silicon, Journal of Crystal Growth 311(1)(2008)20–25.

[9]T. Buonassisi, A.A. Istratov, M.D. Pickett, J.-P. Rakotoniaina, O. Breitenstein,

M.A. Marcus, S.M. Heald, E.R. Weber, Transition metals in photovoltaic-grade ingot-cast multicrystalline silicon:Assessing the role of impurities in silicon nitride crucible lining material, Journal of Crystal Growth 287(2)(2006)402–407.

[10]A. Appapillai, E. Sachs, The effect of substrate material on nucleation behavior

of molten silicon for photovoltaics, Journal of Crystal Growth 312(8)(2010)1297–1300.

[11]I. Brynjulfsen, A. Bakken, M. Tangstad, L. Arnberg, Influenceof oxidation on

the wetting behavior of liquid silicon on Si3N4-coated substrates, Journal of Crystal Growth 312(16–17)(2010)2404–2410.

[12]K. Fujiwara, Y. Obinata, T. Ujihara, N. Usami, G. Sazaki, K. Nakajima, Grain

growth behaviors of polycrystalline silicon during melt growth processes, Journal of Crystal Growth 266(4)(2004)441–448.

[13]D.A. Porter, K. Easterling, Phase Transformations in Metals and Alloys,

Chapman &Hall.

[14]J. Koh, A.S. Ferlauto, P.I. Rovira, C.R. Wronski, R.W. Collins, Evolutionary phase

diagrams for plasma-enhanced chemical vapor deposition of silicon thin filmsfrom hydrogen-diluted silane, Applied Physics Letters 75(15)(1999)2286–2288.[15]E. Vallat-Sauvain, J. Bailat, J. Meier, X. Niquille, U. Kroll, A. Shah, Influenceof the

substrate’ssurface morphology and chemical nature on the nucleation and growth of microcrystalline silicon, Thin Solid Films 485(1–2)(2005)77–81.