Dissolution in Ceramic Systems, I1
Table V. AH for Solid-Vapor Reactions
I free energies of formation.
10 1.2 14 1.6
+ x 103 (OK-’)
Fig. 7. Comparison of observed univariant curves (solid lines) with curves
calculated from thermochemical data (dashed lines).
A t 700°K the experimental Po2 is 10 bars; the thermochemi-
cal Po2 is 0.1 bar. A much greater discrepancy is found
in the reaction 4CrO2 2Cr203 02,21 +
At 700’K the
experimental Po2 is 200 bars; the thermochemical value is
186,000 bars.
The explanation for these seemingly large anomalies lies
in the nature of the calculation. To obtain the equilibrium
pressure a t some temperature one takes the antilogarithm of
Dissolution in Ceramic Systems:
Al uminu, MuIIite, Anorthite, and Silica in a
CaIcium-Aluminum-SiIicate SIag
by B. N. SAMADDAR, W. D. KINGERY, and A. R.
Ceramics Division, Department of Metallurgy, Massachusetts Institute of Technology, Cambridge, Massachusetts
Experimental measurements have been made on
the rate of dissolution of sapphire, dense poly-
crystalline alumina, mullite, fused silica, and
anorthite in a 20A1203-40Ca0-40Si02 slag at
1350’ to 150OOC under conditions of free and
forced convection. Dissolution rates and their
temperature dependence were found to be con-
trolled by transport within the liquid boundary
layer. At low temperatures alumina is by far
the most corrosion resistant ; at higher tempera-
tures differences between the refractories tested
are less pronounced.
1. Introduction
N A previous paper1 the writers analyzed the factors that
I govern the rate of solution, described an apparatus
suitable for precise measurement of solution rates at high
249
Slope AH
Reaction (bars/’ C ) (callmole)
24Pb02 S 2Pb12019 502‘
2PbizOis 8Pb304 302
+
+
-5.86
-7.55
26,900
34,600
2Pbj04 6PbO f 0 2 -7.35 33,700
the small difference term obtained by subtracting the free
energies of formation of the solids. An error of 5% in the
thermochemical data can make a difference of 10 orders of
magnitude in the oxygen pressure. Predicting a pressure-
temperature curve to within experimental error ( +1O0C,
*5% in pressure) would require an accuracy of 0.01% in the
Heats of reaction calculated from the slopes of the curves
in Fig. 7 using the Van’t Hoff equation are given in Table
v.
Acknowledgment
The work encompassed in the synthesis phase of this work was
supported by the Lead Industries Association Expanded Re-
search Program through a fellowship to W. B. White.
Research support for the study of physical properties of insula-
tor and semiconductor materials was provided by the Division
of Research, United States Atomic Energy Commission, Contract
AT(30-1)-2581.
2 1 Buichi Kubota, “Decomposition of Higher Oxides of Chro-
mium Under Various Pressures of Oxygen,“ J . Am. Ceram. SOC.,
44 [5] 239-48 (1961).
11, Dissolution of
COOPER, JR.
temperatures, and reported the dissolution rate of single-
crystal sapphire in a calcium-aluminum-silicate slag. Analy-
sis of literature data indicates that dissolution kinetics dur-
ing corrosion are generally controlled by transport processes.
Received June 3, 1963; revised copy received November 6,
1963.
This research was carried out at the Massachusetts Institute of
Technology with support from the United States Atomic Energy
Commission under Contract No. AT(30-1)-.1852.
A t the time this work was done, the writers were, respectively,
research associate, professor, and associate professor, Ceramics
Division, Department of Metallurgy, Massachusetts Institute of
Technology. B. N. Samaddar is presently with the Department
of Applied Chemistry, University of Calcfjtta, Calcutta, India.
A. R. Cooper, Jr., and W. D. Kingery, Dissolutionin Ceramic
Systems: I, Molecular Diffusion, Natural Convection, and
Forced Convection Studies of Sapphire Dissolution in Calcium
Aluminum Silicate,” J . Am. Ceram. Soc., 47 [l] 37-43 (1964).
250 Journal of The A merican Ceramic Society-Samaddar
Fig. 1. Isothermal cut through the lime-alumina-silica phase equilibrium
diagram a t 1400' and 1 5 0 0 ° C illustrating solid phases in equilibrium with
the liquid slag.
et al. Vol. 47, No. 5
(2) Measuremenfs
Experimental measurements were made with the apparatus
and techniques which have been described previous1y.l
Free convection measurements were made with a specimen
dipped into the molten slag to a depth of in. Forced
convection measurements were made by dipping a 0.3 to
0.35-in.-diameter rod into the molten slag to a depth of in.
and rotating it at 1200 rpm. After the specimen was re-
moved from the melt, adhering slag was dissolved by im-
mersing the specimen overnight in 1/1 hydrochloric acid.
For determination of corrosion diameters, slag-free specimens
were measured at intervals of in. with a micrometer having
a l/M-in.-diameter point. For anorthite, itself soluble in
HC1, this technique could not be used. Anorthite specimens
were cut and polished and the specimen cross section was
measured directly with a filar micrometer microscope.
Measurements were made at constant temperature over
the range 1350" to 1500°C.
111. Results
(I) Free Convection
Experimental results for the rate of corrosion under con-
ditions of free convection from a cylindrical specimen are
illustrated for sapphire, polycrystalline alumina, mullite,

In thc case of single-crystal sapphire, experimental results

were consistent with transport in a liquid boundary layer as
the rate-controlling step ; the calculated diffusion coefficients
0.225

rr 0
1350 "C
Vitreous silica
A Mullite
for dissolution were internally self-consistent and coincided x Sapphire -
at higher temperatures with measured values for tracer v Polycryst. Ale03
diffusion coefficients of Si4+.
In the present investigation these studies have been ex-
tended by determining the dissolution rates of polycrystal-
line aluminum oxide in comparison with sapphire single
crystals and by determining corrosion rates for mullite,
anorthite, and fused silica, solid ceramic materials which are
primary phases in equilibrium with liquids in the system
lime-aluminasilkate. Isothermal sections through the
lime-alumina-silica phase diagram at 1400" and 1500°C
are shown in Fig. 1.
II. Experimental
(I) Moferiuls
Single-crystal sapphire specimens were supplied by the
Linde Division of Union Carbide Corporation. Polycrystal-
line aluminum oxide specimens were prepared by hydrostatic
pressing of Linde Type A-5175 alumina powder and sintering
at 1800°C to form a solid specimen having 97.2y0 of theo- Time ( h r )
retical density.
Mullite specimens were prepared from fused mullite grain
supplied by Babcock & Wilcox Company. The grains were
crushed, milled, leached with hydrochloric acid until free
from iron, pressed, and sintered to form a polycrystalline
material having 92.5y0 of theoretical density,
Anorthite specimens were prepared by sintering the
stoichiometric amounts of analytic reagent grade calcium
carbonate, silicic acid, and alumina. Anorthite formation
was confirmed by X-ray examination. Specimens were sub-
sequently pressed and sintered to 85% of theoretical density.
Vitreous silica specimens were cut from fused silica rods
supplied by the General Electric Company.
The slag was made by melting together proper amounts of
analytic reagent grade calcium carbonate, silicic acid, and
alumina to give the composition 20 wt% A1~03, 40 wt% Time (second)
SOz, 40 wt% CaO. The slag was chemically analyzed be- Fig.. 2. Free Convection corrosion of sapphire, polycrystalline alumina,
fore and during testing, and the composition was adjusted i O ~ ( A ) 1 35OoC
mullite, and vitreous silica in the 4 0 C a 0 - 2 0 A l ~ O ~ - 4 0 Sslag.
by small additions as required. and (5) 15OO0C.
May 1964 Dissolufion in Ceramic Systems, II 251
0 150

A
1350 O C
-__
Mullite I 0
1460°C
Sapphire
(6,
I
Poly A1203
I 0.100

-:
c
0
- ~ _ _ _ - e
b 0050
u

-
0
0 3600 7200 10,800 14, 00
Time (second)
0,200-
I390 O C
0.25(

Anorthite

0.20c

- 0.150

I
5
c
0
?
s 0.100
I500 "C
0 2000 3000 4000 51 x Sapphire
Time (second) 0.05C v Polycryst A1203
A Mullite
Vitreous SiOz

0 I
Sapphire
1000 21 0 3000 4000
Palycryst A1203 T i m e (second)
Mull i te
Vitreous silica ___ Fig. 3. Corrosion rate under forced convection condi-
Anorthite 0 ~ of indicated
tions in the 4 0 C o 0 - 2 0 A l ~ 0 ~ - 4 0 S islag

I
specimens of sapphire, polycrystalline alumina, mullite,

I vitreous silica, and anorthite at: ( A ) 1 35OoC, (8)

1390°C, (C) 1425OC, ( D ) 1460°C, and (€1 1500°C.

-0 I I I
0 3000
Time (second)

and vitreous silica in Fig. 2. The rate of corrosion of poly-
crystalline alumina and sapphire was identical within the
precision of experimental measurements. The corrosion
of vitreous silica and of mullite was substantially more rapid
than that oC alumina. Vitreous silica was corroded at twice
the rate for iiiullite a t 1350OC; a t 1500'C the rates ol cor-
rosion were nearly identical.
(2) Forced Convection
Experimental results for forced corrosion at temperatures
from 1350° to 1500°C are shown in Fig. 3. At temperatures
below 1425OC the small corrosion rate for alumina was not
measured. The rate of corrosion of single crystal and poly-
crystalline alumina was essentially identical. (At 1500°C the
initial corrosion of the polycrystalline material was more
rapid than that of the single crystal, probably caused by
greater surface roughness, but after the first 20 minutes the
rates of corrosion of single and polycrystalline alumina be-
came equal.) At the lower temperatures the rates of cor-
rosion of vitreous silica and anorthite were substantially
greater than that of mullite. A t temperatures above 1400°C,
however, the rate of corrosion of niullite became more rapid
than that of vitreous silica. At the higher temperatures
perhaps the most striking result is that alumina, mullite, and
vitreous silica varied by a factor of only 5 in their overall
corrosion resistance. A t lower temperatures there is a much
wider range of observed corrosion rates among the materials
tested.
(3J Microscopic Observcrtions
Smooth interfaces were observed for anorthite, fused
silica, and alumina during dissolution. For mullite dis-
solution a t 1500"C, needlelike crystals were observed a t the
interface as shown in Fig. 4. The concentration of the new
crystalline phase diminished in an air-quenched specimen,
suggesting that these crystals are precipitated during cooling
252 Journal of The American Ceramic Society-Samaddar et al. Vol. 47, No. 5

Fig. 4. Anorthite crystals precipitating a t the interface of mullite,

on left, and slag, on right, after corrosion a t 1 5 0 0 ° C and air
quenching (polished section, unetched, X 100). Note that the
porosity in the mullite specimen remains in the boundary layer of
the slog during corrosion and also that the mullite specimen has a
small fraction of a glass phase present between the mullite crystals.

rather than formed as part of the corrosion process. X-ray

examination showed that anorthite was the main phase
prccipitating, with a-alumina as a minor phase.
At 1350°C the interface formed on mullite contained a
thin layer of anorthite with substantial growth of a-alumina
(Figs 5 ( A ) and ( B ) ) . The phase identification was made
by X-ray diffraction and was confirmed optically by de-
termining indices of refraction.
Fig. 5. Growth of or-alumina a t the bottom of the mullite specimen during
forced convection corrosion a t 135OOC. ( A ) Macroscopic view o f the
specimen. ( B ) Microscopic view of the interface with mullite a t left, and
slag a t right. Major new phase i s m-AlpO~with thin layer of anorthite
between alumina and mullite (polished specimen, unetched, X 160).

IV. Discussion
(I) Corrosion Mechanism
The increased rate of corrosion under conditions of forced
convection as compared with free convection, the shape of
the specimen after corrosion, and the detailed analysis pre-
viously reported for single-crystal sapphire' indicate that
the rate limiting step during corrosion is diffusion through a
boundary layer, the thickness of which is determined by hydro-
dynamics. The transport rate of material through the
boundary layer depends on the effective diffusion coefficient.
From these considerations under conditions of forced cor-
rosion, the rate ol corrosion of a disk face rotating at an
angular velocity, W , is given by the relation1
--3
, ax = ij ci - c, -
- pendently of one another, or perhaps as complex ions or
(1 - ViCi)
~~
In Table I the effective diffusion coefficient has been cal-
culated on the basis of slag characteristics assuming that the
effective viscosity can be represented by the bulk viscosity*
and measured corrosion rates. These calculations must be
viewed with caution, however. The writers, on the basis of
the previous discussion, note that the liquid a t the solid-
slag interface is saturated with the dissolving refractory and
this determines the concentration gradient. However, the
phase diagram does not give a positive indication as to the
precise composition of the saturated liquid which may fall
a t any point along the liquidus curve, depending on the rela-
tive mobilities of the individual ions present in the boundary
layer. With four ions (Ca2+,A13+,Si4+,02-), diffusing inde-

=-
dt $ molecular groups, the composition through the boundary
where j is the dissolution rate (cm3 per cm2 per second),
is the linear distance corroded (cm), t is time (second), D
is the effective binary diffusion coefficient between saturated
and bulk solution (em2per second), is the effective boundary
thickness (cm), v" is the effective kinematic viscosity, C is
volume concentration of refractory, i.e., volume refractory/
volume refractory +
bulk solution, V is the partial volume
(cm3 per em3) of the refractory oxide, and in order that j =
d X / d t , the concentration C must be measured as the volume
fraction of refractory. To have the same base for all solutes,
the volume of refractory plus the volume of bulk slag is the
most convenient base for the concentration. Ci is the satura-
tion volume concentration in these terms, C, = 0 is the
refractory concentration in the bulk, and A? = (C, - C,)/
(1 - VZC,).
layer cannot be predicted from the phase diagram alone.
Thus, the effective diffusion coefficients cannot be
interpreted in terms of tracer diffusion values without a
X complete knowledge of the relative mobilities of the species
*
present and their variation with concentration. The con-
centration change through the boundary layer is large, in-
creasing for SiOz dissolution, for example, froin about 40%
SiOz to near 75% Si02. Over this range of composition
there are substantial changes in the structure of liquid as
indicated by the drastic change in viscosity, and there may be
large variations in chemical activity and mobility as a func-
tion of position within the boundary layer. As the tempera-
* Detailed calculations in the case of silica dissolution reveal
that the one-thousandfold increase in viscosity causes only a fac-
tor of 2 decrease in mass transport rate.
May 1964 Dissolution in Ceramic Systems, 11 253
Table I. Calculated Values of the Effective Diffusion Coefficient for Forced Convection Dissolution of Sapphire, Polycrystalline
Alumina, Mullite, and Silica
Refractory saturated slag Bulk slag
- , 7

Dynamic Dynamic
Temp. jx 106 Density* viscosityt Density* viscosityt
Nature of sample (“C) (cm/sec) (g/cm3) (poise) (g/cm3) (poise) A? 5 X 108

Silica 1500 8.9 2.17 1000 2.83 8 1.o 6.5

1425 4.5 2.34 700 2.86 15 0.70 4.5
1390 3.3 2.40 1000 2.88 21 0.64 3.2
1350 2.1 2.43 800 2.90 35 0.58 1.8
Sapphire 1500 3.5 2.79 12 2.83 8 0.625 3.0
1460 0.9 2.79 23 2.84 13 0.58 0.5
1425 0.3 2.80 31 2.86 15 0.55 0.12
Polycrystalline alumina 1500 3.6 2.79 12 2.83 8 0.64 2.9
1460 0.9 2.79 23 2.84 13 0.59 0.49
1425 0.3 2.80 31 2.86 15 0.56 0.12
Mullite 1500 16.5 2.68 26 2.83 8 1.65 7.8
1425 4.3 2.76 57 2.86 15 1.56 1.3
1390 2.9 2.81 61 2.88 21 1.52 0.8
* Mostly cxtrapolatcd from data given by L. R. Barrett and A. G. Thomas, “Surface Tension and Density Measurements on Molten
Glasses in the Ca0-AI203-SiO2 System,” J . Soc. Glass Technol., 43 [211] 179-90T (1959); Ceram. Abstr., 1960, February, p. 32e.
J. S. Machin and Tin Boo Y c e , “Viscosity Studies of System Ca0-Mg0-A1203-Si02: 11, CaO-A1203-SiO~,” J . Am. Ceram Soc., 31
[7] 200-204 (1948).

ture is increased, the position of the liquidus and consequently

the saturation concentration in the boundary layer also
changes. 20
h

(2) Crystalline Formation at the Interface

In the case of S O z and anorthite, direct exchange of the
dissolving species with the slag leads to a point on the liqui-
dus curve in the equilibrium diagram. For mullite and
alumina, however, direct dissolution does not correspond to
an equilibrium interface with the primary crystalline phase.
The evidence that, at higher temperatures, new phases are
c
not formed a t the interface during the corrosion process in- .0
_
dicates either that the change in composition of the boundary
0.8
layer is such as to give an interface on the liquidus or that
nonequilibrium conditions are attained in which an extension
of the liquidus functions as a metastable interface. To as- 0.4
certain the details of the process and to properly evaluate the
meaning o€ calculated diffusion coefficients, it is essential 0.2
to have a knowledge of the actual concentration distribution
of individual species within the boundary film. 0.1
In the case of mullite and alumina there is not an equi- 5.4 5.6 5.8 6.0 6.2 6.4
librium interlace corresponding to direct solution of the re- 104 I T (OK)
fractory phase and its transference across the boundary
layer. At the higher temperatures experimental observa- Fig. 6. Temperature dependence of forced convection corrosion
i 0 ~ of alumina, mullite, and fused
in the 4 0 C a 0 - 2 0 A l ~ 0 ~ - 4 0 S slag
tions indicated that no new solid phase was lormed during silica.
the corrosion process With mullite, anorthite precipitated
on cooling; this suggests that the concentration profile
in the boundary corresponds to a liquid composition within
the primary field of anorthite. i.e , that the liquidus lies on an
extrapolation of the mullite liquidus surface. At the lowest
temperature studied with mullite, crystallization of alumina tion energies observed is large and there is considerable dif-
was observed, corresponding to a concentration below the ficulty in interpreting them in terms of the physical charac-
liquidus temperature of alumina a t the interface. teristics of the process.
The difficulty in interpretation of refractory corrosion in
(3) Corrosion as an Activated Process terms of any activation energy results from several factors.
The temperature dependence of the corrosion process has First, the liquidus composition and consequently the driving
often been represented in the past by an equation of the force for the diffusion process changes with temperature.
Arrhenius form When the bulk composition approaches the liquidus, this
change can be large. In addition, activity coefficients and
j = A exp(E,/RT) (2)
where E , corresponds to an activation energy for the cor-
rosion process. Plots of the logarithm of the corrosion rate as ( a ) H. H. Blau and C. D. Smith, “Refractory Problems in
a function of the inverse temperature for the present results Glass Manufacture,” Am. Ceram. SOC.Bull., 29 [I].6-9 (1950).
follow this behavior as illustrated in Fig. 6. ( b ) F. H. Norton, Refractories, Chapter XV. McGraw-Hill
Previous investigators2 have frequently interpreted their Book Co., Inc., New York, 1942. 798 p p . ; Ceram. Abstr., 22 [6]
data as an activated process and attempted to correlate 102 (1943).
(c) N. McCallum and L. R. Barrett, “Some Aspects of the
experimental activation energies with other characteristics Corrosion of Refractories,” Trans. Brit. Ceram. Sac., 51 [I11 523-
of the materials being studied; however, the range of activa- 48 (1952); Ceram. Abstr., 1954, March, p. 51d.
254 Journal of The American Ceramic Society-Samaddar
diffusion coefficients in silicate systems depend greatly on the
composition and liquid structure. Since the liquidus com-
pmition changes with temperature, there will be correspond-
ing changes in the effective diffusion coefficient with tempera-
ture that have no relation to any activated process. Finally,
the relative mobilities of the ions participating in the dif-
fusion process determine the actual liquidus composition.
Consequently, the point on the liquidus curve corresponding
to the saturation condition is a function of temperature.
The combination of these variables cannot be interpreted
without having a detailed knowledge of the relative mobilities
and actual interface concentration. All these factors com-
bine to niake interpretation of corrosion behavior on the
basis of activation energy ineffective at best and misleading
in general.
Although an Arrhenius plot is generally suitable for em-
pirically reporting and interpolating data, some systems
show far from exponential temperature dependence as
found €or mullite in Fig. 6. Extrapolation of corrosion data
should be done only by including the known or estimated
changes in all the temperature defiendent factors of equation ( I ) ,
1
ie , I;2/a, ~ - ' / 6 , AC.
(4) General Observations
With regard to the experimental results found here, some
qualitative comments are of interest. One of the main
observations is that the temperature dependence of the cor-
rosion rate is much smaller for €used silica than for alumina
or mullite. This probably results because both the dif-
fusion coefficient and the viscosity change rapidly with silica
et al. Vol. 47, No. 5
content a t high silica concentrations. As the temperature
is raised, the silica concentration at the interface is increased
with a resultant change in the effective diffusion coefficient
and viscosity opposite to that induced by temperature
changes. Consequently, although silica is a rather poor
refractory at low temperatures, it becomes an increasingly
satisfactory corrosion-resistant refractory as the tempera-
ture level is increased, and for very high-temperature appli-
cations may be the most suitable one for corrosion resistance
in this system.
V. Conclusions
Experimental studies of the dissolution rate of alumina,
fused silica, mullite, and anorthite in a lime-alumina-silica
slag have shown that the temperature dependence and cor-
rosion rate are determined by the composition and charac-
teristics of the liquid boundary layer. The concept of an
activation energy for the corrosion process may be useful
in some cases for reporting data, but cannot be interpreted
in terms of the processes occurring without much more de-
tailed information with regard to diffusion characteristics
and the boundary layer concentration than is now available.
The corrosion rates of single-crystal and polycrystalline
aluminum oxide were similar, indicating that the presence
of grain boundaries and crystal orientation does not have a
significant effect on the corrosion rate of dense single-phase
refractories in a lime-alumina-silica slag.
At low temperatures alumina is by far the most corrosion-
resistant refractory investigated, but as the temperature level
is raised the differences between the materials decrease :
they differ by a factor of only 5 a t 1500°C.

CORRECTION
Estimate of Free Energy of Formation of Kyanite
by R. W. Haskell and R. C . DeVries
47 [4] 202-203 (1964).
Publishcd in J . Am. Ceram. SOC.,
Point A in Fig. 1 should be placed at the intersection of the
liquidus lines of the liquid plus corundum and kyanite plus liquid
fields.