VOL. 39, NO.
1
JOURNAL
of the
American Cerumic Society
Reactions Between Aluminum Oxide and Carbon
The AI,O,-Al,C,
by I.. M. FOSTER, G. LONG, and M. S. HUNTER
Aluminum Research laboratories, Aluminum Company of America, New Kensington, Pennsylvania
A phase diagram of the system Al2O3-A14C3is pro-
posed. Two intermediate oxycarbides, A14O4C
and A120C, were established. Eutectic melt-
ing between alumina and A1404C occurred at
1840°C. No other low melting was observed.
The alumina phase was not corundum but was
similar to delta-alumina. Because of the high
reactivity of aluminum carbide and all the inter-
mediate compounds with moisture and oxygen,
use of refractories based on the system Alz03-
A14C3must be limited to applications where these
agents are excluded. The behavior of high-
alumina refractories in the presence of carbon is
explained.
I. Literature Review
GREAT deal of work on the system aluminum oxide-
A carbon is recorded in the technical and patent litera-
ture. Moissan’ prepared aluminum carbide from alu-
minum and carbon in a high-temperature resistance furnace,
and there is other early evidence that A14C3results sponta-
neously from its elements and is the preferred product from the
reaction of alumina and carbon up to approximately 2000°C.
Prescott and Hincke,2 B r ~ n n e rand
, ~ Treadwell and Gyger4
made equilibrium pressure measurements of the system
Al2O3-C. The results were very discordant, undoubtedly be-
cause of different assumptions about the nature of the re-
Presented at the Fall Meeting of the Basic Science Division,
The American Ceramic Society, The Pennsylvania State Uni-
versity, University Park, Pennsylvania, October 17, 1955. Re-
ceived June 9, 1955.
The authors are, respectively,chief, Physical Chemistry Divi-
sion; research engineer, Physical Chemistry Division; and as-
sistant chief, Metallography Division, Aluminum Research
Laboratories, Aluminum Company of America.
1H. Moissan, “Aluminum Carbide,” Compt. rend., 119, 16-20
(1894).
2 C. H. Prescott, Jr., and W. B. Hincke, “High-Temperature
Equilibrium Between Aluminum Oxide and Carbon,” J. A m .
Chem. SOC.,49,27,53-59 (1927).
3 R. Brunyyr, Equilibrium in Systems of SOz,CaO, and A1208
with Carbon, Z . Elektrochem., 38, 55-68 (1932); Ceram. Abstr.,
12 [9] 342 (1933).
4 W. D. Treadwell and A. Gyger, “Reduction of Alumina with
Carbon,” Helv. Chim. Acta. 16, 1214-25 (1933).
1
JANUARY 1, 1956
Phase Diagram
actants and products. Prescott and Hincke assumed that the
reaction was that to produce aluminum carbide :
2AlzOa + 9C = A14C3 + 6 C 0
Brunner claimed to have, in addition, observed the pressure
of the reaction to produce aluminum metal:
A1203 + 3C = 2A1 + 3CO
By analogy to other oxide-carbon systems, the pressure curves
of these two reactions should intersect at a quintuple point
where the oxide, carbide, metal, carbon, and gas coexist.
Baur and Brunner5 were unable to observe the quintuple
point experimentally because of prefusion and disappearance
of some phases at temperatures below the quintuple point.
They observed the melting points of approximately 2-gm.
intimately mixed charges of alumina, aluminum, aluminum
carbide, and carbon. Low melting was observed with the
Al2O1-A1mixtures, with a eutectic at 12% aluminum and
2016OC. At 15% aluminum and 205OoC. there was a smooth-
melting maximum that was attributed to the compound
Alg09. Low melting was observed with AI203-A14C3mixtures,
with a eutectic at 13.5y0carbide and 2007OC. Since yellow
carbide was not in evidence at fairly high carbon contents,
Baur and Brunner assumed that an oxycarbide of unknown
composition had been formed. Their data indicated a ear-
bide of the composition A19C3,and a eutectic a t somewhat
more than 50% A14c3, at 2450°C. Various ternary melts were
prepared, and a ternary diagram was constructed with alumi-
num, oxygen, and carbon at the apexes.
Brunner3 observed condensate in the cold parts of his ap-
paratus that, from gain on ignition and total acid-soluble
aluminum, appeared to be the compound A14C303,an alumi-
num carbonyl.
The greater part of the early work was carried out with
very small charges and suffered from incomplete cheniical
analyses.
The work of Kohlmeyer and Lundquist6 most closely ap-
proached the work reported here. Their charge (approxi-
Emil Baur and Roland Brunner, “Fusion Area iy, the
Aluminum, Alumina, and Muminum Carbide System, Z .
Elektrochem., 40, 154-58 (1934).
6 E. J. Kohlmeyer and S. Lundquist, “Thermal Reduction of
Argillaceous Earth; Importance of Rate of Heating to the Re-
duction Process,” Z . aeorg. Chem., 260,208-30 (1949).
2 Journal of The American Ceramic Society-Foster, Long, and Hunter
mately 1000 gm.) consisted of briquetted alumina, which was
thrown into a carbon crucible that was preheated to 2000' to
2300'C. The crucible served as the source of carbon. The
products were analyzed for acid-insoluble alumina, acid-
soluble aluminum, insoluble carbon, and total carbon, which
were then calculated to Al2O3,Al, and A14Ca. The melts ap-
proached the composition A10 as reaction progressed. This
was offered as evidence for the compound Alsogthat had been
reported by Baur and Brunner. Only after that composition
was reached did the melt begin to absorb carbon, as was evi-
denced by a yellow color attributed to aluminum carbide.
They found that the cold A10 melt evolved gas in acid. This
was assumed to be hydrogen liberated by aluminum in solid
solution.
The unusual bonding required in the compounds &C3,
Also9, and Al4C3Oa,and the extensive solubility of normally
immiscible aluminum and alumina, that was proposed by the
early workers, indicated that further clarification of the sys-
tem was required.
II. Introduction
Preliminary experiments confirmed the gross observations
of the earlier workers. When aluminum oxide and carbon
were heated together a vigorous reaction occurred near the
alumina melting point. Large amounts of carbon monoxide
were evolved, laden with particulate matter. If the alumina
content of the original charge was SOT0 or more, by weight,
the charge melted readily, at perhaps 100'C. below the alu-
mina melting point, to a very fluid pool, as shown by a graph-
ite probing stick. Appreciable volatilization occurred at
higher temperatures. If the charge temperature was not
allowed to exceed approximately ZOOO'C., however, and was
held only briefly at that temperature, carbon monoxide
evolution diminished greatly after the initial vigorous reac-
tion and essentially ceased after a few minutes.
Solidified melts were homogeneous to the naked eye, were
gray to yellow in color, and appeared to be devoid of free
carbon. It was established by X-ray analyses that, con-
trary to the findings of Kohlmeyer and Lundquist, alpha-
alumina and metallic aluminum were notably absent in this
fused portion. The gas that was evolved in acid was methane
from the reaction of aluminum carbide or related products,
and not hydrogen from metallic aluminum, as those authors
had assumed. Moreover, great quantities of methane were
liberated from charges that, by external appearance, con-
tained no free aluminum carbide.
It was found by chemical analyses that oxygen-containing
compounds still comprised a substantial part of the charge
after the reaction had essentially ceased, even though carbon
was still available to the melt. Clearly, either equilibrium
was not achieved in these experiments or carbon was not a
primary participant. The latter was shown to be the case by
observing the behavior of mixtures of alumina and aluminum
carbide. When sufficient aluminum carbide was employed,
the attack on the graphite crucible was negligible. More-
over, when aluminum cqbide was used, there was essen-
tially no carbon monoxide evolution, as long as the temper-
ature did not reach the point where volatilization of the charge
was significant. I t was apparent that the system under
study was the system aluminum oxide-aluminum carbide and
that the graphite crucible served for the greater part as an
inert container. In experiments where the graphite crucible
was the sole source of carbon, it participated in the reaction,
again only up to a certain point, and then ceased to be at-
tacked. When a solidified charge residue was broken up af-
ter an experiment, a thin yellow layer of aluminum carbide
could generally be found at the interface between the residue
and the graphite. Possibly this high-melting material served
as a barrier that minimized attack on the crucible.
The present investigation was concerned with the isolation
and identification of these reaction products, the conditions
Vol. 39, No. 1
under which they are formed, and the implications of these
findings in the ceramic field.
111. Experimental
(1) Reactor Construcfion
Induction heating was employed. The susceptor was a
heavy-walled graphite tube or crucible, insulated from a fused-
silica retaining tube by about 3/4 in. lampblack. The high-
frequency power source was a 50-kw. 9600-cycle motor-gener-
ator set. The work coil consisted of ten turns of 3/s-in.
water-cooled copper tubing, spread over a I-ft. length. This
unit is shown in Fig. 1.
The unit was not vacuum tight. No oxygen could reach
the charge because of the immediate reaction with the hot
lampblack or graphite, and since the unit always operated
under a slight pressure, there should have been no tendency
to draw in nitrogen. A pad of fine alumina was maintained
in the space between the silica tube and Transite box to reduce
air convection and diffusion further.
The charge crucibles were machined from solid graphite.
When high-alumina charges were fused, these crucibles oc-
casionally eroded through. This was alleviated in part by
using a special high-density graphite, or by impregnating or-
dinary graphite with a colloidal graphite dispersion. Finally,
when the charges were made up with aluminum oxide-alumi-
num carbide mixtures, the attack on either type of crucible
was insignificant.
(2) Ternperdure Measurement
Temperature measurements in most of the experiments
were made with a Leeds and Northrup potentiometer-type
optical pyrometer that was calibrated by comparison with a
standard tungsten-strip lamp. The maximum calibration
temperature was 1700'C. Extrapolated corrections were
applied at higher temperatures. A 2OOC. sight-tube window
correction was added to all temperatures measured with this
pyrometer .
The sight tube is shown in Fig. 2. Argon was employed as
a flushing gas at a flow rate of approximately three liters per
minute. It had a free path through the entire length of the
central tube, up through the annular space between the two
tubes, and out through the ports near the top,
A modified Radiamatic radiation pyrometer, made by the
Brown Instrument Company, was employed in some experi-
ments. This instrument focused the radiant energy from the
target (the inside flat bottom of the graphite sight tube) onto
a thermopile. The output was recorded continuously.
Thermal arrests recorded by the radiation pyrometer were
of a more logical shape than those determined with the optical
pyrometer, possibly because of the subjective element in the
manual measurements. However, the absolute temperature
was probably less accurately known with the radiation pyrom-
eter, since the amount of radiation collected was very sen-
sitive to alignment of its optical system.
(3) Charge Composition
The initial charge composition was not always indicative
of the composition of the final melt, because of volatilization
of material at the higher temperatures, and because of rather
extensive reaction with the graphite crucible in some experi-
ments. It was possible to attain a melt of appreciable car-
bon content simply by fusing alumina in a thick-walled
graphite crucible. However, in this case, little occurred
until the melting point of the alumina was reached; at this
point the reaction started atmost explosively.
Two types of charge were satisfactory. The first was an
alumina-pitchcoke briquette, crushed to about minus 10
mesh, that was made by calcining an intimate mixture of
alumina and low-melting pitch at about 1000°C., in the ab-
sence of air. The second was a mechanical mixture of minus
50-mesh aluminum carbide and calcined alumina. The
January 1956 Reactions Between Aluminum Oxide and Carbon 3
aluminum carbide was made in a separate experiment by re-
acting a ball-milled mixture of minus ZOO-mesh atomized
aluminum powder and lampblack at approximately 18OOOC.
in a reactor similar to that shown in Fig. 1.

(4) Procedure
The procedure for a typical experiment follows. Varia-
tions of this are noted in the text.
.OK€\ The charge was a 9-in.-deep bed (1000 gm.) of minus 20-
SILICA TUBE
mesh 83.3% A1203-16.7~0C crushed briquette. This was
heated slowly to about 1900°C., where fusion began. Charge
TRANS I TE additions were then made slowly, a few hundred grams at a
SUSCEPTOR time, until about 1600 gm. more had been added. This added
material was generally of somewhat higher carbon content,
e.g. 6870 A1203-32% C. The charge was made up in this
manner so that a low-melting, fluid pool could be made, to
which the high-carbon material could be added without dan-
ger of freezing the melt.
The over-all composition in the example above was about
74y0 &03-2670 C. This was varied over fairly wide limits
in different experiments.
Melts containing more than about 85% by weight of
REFRACTORY alumina were very “carbon-hungry” and difficult to main-
tain in a carbon crucible without serious composition changes.
It was possible to make these melts, however, by dissolving
portions of alumina in the fluid mixes immediately before
Fig. 1. Induction-heated reactor. reducing the power.
It also was difficult to make melts high in carbon, since
these were very high melting, and it was impossible to attain
complete fusion without excessive volatilization from second-
ary reactions.
The consistency of the charge was determined from time to
time by probing with a graphite rod. After the charge was
completely liquid, the sight tube was immersed to a point
WINOOW approximately 1 in. from the bottom of the crucible. The
power was decreased until the reactor began to cool at a
B R A S S HEAD slow, uniform rate, and temperature measurements were
begun.
For cooling curves to be meaningful, it was necessary that
ARGON the over-all composition of the melt remain fixed throughout
ca
IN the cooling range. (Cessation of the reaction had already
been indicated by the absence of CO evolution throughout the
cooling period.) That this situation indeed obtained is shown
by the following analysis of a typical melt:
Sample C 0 Al
NO. (%) (%) (%)
1 7.7 32.0 60.3
G R A P H I T E CEMENT 2 7.2 33.3 59.5
3 8.5 31.0 60.5

ARGON O U T Sample No. 1 was quenched from 20OO0C. by inserting a

graphite rod into the melt and rapidly removing it. Sample
No. 2 was quenched in a similar fashion from 1960OC. Sam-
ple No. 3 was part of the residual melt, cooled slowly to room
temperature.

(5) Analysis of Products

The nature and extent of the reaction products were de-
termined by X-ray powder analyses, metallographic exami-
nation, and chemical analyses.
( A ) X-Ray IdentiJicution: Debye-Scherrer diffraction
patterns were made in a 14-cm. General Electric powder
MACH I NED GR‘A P HI T E diffraction camera, using filtered CuK radiation.
PAR T S ( B ) Metallographic Examination: The sample was em-
bedded in Lucite or Bakelite in the usual manner. The
customary sequence of file, metallographic papers, and Alun-
dums generally gave good definition to the constituents when
illuminated with reflected light. Oblique illumination was
employed when it was desired to show the color of the phases.
Relief of the various phases resulted naturally from their
different hardnesses. This could be enhanced by a final
Fig. 2. Sight tube. polish with magnesia.- The carbon content of phases could
4 Journal of The American Ceramic Society-Foster,
Fig. 3. Microst:uctures of two typical solidiAed melts. ( A )
29.1 % Al& 55.2% acid-soluble Al, 7.A% insoluble residue, and
63.1% soluble A I Z O ~ ;sample accounted for, 99.6%. ( B ) A2.0%
AI&, O.AY0 metallic Al, 60.6y0 acid-soluble Al, 2.570 insoluble
residue, and 5A.3% soluble A1203; sample accounted for, 99.2%.
(250X.) (Soluble A1203 calculated from excess acid-soluble aluminum
left after subtracting the metallic aluminum and the aluminum
part o f the aluminum carbide.)
be estimated metallographically from their relief on polishing
since the hardness of a phase decreased with increasing carbon
content. Moreover, the brightness or reflectivity of a phase
was a good indication of its carbon content. (The mean re-
fractive index of aluminum carbide has been reported as
2.70; alpha-alumina is 1.X.)
( C ) CheKicaZ Analyses: Samples were analyzed for ( a )
aluminum carbide, as determined from methane evolution in
dilute HCI, ( b ) metallic aluminum, as determined from hy-
drogen evolution, (c) total acid-soluble aluminum, and ( d )
acid-insoluble residue. The metallic aluminum figure was
always low. Actually, aluminum was probably absent in the
sample, and the reported figure represents the precision of the
analysis, since hydrogen in excess of that required for methane
was calculated to free aluminum.
In a typical analysis, 1: 1hydrochloric acid was added t o the
weighed finely ground sample in a closed train. The acid
Long, and Hunter Vol. 39, No. 1
was brought to boiling, and the evolved gas was swept through
the train with oxygen. Methane and hydrogen were burned
over copper oxide at 90O0C., and the water and carbon dioxide
were collected for weighing on Anhydrone (anhydrous mag-
nesium perchlorate) and Ascarite, in train.
After the sample had reacted until there was no further
weight gain in the absorber bulbs, any insoluble residue was
filtered off, dried, and weighed. Total acid-soluble aluminum
was determined in the filtrate by precipitation with ammonia.
The determination of total carbon by combustion, as at-
tempted by several previous workers, was not satisfactory.
Combustion was incomplete after several hours at 1400'C. in
moist oxygen.
IV. Results
(I) Identification of Phoset
( A ) A l u m i n u m Oxycarbide, A1404C: The most striking
feature of the early work was the almost complete absence of
free aluminum carbide, as ascertained by X-ray and micro-
scopic examination, in many solidified melts that contained
up to 35% by weight of aluminum carbide, calculated from
methane evolution. Nevertheless, if this methane carbon
and the appropriate amount of aluminum were expressed as
A14C3,and the remainder of the acid-soluble aluminum were
calculated to A1203,essentially 100% of the acid-soluble part
of the sample could be accounted for.
Figure 3 shows microstructures and analyses of two typical
melts. The yellow constituents (white in the photographs)
were subsequently identified as A14C3. Clearly those con-
stituents did not constitute 29.1% in the first case, and 42.0%
in the second case, of the entire sample, as indicated by the
analyses, so additional carbide must have been present as an
acid-soluble oxycarbide.
When the melts were essentially white, and showed little or
no yellow carbide under the microscope, the mole ratio of the
acid-soluble alumina to the aluminum carbide calculated from
methane evolution was consistently close to 4: 1. Table I gives
the analyses of a number of such residues. The first part of the
table gives the analyses as reported. The last part gives the
calculated analyses, assuming that all the acid-soluble alumina
and an appropriate amount of the aluminum carbide were
present as an oxycarbide, 4A1203.A14C3. These calculated
analyses were consistent with the microstructures, assuming
that the glasslike material that made up the bulk of the melt
was the oxycarbide. This oxycarbide can be written A1,04c.
The two forms will be used interchangeably in the text.
No solidified melt was completely free of both aluminum
carbide and acid-insoluble residue, from which a standard
X-ray pattern could be made. However, similar patterns
were obtained from light, fluffy material that frequently con-
densed in the cold parts of the reactor. About 0.2 gm. of
this material, that had been separated under the microscope,
gave the following chemical analysis:
(%)
73.8 acid-soluble A1203
25.7 A14C3
0.2 metallic aluminum
1.1 acid-insoluble residue
__
100.8 total
with an A1203/A14C3 mole ratio of 4.05 : I, corresponding essen-
tially to 4A1203.A14C3(Al4O4C). . Another very small sample
(0.0058 gm.) of such condensate that was perfectly snow-
white and clean in appearance gave 58.7y0 acid-soluble alu-
minum, corresponding exactly to that compound.
The X-ray powder data for this new compound are given
in Table 11. A single-crystal examination showed that the
structure is orthorhombic, with a unit cell of dimensions a =
9.23 a.u., b = 8.64 a.u., c = 5.77 a.u. and vol. = 460 a.u3,
containing two molecules of A1404C. This gives a calculated
density of 2.65 gm. per cc., in agreement with a rough esti-
January 1956 Reactions Between Aluminum Oxide and Carbon 5
Table 1. Analyses of White Solidified A1203-A14CsMelts
Calculated analysest
______-______h_______-_--__-

Reported analyses
Acid- Insoluble Free Sample
Experiment AlrC, soluble Metallic residue Alr01C AlrCs accounted
No. (%) Alto3 (%)* A] (70) (%I
~~ ~
(%I (70) for (%)
~~

57 22.9 64 6 0.0 12.8 87.5 0.0 100.3

93 21.8 60 4 1.1 17.8 81.7 0.5 101.1
19 22.1 58 3 4.0 16.3 78.9 1.5 100.7
85 23.3 59 I 0.3 17.1 79.8 2.6 99.8
92-B 24.5 64 3 0.1 9.8 87.0 1.8 99.7
27-B 23.6 66 6 1.4 9.1 90.2 0.0 100.7
39 14.6 39 0 0.0 46.4 52.7 0.9 100.0
09-B 23.0 62 5 0.4 14.8 84.6 0.9 100.7
10-B 22.4 fjl 5 0.4 16.4 85.2 0.7 100.7
12-R 25.6 65 7 1.1 10.0 88.9 2.4 102.4
20-B 15.7 44 4 0.1 40.3 60.0 0.1 100.5
* Calculated from the acid-soluble aluminum after the metallic aluminum and the aluminum part of the aluminum carbide had been
subtracted.
t By expressing all the acid-soluble alumina (column 3) and an appropriate amount of aluminum carbide (from column 2) as A1404C.

Table II. X-Ray Data for Compounds A1404C and AltOC Table 111. Analyses of Brown Solidified AltO:i-Al~C3Melts
Compound AI4O4C I nsolu-
Free ble Sample
Sample Metallic Alr01C AlrC3 residue accounted
d
Relative
intensity I d
Relative
intensity
NO. A](%) (9%) (%I (%) for(%)

83 0.5 88.6 7.6 3.3 100.0

3.22 100 1.65 20 67 0.1 86.9 11.0 1.2 99.2
3.87 90 1.61 30 38 0.7 83.2 11.4 3.7 99.0
3.30 70 1.56 20 84-2 0 8 811 147 3 8 100.4
3.11 80 1.49 60 43-2 0 6 81 8 15 9 3 0 101.3
2 57 30 1.45 70 71-2 2 9 778 174 1 8 99.9
1.41 30 07 0 0 777 190 3 4 100. 1
1.39 20 30-2 0 0 813 192 0 4 100.9
I .36 30 54-1 1 0 0 800 194 2 4 101.8
1.34 30 27-2 0 6 76 4 21 4 10 99.4
2.02 30 1.31 40 55 0 4 737 225 2 5 99.1
1.96 5n 1.28 20 91 0 6 744 247 0 6 100.3
1.88 20 1.26 20 13 0 9 601 394 1 8 102.2
1.84 20 1.24 40
1.78 10 1.18 10
1.73 10 1.11 20
1 68 39
Compound AIZOC
Relative Relative The manner of formation of this new phase tixed its com-
d intensity d intensity
position somewhere between that of the oxycarbide 4AI203.-
2 75 100 1 45 10 A14C3 and pure aluminum carbide. The most likely com-
2.54 1.38 20 pound, based on the simplicity of its formula, was the oxy-
2.42 I 3.5 30
carbide AI2O3.Al4C3,or 3i\12oC. Attempts to prepare this
~ ~~

1.87 40 1.33
1.58 80 1.21
mate from a flotation experiment. The crystals were needle-
shaped. They were unusual in that the needle axis did not
correspond to a simple crystallographic axis. I n this re-
spect they resembled the oxycarbide 91,0C, described below.*
With the A1404C phase established, it was possible to
characterize the grpup of solidified melts that contained
free aluminum carbide, as evidenced by their brown color.
By expressing the appropriate amount of aluminum carbide
as A1404C,the free aluminum carbide could be obtained by
difference. Typical analyses are given in Table 111.
( B ) Aluminum Orycarbide, A120C: Further micrographs
revealed another crystalline phase that appeared in increasing
amounts with increasing carbon content. Figure 4 shows this
phase in four successive stages of formation. I t is seen that
this phase resulted from the reaction of the aluminum carbide
constituents with the matrix.
*The authors are indebted to G. A . Jeffrey of the University
of Pittsburgh for obtaining the crystal data for AI4O4C.
20
10 in a pure state were only partially successful. Equiniolar
mixtures of alumina and aluminum carbide showed little
change when heated to just below the fusion point, and ex-
tensive volatilization with accompanying composition change
occurred when the mixture was fused. However, when a
1:3 molar mixture of 4A1203.M4C3 and A14C3 was held at
18OOOC. for 10 hours, reaction was complete to form a new
compound, as shown by X ray. I t is isomorphous with alu-
minum nitride, but is readily distinguished from that by an
appreciable lattice-parameter shift and by the very strong
reflections at 2.75 a.u. and a t 1.58 a.u. This material was
contaminated with nitrogen from a leak in the furnace. The
aluminum nitride thus formed apparently served to seed the
formation of the new compound, since, in other experiments
where nitrogen was excluded, only traces of this material were
found in the residual charge. In such experiments, however,
white, fluffy, nitrogen-free condensate that gave the pure pow-
der pattern could be recovered from cold regions of the fur-
nace. X-ray data for the compound AI,OC are given in Table
11. The crystal is hexagonal, with a = 3.19 a.u. and c =
5.09 a.u. The space group is PG3mc. With one ,41,0C mole-
cule per unit cell, the calculated density is 3.00 gm. per cc.
It was not possible to recover enough of this material free of
occluded impurities to obtain a reliable chemical analysis.
X-ray patterns of solidified melts, whose micrographs showed
much of this new phase, showed lines corresponding to those
6 Journal of The American Ceramic Society-Foster, Long, and Hunter Vol. 39, No. 1

Fig. 4. Microstructures of solidified A1203-AI& melts, showing successive stages in the formation of AleOC from the peritectic reaction
of Al& with the melt (250 X).
January 1956 Reactions Between Aluminum Oxide and Carbon 7

Fig. 6 . Microstructures of solidified AI2O3-AI4C3melts showing increasing acid-insoluble residue. ( A ) 32.0% A G , 61.6% acid-
soluble AlnOa, 5.7% insoluble residue; sample accounted for, 99.3%. (B) 29.1% A14C3, 63.1% acid-soluble AlzOs, and 7.4% in-
soluble residue; sample accounted for, 99.6%. (C) 23.3% A&, 62.5% acid-soluble AIZO:~,and 14.1 % insoluble residue; sample
accounted for, 99.9%. ( D ) 0.9% metallic Al, 16.9% AI4C3,53.0% acid-soluble AIz03, and 29.0% insoluble residue; sample accounted
for, 99.8%. (250 X.)

of Table 11, and good correlation was obtained between the
microstructures, X-ray patterns, and chemical analyses of a
large group of melts if the composition izlzOC was assigned to
the new phase.
Explicit chemical analyses of samples, such as those whose
micrographs are shown in ( A ) , ( B ) , and (C) of Fig. 4, were
not possible, since all three phases, Al4O4C,AlzOC, and A14C3,
were present. However, the calculation was possible for
samples, such as those shown in Fig. 5 , where free carbide was
absent.
(C) Alumina Phase: Most of the solidified melts con-
tained some material that was insoluble in boiling dilute
mineral acid. Figure 6 shows photomicrographs and analyses
of four melts, arranged in order of increasing amounts of this
phase. The aluminum carbide and soluble alumina are giver
separately. It should be understood that these were com
bined for the greater part as oxycarbides in the samples
Since both of the oxycarbides and free aluminum carbide wert
present concurrently, their proportions could not be statec
explicitly.
The acid-insoluble residues gave powder patterns veq
similar to that of the delta-alumina of Stumpf et al.? that war
prepared by controlled calcination of alpha-alumina monohy
drate (boehmite). It is known that the delta-like alumin:
structure can be stabilized with nitrogen, since that phast
alone results from the fusion of alpha-alumina with as littlc
as 0.34% nitrogen (as added A1N) in a molybdenum boat
Since the analysis for traces of nitrogen in such a refractoq
material was not reliable, i t remained to be demonstrated t h a

-
Fig. 5. Microstructures of A1z03-AlrC3 melts solidified from the
Al404C-AIzOC field. ( A ) 80.0% A1404C, 9.970 AIzOC, and
10.6% insoluble residue; sample accounted for, 100.5%. (8)
7 H. C. Stumpf, A. S. Russell, J. W. Newsorne, and C. M
Tucker, “Thermal Transformations of Aluminas and Alumin;
Hydrates,” Ind. Eng. Chem., 42 [7] 1398-1403 (1950); Ceram
Abstr., 1951, July, p. 131d.
8 Journal of T h e American Ceramic Society-Foster,
I boundaries could not be plotted.
2100 I
I No solid solubility is shown, as no estimate of its extent could
I
I
I be made.
I
I ( A ) Equilibrium Structures: Referring to Fig. 7, melts
/ A14C3+
2000
-u'
Q more than 50 mole yo aluminum carbide should consist of
a A120C and A14C3. The fact that the alumina phase was not
5
I-
190C
180C
1 1
'I*OS 20 50
A14C3(Mole %)
Fig. 7. The system Als3;rA14Ca.
delta alumina could be prepared under conditions where
nitrogen contamination was not possible. This was ac-
compliqhed by maintaining a direct-current arc between two
high-purity aluminum electrodes, in a completely tight
vessel, containing an atmosphere of SO% argon and 20%
oxygen. The delta-alumina was deposited as a fine white
powder on the cold walls of the vessel.
(2) Construction of the Diagram
Since the two oxycarbides can be considered as intermediate
compounds of aluminum oxide and aluminum carbide, and
since the delta-alumina phase was shown to be a possible
structure of pure aluminum oxide, all the data can be pre-
sented on a binary A1203-iZ14C3 diagram.
Aluminum carbide is very high melting and decomposes
into its elements before the melting point. It has been re-
ported to melt a t 2700OC. under the saturation pressure of its
elements, but this must be taken with great reservation.
The alumina melting point was taken as 2040°C., from the
"Selected Values of Chemical Thermodynamic Properties. "*
There are compound lines at 20 and 50 mole yo aluminum
carbide, corresponding to 4A1203.A14C3 and AlzO3. A14C3,and
the latter is formed in a peritectic reaction between the
A14C3constituents and the liquid melt.
The only eutectic structure that was apparent in the photo-
micrographs was that between 4A1203.Al& and the alumina
phase. This located the eutectic on the alumina side of the
4A1203.A14C3 compound line. No low melting was observed
in any other composition range.
The proposed diagram is shown in Fig. 7 .
The phase rule requires that no more than two phases exist
concurrently at equilibrium. Only rarely did this situation
obtain. Usually, eutectic, primary 4Al2O3.A14C3, Al203. -
A14C3,and A14C3were all present to some extent. In general,
the presence of nonequilibrium phases confused the interpre-
8 Selected Values of Chemical Thermodynamic Properties,
Circular No. 500, National Bureau of Standards (1952).
Long, and Hunter Vol. 39, No. 1
tation of the gross chemical analyses so that exact liquidus
LlQ. solidified under equilibrium conditions should consist of one of
three binary compositions. Those with less than 20 mole yo
aluminum carbide should consist of alumina and A1404C;
those in the composition range 20 to 50 mole % ' aluminum
carbide should consist of A1404C and A120C, and those with
alpha-alumina means that the extrapolation of the liquidus
line at the left-hand ordinate to 2040OC. (the melting point
of alpha-alumina) is not valid. The liquidus line is dotted
in to indicate this uncertainty.
It was impossible to maintain enough alumina in the melt
t o get to the left of the eutectic minimum and to obtain alu-
mina as a primary phase. Consequently, the composition
of the eutectic is not known with certainty. Its composition
is less than 11.8 mole yo aluminum carbide, the lowest car-
bide melt for which reliable analyses were obtained. This
melt, which appeared to have been solidified under equilib-
rium conditions, is shown in Fig. 8 ( A ) . Similar structures
I I with decreasing amounts of eutectic were found in other melts
AI4'3 up t o approximately 20 mo!e yoaluminum carbide.
The nature of the diagram between 20 and 50 mole yo
carbide is fairly well established, although the highest carbide
charge that was employed in this region was about 27 mole yo.
Figure 8 ( B ) shows a sample from a melt with 22.1 mole yo
carbide. Since both aluminum carbide and AlzOC were pres-
ent to a small extent in this sample, an explicit analysis
cannot be given. However, it is clear from the photomicro-
graph that the solidified melt consisted largely of A1404C.
Its gross analysis places it slightly to the right of the &04C
compound line, so a small amount of A1,OC should be present
as shown, but the carbide and alumina should be absent in a
truly equilibrium sample.
Figure 8(C) shows a sample from a melt whose gross analy-
sis was 24.1 mole yo aluminum carbide. With the exception
of the small amount of alumina constituent, the structure is
consistent with this composition range.
( B ) NonequilibrimB Structures: The sample of Fig. 9(A)
came from a melt whose gross composition was 25.7 mole 7.
carbide. I t is principally A1404C, but contains a few large
aluminum carbide crystals that are undergoing reaction with
the matrix to form A1,OC. There are no small carbide crys-
tals, and the large crystals do not appear to be a precipitated
phase because of their size. More likely this carbide was
never in solution a t the high temperature. On cooling to a
region such as that marked ( A ) in the diagram of Fig. 7, pri-
mary AI,OC came out, but, at the same time, these large car-
bide crystals continued to react with the liquid to form more
A1,OC. Some were not completely reacted because of their
large size.
Figure 9 ( B ) , which shows a sample from a melt that was
quenched from the high temperature, should be compared
with Fig. 9(A), which shows a sample from a melt that was
slowly cooled. Its gross composition was 26.6 mole yo car-
bide. There are large carbide crystals that were never com-
pletely in solution. Also present, however, are some very
small carbide crystals, and particularly, a large amount of
fine A1,OC of such shape as to clearly have come from the
peritectic reaction of the small A14C3crystals. Considerable
fine eutectic is present, because the composition of the melt,
discounting the large carbide crystals, was probably on the
eutectic side of the Al,04C compound line. This could not
be established analytically because it was impossible to sepa-
rate the large carbide crystals from the remainder of the
sample. This structure is typical of a large number of melts
quenched from the high temperature.
January 1956 Reactions Between Aluminum Oxide and Carbon 9

Fig. 8. Equilibrium structures (250 X). Fig. 9. Nonequilibrium structures (250 X).
10 Journal of The American Ceramic Society- Foster, Long, and Hunter Vol. 39, No. 1
V

200(

-
0
Q
a 1901
+E

1801

Arbitrary Time

Fig. 10. Cooling curves (see text for descriptions).

Figure 9(C) shows A1404C and considerable eutectic, as

would be expected from its gross composition of 16.1 mole yo
carbide, and also A120C in two forms. The large A120C
crystals appear to be the primary compound that was re-
tained from the A1,OC-plus-liquid field. The small dendritic
A1,OC crystals appear to be of a different origin, and probably
resulted from the conversion of thin A14C3plates or needles
that precipitated out during nonequilibrium solidification of
the sample.
Beyond 50 mole yo,the diagram is uncertain. Charges in
this composition range were so high melting that it was im-
possible to obtain homogeneous melts without extensive
volatilization to change the composition. Occasionally,
foreign lines appeared in the X-ray patterns, and traces of
phases appeared in the micrographs that did not seem to
correspond to any of the established compounds. Conse-
quently, in the diagram this region is dotted in.

(3) Cooling Curves

Figure 10 shows typical cooling curves from which the
diagram of Fig. 7 was constructed. The curves are accurate
tracings of the experimental points. Curves I, 11, and 111
are representative of cooling curves of melts similar to that of
Fig. 3 ( A ) , which consisted almost entirely of an A1404C ma-
trix and large A14C3 crystals. If the A14C3 was never in solu-
tion, and hence contributed no thexmal effects to the cooling
process, the curves should represent solidification coming
down very near the Al404C compound line. Al2OC precipi-
tation appeared to start at about 196OoC., and the peritectic
conversion of AI,OC into A1404C started at about 1870' to
1890OC. No eutectic freezing is indicated in the curves, and
no eutectic was found in the micrographs.
Cooling curves showing distinct eutectic arrests were in-
frequent. However, by maintaining the initial charge high
in alumina, and cooling rapidly after complete fusion had
taken place, curve IV was obtained which showed a distinct
break lower than that of the peritectic. The microstructure
showed primarily eutectic.
The eutectic structure also could be developed if the melt
was held for a considerable period immediately below the
peritectic transformation temperature. Curve V is the cool-
ing curve of the melt shown in Fig. ll(A). Curve VI is
from an identical melt held for 1 hour at a temperature 10°C.
below the peritectic reaction temperature. This solidified
melt is shown in Fig. 11(B). I t is seen that, in the latter
case, conversion of the AI20Cinto A1404cby reaction with the
liquid had gone to completion, and only the eutectic struc-
ture with A1404C primary crystals remained. Thus, the
eutectic temperature was established at about 184OOC.
Fig. 1 1. Microstructures of A120~-A14C~melts, showing [ A )
incomplete and ( B ) complete conversion of AltOC to Al404C. The
microstructure of ( 6 ) was obtained by holding (A) for one hour,
1O°C. below the AlzOC-Al4O~Cperitectic temperature. (250 X.)
One more peritectic, that of A14C3 plus liquid going to
ALOC plus liquid, is indicated on the phase diagram. This
cannot be located with certainty. The fact that such a peri-
tectic exists is shown primarily by the photomicrographs.
Many of the cooling curves showed brief thermal arrests and
irregularities near 2000°C., and tentatively this temperature
has been assigned to the AI4C8-Al20Cperitectic. This line
is dotted in the diagram.
V. Discussion
The fact that high-alumina melts could be contained in
graphite without extensive carbon pickup was a surprising
result of this investigation. This was probably due in part to
the presence of a high-melting aluminum carbide layer at the
graphite-melt interface. However, this layer did not prevent
the melt from rapidly picking up sufficient carbon to convert
it to Al4O4c. This behavior could be explained by a high
stability of this oxycarbide toward further reduction by car-
bon because of the lowered alumina activity. This sta-
bility has considerable significance in that this oxycarbide is
January 1956 Physical and X-RayStudy of Disilicides of Ti,Zr,and Hf
certainly the primary product from the slagging of alumina in
the presence of carbonaceous material, and must have been
formed.extensively, although not recognized, during various
metallurgical operations.
Aluminum oxycarbide, A1404C,was undoubtedly the prod-
uct obtained by Kohlmeyer and Lundyuist, and by Baur
and Brunner, who reported the preparation of melts whose
composition approached 410, as deduced by acid-soluble
aluminum content. It might also have been the compound
reported by Brunner as an aluminum carbonyl. (A1404C
contains 59y0 aluminum; A1809, C;O%; -410, 63%; and
-4l4c303, 58%.)
I t was gratifying that the system could be defined without
resorting to compounds of unusual valence. The two oxy-
carbides have logical positions in a series beginning with
aluminum oxide and ending with aluminum carbide, where
successive pairs of oxygens are replaced by carbons, i.e.,
-\I4o6,A1404C, A1402C2, and AI4C3. Moreover, the com-
pleted series is evidence that no compounds remain to be
identified.
The second oxycarbide, Al,OC, is isomorphous with AIN.
This is perhaps not surprising, since the average crystal
radius and electronegativity of oxygen and carbon are just
those of nitrogen. -4lthough the detailed crystal structure of
-4I4O4C has not been established, it has been suggested*
that AlrOC is actually the basic structural unit, and that
A1,O4C is made up of alternating AI203 and AlzOC layers.
This would be analogous to the structural relationship of the
carbonitride, A15C3N,to A14Cs and A1N,9although there must
be considerable differences in detail.
The fact that the alumina phase was not alpha-alumina was
unexpected. However, the conditions for stability of this
Physical and X-Ray Study of the Disilicides of Titanium,
Zirconium, and Hafnium
by PERRY G. COTTER, J. A. KOHN, and R. A. POTTER
Electrotechnical laboratory, Bureau of Mines, Norris, Tennessee
A method of preparing the disilicides of titanium,
zirconium, and hafnium is outlined. Unit-cell
dimensions, density, Knoop microbardness num-
bers, chemical analyses, and solubility in various
reagents are given for the three compounds.
It is concluded that these disilicides have no value
as industrial hard materials.
1. Introduction
s A PORTION of a program designed to secure fundamental
A information on the so-called hard materials, the di-
silicides of titanium, zirconium, and hafnium were in-
vestigated. These compounds were prepared by alumino-
thermic reduction of the metallic oxides and silica. Titanium
Received June 15, 1955; revised copy received August 1, 1955.
A t the time this work was done the authors were, respectiveIy,
powder metallurgist, mineralogist, and ceramic engineer. J. A .
Kohn is now with the Chemical-Physics Branch, Signal Corps
Engineering Laboratory, Fort Monmouth, New Jersey.
‘ 0 . Honigschmid, Comfit. rend., 143, 224 (1906); Monatsh.,
27, 1069 (1906).
11
delta-like phase are not known. In addition to the prepara-
tion of pure delta-alumina in an arc, that was described, it
resulted from burning aluminum powder in air; it was the
alumina phase that was recovered from a spent photoflash
bulb, and it was made by exploding aluminum carbide in a
bomb in 450 Ib. per sy. in. of oxygen. Whether, in the pres-
ent investigation, it precipitated from the melt under equilib-
rium conditions, or was quenched or stabilized by carbon or
traces of impurities, is not known. That the latter might be
the case is suggested by the work of Schneider and Gattow,’@
who found that the product from the burning of aluminum in
the presence of a carbonaceous promoter was not alpha-
alumina, but something that is identified as lying between the
eta- and delta-aluminas of Stumpf et a1.’
Because of the high reactivity of aluminum carbide and
all the intermediate compounds with moisture and oxygen,
use of refractories based on the system Al2O3-AI4C3must be
limited to applications where these agents are excluded.
The behavior of high-alumina refractories in the presence of
carbon at high temperature is explained by the formation of
AI404C.
Acknowledgment
The authors express their appreciation to R. L. Parsons and
G. Bruno for the analytical part and to H. C. Stumpf for the X-
ray analyses.
*G. A. Jeffrey, private communication.
M. v. Stackelberg, E Schnorrenberg, R Paulus, and K. F.
Spiess, “Aluminum Carbide, ALCn, and A luminurn Carbonitride,
AlEpN,” 2. physik. Chem., A175, 127-39 (1935).
lo A. Schneiier and G. Gattow, “Heat of Formation of Alu-
minum Oxide, 2.anorg. Chem., 277, 41-48 (1954).
and zirconium disilicides also were prepared by siriteriiig the
metal hpdrides with silicon to compare the resultant phases
with those obtained by the aluminothermic method.
II. Preparation
The aluminothermic process was first applied to the prep-
aration of titanium and zirconium disilicides by Honig-
schmid.’ The process used in the present experiments
was essentially the same as that of Honigschmid’s. il mix-
ture of iron-free silica sand, sulfur, aluminum powder, and the
metallic oxide was placed in a fire-clay crucible, covered with
magnesium powder, and ignited by means of a blowtorch.
The disilicide resulting from this reaction collected a t the
bottom of the crucible in the form of a metallic-appearing
button. The button was freed from a small amount of ad-
hering slag by brief immersion in dilute hydrochloric acid.
It was then finely crushed and subjected to successive treat-
ments with dilute hydrochloric acid and 10% potassium
hydroxide solution until free from sulfides and elemental silicon.
The material was washed and separated by centrifuging be-
tween treatments to avoid contamination with filter-paper
fibers.