Thermodynamics of the Si-C-O System for the

Production of Silicon Carbide and Metallic Silicon

M. NAGAMORI, I. MALINSKY, and A. CLAVEAU

The predominance diagrams of the Si-C-O system for 1350 to 2200 ~ have been constructed thermo-
dynamically by taking into account the presence of liquid silicon monoxide. The high-temperature
behavior of various feeds, such as the SiO2-C and SiO2-SiC systems, have been calculated as functions
of mixing ratio and reaction temperature by using the method of equilibrium mass-balance. For an
efficient SiC-making, a charge having a C/Si ratio of 2.91 should be heated between 2035 and
2045 ~ where SiC crystals can be grown mainly via gas-phase reactions of SiO(g). Metallic silicon
cannot be produced unless a charge of the Si-C-O system is heated above 2037 ~ In Si-making,
liquid SiO plays an important role in that the yield can be improved substantially by enhancing the
condensation of SiO gas in the upper cooler part of the furnace. This work also lends thermodynamic
proof and support to the double-reactor model, in which a greater fixation of ascending SiO gas on
carbon as SiC is considered essential for obtaining a higher silicon yield in industrial furnaces. Some
fundamental strategies useful to the improvements of commercial Si- and SiC-making operations have
been described quantitatively.

I. INTRODUCTION published hitherto, the effect of temperature has never been

THE invention of predominance diagrams by Kellogg and
Basu 1 marked a giant leap in the pyrometallurgical under-
standing of the metal-sulfur-oxygen systems. The predom-
inance diagrams permit description of thermodynamic
stability conditions of various substances and their inter-
relations as functions of 02 and SO2 partial pressures as well
as temperature. Many process improvements and new pro-
cess developments have resulted therefrom.
The electric furnace production of silicon carbide and
metallic silicon with silica and carbon as raw materials have
been practiced commercially for many decades. 2'3'4 In both
cases, the reactions are of the Si-C-O system.
It is known that heating a mixture of SiO2 + 2C in a
laboratory furnace fails to yield metallic silicon. Thus, the
thermochemistry for metallic silicon production remains a
mystery today despite the successful industrial operation. 5
Although the predominance diagrams of the Si-C-O system,
analogous to those of the metal-sulfur-oxygen systems, are
felt most useful for metallurgical practice, such diagrams are
yet to be published.
Several attempts were made in the past to draw a coherent
thermodynamic picture of the Si-C-O system, but without
much progress in theory. 5'6'7 Although the Si-C-O system
may appear to be a simple ternary at first glance, further
approach reveals its rather complex nature due especially to
the formation of SiO gas. There is also the presence of liquid
(amorphous) silicon monoxide, as described in detail by
Kubaschewski and Alcock, 8 which was, strangely enough,
never included in the previous studies of the Si-C-O system.
In all the previous works, the SiC- and Si-making systems
were not properly recognized as being governed by two
independent variables of carbon and oxygen activities at a
given temperature, as seen in the frequent errors of looking
upon the SiO partial pressure as an independent parame-
ter. 5'6'7 Also in all of the so-called stoichiometric models
M. NAGAMORI, I. MALINSKY, and A. CLAVEAU are with the
Centre de Recherches minrrales, Quebec Government, 2700 Einstein, Ste-
Foy, Quebec G1P 3W8, Canada.
Manuscript submitted July 11, 1985.
included nor defined thermodynamically.
The present work has been undertaken to construct the
predominance diagrams of the Si-C-O system at electric
furnace temperatures such as 1350 to 2200 ~ with the inclu-
sion of liquid as well as gaseous SiO. The work has been
extended to cover the equilibrium mass-balance analyses of
various silica-carbon feeds with the aim of laying a ther-
modynamic foundation for the practical control of electric
furnace operations.
II. THERMODYNAMIC DATA
The selected thermodynamic data required for calculating
the predominance diagrams are listed in Table I. It appears
that some thermodynamic properties of silicon monoxide,
such as melting point, remain unsettled as yet. According to
the detailed analyses by Kubaschewski and Alcock,8 how-
ever, SiO(s, 1) is stable between about 1000 and 2150 ~
Solid SiO decomposes into SiO2 and Si below about
1000 ~ whereas the vapor pressure of liquid SiO reaches
an atmospheric pressure at about 2150 ~ The two-term
free energy equation for SiO(1) = SiO(g) listed in Table 19
has been calculated from the three-term equation.8 Although
the notation SiO(1) has been adopted by assuming that the
melting point of SiO(s) is somewhere below 1350 ~ the
validity of this work remains intact regardless of where
the melting point is, so long as SiO exists in a condensed
state and its vapor pressure can be expressed by the
Kubaschewski-Alcock equation above 1350 ~
III. CHOICE OF COORDINATES
In calculating the thermodynamic stability of various
phases of the Si-C-O system, two chemical potentials must
be taken into account at a given temperature: one repre-
senting the activity of carbon, another that of oxygen. By
analogy with metal-sulfur-oxygen systems, a first choice
may be the coordinates of CO and 02 gases. However, from

METALLURGICALTRANSACTIONS B VOLUME 17B, SEPTEMBER 1986--503

 Table I. Free Energy Data for the Si-C-O System. After Turkdogan 9 Except the Reaction [1]
Number Reaction (c~ AG o, cal(a) (d)
1 SiO(l) = SiO(g) 71180 - 29.47 T ~b)
2 Si(s) = Si(1) 12080 - 7.17 T
3 SiO2(s) = SiO2(1) 2290 - 1.15 T
4 Si(s) + C(s) = SiC(s) - 17220 + 1.68 T
5 Si(1) + C(s) = SiC(s) - 29300 + 8.85 T
6 Si(s) + O2(g) = SiO2(s) -216500 + 41.95 T
7 Si(1) + O2(g) = SiO2(s) -228580 + 49.12 T
8 si(1) + O2(g) = SiO2(1) -226290 + 47.97 T
9 Si(s) + 1AO2(g) = SiO(g) - 24900 - 19.72 T
10 Si(1) + V202(g) = SiO(g) - 36980 - 12.55 T
11 Si(s) + V2O2(g) = SiO(1) - 96080 + 9.75 T
12 Si(1) + V202(g) = SiO(1) -108160 + 16.92 T
13 C(s) + V202(g) = CO(g) - 27340 - 20.50 T
14 C(s) + O2(g) = CO2(g) - 94490 - 0.13 T
15 Si(1) = Si(g) 94500 - 26.62 T
16 C(s) = C(g) 170520 - 37.16 T
(a) 1 cal = 4.19 J; T = temperature in Kelvin.
(b) Recalculated from the three-term equation of Kubaschewski and Alcock:SAG~' = 81200 + 12.65 T log T - 76.25 T. The present equation places
the boiling point of SiO(1) at 2142 ~
(c) Melting point of SiO < 1350 ~ melting point of Si = 1412 ~ melting point of SiO2 = 1722 ~
(d) Due to the lack of the published data for ultra-high temperatures, the linear equations have been used up to 2300 ~ by extrapolating them beyond
their upper limits 2000 ~ As a consequence, more precise free energy data may alter somewhat the lines and temperatures in the phase diagrams, but
without affecting their relative positions to any significant degree.

the viewpoint of thermodynamic rigor, the coordinates of C
and CO potentials should be more appropriate. The condi-
tions of reducing (or oxidizing) and carburizing (or decar-
burizing) are qualitatively indicated in Figure 1.
IV. PREDOMINANCE DIAGRAMS
OF THE TYPE CO-O2
In constructing the predominance diagrams with the use
of CO and O2 potentials, all the equilibrium reactions must
be expressed in terms of the partial pressures of CO and 02
gases, as shown in Table II. In all the previous studies of the
Si-C-O system, the formation of SiO(1) was overlooked.
Such a nonequilibrium viewpoint is called metastable sys-
tem in this paper. In the metastable system, there are
three condensed phases: Si, SiC, and SIO2. This approach
is inconsistent with the elaborate study on SiO(1) by
Kubaschewski and Alcock. s Correctly, the four condensed
phases of Si, SiC, and SiO2, and SiO(1) should be included
in the system above 1000 ~ The system including the four
condensed phases is called stable in this work.

CARBON SATURATION
REDUCING 0
REDUCING
0

8
,-I

-4
-4
OXIDIZING OXIDIZING
] I I ! ] ! ~ I
-20 -15 -10 -4 -3 -2 -1 0
Log P02 Log PCO
Fig. 1 - Various conditions in the predominance diagrams.

504--VOLUME 17B, SEPTEMBER 1986 METALLURGICAL TRANSACTIONS B

Table II. Standard Gibbs Free Energies (AG*), and Equilibrium Constants (K) for the Predominance Diagram of the Type CO-O2
log K (b)
Number Reaction A G ~ cal (a) 1600 ~ 1812 ~
21 Si(1) + O2(g) = SiO2(s) -228580 + 49.12 T 15.95 --
21 Si(l) + O2(g) SiOz(1) -226290 + 47.97 T -- 13.24
22 Si(1) + CO(g) SiC(s) + 0.5Oz(g) - 1960 + 29.35 T -6.19 -6.21
23 SiC(s) + 1.502(g) SiO2(s) + CO(g) -226620 + 19.77 T 22.14 --
23 SiC(s) + 1.5Oz(g) SiO2(1) + CO(g) -224330 + 18.62 T -- 19.46
24 SiO(1) + 0.502(g) SiO2(s) -120420 + 32.20 T 7.02 --
24 SiO(1) + 0.5Oz(g) SiO2(1) -118130 + 31.05 T -- 5.60
25 Si(1) + 0.502(g) SiO(1) -108160 + 16.92 T 8.93 7.64
26 SiC(s) + O2(g) SiO(1) + CO(g) -106200 - 12.43 T 15.12 13.86
27 CO(g) + 0.502(g) CO2(g) - 67150 + 20.37 T 3.39 --
13 C(s) + 0.502(g) CO(g) - 27340 - 20.50 T 7.67 7.35
(a) lcal = 4.19 J; T = temperaturein Kelvin.
(b) K = exp(- AG~ T).

The stability zones of various condensed phases in both maximum along the boundary Si-SiO2 as well as at the
stable (full lines) and metastable (dotted lines) systems are univariant point C where the three phases of Si, SiC, and
shown in Figure 2(a) together with the isobars of COz gas. SiO2 are assumed to coexist. Two univariant points are to be
The partial pressure of CO2 gas is always negligibly small noted in the stable system: the point A for the coexistence
insofar as the industrial production of SiC and Si is con- of Si, SiC, and SiO (1), and the point B for that of SiC, SIO2,
cerned. The metastable system is singled out and shown in and SiO(1).
Figure 2(b) to illustrate the isobars of SiO(g), which become It is important to note that both an increase and a de-
crease in oxygen or carbon activity away from the meta-
stable univariant point C result in a lower SiO pressure, as
' ,"-
indicated by the contours of SiO (g) isobars, Figure 2(b). In
other words, for a given value of SiO gas pressure at a
"z'--. Z - J temperature, there are an infinite number of the different
pairs of oxygen and carbon potentials. Some previous re-
0
searchers 7 were misled by the oversight of the fact that the
en
SiO partial pressure cannot be chosen as an independent
O)
O
.J
variable of the Si-C-O system.
-2 Thermodynamically, under no circumstances can the
activity of carbon be greater than unity. Yet, as shown in
Figure 2(a), the impossibility of the univariant point B at
1600 ~ is not immediately evident unless the carbon iso-
~ j.-/(~) PSio=O,014/. j Si02 IS)
activity lines are also shown. This is a serious shortcoming
-4 ~ T of the coordinates of the type CO-O2. This fault can be
s,(,, ,I J # remedied by choosing the coordinates of the type C-CO,
-20 -18 -16 -14 -12 -10 where the unit carbon activity can be clearly marked and
Log P02 checked. Although the carbon potential, or log ac, is some-
(a) what more abstract than the partial pressure of a gas and
2 therefore more difficult to visualize, the choice of the type
1812"C ' , //r // /V~I- C-CO is more rigorous and convenient than that of the type
/" /'/ /" I
CO-O2.

0 SiC(S) OO~/" / ~ ' ~ / : ~ / ' / i {

8
e~
I I
u PREDOMINANCE DIAGRAMS
Z .," OF THE TYPE C-CO
,.J ~5--"/" ,'/ ! i
-2
In order to calculate the predominance diagrams of the
/ I I I r type C-CO, all the equilibrium reactions must be expressed
I pQ~=9,1 i in terms of C and CO, as listed in Table III. The predomi-
I ~ 0,01 I nance diagrams thus calculated are shown in Figures 3(a) to
I4
0,~)01 Si ($.) I (f). The metastable system Si-SiC-SiO2 is also shown in
I I each diagram with dotted lines.
I I I I I I
-20 -18 -16 -14 -12 -10 The industrial furnaces for the production of SiC or
Log P02 metallic Si are operated usually under atmospheric pres-
(b) sure (1 atm = 101.3 kPa), and the furnace feeds are com-
Fig. 2--Predominance diagrams of the type CO-O2. The reaction numbers
posed primarily of SiO2 and C. The furnace atmosphere
indicated in (a) correspond to those in Table II. Only the metastable system consists mainly of CO and SiO gases in both cases, while
is shown in (b). the SiO partial pressure cannot be higher than 0.75 atm

METALLURGICAL TRANSACTIONS B VOLUME 17B, SEPTEMBER 1986--505

Table III. Standard Gibbs Free Energies (AG ~ and Equilibrium Constants (K) for the Predominance Diagram of the Type C-CC
log K (b)
Number Reaction AG ~ cal (a) 1350 ~ 1779 ~
31 Si(s) + 2CO(g) = SiO2(s) + 2C(s) -161820 + 82.95 T 3.66 --
31 Si(1) + 2CO(g) = SiO2(s) + 2C(s) -173900 + 90.12 T -- --
31 Si(1) + 2CO(g) = SiO2(1) + 2C(s) -171610 + 88.97 T -- 1.17
32 Si(s) + C(s) = SiC(s) - 17220 + 1.68 T 1.95 --
32 Si(1) + C(s) = SiC(s) - 29300 + 8.85 T -- 1.19
33 SiC(s) + 2CO(g) = SiO2(s) + 3C(s) -144600 + 81.27 T 1.71 --
33 SiC(s) + 2CO(g) = SiO2(1) + 3C(s) -142310 + 80.12 T -- -2.35
34 SiO(1) + CO(g) = SiO2(s) + C(s) - 93080 + 52.70 T 1.02 --
34 SiO(1) + CO(g) = SiO2(1) + C(s) - 90790 + 51.55 T -- -1.60
35 Si(s) + CO(g) = SiO(1) + C(s) - 68740 + 30.25 Z 2.65 --
35 Si(1) + CO(g) = SiO(1) + C(s) - 80820 + 37.42 T -- 0.43
36 SiC(s) + CO(g) = SiO(1) + 2C(s) - 51520 + 28.57 T 0.69 -0.76
(a) 1 cal = 4.19 J; T = temperature in Kelvin.
(b) K = e x p ( - A G ~ T).

(see Eq. [8]), or c o n v e r s e l y the C O partial pressure remains T h o s e conditions p r e v a i l i n g in industrial electric furnaceL
always a b o v e 0.25 atm w i t h o u t e x c e e d i n g 1.0 atm. H e n c e , are indicated by the hatched zone in F i g u r e 4. This is a range
of the conditions to be w a t c h e d closely at each temperature.
Pco(ind) + Ps,o(ind) = 1 atm [1]
B e t w e e n 1449 and 2200 ~ the stability zones o f Si and
0 < Ps~o(ind) < 0 . 7 5 [2] SiC are n e v e r adjacent to that o f SiO2. Accordingly, it is an
error, for e x a m p l e , to write the f o l l o w i n g c h e m i c a l reactionL
0 . 2 5 < Pco(ind) < 1 [3] as a reaction m e c h a n i s m at these temperatures:

D O, F E
~o.~ ' ' y // 1506"C // ..,

-1

1~,,,,/ ,,=o,,, ~, o o;
.3t / , , ~ . 6:~,, / /s~,=,/ ..7
- ./, / , / ,/ ,/ ,
-6 -5 -4 -3 -2 -1 -5 -4 -3 -2 -1 0
Log PCO LogPCO
(a) (c)
D F
1449 "C ' ' ' ~ ~ B

-1
.../ ..... .../,,,..,~____~g.~ / ..-
/ / ,/ / ,, , / /
/ / / ," /. ,/
2 ~,/ 0~/,/ / Si6(1) / s.~O~// -
-2
e=,,,f / .,," / r /',
,// / / p,~O~'O,OS
//
/ / /SiO,
, / . . . or
-3
a l.ll.~" "4I

-6
/,/"
-5
,/
-4 -3
, ,

-2
,

-1 -5
/,,<" -4
2" / / :
-3 -2 -1
/: 0
I
1
Log PcO LogPCC)
(b) (d)
Fig. 3 - - P r e d o m i n a n c e diagrams of the type C-CO. The full lines delineate the stable system which includes SiO(1), while the dotted lines the meta-
stable system where SiO is always gaseous. The value of ps,o written in the stability zone of SiO(1) indicates the SiO vapor pressure over pure liquk
SiO at the temperature. The SiO isobars are the same for both stable and metastable systems. The reaction numbers indicated in (b) correspond to those ir
Table III.

506--VOLUME 17B, SEPTEMBER 1986 METALLURGICAL TRANSACTIONS B

 sumed, or as a global mass balance. By the same token,
many of the reaction mechanisms previously proposed were
in error in the light of the present predominance diagrams
which provide a directly visible criterion for judging what
reactions are allowed and what are not.
/ / / / I
3-2 VI. MUTUAL SOLUBILITIES
,, o< /- _J ,
The phase boundaries in Figures 2 and 3 have been calcu-
-3 / / lated by assuming the coexistence of two phases of unit
activity. When a mutual solubility becomes appreciable,
/, ..',/ ,/ ,/ however, the boundary line may be drawn as a broad band
-5 -4 -3 -2 -1 0
Log PCO corresponding to gradually lowering activities.
(e) The solubility of SiO(1) in Si(s, 1) cannot be large due to
the fact that SiO(1) is an oxidic slag as opposed to Sill)
=.c
0 I I I I ~ I / /
which is a metallic (or covalent) melt. By contrast, the
sic ISI mutual dissolution between SiO(1) and SiO2 can be appre-
-1 // ,/ ,~///A//// ciable especially above 1722 ~ (m.p. of silica), because
both are then liquid oxide. SiO(1) may have a small solu-
bility of C or SiC at higher temperatures. Also, there might
O .,,, / ,/.// ,," be formation of other carbides such as Si2C, or oxycarbides
o, / // yo (,, such as Si2C202, which have been neglected in the pres-
--I +o;7" , ,/ ,// / / ent study.
x y/ ,,," / /
" Psto = ~,so sio, tl

/ , / i// ~ / ~ // i '~7 VII. DISCUSSIONS ON

-5 -4 -3 -2 -1 0 PREDOMINANCE DIAGRAMS
L o g PCO
All the discussions in this section refer to the predomi-
(f) nance diagrams of the type C-CO, or Figure 3.
Fig. 3 C o n t . - Predominance diagrams of the type C-CO. The full lines
dlineate the stable system which includes SiO(1), while the dotted lines the
metastable system where SiO is always gaseous. The value ofps,o written A. Below 1412 ~
in the stability zone of SiO(I) indicates the SiO vapor pressure over pure
liquid SiO at the temperature. The SiO isobars are the same for both stable Below the melting point of Si (1412 ~ SiO(1) is the
and metastable systems. only stable liquid, and both the univariant points A and B
can be realized, as shown in Figure 3(a). Since both points
A and B are far away from the industrial furnace conditions
(Figure 4), neither metallic Si nor SiC can be produced at
CARBON SATURATION
1
these temperatures, unless the system is flushed with a neu-
tral gas or put under a sufficiently reduced pressure. Metal-
9 FURNACE lic Si can be produced if the partial pressure of CO gas is
SIC IS)
kept below the value indicated by the univariant point A,
0,1 .-
viz., 2 • 10-5 atm at 1350 ~ (Figure 3(a)).
Under the conditions close to carbon saturation, or log
ac = 0, silica can be converted directly to SiC(s) without
going through SiO(1), provided the CO partial pressure is
0,01 ~ S i O 2(11 kept between 0.13 and 0.046 atm at 1350 ~ Such a direct
conversion is no longer possible above 1449 ~ where the
0,001 carbon activity at the univariant point B becomes unity, as
0,0001 0,001 0,01 0,1 10
shown in Figure 3(b). Incidentally, the SiC crystals pro-
PCO' aim
duced by solid-solid or solid-liquid reactions are bound to be
Fig. 4 - - T h e hatched zone which corresponds to 0.25 < Pco < 1.0 atm contaminated by unreacted reactants, whereas those formed
represents the conditions prevailing in industrial furnaces operating under
atmospheric pressure (1 atm = 101.3 kPa). This is the zone to watch at
via SiO gas can be purer.
each temperature. The background lines represent the conditions at
1860 ~ B. Between 1412 and 1722 ~
As temperature rises, the univariant point A moves gradu-
ally toward carbon saturation, at the same time as the CO
SiO2(s) + 2C(s) = Sill) + 2CO(g) [4] partial pressure of the point A, Pco(A), increases. The Si-
making reaction at the point A or C still remains dormant,
SIO2(1) + 3C(s) = SiC(s) + 2CO(g) [5]
unless the CO partial pressure is forcefully reduced below
They are allowed as a means of estimating the carbon activ- 10 -3 to 10 -4 atm. In the presence of excess carbon under
ity, for example, after carbon has been completely con- atmospheric pressure, the conversion of SiO2 to SiO(l), or
METALLURGICAL TRANSACTIONS B VOLUME 17B, SEPTEMBER 1986--507
that of SiO(1) to SiC, becomes possible at 1493 and 1530 ~
respectively:
Pco(E) + Ps~o(E) = 1 atm at 1493 ~
pco(F) + ps,o(F) = 1 atm at 1530 ~
Since the univariant point B which favors formation of SiC
comes within the industrial furnace conditions (Figure 4),
the surface o f charged carbon is likely to be coated by a layer
of SiC at these temperatures, insofar as silicon is available
from SiO gas (or liquid) or solid SiO2 nearby. It is to be
noted that under the stable system, SiO2 and SiC cannot
coexist above 1449 ~ because of the mutual reaction form-
ing SiO(1) (Figure 3(b)).
C. Above 1722 ~
A b o v e the melting point o f SiO2, there are two liquid
oxides: SiO(1) and SiO2(1), whose mutual solubilities might
be appreciable, as discussed earlier. It is possible that silica
crystals are in stable contact with SiC at lower temperatures.
While such an interface should disappear above 1449 ~ in
the stable system, the interface could exist up to 1779 ~ in
the hypothetical metastable system. However, such an inter-
face will be totally gasified at 1779 ~ by virtue of the
following reaction:
2SIO2(1) + SiC(s) = 3SiO(g) + CO(g)
where Ps,o(C) and pco(C) reach the critical values of 0.75
and 0.25 atm, respectively. Owing to this gasification reac-
tion, there is no formation of metallic silicon up to 2037 ~
although the univariant point C which represents the meta-
stable Si-making reaction is located well within the industrial
furnace conditions above 1779 ~
[6]
If heating is continued till metallic silicon can be pro-
[7] duced as a result of a boiling-like evolution of CO gas, i.e.,
Pco(A) = 1 atm, then the temperature must reach 2107 ~
as shown in Figure 3(f). At 2107 ~ metallic silicon can
coexist stably with a gas phase characterized by the point H
in Figure 3(f), where
Psio(H) = 0 . 5 0 atm
Pco(H) = 0.50 atm
In reality, metallic silicon can be produced slightly below
2107 ~ due to the presence of SiO gas in the atmosphere,
although the production yield is poor for the very same
reason.
D. Partial Pressures of SiO, CO, Si, and C
The isobars o f SiO gas are shown in some of the Figures
2 and 3. At a given temperature, the value of SiO vapor
pressure is the same all over the SiO(1) stability zone, while
it varies over other condensed phases as functions of C and
CO potentials. The SiO isobars have been calculated over
each zone of Si, SiC, and SiO2 where the unit activity of the
respective condensed phase prevails. The free energy data
[8]
and equations are listed in Table IV.
The partial pressures of SiO and CO at various univariant
points are listed in Table V. The points A and C are particu-
larly important from the Si-making point of view, and they

Table IV. Partial Pressure of SiO over the Condensed Phase of Si, SiC, and SiO2

Over Condensed
No. Phase of Reaction AG ~ cal <a) SiO Partial Pressure, arm. <b)
41 Si Si(s) + 0.502(g) = SiO(g) - 24900 - 19.72 T log pslo = log K41 + 0.5X
41 Si Si(1) + 0.502(g) = SiO(g) - 36980 - 12.55 T log ps,o = log K41 + 0.5X
42 SiC SiC(s) + O2(g) = SiO(g) + CO(g) - 35020 + 41.90 T log ps~o = log K42 + X - Y
43 SiO2 SiO(g) + 0.502(g) = SiO2(s) -191600 + 61.67 T log ps,o = - log K43 - 0.5X
43 SiO2 SiO(g) + 0.502(g) = SiO2(1) -189310 + 60.52 T log ps,o = - log K43 - 0.5X
44 Si Si(s) + CO(g) = SiO(g) + C(s) 2440 + 0.78 T log ps,o = log K44 + Y - Z
44 Si Si(1) + CO(g) = SiO(g) + C(s) - 9640 + 7.95 T log pslo = log K44 + Y - Z
45 SiC SiC(s) + CO(g) = SiO(g) + 2C(s) 19660 - 0.90 T log ps,o = log K45 + Y - 2Z
46 SiO/ SiO2(s) + C(s) = SiO(g) + CO(g) 164260 - 82.17 T log ps~o = log K46- - Y + Z
46 SiO2 SiO2(1) + C(s) = SiO(g) + CO(g) 161970 - 81.02 T log pslo = log K46- - Y + Z
(a) 1 cal = 4.19 J; T = temperature in Kelvin.
(b) 1 atm = 101.3 kPa; X = log Po2; Y = log Pco; Z = log ac; p = partial pressure in atm; K = exp(-AG~ T).

Table V. Partial Pressures of SiO and CO at the Univariant Points A, B, C, E, and F ~*) of the Si-C-O System

Number Temperature Range, ~ Partial Pressure, atm ~)

51 1200 to 2400 ps,o(A) = exp [ - ( 71180 - 29.47 T)/ 1.986 T]
52 1200 to 2400 Pco(A) = exp [ - (110120 - 46.27 T)/ 1.986 T]
53 below 1449 pco(B) = exp [ - (134640 - 76.83 T)/ 1.986 T]
54 below 1722 ps,o(C) = exp [ - ( 77310 - 37.11 T)/ 1.986 T]
55 below 1722 pco(C) = exp [ - (116250 - 53.91 T)/ 1.986 T]
56 above 1722 ps,o(C) = exp [ - ( 76165 - 36.535 T)/1.986 T]
57 above 1722 pco(C) = exp [ - (115105 - 53.335 T)/1.986 T]
58 below 1493 pco(E) = exp [ - ( 93080 - 52.70 T)/ 1.986 T]
59 1200 to 2400 Pco(F) = exp [ - ( 51520 - 28.57 T)/ 1.986 T]
(a) 1 atm = 101.3 kPa; T = temperature in Kelvin.
(*) The point A for the coexistence of Si, SiC, and SiO(1); the point B for the coexistence of SiC, SiO(1), and SiO2; the point C for the coexistence ot
Si, SiC, and SiO2; the point E for the coexistence of C, SiO(1), and SIO2; the point F for the coexistence of C, SiC, and SiO(1).

508--VOLUME 17B, SEPTEMBER 1986 METALLURGICALTRANSACTIONSP

are graphically shown in Figure 5. The intersections of A. Si-Making Reactions
Pco = Psio occur at 2045 ~ for both the stable and meta-
stable systems. This implies, as will be demonstrated in the Suppose that a charge consists of 1 mole of SiO2 and a
next section, that metallic silicon cannot be produced below given number of moles of C, and that the charge is heated
to various temperatures with the aim of producing metallic
about 2045 ~ (to be precise, 2037 ~ A minimum tem-
perature of 2037 ~ required for producing silicon metal is silicon. Knowing that the most favorable condition (i.e., the
substantially higher than previously believed. highest Pco) for Si-making is represented by the univariant
The partial pressures of atomic Si(g) and atomic C(g) point A, where Si, SiO(1), and SiC coexist, the overall
remain negligibly small under all the conditions of the reaction may be expressed as
present study. For example, even at such a high temperature Low temperature High temperature
as 2107 ~ they are as low as O O
nsio2, n c "'> nsl, n s i c , ns~o(1), nsio(g), nco
psi(over pure Si) = 1.4 • 10 -3 atm
where n's refer to the number of moles of the subscripted
pc(over pure C) = 2.9 • 10 -8 atm. species. Let nslo0) = n m for SiO melt, and ns,o(g) = nv for
SiO vapor. When all the SiO melt is converted to Si and/or
S i C , nm becomes zero. So, as mass balances before and after
VIII. FEEDS OF THE SYSTEM SiO2-C the reaction, we have
In industrial practice, raw materials are mainly silica Si bal.: n~,o2 = 1 = nsl + nsic + nv [9]
and coke for both Si- and SiC-making operations. Accord-
C bal.: nc~ = ns,c + nco [10]
ingly, one gram atom of silicon is inevitably accompanied
by two gram atoms of oxygen in the feeds. This brings O bal.: 2n],o~ = 2 = nv + nco [11]
about an important limitation on the thermochemistry of the
Si-C-O system. Since there are 4 species (Si, SiC, SiO(g), CO) and 3 mass
balances, the reaction products are interrelated by 1 equilib-
rium; i.e., 4 - 3 = 1. This equilibrium may be chosen as
Temperature, *C
1500 1600 1800 2000 2200 2400 SiC(s) + SiO(g) = 2Si(1) + CO(g) [12]
I I I I f I I I I I /11
100
K~E = e x p [ ( - 3 8 9 4 0 + 16.80T)/1.986T]
Pco(t _ a 2
K12 sP
ic_________~o_ Pco _ nco [13]
asicPs~o PSlO nv
lO
If there is a condition where metallic Si is the only product
as condensed phase, the value of ns,c must become zero.
Under those conditions,
nsi > 0, ns1c = 0; nv > 0; nco > 0
[141

i
a. o,1 PsiolC) ~ /r ~
b.p.
SiO(=)
On the other hand, below 2045 ~
we have
Psio > Pco or
as shown in Figure 5,

nv > nco [15]

(A}
Rewriting Eqs. [9] to [11], and using the Si-making condi-
tions of the inequalities [14] and [15], we find
O O
ns,= nc- 1 > 0 or nc > 1 [16]
0
m.p. nr = 2 - nc > 0 or 2 > n~ [17]
SiO2
0 0
nco= nc>0 or nc>0 [18]
nv= 2- n co > n ~ or 1 >nco [19]

The inequalities [16] and [19] are incompatible, and there-

I/ I I I I
fore the simultaneous Eqs. [9] through [13] do not have a
0,55 0,50 0,45 0,40 solution below 2045 ~ In other words, metallic Si cannot
IO00/T be produced unless temperature is raised above 2045 ~
Fig. 5 - Partial pressures of SiO and CO at the univariantpoints A and C. regardless of stable or metastable system, and whenever
The value ofps,o(A)is identicalwith that ofps,ooverpure SiO(1). The point metallic Si is being produced, the evolving gas must contain
A represents the three-phase coexistence of Si, SiC, and SiO(1), whereas more SiO(g) than CO(g).
the point C that of Si, SiC, and SiO2. The value ofpco(A, C) represents the
most favorableSi-makingconditions (i.e., the highest allowed CO partial In order to find the solution for temperatures above
pressure) at each temperature. Note Ps,o(A, C) > Pco(A, C) below 2045 ~ nslc should be restored in Eqs. [9] and [10] such
2045 ~ 1 atm = 101.3 kPa. that the conditions correspond to the unit activity of SiC as

METALLURGICAL TRANSACTIONS B VOLUME 17B, SEPTEMBER 1986--509

assumed in Eq. [13]. The solution is given as: B. Conversion o f SiO(l) to SiC
nv = 2/(1 + K12) [20] In SiC-making operations, an incomplete conversion 0
SiO(1) to SiC can be avoided by heating a mixture 0
ns, = 3 - n~ - [4/(1 + KI2)] [21] (SiO2 + nC) above a certain temperature. The relationshil;
ns,c = n~ - 2 + [2/(1 + Kz2)] [22] between the carbon content of the mixture and the minimmr
temperature may be calculated in the following manner
n c o = 2 - [2/(1 + K,2)] [23] Suppose that a feed of (1.0SiO2 + n~C) is heated tc
obtain a maximum amount of SiC under the stable equi
The equations [20] through [23] have been calculated at
librium conditions. The reaction may be written as
various temperatures, and shown in Figure 6 as a zone of
Si + SiC + Gas, and also in Figure 9 to indicate varying Low temperature High temperature
amounts of each product at 2200 ~ The highest recovery
of metallic Si at 2200 ~ is 26 pet, and it occurs when the n~
s~o2, n C
~~ ns~c, ns~o(1), ns,o(g), n c o
cold charge consists of (1.0SIO2 + 1.26C) (Figure 9). The The mass balances before and after the SiC-making reac
results also indicate that the production of metallic Si at tion are
2100 ~ for example, can be done from a mixture of
(l.0SiO2 + 1.1C) but not from that of (1.0SiO2 + 1.2C), Si bal.: n~,o2 = 1 = ns~c + n,,, + nv [24
the latter yielding no Si but only SiC. This is quite con- C bal.: n~ = nsic + rico [25
tradictory to the conventional understanding that for Si-
making operations, the feed composition should be close to O bal.: 2n~1o2 = 2 = n,, + nv + nco [26
(1.0SiO2 + 2C). It is to be noted that this rather surprising
There are 4 species (SIC, SiO(1), SiO(g), CO) and 3 mas'~
outcome has resulted straightforward from the reliable free
balances, so that among the products there should exis
energy data, and it is not a consequence of any assumption.
1 equilibrium, which may be chosen as
SiO(1) = SiO(g) [1
K~ = e x p [ ( - 7 1 1 8 0 + 2 9 . 4 7 T ) / 1 . 9 8 6 T ]
~ I I I I I I I
[27
3,0 C+SiC§
K1 = Ps,o = Pnv/(nv + n c o ) [28
where the total pressure of the system P may be assumed t(
be atmospheric, or P = 1 atm. The solution of the simulta
neous Eqs. [24] to [28] can be expressed as
2,5
SiC+Gas
nv = 0.5K~(1 + n~)/(1 - KI) [29
n,, = 2 - 0.5[(1 + n~)/(l - K1)] [30
nslc = 0.5(n~ - 1) [31
~ 2,0 SiC§
(J nco = 0.5(n~ + 1) [32
C Thus, the number of moles of the four products can b~
§
calculated as a function of the molar ratio, or C/SiG
~1,5 (= n~), in the feed so long as n~ -> 1. The values of n~ tha
._= complete the conversion of SiO(l) to SiC are plotted ii
Figure 6 as a function of temperature.
Between 1493 and 2142 ~ (b.p. of liquid SiO), SiO
1,0 disappears completely at n~ = 1 and is converted to SiO(1
Si+SiO(J) + G u and SiO(g). Above 2035 ~ (but below 2045 ~ SiO(1
does not exist in contact with SiC, as shown in Figure 6
Thus, above 2035 ~ SiC can be formed exclusively frorr
SiO2+Gas the reaction of SiO gas with solid C.
0, 5 ~ SiO2+SiO( II +Gas

C. SiC-Making at C-Saturation
In view of avoiding the contamination of SiC products b"
0 ~ graphite, it is important to know the maximum allowec
1600 1800 2000 2200
Reaction temperature, ~ content of C in the feed.
Suppose that a cold feed of (1.0SiO2 + n~C) is heatec
Fig. 6--Various products under the stable systemresulting from the high- to a high temperature where the products include exces
temperature reaction of a room-temperaturemixture consisting of 1 mole
of silica and a varying number of moles of carbon. The Gas refers to carbon as well as SiC, or ac = 1. The reaction may be ex
SiO(g) + CO(g). pressed as

510--VOLUME 17B, SEPTEMBER 1986 METALLURGICAL TRANSACTIONS i

 Low temperature High temperature The low-temperature values of n~ which give zero re-
n ~s,o2, n~ ~ ns,c, nsio(g), n c o , n c sidual carbon, or nc = 0, at high temperatures are plotted in
Figure 6 as a function of temperature. It is noteworthy that
The mass balances are the C/SiO2 ratio ( = n ~ ) in the cold mixture should be appre-
ciably less than 3 if the contamination by graphite is to be
Si bal.: n~io2 : 1 = ns,c + nv [33]
prevented; e.g., 2.91C vs' 1.0SIO2 for heating at 2045 ~
C bal.: n~ = ns,c + nco + nc [34]
D. Other Considerations
O bal.: 2n],o2 = 2 = nv + nco [35]
The thermodynamic analyses for the hypothetical meta-
In view of 4 species (SIC, SiO(1), C, CO) and 3 mass bal- stable system are summed up in Figure 7, where the condi-
ances, there is 1 independent equilibrium, which may be tions for Si-making and the SiC-making at C-saturation
chosen as remain the same as in the stable system.
SiO(g) + 2C(s) = SiC(s) + CO(g) [36] The variations of each product in the stable system at
1800 ~ are shown in Figure 8, featuring the possible con-
K36 = exp[(19660 - 0.90T)/1.986T] [37] tamination of SiC by SiO(1) for most of the SiC-making
conditions at this temperature.
as,cPco _ n c o [38] SiC is known to decompose into C(s) and Si(l) above
K36- a~Psio nv
2730 to 2900 ~
The solution of Eqs. [33] to [38] is given by
nv = 2/(1 + K36) [39] IX. DISCUSSIONS ON I N D U S T R I A L P R A C T I C E
nsic = (K36 - 1)/(1 + K36) [40] A. SiC-Making
n c o = 2 K 3 6 / ( 1 + K36) [41] In regard to SiC-making, perhaps the most important
outcome of the present study is the fact that the heating must
nc = n~ - (3K36 - 1)/(1 + K36) [42]
be controlled ideally between 2035 and 2045 ~ throughout
a heat. A slight underheating may be more acceptable than
1 T ~ T r TmT-~--
an overheating provided the heating is prolonged in accor-
dance with increasing carbon content m the cold charge
C+SiC+Gas
3,0 - (Figure 6). On the other hand, overheating above 2045 ~
inevitably brings about the formation of metallic Si. This not
only is a waste of electrical energy, but can cause molten
metal to drip down through the granular solid charge to the
furnace bottom, which is then to be dug and recemented
2,5
anew to prevent electrical short circuit.
Between 2035 and 2045 ~ SiC can form solely as a
SiC+Gas result of the carburization of SiO gas without attendant
formation of liquid SiO or metallic Si. The SiC crystals
10 2,0 resulting from gas-phase reactions can be undoubtedly purer
8 than those formed from condensed-phase reactions. The
stability zone of SiC quickly narrows below about 1900 ~
r- SiC+SiO2+Gas
§
as indicated by the converging boundary curve between
SiC + Gas and SiC + SiO(1) + Gas zones (Figure 6).
c
e- ,D i i 1 i i ..... 2,0
1800"C
Si+ Gas SIC
1,0

SiO2+G,as

0,5

~ ~ ~ t .L-.--.
1600 1800 2000 2200 ~00 05 10 15 20 2,5 3,0 0
Reaction temperature, "C n =n(SiO~+n.C)cold
Fig. 7--Various products under the metastable system (= absence of Fig 8 - - Number of moles of SiO:, SiO([, g), SiC, CO, and C at 1800 ~
liquid SiO) resulting from the high-temperature reaction of a room- resulting from a cold feed consisting of (1.05iO2 + nC), in the stable
temperaturemixtureconsistingof 1 mote of silica and a varyingnumberof system under 1 arm (= 101.3 kPa). The right-hand side scale is for CO
moles of carbon. gas only.

METALLURGICAL TRANSACTIONS B VOLUME I7B, SEPTEMBER 1986--511


1,(~ 2200~ I I I I
i SiOlg)
0,5 1,0 1,5 2,0 2,5
n in (SlO~ + n'C)eold
Fig. 9 - - N u m b e r of moles of high-temperature reaction products at
2200 ~ resulting from a cold feed of (1.0SiO2 + nC). Note the for-
mation of metallic Si peaking at n = 1.26. Under the stable system at a
total pressure of 1 atm (= 101.3 kPa). The right-hand side scale is for CO
gas only.
Thus, the portion of the furnace feed which has not under-
gone a heating above 1900 ~ is unlikely to become quality
crystals of SiC. Although these semi-finished crystals can
still be reused as a charge in later heats by recycling, efforts
should be made to operate a furnace in such a way that the
largest possible cylinder heated above 1900 ~ may result
therefrom.
The stoichiometric mixture of (1.0SiO2 + 3.0C) infal-
libly results in the contamination of SiC by C irrespective of
heating temperature. This is due to the effect of SiO gas.
Such a contamination may be averted in practice by using
slightly less carbon in the feed than 3/1 molar ratio de-
pending on heating temperature, with Figure 6 as guide. For
example, the C/Si ratio must be set to 2.91 for the heating
at 2045 ~
If the metastable system is assumed, the conditions fa-
vorable to SiC-making widen considerably, as shown in
Figure 7, suggesting a very easy control of temperature in
practice. Nonetheless, the most efficient heating can still
be done just below 2045 ~ as in the case of the stable re-
action system.
B. Si-Making
As mentioned earlier, in order to produce metallic Si, a
feed consisting of SiOz and C in whatever ratios must be
heated to a minimum temperature of 2045 ~ (Figures 5
and 6). (Incidentally, if the presently used free energy data
for reaction [12] is subject to an error of 2 kcal/mole, the
critical temperature of 2045 ~ may lower to 1926 ~ If a
feed consists of (1.0SiO2 + 2.0C), temperature must be
raised above 2400 ~ to produce metallic Si.
In practice, the temperature of electrode surface and the
cavity walls surrounding the electrode has been measured to
be about 2000(---50) ~ as reported by Schei and Larsen. 6
The theoretical maximum yield of Si-making from SiO2
is merely 0.9 pct at 2050 ~ (with n~ = 1.01, Figures 6
and 9), and even with the lower free energy giving 1926 ~
as the minimum temperature, the yield at 2000 ~ is only
13.7 pct with n~ = 1.137. This is a great contradiction with
the fact that industrial electric furnaces actually produce
metallic Si at a fairly respectable yield of 65 to 90 pct at an
electrode temperature of about 2000 ~
512--VOLUME 17B, SEPTEMBER 1986
2,O
m
This theoretical impossibility of Si-making despite suc-
cessful industrial operations has been well known to the
industry, although in none of the previous theoretical stud-
co
ies, the approach was so mathematically rigorous as is in
this work, nor the presence and role of liquid (amorphous)
SiO was recognized as such. The industry has been also
1,O
aware of the experimental impossibility of a small-scale
Si-making in a laboratory furnace. 5 As described in the
discerning paper by Selmer-Olsen, 5 a fascinating series ot
experiments carried out by W. M. Kelly led the engineers in
the field to postulate a double-reactor model to account for
the mysterious thermochemistry in commercial Si-making
3,0 furnaces.
The double-reactor model was first proposed by Sandber5
et al. 4.1oThe model assumes that ascending SiO gas react.
with the coke situated in the upper cooler parts of an electric
shaft furnace to form SiC, which in turn descends to a lower
hotter zone in the vicinity of electrode tips, and reacts with
SiO2 or SiO(g, 1) to yield metallic Si. In other words, a
furnace has two distinct functions of reduction (lower zone;
and carburization (upper zone). Schei and Larsen 6 further
developed and refined the double-reactor model, while
Selmer-Olsen5 extended the concept further to formulate
multiple-layer reactor model.
According to the present thermodynamic study where the
effect of SiO(1) has been established, an additional im-
portant mechanism to be considered is the simple physical
condensation of SiO(g) as SiO(1) on cooler surfaces of the
furnace charge. When temperature lowers from 1900 tc
1500 ~ for example, the SiO vapor pressure over pure
SiO(1) becomes as low as 1/75 (Figure 5). This suggests tha,
98.5 pct of the SiO gas coming from the hotter zone car
condense out on the cooler charges (SiO2 and SiC) situatec
in a 1500 ~ zone under an ideal gas-solid contact which ha~
a high condensation efficiency. Thus, the present thermo
dynamic model wherein SiO(1) has been included as a stabl
phase explains well the condensation of glassy SiO observec
in Kelly's experiments, 5 as well as a high yield such as 7f
to 85 pct that is observed in industrial Si-making operations.
Suppose that the condensation of SiO(1) and the carbu-
rization bring about a compositional change such that a feec
at the time of the arrival at hot electrode tips can be ex-
pressed by a binary mixture of (SiO(1) + n 9 SIC). The
yield of metallic Si production from such a binary feed k
graphically shown in Figure 10 as functions of mixture com
position and temperature, together with the yields of otheI
likely feed systems (Table VI). These feed systems may be
considered resulting from the following partial reaction:
which the initial charge has undergone:
2SIO2 + 4C = [SiO(1) + SiC] + 3CO ~'
2SIO2 + 4C = [SiO2 + SiC + C] + 2CO 1'
2SIO2 + 4C = [SiO(1) + SiO2 + 3C] + CO 1'
2SIO2 + 3C -- [SiO2 + SiC] + 2CO 1'
where [ ] refers to a feed system at the moment of reaching
the hot reaction zone, and CO 1' to carbon monoxide ga~
that can be eliminated possibly before the charge descend~
to the hot electrode tips. A high yield of Si-making ir
industrial furnaces can be explained in terms of the for
mation of SiC and SiO(1) in the upper zone of electric
furnaces, as shown in Figure 10. It is to be noted that on the
METALLURGICAL TRANSACTIONS i
 Table VI. Solutions for Equilibrium Mass-Balances of Various Feed Systems under Atmospheric Pressure
Feed system SiO(1) + n ' SiC SiO2 + n 9 SiC SiO2 + SiC + n 9 C SiO(1) + SiO2 + n 9 C

Reactants n~,oo) = 1, n s,c

~ n s~o2
~ = 1, n SlC
~ n slo2
~ = 1, n s,c
~ = 1, n ~ o
~SlO~[) = 1 , n ~SlO 2 = 19 ~

Products ns,c(---* 0), nv, n s , nco

Equilibrium SiO(g) + SiC(s) = 2 Si(1) + CO(g) K = e x p [ ( - 3 8 9 4 0 + 16.80 T)/1.986 T] = nco/nv

Solution nv = 1/(1 + K) nv = 2/(1 + K) n~ = 2/(1 + K) n~ = 3/(1 + K)

nco = K/(1 + K) nco = 2K/(1 + K) nco = 2 K / ( 1 + K) nco = 3K/(1 + K)
ns,c = n~lc [K/(1 + K)]
- - ns,c = n~,c - [2K/(1 + K)] ns,c = [(1 - K)/(1 + K)] + n~ ns,c = n~ - [3K/(1 + K)]
ns, = 2K/(1 + K) ns, = (3K - 1)/(1 + K) ns, = [(3K - 1)/(1 + K)] - n~ ns, = [(5K - 1)/(1 + K)] - n~
Note: n refers to the number of moles of the subscripted species, and n~, = nslo(g).

lOG I I i ! i to direct interactions between charged silica and coke, but

also to the carburization of ascending SiO gas on coke
surfaces. The fixation of SiO gas by coke depends greatly on
the surface area per volume and reaction kinetics of the
coke. This important parameter was astutely recognized
! by Tuset and Raaness, 7 who developed a reliable experi-
mental method to measure and quantify such a surface reac-
tivity. The method has been put in commercial practice in
70 I j,~,55 / 0,40~=~ S i O 2 * S i C * n ' C Scandinavia. 7
It is surprising that the double-reactor model or coke
60 , //o,2e reactivity were well recognized and incorporated in the in-
i ~ ,~o// sio(~)+sio...-c dustrial practice by those watchful engineers even when a
~, , /j 2,o~

/
systematic thermodynamic analysis such as this study was
still unavailable. The present thermodynamic analysis has
given a theoretical basis to these technical developments.
.'~ -I, I /1,89 Si02+n.C More importantly, however, the present work has clarified
and unified all the thermochemical phenomena of the Si-C-
O system, and has laid down the general and basic strategy
for future process improvements. An emphasis has been

20 i 1,48
placed in this study on the enhancement of the physical
condensation of ascending SiO gas on the charge in the
upper and cooler part of the shaft furnace. Such an enhanced
condensation is essential to the increase in Si-making effi-
ciency, and it can be done by increasing the layer thickness
lO ! /1,1o of upper charge or making a shaft furnace taller as well as
by cooling the upper part of the furnace.
0 I ~Jl,OO i I I
2000 2100 2200 2300
Temperature, ~ ACKNOWLEDGMENTS
Fig. 1 0 - - T h e theoretical maximum yields of Si-making from various feed
systems. The numbers along a curve indicate the number of moles, n, of The authors are grateful to Mr. R. Morissette, Plant
a variable constituent in the feed system. Note the high Si-making yields General Manager, and Mr. R. Leblanc, Quality Control
with the feed systems containing SiO(1), implying that the condensation of Manager, of Norton Company, Quebec Plant, for invaluable
SiO(1) is a key factor influencing the production efficiency in industrial
Si-making furnaces. This graph also lends thermodynamic support to the
suggestions and inspiring discussions. The authors also ex-
double-reactor model of Sandberg et al. 1o press their gratitude to Mr. J. Gobeil, Division Head, Met-
allurgical Research Laboratory, CRM, for encouragement.

basis of the predominance diagrams, the coexistence of S i O 2

and SiC at the time of entering the hot reaction zone is not
allowed and therefore unlikely to take place actually. Never-
theless, the calculated yields are valid insofar as the chem-
ical (not necessarily physical) composition of a feed may be
so approximated.
The high Si-making yields in various feed systems (Fig-
ure 10) can be construed as lending firm physical support to
the double-reactor model. ]~ The SiC-forming reactions in
the upper zone of an arc furnace may be attributed not only
REFERENCES
1. H . H . Kellogg and S. K. Basu: TMS-AIME, 1960, vol. 218, pp. 70-81.
2. M. Riss and Y. Khodorovsky: Production of Ferroalloys, translated by
I.V. Savin, MIR Publisher, Moscow, USSR, 1967.
3. R. Duffer and G. Volkert: Metallurgie der Ferrolegierungen, 2nd ed.
by G. Volkert and K . D . Frank, Springer-Verlag, New York, NY,
1972.
4. A . G . E . Robiette: Electric Smelting Processes, John Wiley & Sons,
New York, NY, 1973.
5. S. Selmer-Olsen: 39th Electric Furnace Conference Proceedings, ISS-

METALLURGICAL TRANSACTIONS B VOLUME 17B, SEPTEMBER 1986--513

 AIME, Warrendale, PA, 1981, pp. 310-18.
6. A. Schei and K. Larsen: 39th Electric Furnace Conference
Proceedings, ISS-AIME, Warrendale, PA, 1981, pp. 301-09.
7. J. Kr. Tuset and O. Raaness: 34th Electric Furnace Conference
Proceedings, ISS-AIME, Warrendale, PA, 1976, pp. 101-07.
8. O. Kubaschewski and C. B. Alcock: Metallurgical Thermochemistry,
514--VOLUME 17B, SEPTEMBER 1986
5th ed., Pergamon Press, New York, NY, 1979, pp. 221-26.
9. E.T. Turkdogan: Physical Chemistry of High Temperature
Technology, Academic Press, New York, NY, 1980.
10. O. Sandberg, H. Natvig, and O. Braaten: Symposium on Ferroalloy
Industry in India, The Technical Journals of India Private Ltd., 1962,
pp. 158-60.
METALLURGICAL TRANSACTIONS P