- Email: [email protected]
Undercooling measurement and nucleation study of silicon droplets on various substrates
Author’s Accepted Manuscript Undercooling measurement and nucleation study of silicon droplets on various substrates M.G. Tsoutsouva, T. Duffar, D. Chaussende, M. Kamguem www.elsevier.com/locate/jcrysgro
PII: DOI: Reference:
S0022-0248(16)30373-6 http://dx.doi.org/10.1016/j.jcrysgro.2016.07.018 CRYS23463
To appear in: Journal of Crystal Growth Received date: 27 January 2016 Revised date: 22 June 2016 Accepted date: 16 July 2016 Cite this article as: M.G. Tsoutsouva, T. Duffar, D. Chaussende and M. Kamguem, Undercooling measurement and nucleation study of silicon droplets on various substrates, Journal of Crystal Growth, http://dx.doi.org/10.1016/j.jcrysgro.2016.07.018 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.
Undercooling measurement and nucleation study of silicon droplets on various substrates M.G. Tsoutsouvaa, T. Duffara*, D.Chaussendeb, M. Kamguema a
CNRS, Université Grenoble Alpes, SIMaP–EPM, 38402 Saint Martin d’Héres, France b
CNRS, Université Grenoble Alpes, LMGP, 38000 Grenoble, France
* Corresponding author: [email protected]
Abstract The heterogeneous nucleation of solid silicon is studied when molten droplets solidify on various substrates. An experimental installation has been developed in order to record in real time the melting–solidification process, measure the undercooling temperature and look at the solidification of the droplets. Three different categories of substrate materials are studied: oxides (silica, zyarock and sapphire), nitrides (silica+oxidized Si3N4 coating, zyarock + oxidized Si3N4 coating, sintered Si3N4, PBN and HIP – BN) and carbon – containing (isostatic graphite, glassy carbon and SiC). Higher undercooling values are measured in the case of oxide substrates where the solidified droplet is found to be mainly composed of a single – crystal. In the case of nitride substrates, a dissolution/precipitation process takes place and β-phase Si3N4 precipitates are found to act as nucleation centers for the silicon solidification. The nucleating power of Si3N4 is attributed to the good epitaxial fit with silicon. Oxidation of Si3N4 powder at a higher temperature increases the undercooling of the droplet. When the silicon droplet is solidified on BN substrates, BN particles are detected on the surface of the droplet as well as a Si3N4 layer at the substrate/Si interface which promotes nucleation. Carbon – containing substrates are found to favor the nucleation of
1
silicon due to the creation of a SiC layer at the substrate/silicon interface and precipitation of SiC particles in the droplets. However, no explanation of the important nucleating effect of SiC has been found. Keywords: A1. Nucleation; A1. Solidification; A2. Growth from melt; B2. Semiconducting silicon. 1. Introduction An important challenge in mold solidification of photovoltaic (PV) silicon (Si) is the efficient control of the crystalline structure, grain sizes and orientations in the solidified ingot, as it interacts with the solar cell efficiency by the way of carrier recombination at grain boundaries. Several techniques are promoted with various industrial successes in order to control the final grain structure in the grown ingot: -
The standard process, used for many years, applies a sudden cooling at the bottom of the crucible in order to induce an undercooling and associated nucleation [1].
-
A variant of this process introduces some highly heat conducting pieces at the bottom of the crucible in order to promote the nucleation of few dendrites, likely to impose their crystallographic orientation and generate large, well oriented grains [2].
-
A paving of single crystal seeds may also be placed at the bottom of the crucible in order to get a Si single crystal [3]. However spurious nucleation on the crucible walls largely decreases the yield of the process.
-
On the contrary, another variant uses many randomly oriented seeds in order to promote grain selection, which has been proved to improve the material overall quality [4]. However, uncontrolled nucleation on crucible walls is still an issue.
2
Clearly, a better understanding and control of the nucleation and growth phenomena in PV Si casting is necessary. The theory of heterogeneous nucleation is well advanced [5] however, in its classical treatment, involves a contact angle representing the equilibrium of seed – liquid – substrate interfacial energies. This angle is aiming to describe the mechanism promoting the nucleation event on the foreign substrate, which is highly dependent on both the nucleating and nucleant materials. Influence of many different parameters has been invoked, such as roughness [6], cavities in the substrate wall [7], chemical reactions [8] or epitaxial relationships between the two solid materials [9]. Also possible correlation with the wetting angle of the liquid on the crucible material has not been elucidated so far [10]. However, the predictive ability of the current state of the art is low and the only way to study the nucleating ability of a substrate facing cooling liquid Si is to perform experiments. In PV casting technology, the crucible is classically made of sintered silica (called Zyarock) coated with a layer of silicon nitride (Si3N4) powder. This layer is fired under air at various temperatures and durations, depending on the specific process. The Si3N4 powder coating constitutes a barrier to wetting and infiltration due to the presence of non-wettable SiO2 formed on the Si3N4 particles during their firing. This layer also acts as mechanical fuse due to its weak mechanical resistance resulting from its porous microstructure, allowing the easy detachment of the ingot from the crucible. Depending on the particular industrial process, the oxidation temperature of the nitride powder layer varies from 900°C to 1100°C. Besides, other materials, based on silicon carbide layers are presently under study with the aim to get reusable crucibles. Also, boron nitride crucibles are sometimes used for research purpose, in spite of their deleterious
3
boron doping. Therefore the nucleating ability of various kind of oxide, carbide and nitride substrates should be investigated. An efficient method to study the nucleation phenomenon is by measuring the undercooling temperature during the cooling process since it gives information about whether or not heterogeneous nucleation is favored. Some researchers performed undercooling measurements in an electromagnetic levitation setup either concerning pure Si or Si containing a controlled percentage of impurities [11]-[15]. Those experiments have clearly shown the correlation of the measured undercooling temperature with the concentration of gaseous impurities surrounding the droplet, leading to dissolution and precipitation of particles (Si3N4, SiC) in the melt, then solid Si nucleation on the precipitates. The nucleating ability of β-Si3N4 (almost 0 K of undercooling) was shown to be higher than in the case of 3C-SiC precipitates (3 or 4 K of undercooling at high concentration). However, this is only an indirect measurement of the nucleating action, not taking into consideration the role that the substrate (crucible or mold walls) plays. Tsai et al. [16] investigated the firing conditions of the Si3N4 coating on the nucleation and grain structure of solidified Si by sessile drop and directional solidification experiments. It was observed that the preferred growth orientation was affected by the coating when the cooling rate was low, while it was not affected at all in case of high cooling rate. Appapillai et al. [17]-[18] studied nucleation by using differential scanning calorimetry to measure the undercooling below the melting temperature of molten Si encapsulated in silicon oxide and nitride layers. Various coatings on the Si samples were tested for their influence on nucleation. Both dry and wet oxide coatings presented higher interfacial stability resulting in higher undercooling values than the nitrides coatings. However, the undercooling
4
measurements presented significant variations, probably due to the coating nonuniformities and the overall sample deformations during melting and solidification process. Brynjulfsen and Arnberg [19] studied the nucleation undercooling phenomenon during the solidification of Si on SiO2 coated with oxidized Si3N4, and on Al2O3, by the sessile drop method. Substrates provided by the commercial crucible producer for photovoltaic applications, Vesuvius, were also investigated. They concluded that the coating parameters such as thickness, roughness and the amount of oxygen in the coating have no significant influence on the undercooling of Si. However, they used a thermocouple placed below the sample holder in order to measure the undercooling temperature interval and this method cannot give accurate values of the undercooling. Alphei et al. [20] studied the nucleation of Si in alumina crucible and found undercooling values as high as 100 K, however, when amorphous Si3N4 nanoparticles were added to the droplet, the undercooling decreased to 2 – 6 K and nucleation was apparently triggered by nitride crystallites. Fujiwara et al. [21] theoretically studied the effect of the Si – substrate interfacial energy on the orientation of nucleated Si grains. However, as these energies are unknown, it is difficult to draw practical conclusions. The present work has been performed in order to provide additional data on the influence that the material of the substrate, on which Si is solidified, has on the undercooling temperature and the initial nucleation of solid Si. It follows former experiments [22] with new objectives: different materials of the substrate (oxides, nitrides, carbon-containing) are investigated, Si is placed and melted directly on the substrate (this avoids the problems of non-uniformity of the coating and deformation of the sample experienced in [18]), a pyrometer is used for measuring the temperature directly on the Si droplet and in addition the melting – solidification process is
5
visualized in situ and in real time. The final crystalline structure of the solidified Si droplet is also investigated.
2. Experimental An experimental installation has been designed and implemented (Figure 1) in order to measure the undercooling temperature and visualize the Si droplet solidification. An electronic grade Si cube (0.8±0.2 g) is mechanically polished with SiC abrasive papers with increasing fineness (from 600 to 2000 grid), ultrasonically cleaned in acetone, etched in a HF solution for 30 s and rinsed with distilled water. It is then quickly (less than 15 min until vacuum is applied) placed on a substrate which is contained in a cylindrical crucible made of pure fused silica, for which it has already been proven that high undercooling can be obtained [22]. The mass of Si is optimized in relation to the dimensions of the substrates so as, when the Si is fully melted, the liquid droplet is fixed and centered on the substrate edges. This allows the accurate and reproducible measurement of the temperature by the pyrometer. All substrates have the same surface (10 mm × 10 mm). Heating and thus melting of Si is realized by electromagnetic induction with the aid of a graphite susceptor. The whole system is housed inside a water cooled stainless steel vacuum chamber evacuated to a pressure of 3∙10-6 mbar with a turbomolecular pump (TMP), then rinsed and backfilled with 6N Ar to a pressure of 1000 mbar (Figure 1). The sample temperature is monitored in real time and a constant inductive power is maintained as soon as the melting plateau is observed on the pyrometer reading. For all the experiments the cooling rate during solidification is 300 K.min-1, obtained by switching off the inductive current. The Si temperature during the melting –
6
solidification process and undercooling is measured by means of an IRCON Modline 5 5R-1810 bichromatic pyrometer operating in the temperature range 700°C to 1800°C (spectral range 0.75 μm to 1.05 μm). A camera is also installed above the pyrometer as shown in Figure 1 and the whole process is recorded in real time, through the pyrometer viewing window. More details about the experimental installation and the melting – solidification procedure are given in [23]. The output of each experiment is the recorded temperature profile that gives the undercooling temperature measurement, the video film showing the melting and solidification process and the solidified Si droplet samples. In the frame of this research, experiments are performed on different substrates that can be categorized as oxides (silica, sapphire and zyarock), nitrides (silica+ oxidized Si3N4 coating, zyarock+ Si3N4 oxidized coating, sintered Si3N4, Pyrolytic BN -PBN- and hot isostatically pressed BN HIP BN) and carbon-containing (isostatic graphite, glassy carbon and SiC). After solidification, the droplets are embedded in resin and cut in cross section. They are polished using diamond pastes down to 1μm and chemically etched for 5 sec in a solution containing nitric acid, hydrofluoric acid and acetic acid in the ratio 2:1:1 to reveal their microstructure. In certain cases the samples are also polished from the bottom, in order to remove the substrate sticking to the droplet, and the microstructure of the surface that is directly in contact with the substrate is investigated. The microstructural characterization of the solidified droplets is performed with the aid of an optical microscope, a scanning electron microscope (SEM) equipped with an energy dispersive spectrometer (EDS) and synchrotron X-ray Bragg diffraction imaging. In certain cases the observed precipitates are identified by micro – Raman and X-ray diffraction techniques.
7
3. Results and discussion Table 1 gathers the investigated substrates and the obtained experimental results. 3.1. Oxide substrates Figure 2 illustrates a representative temperature profile of the Si sample during its melting and solidification process on a fused silica substrate, as recorded by the bichromatic pyrometer. During melting, the sample absorbs the melting enthalpy and the temperature remains almost constant as it can be seen from the presence of a temperature plateau in the bichromatic pyrometer temperature curve. Once the Si is fully melted the temperature rises (at 20 min on Figure 2), since the enthalpy of fusion is no more absorbed by the sample, until it stabilizes at 1840±10 K. Seven minutes later the electric power is cut abruptly and the Si droplet starts to cool down (cooling rate 300 K.min-1). Rapid non – equilibrium solidification of the undercooled droplet then proceeds in several stages. Nucleation occurs at TN=1564 K, as presented in Figure 2 and then the recalescence phase (quasi-adiabatic solidification) follows. During this period, the temperature of the droplet rapidly increases (abrupt temperature rise observed at 27.25 min) due to the release of latent heat of solidification. In the next period of quasi – isothermal plateau, the temperature of the droplet remains almost constant at the melting point until the solidification process is completed. The isothermal plateau after recalescence corresponds to the solidification temperature (TS), well known as 1687 K for Si. However, this temperature is not in accordance with the Ts (1653 K) measured by the pyrometer after recalescence, as shown in Figure 2. This discrepancy can be due to the fact that during melting the pyrometer focuses on solid Si, with a given emissivity, while before and after
8
recalescence, it focuses on liquid Si with another emissivity. Despite that the two color pyrometer is specially designed so that by means of a mathematical ratio calculation, certain influences on measurement can be eliminated, the error of the temperature measurement is ± 30 K. In this work it is considered that the maximum nucleation undercooling (ΔT) can be approximated as the difference between TS and TN and thus in the present experiment ΔT = 89 K. Considering the pyrometer sensibility (0.1 K) the accuracy of the measured undercooling is estimated to be ± 1 K. Experiments have been repeated three times and the measured undercooling values are in the range of [42 to 89 K]. The final temperature decrease represents the cooling of the solidified droplet. The temperature profile is found to present a similar tendency, when the Si sample is melted and solidified on a (100) sapphire or a zyarock substrate. However, the measured undercooling values are in the range [36 to 67 K] in case of (100) sapphire substrate and in the range [41 to 51 K] in case of zyarock. Zyarock is the standard material used for the industrial crucibles in the photovoltaic industry. It is a sintered silica product with an apparent porosity of 12% and typical composition: silica 99.6%, alumina 0.12%, iron oxide 0.10%, titania 0.027%, magnesia 0.013% sodium oxide 0.008%, calcia 0.007% and potassium oxide 0.005%. At the beginning of solidification, the recorded video shows thin solid strips crossing the whole droplet [23]. As time goes on, they develop to several continuous lines. Z. Jian et al. [24] have observed similar structures during rapid solidification experiments in an electromagnetic levitation setup. They suggested that the crystal growth depends on the nucleation undercooling and for undercooling in the range 40 to 100
10 K (similar to the present study) the initial crystal seed solidifies as dendrites
that form a number of plates. Apparently the process is the same in case of solidification
9
on a substrate. For all the studied oxide substrates, the solidified droplets do not spread but, instead, form at equilibrium a spherical cap resting on the substrate (see cross sections of droplets on Figure 3). The “horn” appearing at the end of solidification, as shown in the zyarock droplet cross section, is due to the 10% decrease of density between molten and solid Si. The grain structures, observed in cross section, reveal that the solidified Si droplets are mainly composed of a single crystal (Figure 3). No nucleation center is found at the silicon / substrate interface, which shows that nucleation event is quite rare and probably unique. The single crystal growth is confirmed by synchrotron white beam Xray Bragg diffraction imaging, not presented in the present paper. In the case of fused silica substrate, a twin crystal is observed on the left side of the droplet as well as a dendrite in its center (discussed in paragraph 3.4). The rest of the droplet is a singlecrystal. The droplet solidified on the sapphire substrate is composed of a single crystal which is highly distorted. Crystal deformation is revealed on the Laue spots of the white beam topographs, by the varying degree of asterism indicative of crystal mosaicity. In the case of zyarock substrate, the droplet is a single crystal as well, except a grain on the left side and multi-crystalline grains on the top of the “horn”. The droplets are highly fractured at the Si/substrate interface due to the difference in thermal expansion coefficients between Si and the substrate material and in the central area as well, due to the rapid solidification conditions. Many electromagnetic levitation experiments of pure Si have shown undercooling larger than 300 K. The contact with pure silica decreases this value to some tens of K (162 K) as measured by [22] but anyhow, it can be concluded that it is a bad nucleating
10
agent for Si. Chemical reaction exists between Si and SiO2, the oxide substrate slowly dissolves into the Si melt and oxygen is transported by diffusion and convection into the liquid. However, much of the oxygen evaporates from the melt free surface and only about 1% is incorporated into the growing crystal that apparently does not contribute to the nucleation process under the present experimental conditions. The somewhat smaller undercooling measured in the case of zyarock and sapphire shows that the chemical and/or roughness of the substrate impacts the nucleation behavior, but not that much. Based on the literature, it would be expected that the porosity of the zyarock substrate should increase drastically the nucleation probability of Si [7]. However, from the present experimental results such a conclusion cannot be drawn: the effect of the 12% open porosity is relatively small; the undercooling is decreased by 20 or 30 K but remains large, 45 K in average. Further, zyarock is not only pure silica and other oxide components might have a nucleating power. The effect of the sapphire substrate on Si nucleation is quite similar. This is consistent with the analysis proposed by Drevet and Eustathopoulos [25]: thermodynamic calculations show that Al2O3 is not reacting with Si but it is slightly dissolved and re-deposited as a SiO2 layer on the sapphire substrate. Therefore, the SiAl2O3 interface is transformed to a Si-SiO2 interface. Taking this into account, all those experiments can be considered as nucleation of solid Si on silicon oxide. This gives a mean undercooling of 55 K with a standard deviation of 20 K under the present experimental conditions, for the 9 experiments on oxide substrates.
3.2. Nitride substrates In contrast to the case of oxide substrates, no, or very low, recalescence is observed
11
on the temperature recording when Si is solidified on nitride substrates. For Si droplets on zyarock substrates coated with Si3N4 oxidized at 900 °C, experiments have been repeated four times and the measured undercooling values are in the range [0 K to 2 K]. Similar undercooling values are measured in the case of the fused silica substrate coated with Si3N4 oxidized at 900 °C and of the sintered Si3N4 substrate. A higher undercooling value of 6 K is observed in two experiments, when the Si is left to solidify on a zyarock substrate coated with Si3N4 oxidized at 1100 °C. This different behavior might be caused by a denser and better oxidized Si3N4 coating when firing is performed at higher temperature. Snapshots of the solidification process of Si on a zyarock substrate coated with Si3N4 powder oxidized at 900 °C have been extracted from the respective recorded video (Figure 4). They reveal the presence of a solid particle that floats on the surface of the liquid droplet and acts as the nucleation center for the Si solidification. At t=0 s the particle is observed on the surface (inside the black spot of the pyrometer). At t=2 s, nucleation of Si occurs on that particle, then thin solid strips cross the whole droplet and crystal grows until the droplet is fully solidified. Figure 5a presents the morphology of the above mentioned particles found on the surface of the Si droplet solidified on a zyarock substrate coated with Si3N4 oxidized at 900 °C (red circle in Figure 5b). They are identified as β-Si3N4 by X-ray diffraction analysis. From the same figure (Figure 5b) it can also be noticed that the sintered Si3N4 substrate is wetted by the molten Si and the solidified droplet is stuck on it. On the contrary, the oxidized Si3N4 powder coating is not wetted by the Si droplet which is easily detached from the substrate. Whalen et al. [26] measured wetting angles of 10° to 43° on hot pressed Si3N4 in vacuum while Drevet et al. [27] found that molten Si on
12
sintered Si3N4 has a wetting angle of 82° which decreases to 49° over time in argon. The non – wetting behavior between Si and oxidized Si3N4 has been investigated in details [19],[28][29] and has been shown to result from the interaction with the SiO2 formed on the Si3N4 during its firing. SiO2 dissolution and infiltration of molten Si in the Si3N4 porosity occur when contact time is prolonged but it does not appear in the present experiments. The micrograph of the solidified droplet in cross section (Figure 6), shows that it is mainly composed of a single – crystal. However, nucleation of small parasitic grains occurs at the bottom of the droplet (Figure 6, view from the bottom), where the liquid Si is in contact with the substrate, leading to the growth of dendrites and twins. Appapillai et al. [17]-[18] have also observed the presence of nucleation sites when they performed melting – solidification experiments of Si coated with a Si3N4 layer. A first conclusion is that nucleation of Si is strongly promoted by the contact with Si3N4. More precisely, it appears that the β-Si3N4 polytype is a better nucleating agent compared to α-Si3N4. By the way, the β-Si3N4 nucleating particle that floats on the surface of the liquid Si obviously has not been detached from the α-Si3N4 coating. It is more probable that nitrogen dissolution – precipitation process took place, eventually favoring the nucleation of Si. M. Beaudhuin et al. [13] intensively investigated the interaction of nitrogen with liquid and solid Si in an electromagnetic levitation set-up. They have found that the increase of nitrogen concentration in the liquid decreases the Si undercooling temperature. They have also observed that the morphologies of the Si3N4 precipitates depend on the growth conditions and that nucleation was in relation with first-to-grow β-Si3N4 needles. Taking into account that epitaxial relationships between two solid materials is likely to promote nucleation [9], the nucleating effect of
13
β-Si3N4 is certainly linked to the fact that the (0001) face of β-Si3N4 shows a good epitaxy with (111) Si at high temperature. This is due to a small misfit (1.1%) between a 2×2 unit cell on (111) Si and a unit cell on the (001) plane of the nitride, as well as atomic arrangement reconstruction at the interface [30]-[31] . In the case of pyrolytic BN (PBN) and hot isostatically pressed BN (HIP BN) substrates, the temperature profiles during melting and solidification are similar to that of the Si3N4 substrates. After five different experiments the measured undercooling values are found to be in the range [0 K to 3 K] for both PBN and BN HIP denoting that BN favors the nucleation of Si. Real time observation of the melting – solidification process shows that several solid particles float on the surface of the liquid droplet. The optical micrographs of those precipitates (Figure 7a), are visible macroscopically on the surface of the solidified droplet, reveal their hexagonal shape and that they are composed of two different phases; a light one in the periphery and a dark one mainly in the center. These structures are identified as BN with the aid of X-ray diffraction analysis. A further investigation of those phases is done by micro – Raman analysis. Raman spectra show a broaden c-Si peak at 520 cm-1, denoting that solidified Si near the BN precipitates is highly doped with B. A second peak set at 620cm−1 is present and it is related to the substitutional boron isotopes (11B) [32]. In addition, the micro – Raman spectra of the dark phase present a peak set at 1370 cm−1 that corresponds to the h-BN phase [33]. However, in contrast to the previous case of the silicon nitride substrates, those particles do not seem to act as nucleation sites for the Si solidification. As illustrated in the optical micrograph of the solidified droplet on a PBN substrate in cross section (Figure 7b), significant nucleation of Si happens on the substrate. This
14
interaction leads to the solidification of a droplet composed of multiple crystal grains, giving several dendritic and twin structures. In addition, the BN substrate is not wetted by the liquid Si and the solidified droplet can be easily detached. A closer look at the interface between BN substrate and solidified Si reveals the formation of a Si3N4 intermediate layer (Figure 7c, layer 2) as identified by EDS analysis. Drevet et al. [25] have observed the formation of large (several micrometers in size) Si3N4 particles at the Si/BN interface due to the reaction: ( )
(( )) .
As already mentioned, such a Si3N4 layer is likely to favor the heterogeneous nucleation of Si on it, rather than on BN, as shown by the absence of nucleation on the floating BN particles. Furthermore, this Si3N4 layer shows that the floating BN particles are most probably detached from the substrate, rather than formed due to a dissolution – precipitation process. As already discussed, the solidification of the droplet occurs from the cold periphery toward the center. However, nucleation does not begin from the cold particles floating on the top, showing that BN has no significant nucleating effect. Similar observations are made when the Si droplet is solidified on a BN HIP substrate.
3.3. Carbon – containing substrates Four different types of materials, isostatic graphite, glassy carbon, 4H-SiC and 6H SiC, have been investigated. For the three first substrates, no undercooling temperature is measured by the bichromatic pyrometer during the solidification of Si. The video recorded during the solidification process (Figure 8) reveals that in all the examined cases, heterogeneous nucleation takes place on solid particles that float on the surface of
15
the liquid droplets and on the carbide substrate as well. This is confirmed by the occurrence of solid strips that appear at the liquid Si/substrate interface (at the edge of the droplet) and grow crossing the surface of the droplet.
These solid particles are macroscopically visible on the surface of the solidified droplet as well, as a dark phase indicated by the red circle in Figure 9a. The morphology of those particles that also act as nucleation sites for the Si solidification can better be seen in Figure 9 b and c. With the aid of a micro – Raman analysis they have been identified as 3C-SiC (characteristic Raman peaks at 796 and 980 cm-1). The nucleation of a Si dendrite on the 4H-SiC substrate is illustrated in Figure 9a (yellow circle) confirming that carbon based substrates strongly favors the Si nucleation. From the same figure it can also be concluded that carbon – containing substrates are wetted by the molten Si: in the case of the isostatic graphite the liquid Si is totally absorbed by the substrate while in the cases of glassy carbon and SiC the solidified droplets are stuck to the substrates. The cross section micrographs of the solidified droplets (Figure 10 a and b) illustrate the nucleation and growth of multiple grains at the droplet/substrate interface. In the case of glassy carbon substrate (a) their growth is blocked by the presence of a longitudinal dendritic structure in the middle part of the droplet (see also Figure 11). While, in the case of 4H-SiC substrate those small Si grains that nucleate at the interface grow and become larger toward the top of the droplet.
16
A closer look at the solidified silicon/glassy carbon interface reveals the presence of an intermediate layer (Figure 10 c). This layer is the result of the liquid Si reaction with carbon forming a SiC phase at the interface according to the reaction: ( )
( )
( ).
The formation of this SiC intermediate layer strongly favors nucleation. It is well known that there is a 20% misfit between 3C-SiC and Si at high temperature [34]. Furthermore, compared to the nitride case, reconstruction of the SiC/Si interface has never been observed and this makes the hetero – epitaxial growth of SiC on Si a challenging task, only solved by complex growth processes [35]-[36]. Therefore the nucleating effect of 3C-SiC cannot be explained in the same way that for nitrides, and remains mysterious. The case of the 6H-SiC substrate is very different as an undercooling of 40 K is measured by the pyrometer. There is no reason that 4H-SiC and 6H-SiC behave in a different way since the two substrates have the same crystalline orientation and the same surface polarity (both being Si-face terminated). The most reasonable explanation is that a thicker silicon oxide layer is present on the 6H-SiC substrates. Therefore, it is hypothesized that this high undercooling is due to a contact with silicon oxide, in agreement with results presented in paragraph 3.1. Incidentally, this allows discovering the fact that, SiC-layered crucibles could be used in the future for Si solidification, being protected against nucleation by an oxide layer.
3.4. Observations common to all substrates From the obtained temperature plots, such as in Figure 2, it appears that the duration of the solidification plateau decreases when the undercooling increases. This
17
shows that the fraction, f, of the material solidified at the end of recalescence is proportional to the undercooling ΔT, following the classical relationship (cp heat capacity, ΔH latent heat of solidification):
f
c p T
H
Comparison of all experiments shows that the duration of solidification of the remaining liquid is not significantly influenced by the nature of the substrate. This indicates that the heat conductivity of the substrate material (increasing from the insulating silica to the conducting graphite) does not affect the droplet solidification duration, clearly showing that droplets cool down from the top, by radiation to the cold surrounding. Independently of the material of the substrate, in many cases dendritic structures are observed in the center of the solidified droplet (Figure 11). This phenomenon is attributed to the collapse of the solidification front coming from the colder surface of the droplet towards its center, following different directions. This observation shows that the inner part of the droplet is at higher temperature in comparison to the surface and thus solidification proceeds from the surface to the center with, however, the last to freeze liquid rejected outside in order to produce the dilatation “horn”. Nucleation events occur either on floating particles or on the substrate, rather than in the center of the liquid.
4. Conclusions Nucleation experiments of Si droplets solidified on various substrates have been performed in a specially designed experimental installation. These experiments
18
investigate many different substrate materials, under identical conditions, in order to allow the comparison of their potent effect on Si nucleation. Also, compared to previous studies on this topic, pyrometric measurement of undercooling, as well as simultaneous camera video recording of the droplet solidification, appear to be important ways of understanding the experimental observations. Generally, experiments have been repeated two to four times, so that statistics remain poor and some conclusions might be taken cautiously. Undercooling is found to be significantly related to the nature of the substrate and the presence or not of precipitates/particles that favor the nucleation of Si. Oxide substrates present the highest undercooling values while heterogeneous nucleation site are not unambiguously observed, neither on the substrate nor on formed precipitates. By the way, because those samples are single crystals, only one nucleation event occurs, the location of which is very difficult to find inside the sample. In the case of nitride substrates, small undercooling values are measured and heterogeneous nucleation takes place on β-Si3N4 precipitates formed by a nitrogen dissolution – precipitation process occurring at the silicon/substrate interface. Surprisingly, β-Si3N4 is a good nucleating agent, while α-Si3N4 appears to be less. This nucleating effect of β-Si3N4 is possibly linked to the relative ease of epitaxy on Si. An important result is that oxidation of the Si3N4 layer at 1100 °C, instead of 900 °C, increases the undercooling from 0 K to 6 K. In the case of BN substrates, BN particles found in the droplet do not act as nucleation sites. Finally, during the Si solidification on carbon – containing substrates, no undercooling is measured and heterogeneous nucleation occurs on an intermediate SiC layer at the substrate/silicon interface and on SiC particles observed on the surface of
19
the droplets. However, this nucleating ability of SiC remains unexplained. One SiC substrate, presumably oxidized, gives a high undercooling. In the literature often clues about the expected undercooling temperature are given based on the wetting behavior of the material. From the present work it becomes clear that a direct correlation between the wetting behavior and the undercooling temperature cannot be made. When Si is solidified on the oxide substrates, the highest undercooling, among the studied substrates, are obtained. However the oxide substrates are rather well wetted by the molten Si (wetting angle lower than 90°). In contrast the silica and zyarock substrates coated with an oxidized Si3N4 layer are not wetted at all by the Si while the measured undercooling temperatures are very small. This is also the case for BN substrates. Also, comparing fused silica and zyarock cases, the roughness and the porosity of the substrate does not have a significant effect on the undercooling values. All in one, it is not possible to find any general behavior able to explain in which cases nucleation of Si is promoted: each substrate material appears to be specific.
Acknowledgments The authors are indebted to the substrate providers and would like to thank G. Fournier and C. Garnier who were involved in the experimental installation development. We would also like to thank Dr. D. Camel, Dr. B. Drevet and Prof. N. Eustathopoulos for inestimably rich discussions. The financial support from the French National Research Agency (ANR-HABISOL-Mosaïque) is also gratefully acknowledged.
References
20
[1] K. Fujiwara, W. Pan, N. Usami, K. Sawada, M. Tokairin, Y. Nose A. Nomura, T. Saishido, K. Nakajima, Acta mat. 54 (2006) 3191. [2] T.Y. Wang, S.L. Hsu, C.C. Fei, K.M. Yei, W.C. Hsu, C.W. Lan, J. Crystal Growth 311 (2009) 263. [3] F. Jay, D. Munoz, T. Desrues, E. Pihan, V.A. Oliveira, N. Enjalbert, A. Jouini, Sol. Energ. Mat. Sol. C. 130 (2014) 690. [4] R. R. Prakash, T. Sekiguchi, K. Jiptner, Y. Miyamura, J. Chen, H. Harada, K. Kakimoto, J. Cryst. Growth 401 (2014) 717. [5] K. F. Kelton, A. L. Greer, Nucleation in condensed matter, Elsevier 2010, Chapter 6. [6] M. Wang, H. Zheng, X. Lin, W. Huang Mat. Sci. Eng. 27 (2011) 012006. [7] D. Turnbull J. Chem. Phys. 18 (1950) 198. [8] H. Tian, D.M. Stefanecsu, P.A. Curreri Met. Trans. A. (1990) 241. [9] Z. Fan Met. Mat. Trans. A 44A (2013) 1409. [10] T. Duffar, P. Dusserre, N. Giacometti, P. Boiton, J.P. Nabot, N. Eustathopoulos J. Cryst. Growth 198/199 (1999) 374. [11] C. Panofen and D.M. Herlach Appl. Phys. Letters 88 (2006) 171913. [12] R. P. Liu, T. Volkmann and D. M. Herlach Acta Mater. 49 (2001) 439. [13] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti J. Cryst. Growth 336 (2011) 77. [14] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti, J. Mater. Sci. 45 (2010) 2218. [15] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti, T. Indian I. Metals 62 (2009) 505. [16] H. W. Tsai, M. Yang, C. Chuck, C.W. Lan J. Cryst. Growth 363 (2013) 242. [17] A. Appapillai and E. Sachs J. Crys. Growth 312 (2010) 1297. [18] A. Appapillai, C. Shachs and E. Sachs J. Appl. Phys. 109 (2011) 084916.
21
[19] I. Brynjulfsen, L. Arnberg J. Cryst. Growth 331 (2011) 64. [20] L. Alphei, A. Braun, V. Becker, A. Feldhoff, J.A. Becker, E. Wulf, C. Krause, Fr.W. Bach, J. Crystal Growth 311 (2009) 1250. [21] K. Fujiwara, K. Maeda, H. Koizumi, J. Nozawa, S. Uda, J. Appl. Phys. 112 (2012) 113521. [22] L. Sylla, A. Hodroj, N. Mangelinck-Noel, T. Duffar, 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, 7-11 Juin 2004, pp.3775. [23] M. G. Tsoutsouva, T. Duffar, C. Garnier and G. Fournier Cryst. Res. Technol. 50 (2015) 55. [24] Z. Jian, K. Nagashio and K. Kuribayashi Metall. Mater. Trans. A 33 (2002) 294. [25] B. Drevet, N. Eustathopoulos J. Mater. Sci. 47 (2012) 8247. [26] T. Whalen, A. Anderson J. Am. Ceram. Soc. 58 (1975) 396. [27] B. Drevet, R. Voytovych, R. Israel, N. Eustathopoulos J. Am. Ceram. Soc. 29 (2009) 2363. [28] B. Drevet, O. Pajani, N. Eustathopoulos Sol. Energ. Mat. Sol. C. 94 (2010) 425. [29] C. Huguet, C. Dechamp, R. Voytovych, B. Drevet, D. Camel, N. Eustathopoulos, Acta Mater. 76 (2014) 151. [30] X. Wang, G. Zhai, J. Yang, N. Cue, Phys. Rev. B 60 (1999) R2146 [31] M. E. Bachlechner, J. Zhang, Y. Wang, J. Schiffbauer, S. R. Knudsen, D. Korakakis, Phys. Rev. B 72 (2005) 094115. [32] M. Becker, U. Gösele, A. Hofmann, S. Christiansen J. Appl. Phys. 106 (2009) 074515. [33] M. Ben el Mekki, M. A. Djouadi, E. Guiot, V. Mortet, J. Pascallon, V. Stambouli, D. Bouchier, N. Mestres, G. Nouet Surf. Sci. Technol. 116-119 (1999) 93.
22
[34] A. Gupta, C. Jacob, Prog. Crystal growth Charact. Mat. 51 (2005) 43. [35] S. A. Kukushkin, A. V. Osipov, N. A. Feoktistov, Phys. Solid State, 56 (2014) 1507. [36] G. Ferro, Crit. Rev. Solid State Mat. Sci. 40 (2015) 56.
23
Figures and captions
Figure 1: Schematic illustration of the apparatus used to perform undercooling measurements and video recording experiments described in this work.
24
Figure 2: Temperature profile of the Si sample during the melting and solidification process on a fused silica substrate. Inlet shows a zoom of the plot during the recalescence – solidification event.
25
Figure 3: Optical micrographs of the solidified droplets on the oxide substrates in cross section showing the final grain structures. All droplets in this study present a “horn”, as illustrated in the zyarock case.
26
Figure 4: Morphologies of growing crystals during the solidification of Si on a zyarock substrate coated with Si3N4 powder oxidized at 900°C. Since the camera looks at the sample through the pyrometer window, the black ring that appears in all pictures is the pyrometer measurement spot. The surrounding dark grey circle is a window reflection on the top of the droplet. The grey parts are the undercooled liquid and bright parts are crystal solidified from the undercooled liquid. The square substrate appears in dark grey.
27
Figure 5: a) Optical micrograph of the β-Si3N4 precipitate found on the surface of the Si droplet solidified on a zyarock substrate coated with Si3N4 oxidized at 900 °C, b) macroscopic side/top views of the solidified Si droplets on a zyarock substrate coated with Si3N4 oxidized at 900 °C and on sintered Si3N4. Note the “horn” that appears at the end of solidification due to the solid – liquid density gap.
28
Figure 6: Optical micrograph in a cross section view and in a view from the bottom (after partial substrate lapping) of the solidified droplet on a silica substrate coated with Si3N4 oxidized at 900 °C.
29
a)
b)
c)
Figure 7: a) Optical micrograph of the precipitates found on the surface of the Si droplet solidified on a PBN substrate, b) optical micrograph, in cross section, of the Si droplet solidified on a PBN substrate, c) SEM cross section of the interface between Si and the PBN substrate. Layers 1, 2, 3 correspond to the solidified Si, Si3N4 intermediate layer and the BN substrate respectively.
30
Figure 8: Crystal growth process during the solidification of silicon on carbon – containing substrates.
31
Graphite
Glassy carbon
4H-SiC
a)
b)
c)
Figure 9: a) Macroscopic side/top views of the solidified Si droplets on carbon – based substrates. b) SiC precipitates on the surface of the solidified Si droplet, c) SiC precipitate that acts as nucleation center for the growth of a faceted dendrite.
32
Figure 10: Optical micrograph of the cross section of Si droplets solidified on (a) glassy carbon and (b) 4H-SiC substrate. c) SEM cross section of the interface between the Si droplet and the glassy carbon substrate where 1) solidified Si droplet, 2 and 3) SiC layer and 4) glassy carbon substrate.
33
Figure 11: Optical microscopy observation (extracted from Figures 3, 6, 7 b and 9 a) of the central part of the droplets cut in cross section.
34
List of tables Table 1: List of the investigated substrates and obtained experimental results. Substrate
Provider
Undercooling ±1 K
Droplet structure
Analyzed precipitates
Fused Silica
Mondia quartz
98, 75, 45
Single-crystal
No precipitate
(100) Sapphire
RSA
67, 42, 36
Zyarock
Vesuvius
49, 47, 41
INES
2, 3
Grains and twins
Si3N4
β-Si3N4 precipitates on top of droplet
INES
0, 0, 0, 0
Grains and twins
Si3N4
β-Si3N4 precipitates on top of droplet
INES
6, 6
Grains and twins
-
β-Si3N4 precipitates on top of droplet
Sintered Si3N4
INES
4
-
-
Pyrolitic BN (PBN)
MCSE
3, 0, 0, 0
Stressed grains
Si3N4
Hot isostatically pressed BN (HIPBN)
MCSE
21, 9
Grains and twins
BN
Isostatic graphite
MERSEN
0, 0
Polycrystalline
3C-SiC
Glassy Carbon
MERSEN
0, 0, 0
Polycrystalline
3C-SiC
6H SiC Si face
SiCrystal AG
44, 12
-
-
4H SiC Si face
SiCrystal AG
0
-
3C-SiC
Carbon/SiC (Carbosil)
MERSEN
5, 2, 0
Grains and twins
3C-SiC
Silica+Si3N4 powder fired at 900°C Zyarock+Si3N4 powder fired at 900°C Zyarock+Si3N4 powder fired at 1100°C
Distorted single-crystal Few grains and twins
No precipitate No precipitate
Location of nucleation Substrate Substrate Substrate
β-Si3N4 precipitates on top of droplet Si3N4 layer grown on substrate Si3N4 layer grown on substrate 3C-SiC precipitates on top of droplet and at substrate surface 3C-SiC precipitates on top of droplet and substrate surface Unknown 3C-SiC precipitates on top of droplet and substrate surface 3C-SiC precipitates on top of droplet and substrate surface
35
Highlights 1 – Set-up and process have been developed for study of Si nucleation on substrates. 2 – Oxides, nitrides and carbon-based substrates have been studied. 3 – Oxides show high undercooling and only one nucleation event. 4 – Nitrides promote nucleation via precipitates according to good epitaxial fit with Si. 5 – Carbon based substrates are very potent substrates but the reason is not clear.
36
PII: DOI: Reference:
S0022-0248(16)30373-6 http://dx.doi.org/10.1016/j.jcrysgro.2016.07.018 CRYS23463
To appear in: Journal of Crystal Growth Received date: 27 January 2016 Revised date: 22 June 2016 Accepted date: 16 July 2016 Cite this article as: M.G. Tsoutsouva, T. Duffar, D. Chaussende and M. Kamguem, Undercooling measurement and nucleation study of silicon droplets on various substrates, Journal of Crystal Growth, http://dx.doi.org/10.1016/j.jcrysgro.2016.07.018 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.
Undercooling measurement and nucleation study of silicon droplets on various substrates M.G. Tsoutsouvaa, T. Duffara*, D.Chaussendeb, M. Kamguema a
CNRS, Université Grenoble Alpes, SIMaP–EPM, 38402 Saint Martin d’Héres, France b
CNRS, Université Grenoble Alpes, LMGP, 38000 Grenoble, France
* Corresponding author: [email protected]
Abstract The heterogeneous nucleation of solid silicon is studied when molten droplets solidify on various substrates. An experimental installation has been developed in order to record in real time the melting–solidification process, measure the undercooling temperature and look at the solidification of the droplets. Three different categories of substrate materials are studied: oxides (silica, zyarock and sapphire), nitrides (silica+oxidized Si3N4 coating, zyarock + oxidized Si3N4 coating, sintered Si3N4, PBN and HIP – BN) and carbon – containing (isostatic graphite, glassy carbon and SiC). Higher undercooling values are measured in the case of oxide substrates where the solidified droplet is found to be mainly composed of a single – crystal. In the case of nitride substrates, a dissolution/precipitation process takes place and β-phase Si3N4 precipitates are found to act as nucleation centers for the silicon solidification. The nucleating power of Si3N4 is attributed to the good epitaxial fit with silicon. Oxidation of Si3N4 powder at a higher temperature increases the undercooling of the droplet. When the silicon droplet is solidified on BN substrates, BN particles are detected on the surface of the droplet as well as a Si3N4 layer at the substrate/Si interface which promotes nucleation. Carbon – containing substrates are found to favor the nucleation of
1
silicon due to the creation of a SiC layer at the substrate/silicon interface and precipitation of SiC particles in the droplets. However, no explanation of the important nucleating effect of SiC has been found. Keywords: A1. Nucleation; A1. Solidification; A2. Growth from melt; B2. Semiconducting silicon. 1. Introduction An important challenge in mold solidification of photovoltaic (PV) silicon (Si) is the efficient control of the crystalline structure, grain sizes and orientations in the solidified ingot, as it interacts with the solar cell efficiency by the way of carrier recombination at grain boundaries. Several techniques are promoted with various industrial successes in order to control the final grain structure in the grown ingot: -
The standard process, used for many years, applies a sudden cooling at the bottom of the crucible in order to induce an undercooling and associated nucleation [1].
-
A variant of this process introduces some highly heat conducting pieces at the bottom of the crucible in order to promote the nucleation of few dendrites, likely to impose their crystallographic orientation and generate large, well oriented grains [2].
-
A paving of single crystal seeds may also be placed at the bottom of the crucible in order to get a Si single crystal [3]. However spurious nucleation on the crucible walls largely decreases the yield of the process.
-
On the contrary, another variant uses many randomly oriented seeds in order to promote grain selection, which has been proved to improve the material overall quality [4]. However, uncontrolled nucleation on crucible walls is still an issue.
2
Clearly, a better understanding and control of the nucleation and growth phenomena in PV Si casting is necessary. The theory of heterogeneous nucleation is well advanced [5] however, in its classical treatment, involves a contact angle representing the equilibrium of seed – liquid – substrate interfacial energies. This angle is aiming to describe the mechanism promoting the nucleation event on the foreign substrate, which is highly dependent on both the nucleating and nucleant materials. Influence of many different parameters has been invoked, such as roughness [6], cavities in the substrate wall [7], chemical reactions [8] or epitaxial relationships between the two solid materials [9]. Also possible correlation with the wetting angle of the liquid on the crucible material has not been elucidated so far [10]. However, the predictive ability of the current state of the art is low and the only way to study the nucleating ability of a substrate facing cooling liquid Si is to perform experiments. In PV casting technology, the crucible is classically made of sintered silica (called Zyarock) coated with a layer of silicon nitride (Si3N4) powder. This layer is fired under air at various temperatures and durations, depending on the specific process. The Si3N4 powder coating constitutes a barrier to wetting and infiltration due to the presence of non-wettable SiO2 formed on the Si3N4 particles during their firing. This layer also acts as mechanical fuse due to its weak mechanical resistance resulting from its porous microstructure, allowing the easy detachment of the ingot from the crucible. Depending on the particular industrial process, the oxidation temperature of the nitride powder layer varies from 900°C to 1100°C. Besides, other materials, based on silicon carbide layers are presently under study with the aim to get reusable crucibles. Also, boron nitride crucibles are sometimes used for research purpose, in spite of their deleterious
3
boron doping. Therefore the nucleating ability of various kind of oxide, carbide and nitride substrates should be investigated. An efficient method to study the nucleation phenomenon is by measuring the undercooling temperature during the cooling process since it gives information about whether or not heterogeneous nucleation is favored. Some researchers performed undercooling measurements in an electromagnetic levitation setup either concerning pure Si or Si containing a controlled percentage of impurities [11]-[15]. Those experiments have clearly shown the correlation of the measured undercooling temperature with the concentration of gaseous impurities surrounding the droplet, leading to dissolution and precipitation of particles (Si3N4, SiC) in the melt, then solid Si nucleation on the precipitates. The nucleating ability of β-Si3N4 (almost 0 K of undercooling) was shown to be higher than in the case of 3C-SiC precipitates (3 or 4 K of undercooling at high concentration). However, this is only an indirect measurement of the nucleating action, not taking into consideration the role that the substrate (crucible or mold walls) plays. Tsai et al. [16] investigated the firing conditions of the Si3N4 coating on the nucleation and grain structure of solidified Si by sessile drop and directional solidification experiments. It was observed that the preferred growth orientation was affected by the coating when the cooling rate was low, while it was not affected at all in case of high cooling rate. Appapillai et al. [17]-[18] studied nucleation by using differential scanning calorimetry to measure the undercooling below the melting temperature of molten Si encapsulated in silicon oxide and nitride layers. Various coatings on the Si samples were tested for their influence on nucleation. Both dry and wet oxide coatings presented higher interfacial stability resulting in higher undercooling values than the nitrides coatings. However, the undercooling
4
measurements presented significant variations, probably due to the coating nonuniformities and the overall sample deformations during melting and solidification process. Brynjulfsen and Arnberg [19] studied the nucleation undercooling phenomenon during the solidification of Si on SiO2 coated with oxidized Si3N4, and on Al2O3, by the sessile drop method. Substrates provided by the commercial crucible producer for photovoltaic applications, Vesuvius, were also investigated. They concluded that the coating parameters such as thickness, roughness and the amount of oxygen in the coating have no significant influence on the undercooling of Si. However, they used a thermocouple placed below the sample holder in order to measure the undercooling temperature interval and this method cannot give accurate values of the undercooling. Alphei et al. [20] studied the nucleation of Si in alumina crucible and found undercooling values as high as 100 K, however, when amorphous Si3N4 nanoparticles were added to the droplet, the undercooling decreased to 2 – 6 K and nucleation was apparently triggered by nitride crystallites. Fujiwara et al. [21] theoretically studied the effect of the Si – substrate interfacial energy on the orientation of nucleated Si grains. However, as these energies are unknown, it is difficult to draw practical conclusions. The present work has been performed in order to provide additional data on the influence that the material of the substrate, on which Si is solidified, has on the undercooling temperature and the initial nucleation of solid Si. It follows former experiments [22] with new objectives: different materials of the substrate (oxides, nitrides, carbon-containing) are investigated, Si is placed and melted directly on the substrate (this avoids the problems of non-uniformity of the coating and deformation of the sample experienced in [18]), a pyrometer is used for measuring the temperature directly on the Si droplet and in addition the melting – solidification process is
5
visualized in situ and in real time. The final crystalline structure of the solidified Si droplet is also investigated.
2. Experimental An experimental installation has been designed and implemented (Figure 1) in order to measure the undercooling temperature and visualize the Si droplet solidification. An electronic grade Si cube (0.8±0.2 g) is mechanically polished with SiC abrasive papers with increasing fineness (from 600 to 2000 grid), ultrasonically cleaned in acetone, etched in a HF solution for 30 s and rinsed with distilled water. It is then quickly (less than 15 min until vacuum is applied) placed on a substrate which is contained in a cylindrical crucible made of pure fused silica, for which it has already been proven that high undercooling can be obtained [22]. The mass of Si is optimized in relation to the dimensions of the substrates so as, when the Si is fully melted, the liquid droplet is fixed and centered on the substrate edges. This allows the accurate and reproducible measurement of the temperature by the pyrometer. All substrates have the same surface (10 mm × 10 mm). Heating and thus melting of Si is realized by electromagnetic induction with the aid of a graphite susceptor. The whole system is housed inside a water cooled stainless steel vacuum chamber evacuated to a pressure of 3∙10-6 mbar with a turbomolecular pump (TMP), then rinsed and backfilled with 6N Ar to a pressure of 1000 mbar (Figure 1). The sample temperature is monitored in real time and a constant inductive power is maintained as soon as the melting plateau is observed on the pyrometer reading. For all the experiments the cooling rate during solidification is 300 K.min-1, obtained by switching off the inductive current. The Si temperature during the melting –
6
solidification process and undercooling is measured by means of an IRCON Modline 5 5R-1810 bichromatic pyrometer operating in the temperature range 700°C to 1800°C (spectral range 0.75 μm to 1.05 μm). A camera is also installed above the pyrometer as shown in Figure 1 and the whole process is recorded in real time, through the pyrometer viewing window. More details about the experimental installation and the melting – solidification procedure are given in [23]. The output of each experiment is the recorded temperature profile that gives the undercooling temperature measurement, the video film showing the melting and solidification process and the solidified Si droplet samples. In the frame of this research, experiments are performed on different substrates that can be categorized as oxides (silica, sapphire and zyarock), nitrides (silica+ oxidized Si3N4 coating, zyarock+ Si3N4 oxidized coating, sintered Si3N4, Pyrolytic BN -PBN- and hot isostatically pressed BN HIP BN) and carbon-containing (isostatic graphite, glassy carbon and SiC). After solidification, the droplets are embedded in resin and cut in cross section. They are polished using diamond pastes down to 1μm and chemically etched for 5 sec in a solution containing nitric acid, hydrofluoric acid and acetic acid in the ratio 2:1:1 to reveal their microstructure. In certain cases the samples are also polished from the bottom, in order to remove the substrate sticking to the droplet, and the microstructure of the surface that is directly in contact with the substrate is investigated. The microstructural characterization of the solidified droplets is performed with the aid of an optical microscope, a scanning electron microscope (SEM) equipped with an energy dispersive spectrometer (EDS) and synchrotron X-ray Bragg diffraction imaging. In certain cases the observed precipitates are identified by micro – Raman and X-ray diffraction techniques.
7
3. Results and discussion Table 1 gathers the investigated substrates and the obtained experimental results. 3.1. Oxide substrates Figure 2 illustrates a representative temperature profile of the Si sample during its melting and solidification process on a fused silica substrate, as recorded by the bichromatic pyrometer. During melting, the sample absorbs the melting enthalpy and the temperature remains almost constant as it can be seen from the presence of a temperature plateau in the bichromatic pyrometer temperature curve. Once the Si is fully melted the temperature rises (at 20 min on Figure 2), since the enthalpy of fusion is no more absorbed by the sample, until it stabilizes at 1840±10 K. Seven minutes later the electric power is cut abruptly and the Si droplet starts to cool down (cooling rate 300 K.min-1). Rapid non – equilibrium solidification of the undercooled droplet then proceeds in several stages. Nucleation occurs at TN=1564 K, as presented in Figure 2 and then the recalescence phase (quasi-adiabatic solidification) follows. During this period, the temperature of the droplet rapidly increases (abrupt temperature rise observed at 27.25 min) due to the release of latent heat of solidification. In the next period of quasi – isothermal plateau, the temperature of the droplet remains almost constant at the melting point until the solidification process is completed. The isothermal plateau after recalescence corresponds to the solidification temperature (TS), well known as 1687 K for Si. However, this temperature is not in accordance with the Ts (1653 K) measured by the pyrometer after recalescence, as shown in Figure 2. This discrepancy can be due to the fact that during melting the pyrometer focuses on solid Si, with a given emissivity, while before and after
8
recalescence, it focuses on liquid Si with another emissivity. Despite that the two color pyrometer is specially designed so that by means of a mathematical ratio calculation, certain influences on measurement can be eliminated, the error of the temperature measurement is ± 30 K. In this work it is considered that the maximum nucleation undercooling (ΔT) can be approximated as the difference between TS and TN and thus in the present experiment ΔT = 89 K. Considering the pyrometer sensibility (0.1 K) the accuracy of the measured undercooling is estimated to be ± 1 K. Experiments have been repeated three times and the measured undercooling values are in the range of [42 to 89 K]. The final temperature decrease represents the cooling of the solidified droplet. The temperature profile is found to present a similar tendency, when the Si sample is melted and solidified on a (100) sapphire or a zyarock substrate. However, the measured undercooling values are in the range [36 to 67 K] in case of (100) sapphire substrate and in the range [41 to 51 K] in case of zyarock. Zyarock is the standard material used for the industrial crucibles in the photovoltaic industry. It is a sintered silica product with an apparent porosity of 12% and typical composition: silica 99.6%, alumina 0.12%, iron oxide 0.10%, titania 0.027%, magnesia 0.013% sodium oxide 0.008%, calcia 0.007% and potassium oxide 0.005%. At the beginning of solidification, the recorded video shows thin solid strips crossing the whole droplet [23]. As time goes on, they develop to several continuous lines. Z. Jian et al. [24] have observed similar structures during rapid solidification experiments in an electromagnetic levitation setup. They suggested that the crystal growth depends on the nucleation undercooling and for undercooling in the range 40 to 100
10 K (similar to the present study) the initial crystal seed solidifies as dendrites
that form a number of plates. Apparently the process is the same in case of solidification
9
on a substrate. For all the studied oxide substrates, the solidified droplets do not spread but, instead, form at equilibrium a spherical cap resting on the substrate (see cross sections of droplets on Figure 3). The “horn” appearing at the end of solidification, as shown in the zyarock droplet cross section, is due to the 10% decrease of density between molten and solid Si. The grain structures, observed in cross section, reveal that the solidified Si droplets are mainly composed of a single crystal (Figure 3). No nucleation center is found at the silicon / substrate interface, which shows that nucleation event is quite rare and probably unique. The single crystal growth is confirmed by synchrotron white beam Xray Bragg diffraction imaging, not presented in the present paper. In the case of fused silica substrate, a twin crystal is observed on the left side of the droplet as well as a dendrite in its center (discussed in paragraph 3.4). The rest of the droplet is a singlecrystal. The droplet solidified on the sapphire substrate is composed of a single crystal which is highly distorted. Crystal deformation is revealed on the Laue spots of the white beam topographs, by the varying degree of asterism indicative of crystal mosaicity. In the case of zyarock substrate, the droplet is a single crystal as well, except a grain on the left side and multi-crystalline grains on the top of the “horn”. The droplets are highly fractured at the Si/substrate interface due to the difference in thermal expansion coefficients between Si and the substrate material and in the central area as well, due to the rapid solidification conditions. Many electromagnetic levitation experiments of pure Si have shown undercooling larger than 300 K. The contact with pure silica decreases this value to some tens of K (162 K) as measured by [22] but anyhow, it can be concluded that it is a bad nucleating
10
agent for Si. Chemical reaction exists between Si and SiO2, the oxide substrate slowly dissolves into the Si melt and oxygen is transported by diffusion and convection into the liquid. However, much of the oxygen evaporates from the melt free surface and only about 1% is incorporated into the growing crystal that apparently does not contribute to the nucleation process under the present experimental conditions. The somewhat smaller undercooling measured in the case of zyarock and sapphire shows that the chemical and/or roughness of the substrate impacts the nucleation behavior, but not that much. Based on the literature, it would be expected that the porosity of the zyarock substrate should increase drastically the nucleation probability of Si [7]. However, from the present experimental results such a conclusion cannot be drawn: the effect of the 12% open porosity is relatively small; the undercooling is decreased by 20 or 30 K but remains large, 45 K in average. Further, zyarock is not only pure silica and other oxide components might have a nucleating power. The effect of the sapphire substrate on Si nucleation is quite similar. This is consistent with the analysis proposed by Drevet and Eustathopoulos [25]: thermodynamic calculations show that Al2O3 is not reacting with Si but it is slightly dissolved and re-deposited as a SiO2 layer on the sapphire substrate. Therefore, the SiAl2O3 interface is transformed to a Si-SiO2 interface. Taking this into account, all those experiments can be considered as nucleation of solid Si on silicon oxide. This gives a mean undercooling of 55 K with a standard deviation of 20 K under the present experimental conditions, for the 9 experiments on oxide substrates.
3.2. Nitride substrates In contrast to the case of oxide substrates, no, or very low, recalescence is observed
11
on the temperature recording when Si is solidified on nitride substrates. For Si droplets on zyarock substrates coated with Si3N4 oxidized at 900 °C, experiments have been repeated four times and the measured undercooling values are in the range [0 K to 2 K]. Similar undercooling values are measured in the case of the fused silica substrate coated with Si3N4 oxidized at 900 °C and of the sintered Si3N4 substrate. A higher undercooling value of 6 K is observed in two experiments, when the Si is left to solidify on a zyarock substrate coated with Si3N4 oxidized at 1100 °C. This different behavior might be caused by a denser and better oxidized Si3N4 coating when firing is performed at higher temperature. Snapshots of the solidification process of Si on a zyarock substrate coated with Si3N4 powder oxidized at 900 °C have been extracted from the respective recorded video (Figure 4). They reveal the presence of a solid particle that floats on the surface of the liquid droplet and acts as the nucleation center for the Si solidification. At t=0 s the particle is observed on the surface (inside the black spot of the pyrometer). At t=2 s, nucleation of Si occurs on that particle, then thin solid strips cross the whole droplet and crystal grows until the droplet is fully solidified. Figure 5a presents the morphology of the above mentioned particles found on the surface of the Si droplet solidified on a zyarock substrate coated with Si3N4 oxidized at 900 °C (red circle in Figure 5b). They are identified as β-Si3N4 by X-ray diffraction analysis. From the same figure (Figure 5b) it can also be noticed that the sintered Si3N4 substrate is wetted by the molten Si and the solidified droplet is stuck on it. On the contrary, the oxidized Si3N4 powder coating is not wetted by the Si droplet which is easily detached from the substrate. Whalen et al. [26] measured wetting angles of 10° to 43° on hot pressed Si3N4 in vacuum while Drevet et al. [27] found that molten Si on
12
sintered Si3N4 has a wetting angle of 82° which decreases to 49° over time in argon. The non – wetting behavior between Si and oxidized Si3N4 has been investigated in details [19],[28][29] and has been shown to result from the interaction with the SiO2 formed on the Si3N4 during its firing. SiO2 dissolution and infiltration of molten Si in the Si3N4 porosity occur when contact time is prolonged but it does not appear in the present experiments. The micrograph of the solidified droplet in cross section (Figure 6), shows that it is mainly composed of a single – crystal. However, nucleation of small parasitic grains occurs at the bottom of the droplet (Figure 6, view from the bottom), where the liquid Si is in contact with the substrate, leading to the growth of dendrites and twins. Appapillai et al. [17]-[18] have also observed the presence of nucleation sites when they performed melting – solidification experiments of Si coated with a Si3N4 layer. A first conclusion is that nucleation of Si is strongly promoted by the contact with Si3N4. More precisely, it appears that the β-Si3N4 polytype is a better nucleating agent compared to α-Si3N4. By the way, the β-Si3N4 nucleating particle that floats on the surface of the liquid Si obviously has not been detached from the α-Si3N4 coating. It is more probable that nitrogen dissolution – precipitation process took place, eventually favoring the nucleation of Si. M. Beaudhuin et al. [13] intensively investigated the interaction of nitrogen with liquid and solid Si in an electromagnetic levitation set-up. They have found that the increase of nitrogen concentration in the liquid decreases the Si undercooling temperature. They have also observed that the morphologies of the Si3N4 precipitates depend on the growth conditions and that nucleation was in relation with first-to-grow β-Si3N4 needles. Taking into account that epitaxial relationships between two solid materials is likely to promote nucleation [9], the nucleating effect of
13
β-Si3N4 is certainly linked to the fact that the (0001) face of β-Si3N4 shows a good epitaxy with (111) Si at high temperature. This is due to a small misfit (1.1%) between a 2×2 unit cell on (111) Si and a unit cell on the (001) plane of the nitride, as well as atomic arrangement reconstruction at the interface [30]-[31] . In the case of pyrolytic BN (PBN) and hot isostatically pressed BN (HIP BN) substrates, the temperature profiles during melting and solidification are similar to that of the Si3N4 substrates. After five different experiments the measured undercooling values are found to be in the range [0 K to 3 K] for both PBN and BN HIP denoting that BN favors the nucleation of Si. Real time observation of the melting – solidification process shows that several solid particles float on the surface of the liquid droplet. The optical micrographs of those precipitates (Figure 7a), are visible macroscopically on the surface of the solidified droplet, reveal their hexagonal shape and that they are composed of two different phases; a light one in the periphery and a dark one mainly in the center. These structures are identified as BN with the aid of X-ray diffraction analysis. A further investigation of those phases is done by micro – Raman analysis. Raman spectra show a broaden c-Si peak at 520 cm-1, denoting that solidified Si near the BN precipitates is highly doped with B. A second peak set at 620cm−1 is present and it is related to the substitutional boron isotopes (11B) [32]. In addition, the micro – Raman spectra of the dark phase present a peak set at 1370 cm−1 that corresponds to the h-BN phase [33]. However, in contrast to the previous case of the silicon nitride substrates, those particles do not seem to act as nucleation sites for the Si solidification. As illustrated in the optical micrograph of the solidified droplet on a PBN substrate in cross section (Figure 7b), significant nucleation of Si happens on the substrate. This
14
interaction leads to the solidification of a droplet composed of multiple crystal grains, giving several dendritic and twin structures. In addition, the BN substrate is not wetted by the liquid Si and the solidified droplet can be easily detached. A closer look at the interface between BN substrate and solidified Si reveals the formation of a Si3N4 intermediate layer (Figure 7c, layer 2) as identified by EDS analysis. Drevet et al. [25] have observed the formation of large (several micrometers in size) Si3N4 particles at the Si/BN interface due to the reaction: ( )
(( )) .
As already mentioned, such a Si3N4 layer is likely to favor the heterogeneous nucleation of Si on it, rather than on BN, as shown by the absence of nucleation on the floating BN particles. Furthermore, this Si3N4 layer shows that the floating BN particles are most probably detached from the substrate, rather than formed due to a dissolution – precipitation process. As already discussed, the solidification of the droplet occurs from the cold periphery toward the center. However, nucleation does not begin from the cold particles floating on the top, showing that BN has no significant nucleating effect. Similar observations are made when the Si droplet is solidified on a BN HIP substrate.
3.3. Carbon – containing substrates Four different types of materials, isostatic graphite, glassy carbon, 4H-SiC and 6H SiC, have been investigated. For the three first substrates, no undercooling temperature is measured by the bichromatic pyrometer during the solidification of Si. The video recorded during the solidification process (Figure 8) reveals that in all the examined cases, heterogeneous nucleation takes place on solid particles that float on the surface of
15
the liquid droplets and on the carbide substrate as well. This is confirmed by the occurrence of solid strips that appear at the liquid Si/substrate interface (at the edge of the droplet) and grow crossing the surface of the droplet.
These solid particles are macroscopically visible on the surface of the solidified droplet as well, as a dark phase indicated by the red circle in Figure 9a. The morphology of those particles that also act as nucleation sites for the Si solidification can better be seen in Figure 9 b and c. With the aid of a micro – Raman analysis they have been identified as 3C-SiC (characteristic Raman peaks at 796 and 980 cm-1). The nucleation of a Si dendrite on the 4H-SiC substrate is illustrated in Figure 9a (yellow circle) confirming that carbon based substrates strongly favors the Si nucleation. From the same figure it can also be concluded that carbon – containing substrates are wetted by the molten Si: in the case of the isostatic graphite the liquid Si is totally absorbed by the substrate while in the cases of glassy carbon and SiC the solidified droplets are stuck to the substrates. The cross section micrographs of the solidified droplets (Figure 10 a and b) illustrate the nucleation and growth of multiple grains at the droplet/substrate interface. In the case of glassy carbon substrate (a) their growth is blocked by the presence of a longitudinal dendritic structure in the middle part of the droplet (see also Figure 11). While, in the case of 4H-SiC substrate those small Si grains that nucleate at the interface grow and become larger toward the top of the droplet.
16
A closer look at the solidified silicon/glassy carbon interface reveals the presence of an intermediate layer (Figure 10 c). This layer is the result of the liquid Si reaction with carbon forming a SiC phase at the interface according to the reaction: ( )
( )
( ).
The formation of this SiC intermediate layer strongly favors nucleation. It is well known that there is a 20% misfit between 3C-SiC and Si at high temperature [34]. Furthermore, compared to the nitride case, reconstruction of the SiC/Si interface has never been observed and this makes the hetero – epitaxial growth of SiC on Si a challenging task, only solved by complex growth processes [35]-[36]. Therefore the nucleating effect of 3C-SiC cannot be explained in the same way that for nitrides, and remains mysterious. The case of the 6H-SiC substrate is very different as an undercooling of 40 K is measured by the pyrometer. There is no reason that 4H-SiC and 6H-SiC behave in a different way since the two substrates have the same crystalline orientation and the same surface polarity (both being Si-face terminated). The most reasonable explanation is that a thicker silicon oxide layer is present on the 6H-SiC substrates. Therefore, it is hypothesized that this high undercooling is due to a contact with silicon oxide, in agreement with results presented in paragraph 3.1. Incidentally, this allows discovering the fact that, SiC-layered crucibles could be used in the future for Si solidification, being protected against nucleation by an oxide layer.
3.4. Observations common to all substrates From the obtained temperature plots, such as in Figure 2, it appears that the duration of the solidification plateau decreases when the undercooling increases. This
17
shows that the fraction, f, of the material solidified at the end of recalescence is proportional to the undercooling ΔT, following the classical relationship (cp heat capacity, ΔH latent heat of solidification):
f
c p T
H
Comparison of all experiments shows that the duration of solidification of the remaining liquid is not significantly influenced by the nature of the substrate. This indicates that the heat conductivity of the substrate material (increasing from the insulating silica to the conducting graphite) does not affect the droplet solidification duration, clearly showing that droplets cool down from the top, by radiation to the cold surrounding. Independently of the material of the substrate, in many cases dendritic structures are observed in the center of the solidified droplet (Figure 11). This phenomenon is attributed to the collapse of the solidification front coming from the colder surface of the droplet towards its center, following different directions. This observation shows that the inner part of the droplet is at higher temperature in comparison to the surface and thus solidification proceeds from the surface to the center with, however, the last to freeze liquid rejected outside in order to produce the dilatation “horn”. Nucleation events occur either on floating particles or on the substrate, rather than in the center of the liquid.
4. Conclusions Nucleation experiments of Si droplets solidified on various substrates have been performed in a specially designed experimental installation. These experiments
18
investigate many different substrate materials, under identical conditions, in order to allow the comparison of their potent effect on Si nucleation. Also, compared to previous studies on this topic, pyrometric measurement of undercooling, as well as simultaneous camera video recording of the droplet solidification, appear to be important ways of understanding the experimental observations. Generally, experiments have been repeated two to four times, so that statistics remain poor and some conclusions might be taken cautiously. Undercooling is found to be significantly related to the nature of the substrate and the presence or not of precipitates/particles that favor the nucleation of Si. Oxide substrates present the highest undercooling values while heterogeneous nucleation site are not unambiguously observed, neither on the substrate nor on formed precipitates. By the way, because those samples are single crystals, only one nucleation event occurs, the location of which is very difficult to find inside the sample. In the case of nitride substrates, small undercooling values are measured and heterogeneous nucleation takes place on β-Si3N4 precipitates formed by a nitrogen dissolution – precipitation process occurring at the silicon/substrate interface. Surprisingly, β-Si3N4 is a good nucleating agent, while α-Si3N4 appears to be less. This nucleating effect of β-Si3N4 is possibly linked to the relative ease of epitaxy on Si. An important result is that oxidation of the Si3N4 layer at 1100 °C, instead of 900 °C, increases the undercooling from 0 K to 6 K. In the case of BN substrates, BN particles found in the droplet do not act as nucleation sites. Finally, during the Si solidification on carbon – containing substrates, no undercooling is measured and heterogeneous nucleation occurs on an intermediate SiC layer at the substrate/silicon interface and on SiC particles observed on the surface of
19
the droplets. However, this nucleating ability of SiC remains unexplained. One SiC substrate, presumably oxidized, gives a high undercooling. In the literature often clues about the expected undercooling temperature are given based on the wetting behavior of the material. From the present work it becomes clear that a direct correlation between the wetting behavior and the undercooling temperature cannot be made. When Si is solidified on the oxide substrates, the highest undercooling, among the studied substrates, are obtained. However the oxide substrates are rather well wetted by the molten Si (wetting angle lower than 90°). In contrast the silica and zyarock substrates coated with an oxidized Si3N4 layer are not wetted at all by the Si while the measured undercooling temperatures are very small. This is also the case for BN substrates. Also, comparing fused silica and zyarock cases, the roughness and the porosity of the substrate does not have a significant effect on the undercooling values. All in one, it is not possible to find any general behavior able to explain in which cases nucleation of Si is promoted: each substrate material appears to be specific.
Acknowledgments The authors are indebted to the substrate providers and would like to thank G. Fournier and C. Garnier who were involved in the experimental installation development. We would also like to thank Dr. D. Camel, Dr. B. Drevet and Prof. N. Eustathopoulos for inestimably rich discussions. The financial support from the French National Research Agency (ANR-HABISOL-Mosaïque) is also gratefully acknowledged.
References
20
[1] K. Fujiwara, W. Pan, N. Usami, K. Sawada, M. Tokairin, Y. Nose A. Nomura, T. Saishido, K. Nakajima, Acta mat. 54 (2006) 3191. [2] T.Y. Wang, S.L. Hsu, C.C. Fei, K.M. Yei, W.C. Hsu, C.W. Lan, J. Crystal Growth 311 (2009) 263. [3] F. Jay, D. Munoz, T. Desrues, E. Pihan, V.A. Oliveira, N. Enjalbert, A. Jouini, Sol. Energ. Mat. Sol. C. 130 (2014) 690. [4] R. R. Prakash, T. Sekiguchi, K. Jiptner, Y. Miyamura, J. Chen, H. Harada, K. Kakimoto, J. Cryst. Growth 401 (2014) 717. [5] K. F. Kelton, A. L. Greer, Nucleation in condensed matter, Elsevier 2010, Chapter 6. [6] M. Wang, H. Zheng, X. Lin, W. Huang Mat. Sci. Eng. 27 (2011) 012006. [7] D. Turnbull J. Chem. Phys. 18 (1950) 198. [8] H. Tian, D.M. Stefanecsu, P.A. Curreri Met. Trans. A. (1990) 241. [9] Z. Fan Met. Mat. Trans. A 44A (2013) 1409. [10] T. Duffar, P. Dusserre, N. Giacometti, P. Boiton, J.P. Nabot, N. Eustathopoulos J. Cryst. Growth 198/199 (1999) 374. [11] C. Panofen and D.M. Herlach Appl. Phys. Letters 88 (2006) 171913. [12] R. P. Liu, T. Volkmann and D. M. Herlach Acta Mater. 49 (2001) 439. [13] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti J. Cryst. Growth 336 (2011) 77. [14] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti, J. Mater. Sci. 45 (2010) 2218. [15] M. Beaudhuin, K. Zaidat, T. Duffar, M. Lemiti, T. Indian I. Metals 62 (2009) 505. [16] H. W. Tsai, M. Yang, C. Chuck, C.W. Lan J. Cryst. Growth 363 (2013) 242. [17] A. Appapillai and E. Sachs J. Crys. Growth 312 (2010) 1297. [18] A. Appapillai, C. Shachs and E. Sachs J. Appl. Phys. 109 (2011) 084916.
21
[19] I. Brynjulfsen, L. Arnberg J. Cryst. Growth 331 (2011) 64. [20] L. Alphei, A. Braun, V. Becker, A. Feldhoff, J.A. Becker, E. Wulf, C. Krause, Fr.W. Bach, J. Crystal Growth 311 (2009) 1250. [21] K. Fujiwara, K. Maeda, H. Koizumi, J. Nozawa, S. Uda, J. Appl. Phys. 112 (2012) 113521. [22] L. Sylla, A. Hodroj, N. Mangelinck-Noel, T. Duffar, 19th European Photovoltaic Solar Energy Conference and Exhibition, Paris, 7-11 Juin 2004, pp.3775. [23] M. G. Tsoutsouva, T. Duffar, C. Garnier and G. Fournier Cryst. Res. Technol. 50 (2015) 55. [24] Z. Jian, K. Nagashio and K. Kuribayashi Metall. Mater. Trans. A 33 (2002) 294. [25] B. Drevet, N. Eustathopoulos J. Mater. Sci. 47 (2012) 8247. [26] T. Whalen, A. Anderson J. Am. Ceram. Soc. 58 (1975) 396. [27] B. Drevet, R. Voytovych, R. Israel, N. Eustathopoulos J. Am. Ceram. Soc. 29 (2009) 2363. [28] B. Drevet, O. Pajani, N. Eustathopoulos Sol. Energ. Mat. Sol. C. 94 (2010) 425. [29] C. Huguet, C. Dechamp, R. Voytovych, B. Drevet, D. Camel, N. Eustathopoulos, Acta Mater. 76 (2014) 151. [30] X. Wang, G. Zhai, J. Yang, N. Cue, Phys. Rev. B 60 (1999) R2146 [31] M. E. Bachlechner, J. Zhang, Y. Wang, J. Schiffbauer, S. R. Knudsen, D. Korakakis, Phys. Rev. B 72 (2005) 094115. [32] M. Becker, U. Gösele, A. Hofmann, S. Christiansen J. Appl. Phys. 106 (2009) 074515. [33] M. Ben el Mekki, M. A. Djouadi, E. Guiot, V. Mortet, J. Pascallon, V. Stambouli, D. Bouchier, N. Mestres, G. Nouet Surf. Sci. Technol. 116-119 (1999) 93.
22
[34] A. Gupta, C. Jacob, Prog. Crystal growth Charact. Mat. 51 (2005) 43. [35] S. A. Kukushkin, A. V. Osipov, N. A. Feoktistov, Phys. Solid State, 56 (2014) 1507. [36] G. Ferro, Crit. Rev. Solid State Mat. Sci. 40 (2015) 56.
23
Figures and captions
Figure 1: Schematic illustration of the apparatus used to perform undercooling measurements and video recording experiments described in this work.
24
Figure 2: Temperature profile of the Si sample during the melting and solidification process on a fused silica substrate. Inlet shows a zoom of the plot during the recalescence – solidification event.
25
Figure 3: Optical micrographs of the solidified droplets on the oxide substrates in cross section showing the final grain structures. All droplets in this study present a “horn”, as illustrated in the zyarock case.
26
Figure 4: Morphologies of growing crystals during the solidification of Si on a zyarock substrate coated with Si3N4 powder oxidized at 900°C. Since the camera looks at the sample through the pyrometer window, the black ring that appears in all pictures is the pyrometer measurement spot. The surrounding dark grey circle is a window reflection on the top of the droplet. The grey parts are the undercooled liquid and bright parts are crystal solidified from the undercooled liquid. The square substrate appears in dark grey.
27
Figure 5: a) Optical micrograph of the β-Si3N4 precipitate found on the surface of the Si droplet solidified on a zyarock substrate coated with Si3N4 oxidized at 900 °C, b) macroscopic side/top views of the solidified Si droplets on a zyarock substrate coated with Si3N4 oxidized at 900 °C and on sintered Si3N4. Note the “horn” that appears at the end of solidification due to the solid – liquid density gap.
28
Figure 6: Optical micrograph in a cross section view and in a view from the bottom (after partial substrate lapping) of the solidified droplet on a silica substrate coated with Si3N4 oxidized at 900 °C.
29
a)
b)
c)
Figure 7: a) Optical micrograph of the precipitates found on the surface of the Si droplet solidified on a PBN substrate, b) optical micrograph, in cross section, of the Si droplet solidified on a PBN substrate, c) SEM cross section of the interface between Si and the PBN substrate. Layers 1, 2, 3 correspond to the solidified Si, Si3N4 intermediate layer and the BN substrate respectively.
30
Figure 8: Crystal growth process during the solidification of silicon on carbon – containing substrates.
31
Graphite
Glassy carbon
4H-SiC
a)
b)
c)
Figure 9: a) Macroscopic side/top views of the solidified Si droplets on carbon – based substrates. b) SiC precipitates on the surface of the solidified Si droplet, c) SiC precipitate that acts as nucleation center for the growth of a faceted dendrite.
32
Figure 10: Optical micrograph of the cross section of Si droplets solidified on (a) glassy carbon and (b) 4H-SiC substrate. c) SEM cross section of the interface between the Si droplet and the glassy carbon substrate where 1) solidified Si droplet, 2 and 3) SiC layer and 4) glassy carbon substrate.
33
Figure 11: Optical microscopy observation (extracted from Figures 3, 6, 7 b and 9 a) of the central part of the droplets cut in cross section.
34
List of tables Table 1: List of the investigated substrates and obtained experimental results. Substrate
Provider
Undercooling ±1 K
Droplet structure
Analyzed precipitates
Fused Silica
Mondia quartz
98, 75, 45
Single-crystal
No precipitate
(100) Sapphire
RSA
67, 42, 36
Zyarock
Vesuvius
49, 47, 41
INES
2, 3
Grains and twins
Si3N4
β-Si3N4 precipitates on top of droplet
INES
0, 0, 0, 0
Grains and twins
Si3N4
β-Si3N4 precipitates on top of droplet
INES
6, 6
Grains and twins
-
β-Si3N4 precipitates on top of droplet
Sintered Si3N4
INES
4
-
-
Pyrolitic BN (PBN)
MCSE
3, 0, 0, 0
Stressed grains
Si3N4
Hot isostatically pressed BN (HIPBN)
MCSE
21, 9
Grains and twins
BN
Isostatic graphite
MERSEN
0, 0
Polycrystalline
3C-SiC
Glassy Carbon
MERSEN
0, 0, 0
Polycrystalline
3C-SiC
6H SiC Si face
SiCrystal AG
44, 12
-
-
4H SiC Si face
SiCrystal AG
0
-
3C-SiC
Carbon/SiC (Carbosil)
MERSEN
5, 2, 0
Grains and twins
3C-SiC
Silica+Si3N4 powder fired at 900°C Zyarock+Si3N4 powder fired at 900°C Zyarock+Si3N4 powder fired at 1100°C
Distorted single-crystal Few grains and twins
No precipitate No precipitate
Location of nucleation Substrate Substrate Substrate
β-Si3N4 precipitates on top of droplet Si3N4 layer grown on substrate Si3N4 layer grown on substrate 3C-SiC precipitates on top of droplet and at substrate surface 3C-SiC precipitates on top of droplet and substrate surface Unknown 3C-SiC precipitates on top of droplet and substrate surface 3C-SiC precipitates on top of droplet and substrate surface
35
Highlights 1 – Set-up and process have been developed for study of Si nucleation on substrates. 2 – Oxides, nitrides and carbon-based substrates have been studied. 3 – Oxides show high undercooling and only one nucleation event. 4 – Nitrides promote nucleation via precipitates according to good epitaxial fit with Si. 5 – Carbon based substrates are very potent substrates but the reason is not clear.
36