Oxidation of Liquid Silicon - Reaction Mechanisms and Kinetics

Kj etil Hildal Johan Kr. Tuset

Institute for materials technology and electrochemistry Norwegian University of Science and Technology

7491 Trondheim, Norway

ABSTRACT

When liquid silicon reacts with water, an oxide layer forms on the surface of the melt. The composition and structure of this oxide layer is determined by the initial compo9-tion of the melt, the temperature and the flow conditions at the surface. We have used two different techniques to oxidise molten silicon, with varying content of impurities (Al and Ca). The results show that the initial oxidation product is gaseous SiO. This gas will not necessarily deposit on the surface of the silicon melt, but may be carried away depending on the local flow conditions. Im­purities in the melt will result in a slag film on the surface of the melt. This slag consists of a mixture of Si02' Alz03 and CaO. De­pending on the composition and tempera­ture, this slag mixture may be solid or liquid. As a consequence of this slag formation, more hydrogen is produced, which has a desirable effect as to prevent large-scale steam explosions.

1. INTRODUCTION

One of the operational steps in commercial production of Si and FeSi today, is the water granulation process. The main principle of this process is to break up a continuos flow of melt into appropriate sized drops and quench these to granules in a water tank. However, this mixture of molten metal and water are in steam explosion theory known as the premixing stage, a necessary condi­tion to obtain a large-scale steam explosion 1.

A laboratory study has been initiated to in­vestigate how to decrease the possibility of a steam explosion during granulation of Si and FeSi. It is known from industrial experience that the composition of the melt is a very decisive factor2. This has also been verified experimentally3

. They released single drops of silicon-rich alloys into a water tank to investigate the effect of. impurities. In agreement with industrial experience, they found that small additions (<0.05%) of Al and Ca practically eliminated the possibility of violent melt-water interactions.

We believe that additions of Al and Ca pro­hibit a steam explosion in two ways. Dis­solved Al and Ca will react very easily with water vapour at this temperature (over 1400 0 C), i.e. the equilibra

2Al(1)+3H20(g) = (Alz03)cs,1)+3H2<g) Ca(li+H20cgl = (CaO)cs,1)+H2cgJ

will be displaced to the right. The hydrogen mixes with vapour in the film surrounding the molten drop, making it more stable against pressure disturbances, as hydrogen is a non-condensable gas under these condi­tions. The other way Al and Ca may prohibit a steam explosion, is the film of slag that forms on the melt-vapour interface, which consists of a mixture . of Si02, Ah03 and CaO, depending on the initial composition of the melt. This film will in general have properties that differ from those of molten

3~

329

silicon, some of which may have a protect­ing effect against fragmentation.

In this work, we have oxidised molten sili­con samples in a spray of water. Four differ­ent alloy compositions have been studied at two different melt temperatures. The treated samples have then been analysed with an X­ray microprobe, where we can get good im­ages of the oxide film that forms on the metal-vapour interface. Finally, the same alloy compositions have been granulated in water and analysed the sam<:". way. We can then compare the results obtained in our ex­perimental apparatus with the granulation experiments described by Nelson3

.

2. EXPERIMENT AL

2.1 APPARATUS Our apparatus allows a rapid heating to a high temperature (1400-1600 °C) and a sub­sequent rapid cooling. We also need to inject a jet of water when the melt has reached the desired temperature in order to oxidize the melt surface.

A schematic cross-section of the apparatus is shown in figure 1. A graphite susceptor, with the shape of an open cylinder, is lo­cated inside a strong, fluctuating magnetic field. Inside the susceptor, we place a small crucible (12 mm high, wall thickness 1.5 mm, radius 12 mm). These crucibles are for the present work made of glassy carbon, but alumina has also been usecf. Under the base of the graphite susceptor an S-type thermo­couple measures the temperature during the experiment. This thermocouple is calibrated to compensate for the temperature; drop be­tween the interior of the crucible and the base of the graphite susceptor. These vital parts are all located inside a transparent sil­ica tube. The tube has double walls, which allow water cooling of the walls. The inte­rior of the tube is evacu\}ted prior to the ex­periment and flushed with inert gas at all time thereafter. During the heating and oxi­dizing stage, argon gas is used, but after the

sample is oxidized, helium is used to achieve rapid cooling.

When the melt is oxidized, a syringe is in­serted from above. Inside the apparatus, the­pressure is slightly higher than atmospheric pressure to avoid any oxygen leakage into the apparatus. The syringe is taken down to the top of the crucible and 0.2 ml of water is inserted during a three-seconds period of time.

The thermocouple is connected to a reader (Hewlett-Packard 34970A) to obtain a con­tinuous temperature reading during the ex­periment.

A high-frequency generator (Philips PH 1012) powers the solenoid, which consists of four loops of water-cooled tubes of Cu. The tubes have an outside diameter of 6 mm.

Figure 1 - Schematic cross-section of the apparatus.

2.2 ALLOY COMPOSITIONS We used four different compositions for our experiments. These were pure Si, Si with 0.4% Al, Si with 0.04% Ca and finally Si with 0.4% Al and 0.04 % Ca. These four compositions were oxidized at two melt temperatures, 1450 and 1550 °C. The con­centration of other elements is negligible (totals less than 0.01 %) for all practical pur­poses.

2.3 MICROPROBE We use a microprobe (JEOL Superprobe JKA-8900M) to analyze the surface of the oxidized silicon samples. The crucibles are filled with a cold mounting epoxy resin (EPO-Fix). This is a two-component resin that penetrates voids in the sample easily. Then, the crucible is cast into another resin, Demotec-20, before we make a cut through the encapsulated crucible. After polishing, the sample is ready for the microprobe. The images given by the microprobe shows con­centrations of the important elements, which in this context are Si, 0 , Al and Ca. Along with the element concentration profiles, a scale giving the concentration (in weight%) for each color is located next to the profiles. Note that different elements have different scales to improve the contrast in the images.

3. RESULTS

3.1 VISUAL OBSERVATIONS A video camera (8 mm Canon UC-X30) was used to record the experiments. The experi­mental layout did not allow us to observe the melt surface directly, but any smoke or gaseous products can be seen as they rise inside the apparatus with the inert gas. In general, the behavior of different composi­tions at different melt temperatures did not seem to change much. Usually, a stream of smoke was observed just before the water jet was sprayed onto the melt surface, due to some evaporation of water from the lower tip of the syringe. Upon water injection, wa­ter drops were thrown out of the crucible onto the transparent walls, with some smoke production.

3.2 WEIGHT CHANGES The crucible is weighed prior and after the experiment to detect weight changes during the experiment. These weight changes origi­nate from reactions inside the hot zone. A number of reactions can take place, some of which increases the weight, while others decrease the weight.. Reactions between the water vapor and the crucible tend to de­crease the weight, as carbon is lost as CO or COrgas. Also, if a considerable amount of SiO-gas is allowed to escape the hot zone, this gas may condense away from the cruci­ble, thereby reducing the mass of the sam­ple. If SiO-gas condenses on the metal sur­face or the crucible, this will increase the weight of the sample. In general, small weight increments were detected, of the or­der 0.01 grams. In two cases, a decrease in weight took place. One of these was negligi­ble, while the other was as big as 0.07 gram, which corresponds to 3% of the total mass (crucible and metal). We can not rule out the possibility of an experimental error in the last case. The water injection is 0.2 ml, which is approximately 0.18 gram oxygen. So comparing with the weight changes, only 10% (at most) of the oxygen can be found on or inside the crucible after the experi­ment. This is expected, as most of the water is thrown out of the crucible and onto the transparent silica walls.

3.3 OXIDISED SILICON SURFACES

3.3.1 PURE SILICON The surface of the specimen is dominated by a white, patchy SiOrlayer. The typical thickness is 100-120 µm, but at several loca­tions, this thickness drops to a few µm. The oxide layer is very porous, which can easily be seen from microprobe analysis of the specimen. This layer also contains several spherical particles, which are rich on Si, typically around 80-90 wt%. The Si-spheres form a band along the silicon matrix, ap­proximately 50 µm from the matrix.

Another interesting feature is that a layer of Si02 is covering most of the crucible wall. This is most likely due to a condensation of

33(

331

SiO-gas. Particles of Si inside this oxide layer confirm this hypothesis. Compared to the oxide layer on the metal surface, the ox­ide on the wall is thinner, around 20-40 µm.

On the effect of increased melt temperature prior to oxidation, the most striking differ­ence is the change in color of the oxide layer, from white to brown. Besides from that, the SiOrlayer does not appear to be any different from the lower melt tempera­ture. A similar layer is found on the crucible walls, although it appears less porous, with more stable concentrations of Si and 0. Clusters of Si-rich particles have formed close to the wall. It is possible that these clusters contain some C, in the form of SiC­particles.

The thickness of the SiOrlayer is larger for the sample oxidized at the highest melt tem­perature. The oxide layer of pure silicon oxidized at 1550 °C reaches a thickness of 300 µm, while a sample oxidized at 1450 °C has a maximum thickness of 130 µm. The density of the oxide layers does not change noticeably when the melt temperature is in­creased. In figure 2, the composition of the oxide layers is plotted against the distance from the oxide surface. The concentration of Si02 is less than 100% for both of the samples, indicating that there is some porn;­ity in the oxide. We also note that the con­centration increases well above 100 % some places. This is an effect of small spheres of Si inside the layer of Si02. The microprobe uses Si02 as standard, and if the probe de­tects an area of pure silicon, it will give us values that are too high. Figure 3 shows the element analysis of the surface of silicon oxidized at 1450 °C. As we can see from the picture, there exist several spheres of pure silicon in the oxide layer, and these will give us wrong estimates on the concentration of Si02.

Another effect of elevated oxidation tem­perature is a somewhat rougher interface between the silicon matrix and the oxide layer, this will be explained later.

~Si02.1450C I ··• ·· Si02, 1550C

250 ·- ·····-······--·-··-·-·······----·-----···· ···················--;

200 r----- ----iit------------'

0 50 100 150 200 250 300

Distance franCDdde slrlace [rricrareter]

Figure 2 - Oxide layer composition plotted as a function of the distance from the Slll"­

face.

Figure 3 includes an image of the same alloy composition, granulated in water. The tem­perature of the melt was at its liquidus tem­perature, approximately 1412 °C, when it was released into the water. Note that the two magnifications differ by a factor of two. The most striking difference between these two images is the absence of the porous layer of Si02 on the drop of melt granulated in water (lower image). The surface is clean, with no traces of any oxide. However, there is evidence in the form of collected hydro­gen that there has been a chemical reaction between silicon and water. Based on the lack of oxygen-containing compounds on the surface of the granule, we conclude that the oxidation products formed in the reacfion between water vapour and silicon have been gaseous, presumably SiO. The gaseous products have been efficiently washed away from the surface of the drop, due -to turbu­lent flow in the gas mixture (water vapour, hydrogen, SiO) around the granule. In these experiments, one can see that small bubbles are regularly released from the gas film sur­rounding the drop of melt. These bubbles will transport silicon oxide away from the surface of the drop.

0.e 0.0 8.0 0.0 0.0

8.9

9.0 9.0 8.0 8.0 0.1

D Cone RreeX so.ee e .0

56.25 °· 1

52.50 e.s •8.75 !:~ 45.ee

8_

4

~~:E ::: :ie.ee 2.9 . "/I ;u: ::~

· · LO

1s.ee 14.86

1

13.11 °·0

12.19 °·0

11 . 25 °·0

10.31 ::: 9. 38 8.H 0.0

0.0

~:~: 8.0

:::~ ::: 3 . 7s

0_0

2.81

1.87 ::~ :::: 9U

flve-e . 18 °-0

St Co ftreaX 90.00 l7 .6

BUB ~:~ 76 . 75 73. 12 I •

9

67 . 50 ~:: ~!:!: 1.3 50. 63 l .. e

~;:: 1~:: 18. 3

::: ~: ::! 22.58 8.3 16.87 9.3 1 !:~~ 9.3

7.5 )~· I :~En 1' Rve16. ll,

32.: •

17·43

I 1.::t04

I L227

1434 J.100 1.038

1245 0.895 0.848

'JS6 0.6~ 0.S':iS

7<7 0.485 0.469

<98 I 0 . 289 I 0.2 ... , .. 0.07 6 0.ese 8 -0. 129 -0.093

RY"l8. 1.29 A..,.0.099

0 Cone Si Cone:: 4.107 7.046 103.44

3.560

I s. 781

I 90.4'3

3.014 4.Sll' 71.S

l. 7. S2. 64.59

1.907 Sl .6 ..

8.722 36.68

1-4.543 I 25.73

-J. 808 12.78

-3.073 -0.ll Rvd. 533 Rv.SJ.51

Gr-oup KJ"""t.11.HildaJ 5:.w.ple : 04~H80.....p 5.1

Hee. V JS.9 kV P.,.ob C 3.671r98R Scan Off Diuel l<•s> 10.ee Poin-.:.:s 512•51.Z f.H't:llU'V• l( U•) X:L00

't:l.00

Figure 3 - Microprobe images of the surface of non-alloyed, oxidized silicon. The upper picture is taken from a sample oxidized with water at a melt temperature of 1450°C, using the method described earlier in this paper. The lower image is taken from a sample at liquidus temperature (-1412 °C), granulated in water using the method described by Nelson 3

l.

333

3.3.2 SILICON WITH 0.5 % Al The specimen containing Al appear similar to the ones without. A white, patchy oxide layer is covering the entire metal surface and most of the interior of the crucible. At the lower melt temperature, the thickness of the oxide is typically 35-55 µm. A layer of mixed oxides of Ah03 and Si02 is deposited as a thin film along the interface between the matrix of silicon and a porous layer contain­ing only Si02. This interface appears to be denser than in the specimen not containing Al, as already described.

The oxide extends up the crucible walls, but here only very small amounts of Al are pre­sent. Instead, the concentration of Si is very high, and microprobe results indicate that some SiC is deposited along the walls, to­gether with pure Si. Away from the wall/oxide interface, the oxygen concentra­tion increases and we find the usual, porous SiOrlayer. Traces of Al can be found along the interface between pure Si and the oxide layer.

The effect of increased melt temperature seems to be an increase in the thickness of the oxide layer. Basically, the thickness of the area containing Al is the same, but there is more Si02 present further out, in the usual porous configuration. The oxide layer is rather patchy, with a maximum thickness of approximately 150 µm. Some Si-particles can be found inside the oxide layer. This was also found, but to a somewhat higher extent, in the pure specimen. There seems to be no appreciable differences in the oxide on the crucible walls for the two different melt temperatures involved.

For the granulated samples, a very thin film (~ 5-10 µm) of slag was found on the sur­face of the solidified drop. The composition of this slag was typically 60% Si02 and 40% Ah03, with no detectable porosity. Similar as for the pure silicon drops granulated in water, no porous Si02 was found on the out­side of the slag. The slag was found only on a very limited part of the drop surface. Ap-

proximately 80 % of the surface contained no oxide at all. Again, we believe the reason for the lack of Si02 relates to the oxidation mechanism. Gaseous SiO, which forms ini­tially, is carried away from the surface due to the flow field around the drop. This may also influence on the oxidation of Al, as the results clearly indicate much thinner, and mostly absent films of Ah03 on the granu­lated samples. For the samples oxidized with the apparatus described earlier, the oxide film contains a thicker mixture of Alz03/Si02 as well as the porous layer of Si02 outside this mixture.

3.3.3 SILICON WITH 0.04 % Ca Earlier results have shown that even very small additions of Ca to the molten silicon have a strong influence on the composition of the oxide that forms upon water injection. This was also found to be the case this time. However, the melt apparently had a higher concentration of Al than we aimed for. This was seen from microprobe pictures, where a thin film of Ah03 and CaO cover the sur­face, together with the expected Si02. This indicates that some contamination of the samples had occurred, and this is most likely to have happened during the preparation of the rods, which these samples are taken from. A chemical analysis, which shows 0.06 % Al, confirms this.

~ The low-temperature melt has a patchy, white oxide on the surface of the metal and on the crucible walls. The layer is distrib­uted as a square-wave, with a high degree of regularity. The thickness is about 80 µm, but there are areas between the "squares" with no detectable oxide at all. The length of these squares is between 150 and 400 µm. Separation distances are of the order 100 µm.

The oxide layer has certain characteristics. Close to the silicon-matrix, this layer con­sists of a mixture of Si02, Ah03 and CaO. The thickness of this oxide film is of the order of 5 µm. This is seen from figure 4, where the composition of the oxide is shown

as function of the distance from the outer surface for the two different melt tempera­tures. We see that the most striking differ­ence is the concentration of Si02 in the outer oxide layer. For the high temperature melt, the concentration of Si02 is stable around 85 %, but for the low temperature melt, the concentration is only 30 % for a major part of the oxide layer before it increases as we move inwards against the silicon matrix. Concentrations of Ali03 are not shown in figure 4, as they are low, of the order of 2% in the slag mixture at the interface towards the silicon matrix, and zero further out.

~cao.1450C

··•- Si02, 1450 c 100 , ...................................................................................................... fo\·~·········································

00+----------;-~-----; I i ' I ' OO +----------r1--+---M-----<

\

! \ r

- 70 +-----------+---;--+--+-----i

! 60 -1---------1--~r-r--;----; c J \ !

\ ~ 50 +----------1--- t-t--it-----i ·;; 0 i

\ I \ \

i : ,,. .............. ,. ..... ?''\ ~,/ v i ( y \I

~ +-1------~·~------1----;

1o r/-------------1-+-'c------i (\ \

j \\ 10 ~ 30 40 50 60 70 ao 00 100

Distance from surface (micrometer]

Figure 4a - Oxide layer composition plotted as function of the distance from the surface. (Increasing distance means approaching the silicon matrix.) Melt oxidized at 1450 °C.

~cao.1ssoc

- • .. Si02. 1550C

~+----------------~\-i.

i I r.l I

20 40 60 ao 100 120

Distance from surface [micrometer]

Figure 4b - Same as for 4a, but the sample was oxidized at 1550 °C.

Some oxide has been deposited on the cru­cible walls. No signs of either Al or Ca can

140

be found, but high concentrations of Si and 0. It appears that some melt has climbed up along the wall, giving pure Si along the graphite wall, with an outer layer of Si02.

However, further up on the wall, only Si02

can be found, indicating that a condensation of SiO-gas has taken place. From the micro­probe pictures ava_ilable, no spheres of sili­con can be detected.

When we compared these results to micro­probe investigations of similar alloys granu­lated in water, we found that there were dif­ferences of the same kind as for the other compositions. In general, the surface of the granules did not contain any oxides. On a very limited (-10 %) part of the surface, we found a thin film of slag. The thickness of this slag film was of the order 1 µm. This made it difficult to obtain good quantitative analysis of the oxide. However, the oxide was very low on CaO, an upper estimate is 5 %. Traces (max 3 %) of Ali03 were also found in this slag film, together with Si02.

Apparently, the lack of SiOrformation on the surface of the drop affects the formation ofCaO.

3.3.4 SILICON WITH 0.4 % Al AND 0.04 % Ca

The effect of alloying the melt with both Al and Ca is the formation of a firm layer of mixed oxides between the metal matrix and the outer layer of porous silica. This is seen from the concentration profiles of the oxide layer in figure 5 and the micrographs show­ing the results of elemental mapping in fig­ure 6. As observed for the previous samples, the silica layer, which varies in thickness; is most porous for the samples at the lowest temperature. Layers of pure silica are also in this case deposited on the crucible walls, but there are no silicon spots seen in any of these layers, indicating that the deposited SiO has been fully oxidized.

The mixed oxide layer, which covers the entire surface of the metal, has the appear­ance of a mineralized substance consisting of small grains that differs in composition, but the shape of the interface towards the

B35

metal matrix reveals that this layer must have been in a molten state when formed.

· According to the chemical analysis given in figure Sa, the mixed oxide layer consists in average of lS % CaO, 3S % Ah03 and SO% Si02, and this gives a liquidus temperature of about lSOO cc. It is therefore not unreal­istic, even with an initial metal temperature of 14SO cc, that a liquid slag has been formed at the interface in this case. The ir­regularity of the slag/metal interface and the presence of silicon droplets in the slag phase as seen from figure 6 is typical for a situa­tion where reactants, due to chemical reac­tions between two liquids, are transferred across the interface at high rates. The inter­facial tension may then drop to zero, interfa­cial turbulence and instabilities are created which may lead to spontaneous emulsifica­tion (Turkdogan5

) as pictured in figure 6. Similar tendencies was also noticed for the Ca-alloyed sample that unintentionally con­tained 0.06 % Al.

The lower image of figure 6 show the ele­mental mappings of the oxide layer on a granulated alloy with the same composition as the oxidized sample shown in the upper set of images. The most striking difference between the two, is also in this case the ab­sence of the porous silica layer of the granu­lated sample, but the mixed oxide layer re­mains and has the same appearance and chemical composition as the corresponding layer for the oxidized sample.

We also note that the interfaces between the silicon matrix and the oxide mixture layer appear different on the two images shown in figure 6. The granulated drop has a rather smooth, well-defined interface. On- the other hand, the water-sprayed sample has a non­uniform interface between matrix and oxide mixture. One explanation for this may be the interfacial turbulence, as described earlier.

1-sm ·=1 ··•·· CaO 100 rm••··········· ················ ······································································ ··· ······mm•••·••·•••••••···••··••····························· •··• ···• •·•••••••••·

Distance fran Oliar sufeca [rricnrneter)

Figure Sa - Oxide layer composition plotted as a function of the distance from the sm­face. Melt oxidized at 14SO cc.

I ~s;o2 ··• ·· A1203 I ·-..·· Cao

~ 10 t---:-f-''\--+:.C'.:_~~~~~~~~\,.----: t 60 t--r~~-r-~~~~~~~~~-\----' ~ 50 t-r~~~~~~~~~~~~-+----'

i40 tt-~~~~~~~~~~~~--+----' u " tr-~~~~~~~~~~~~---4----'

0 .. " II " • ....., • .,___._ II ., • ., .. II • · • -• 11 • · II • 11·" • • ,• •·• -~:.--_:;.__; ~ ~ ~ oo ~ m

Olatance from surface [micrometer]

Figure Sb - Oxide layer composition plotted as a function of the distance from the sm­face . Melt oxidized at lSSO cc.

From figure S, we clearly see the difference in the density of the porous silica layer be­tween samples oxidized at different temperatures. There is no change in composition of the mixed oxide layer. It appears that the thickness of the mixed oxide layer is larger in figure Sa, but this is an effect of where we choose to analyze, as thickness variations exist on all of the samples. In overall, no change in thickness of the mixed layer could be detected.

E><it Area% Ca Cn Orea%

e.e lS.00

9.9

28.JZ 0.1

14.06 0.0

26.25 0.1

I 13.12

0.1

24.38 0.2

12.13 9.2

22.50 O.J

11.25 0.2

20.63 0.4

10.31 0.4

0.5 9.37

0.5 0.6

8.44 0.4

0.6 0.7 7.50

0.8 6.56 S.62

9.6

9.38 0.5

4.69 0. 9

7.50 0.6

3.75 0.9

5.62 0.6

Z.81 0.6

0.9 0.8 3.75 L87 1.88 I. I 0.94 I. I 0.00 92.1 0.00 9

1. 8

Rve J.04 e.O R-.ie0. 35 0.e

0 Cone RreaX Si Cn flreaZ

30.00 0.3

100.00 4

· 7

28.12 0.4

93.75 2~:: 0.6

26.25 o.s 87.50 0.4

24.38 81.25 ZZ.50

1.2 75.00

0.2

20.62. LS

68.75 0.2

18.75 1.8

62.50 0.2

l.S 0.3 16.BB

1.4 56.25

O. l IS.00

1.6 50.00 o.s

13.12 43.75 11.25

3.0 2.1

9.38 S.4

7.50 S. 3 3.0

JS19

I .. J.74

I 13.11

13"2 37.48 ll.23

10.05 31.22 !1.34

868

I 24.95

I 7 .46

651 10.69 5.57

43< 12.43 3 . 68

217 6.17 1.88

e -e .o!t -e.e!! ftye 0.~6 ftvc-0.09

feConc 0 Cone Si Cone

6.080 60.00 100.ee

I 5.243

I 52. 13

I 87.48

..... 85 '44.26 74 . 97

3.728 36.39 62.45

28.St

I -4!LS-4

20.64 37.'42 12.ll

I 24.91

4 . 90 12.39

... z.s7 -0.13 RvflfL56 Hvfi'<f.28

Group : Kjet:11Ht ldel Sample ! 130100.-p 'l..1.

Ree. Y 15.0 kV p,.,,b c: 3.006e-08R Scan OFF Ot:oellh111a) 19.00 Poi"'U> St2.•.Sl2: ltit.,rv.,I<um> X:LOO

Y:l.ee

Figure 6 - Microprobe images of the surface of alloyed (0.4% Al, 0.04% Ca), oxidized silicon samples. The upper image is taken from a sample oxidized with water at a melt temperature of 1450 °C, using the method described earlier in this paper. The lower image is taken from a sample at liquidus temperature (~1412 °C), granulated in water using the method described by Nelson 3).

331

! 337

4. CONCLUSION

Liquid silicon, either pure or with small ad­ditions of Al or Ca has been oxidized with water under two different experimental con­ditions. One method is to heat a small sam­ple of silicon in an inert atmosphere, using an induction-heated furnace. The other method simulates a granulation process, as drops of silicon is melted and released into a water tank. It is found that there are substan­tial differences between the observed oxide layer on drops that have solidified in water (granulated), and samples that have been oxidized with a small pulse of water. Apart from deviations in layer thickness, the most striking difference is the absence of the po­rous silica layer on the drops and fragments that have solidified in water. One possible explanation may be that there is an initial stage of oxidation, also for the alloyed sam­ples, where the silicon is oxidized by a va­por phase reaction involving SiO. Due to turbulence in the vapor film, this oxidation product of SiO-gas is effectively washed away in the case of granulation, but remains on the surface of the oxidized samples, where it may either condense to Si+Si02 or react with oxygen to form Si02.Thus, what we suggest is that the oxidation of silicon takes place in two consecutive steps. The first is described by the overall reaction

which is endothermic and takes place at the surface of the melt. The second step is be­lieved to take place at a small distance away from the surface, and offers two alternative reaction schemes, which are both. strongly exothermic.

2a) 2SiO(gJ = Si(s,IJ + Si02 2b) SiOcgJ + H20 = Si02<sJ + H2(gJ

Our results suggest that both these alterna­tive reactions are active in the oxidation ex­periments. Reaction 2a dominates close to the metal surface where the gas is depleted

in H20, and the sum of reaction 1 and 2b causes the silica deposit to grow from the bottom. Further out where the H20-pressure is higher reaction 2b is favoured. The result of this is clearly seen on microprobe images, where the porous layer of Si02 close to the silicon matrix includes small spheres of sili­con. When Ca and Al are present, the oxida­tion of these, which is very exothermic, has to take place on the interface. If a liquid slag forms, this blocks the evaporation of SiO.

Increased melt temperatures seem to have an effect on the interfacial turbulence. The in­terface between the silicon matrix and the oxide layer appears more irregular for al­loyed melts oxidised at 1550 °C. This is re­lated to the overall faster chemical reactions at elevated temperatures. A high mass trans­fer rate will lower interfacial tension and induce turbulence.

5. REFERENCES

1. D.F. Fletcher and T.G. Theofanous, Ex­plosive melt-water interactions. Ad­vances in Heat Transfer, vol. 29, pp.129-212, 1997.

2. L.Nygaard, B.Andresen and T.Halland, Critical parameters in water granula­tion of silicon, Proc. of Silicon for the Chemical Industry IV, pp.75-83, Gei­ranger, Norway, June 3-5, 1998.

3. L.S. Nelson, P. Brooks, R. Bonazza, M.L. Corradini, K. Hildal and T. Bergstrom, Steam explosions of single drops of molten silicon-rich alloys, Proc. oflnfacon 9, Quebec, Canada, June, 2001. .

4. K.Hildal and J.Kr.Tuset, Oxidation re­actions on liquid silicon, Proc. of Sili­con for the Chemical Industry V, Troms0, Norway, May 29-June 2, 2000.

5. E.T. Turkdogan, Physiochemical prop­erties of slags and glasses, The Metals Soc., London, 1983.