C-248 Communications of the American Ceramic Society September 1985

The great variety of results for phosphorus
diffusion is probably due to’the fact that, in
all instances but the present, measurements
of various sorts were made on oxidized sili-
con. It is difficult to correlate these with
one another or with the present work.
The silicon diffusion was measured in a
bulk sample, and at least the value of E
(6X los J) seems reasonable. The value of
E (l.14X lo” J) found in the present in-
vestigation seems quite low for a network-
forming cation, being more in line with
that found for monovalent modifiers‘ or P
in Na20-P2OS-SiO2glasses’ (Table 1).
As dramatic as is the change in profile
shape for the P20s-Si02core fiber, the use
of diffusion for profile modification has
limited applicability. A major problem is
that, as found in the present study, the time-
temperature conditions necessary for dif-
fusion of the phosphorus are similar to
those for nucleation and crystal growth.
This result indicates that care must be taken
to avoid devitrification of a device being
hcat-treated. Additionally, for many appli-
cations it might be desirable to perform the
heat treatment on the preform from which
the device is to be made. Diffusion is not
sufficiently rapid to make this practical
since the time necessary for diffusion is
related to the distance squared. For ex-
ample, the elimination of the dip shown in
Fig. 1 would take years of heat treatment of
a standard preform on a lathe using a torch
producing a temperature of 2000°C rather
Table I. Preexponential Terms and Activation Energies for Diffusion of Cations
in Silicate Glasses
species DO E (J/molX Ref.
P 5.73~ 2.30 4 (Sah et al.)
6.39X lo-’’ 1.27 4 (Allen et al.)
1.9X10--9 1.1 4 (Li et al.) ’
4.72 4.21 4 (Thurston et al.)
1.86X lo-’ 4.03 4 (Barry and Olofsen)
2.0x lo-’” 1.7 3
8X lo-’ 1.14 Present investigation
Si 3.28X 10’ 6 5
P in Na2O-P2OS-SiO2 5.2X lo-’ 0.96 7
than hours as was observed on the fiber. diffusion under the same conditions is quite
Changes in preform heating time as a result substantial. Phosphorus diffusion could be
of modifications in the preform fabrication used to systematically change the profile
technique will not cause any significant shape in certain optical devices.
change in profile shape through diffusion;
nor will there be any significant change REFERENCES
during fiber drawing. However, heat treat- ‘H. S . Carslaw and J. C. Jaeger, Conduction of
Heat in Solids, 2d ed. Clarendon, Oxford, U.K.. 1962.
ment of devices where diffusion distance is ,.I.Hersener and G. P. Huher, “Semiquantitative
short may have some application. X-Ray Microanalysis on Preforms for Optical Fibers,”
Electron. Lett., 20 [ 5 ] 208-209 (1984).
3 D . R . Campbell, E. I. Alessandrini, K . N. Tu,
SUMMARY and J. E. Lewis, “Phosphorus Diffusion in Partially
A method of determining the diffusion Crystallized Films of SiO,,” J. Electrochem. Soc., 121
[8] 1081-85 (1974).
of elements used in the fabrication of light 4M. Ghezzo and D. M. Brown, “Diffusivity Sum-
guides has been developed. It employs mary of B, Ga, P, As, and Sh in SiO,,” J. Electrochem.
Soc., 120, 146 (1973).
X-ray analysis to discern changes in chem- ’G. Brehec e f al., “Diffusion of Silicon in Amor-
ical profile shape. These data are fit to a phous Silica,” Acta Metall., 28 [ 3 ] 327-33 (1980).
6P. Winchell, “The Compensation of Law for Dif-
solution to Fick’s second law to yield fusion in Silicates,” High Temp. Sci., 1 , 200-15
diffusion data. The data indicate that Ge (19691).
diffusion is very small in these character- P. 0. Perron and H. B. Bell, “Diffusion of Phos-
phorus and Sodium in Na,O-P,O,-SiO,, Liquids,”
istic light guide compositions and that P Trans. Br. Ceram. Soc., 66, 347 (1967). 0

J. Am. Ceram. Soc., 68 [9] C-248-C-251 (1985) EXPERIMENTAL

PROCEDURES
Unidirectionally aligned SiC-fiber-
High-Temperature Mechanical Behavior of reinforced lithium aluminosilicate matrix
composites* were used in this study. The
Fiber-Reinforced Glass-Ceramic-Matrix nominal composition of the fiber was (in
wt%) 56Si*3OC*140.The composites were
Composites tested in four-point flexure and in tension.
The nominal dimensions of the flexural test
T. MAH*AND M. G . MENDIRATTA specimens were 5 by 2.4 by 50.8 mm. The
span of the lower support pins was 3.8 1 cm
Univcrsal Energy Systems, Inc., Dayton, Ohio 45432
and that of the loading pins was 0.317 cm
A. P. KATZ,* R. RuH,* AND K. s. MAZDIYASNI~ and, thus, the support span-to-depth ratio
was 16. The flexure tests were performed
Air Force Wright Acronautical Laboratories, Materials Laboratory, Wright-Patterson Air Force
Base, Ohio 45433-6533
using a universal testing machinet with a
22-kN (5000 lb) capacity load cell and
Lithium uluminosilicate glass-ceramic-matrix composites reinforced with S i c 1.27X lo-’ cm/min (0.005 in./min) cross-
head speed. This machine was equipped
fibers were tested at 900” and 1000°C in.fZexural and tensile configurations. The with a vacuum furnace utilizing either tung-
composites showed severe thermomechanical degradation when tested in oxidizing sten mesh or platinum sheet heating ele-
atmospheres. This degradation was found to depend on oxygen partial pressure. ments for high-temperature testing under
For P0,Z103 Pa the composites failed in a brittle fashion; a single crack initiated vacuum or in inert or oxidizing atmo-
and propagated through the matrix and fibers. At lower Po, multiple matrix spheres. The testing environments in-
cracking was observed, and the fibers bridged the matrix cracks.
EDITOR-J. A. PASK
CONTRIBUTING
IBER-REINFORCED ceramic matrix com- limited data have been published on the
Fposites are becoming increasingly at- high-temperature mechanical properties of Received March 20, 1985; revised copy received
May 16, 1985; approved May 23, 1985.
tractive for high-temperature structural these composites. Supported under USAF Contract No. F33615-
applications. A number of publication^'-^ This communication presents results 84-C;5071.
Member, the American Ceramic Society.
have reported on the processing, micro- on the high-temperature mechanical be- *Composites (SiC/LAS-11) received from United
structure, and room-temperature mechani- havior of glass-ceramic-matrix composites Technologies Research Center, E. Hartford, CT,
cal properties of Sic-fiber-reinforced and their environmental sensitivity at those through the U S . Air Force; Sic fibers were Nicalon,
Nippon Carbon Co., Tokyo, Japan.
ceramic-matrix composites. However, only temperatures. hstron Corp., Canton, MA.
September 1985 Communications of the American Ceramic Society C-249
__
(mm)
0.127 0.254
100

500

200
-z
I
Y

.. .. .- 100

CROSS HEAD DISPLACEMENT (IN.)

Fig. 2. Stress vs crosshead displace-

ment curve for SiC/LAS-I1 composite
tested in tension at 900°C in air.

tension is shown in Fig. 2. The stress in-

creased linearly with ram displacement up
to about 255 MPa (37 ksi), after which it
increased nonlinearly, reaching a maxi-
mum value of 340 MPa (49 ksi) before sig-
nificant drop-off. This behavior is quite
FRONT SIDE
different from the room-temperature ten-
VIEW VIEW sile tests on the same For
those tests the composite stress-strain curve
Fig. 1. Drawings of tensile test specimen and high-temperature tensile testing assemblage; A=semi- exhibits three distinct zones which can
circular reflector, B = 1-kW tungsten-halogen light bulb, C=Inconel frictional grip, D=specimen, be explained by linear elastic behavior,
E=alumina tube, and F=zirconia felt.
progressive multiple matrix cracking, and
final fiber failure. A careful examination of
Table I. High-Temperature Tensile Test Data Obtained in Air the room-temperature test result (Fig. 3 of
for SiC/LAS-I1 ComDosite Ref. 7) revealed that there was a nonlinear
Soaking time at test Ultimate tensile strength region just before a sudden drop in the load
Test temp. ("C) temp. before test (min) [MN/m* (ksi)] (referred to as a blip) at 330 MPa. Also the
900 5 333 (49.8) change from linearity to nonlinearity oc-
900 7 321 (46.6) cured at 245 MPa (35.6 ksi). These stress
900 30 339 (49.1) values agree very well with those of the
1025 3 347 (50.3) above-described 900°C test. The failure
strain of the 900°C test specimen also coin-
cides with the strain (0.24%) at the load
vestigated were 7-mF'a vacuum, lo5 Pa of With this arrangement a temperature of blip during the room-temperature test. Re-
argon, air, and various partial pressures of 1000°C could be reached within 10 min. sults for the tests run at 900" and 1000°C are
oxygen [lo4 to lo2 Pa (lo-' to atm)] The test temperature was measured by plac- comparable. In general then, the load-vs-
in argon. Testing temperatures were 900" ing thermocouples adjacent to the specimen deflection behavior in air at these tempera-
and 1000°C. gage section. Temperature was controlled tures is identical with the behavior at room
The tensile test specimens were ma- by the output of a variable transformer temperature, except that the composite
chined from bars of flexural test geometry wired to the lamps. The temperature varia- fails catastrophically at the point of sudden
using a 220-mesh diamond grinding wheel, tion was held within 210°C by simply load drop.
5.08 cm (2 in.) in diameter, to form the adjusting the variable resistor,' eliminat- Unlike the room-temperature test
gage section. The resulting gage section ing the need for an elaborate temperature specimens which exhibited parallel matrix
measured 2.11 by 1.05 by 15 mm. The controller. cracks (Figs. 4 and 5 of Ref. 7), only one
tensile tests were performed in air and at crack was observed in each specimen tested
temperatures of 900" and 1000°C using a RESULTSAND DISCUSSION at elevated temperature. Careful exami-
servohydraulic machine' at a crosshead The elevated-temperature tensile test nation of the crack sites revealed that the
speed of 1.27X lo-' cm/min (0.005 in./ results are shown in Table I. A typical fibers adjacent to the surface were broken,
min). The heating was achieved by using stress vs crosshead displacement curve and no fibers bridging the matrix were ob-
two diffusely focused tungsten halogen for the composite tested at 900°C in air in served. This behavior can be clearly seen in
lamps. Schematic drawings of the tensile Fig. 3, which contains SEM fractographs
test specimen and the high-temperature ten- of a tensile specimen tested at 900°C in air.
sile testing assemblage are shown in Fig. 1. Variac, GenRad Inc., Concord, MA. The fracture surface consisted of an outer
C-250 Communications of the American Ceramic Society
Fig. 3. Scanning electron frdctographs of Sic/
LAS-I1 composite tested in tension at 900°C in
air.
region with uniform thickness ~ 1 5 p0m
exhibiting completely brittle failure and an
inner region showing fiber pullout. All the
specimens listed in Table 1 showed similar
failure behavior. These observations are
quite different from those of the room-
temperature tests, where equally spaced
multiple cracks and unbroken fibers bridg-
ing thc matrix were found.
On the basis of these results the fol-
lowing phenomenological explanation can
be made. For both room- and elevated-
temperature tests, several noncritical
microcracks form in the vicinity of the
lincar-to-nonlinear portion of the load-
displacement curves in a statistical fash-
ion. With increasing load within the non-
linear region of the curves, the microcracks
grow and join together. A single large
crack forms which covers the whole cross-
sectional area of the gage section, and this
phenomenon is manifested as a blip in the
curve. For room-temperature tests, at the
moment of the blip, the stress is redis-
tributed and generates other large cracks in
the matrix away from the initial crack. The
crack spacing depends on the fibedmatrix
interfacial shear strength (progressive mul-
tiple cracking model). In contrast, at high
temperature in an oxidizing atmosphere
(as discussed below), environmental effects
dominate at the fibedmatrix interface and
fiber failure occurs. This precludes the pos-
sibility of load transfer and thus prevents
the generation of multiple parallel cracks.
The fracture strengths of the com-
posites were also determined in four-point
bending at 900" and 1000°C in air. The load
vs crosshead displacement curve for the
composite tested at 900°C showed linear
Table 11. Four-Point Flexural Strength of SiCILAS-I1
Composites Tested at 900°C
Oxygen partial Flexurdl strength based on
pressure [Pa (atm)] maximum load [MN/m2 (ksi)]
2 x 104 ( 2 x 521 (75.51
1 x 104 (1 x io-lj 586 (8s.oj
I x 103(1x 664 (96.3)
1 x 1 0 ~(1x10.~) 1010 1146.51
*In air.
TENSILE SURFACE
SlOE V I E W
Fig. 4. Macroscopic view of four-point flexure
specimens tested at 900°C in varied Po2's. From
top to bottom, for each view, 20%, lo%, I%,
and 0.1% oxygen.
behavior to 488 MN/mz and slight non-
linearity before it failed in a brittle manner
at 521 MN/m2 (75.5 ksi). The 1000°C fail-
ure strength was 659 MN/m2 (95.6 ksi).
This behavior is the same as that reported5,'
for similar composites -maintenance of
room-temperature strength (931 MN/mz)
up to 600°C, a sharp drop in strength
( ~ 3 0 MN/m2)
0 at 900"C, and an in-
creasing strength (827 MN/m2) at 1000°C.
Unlike room-temperature fracture behavior
where specimens failed in a compression/
shear mode,'.' the high-temperature test
specimens failed in tension in a very brittle
fashion. The failure initiated on the tensile
surface of the flexure bar and proceeded in
a brittle manner through about one-eighth
of the specimen thickness before the crack
propagated in a delamination mode.
To determine whether time at tem-
perature prior to loading contributed to
the high-temperature degradation, tensile
specimens were exposed at 900°C for up to
30 min prior to testing. These exposures
had no effect on the 900°C tensile behavior
(Table I). In addition, a flexure specimen
was exposed at 900°C in air without load
for 20 min. The specimen was subse-
quently tested at room temperature, and the
failure behavior was identical with that of
as-received specimens tested at room tem-
September 1985
~
perature. These observations show that the
high-temperature degradation occurs only
upon loading and that it is a rapid process.
To determine whether the degrada-
tion was indeed environment and not
just temperature dependent, flexure speci-
mens were tested at 900°C under vacuum
(7 ILPa), and in argon and partial pressures
of oxygen. The failure behavior and ulti-
mate strengths of composites tested at
900°C under vacuum and in argon were the
same as for those tested at room tem-
perature. This means that gaseous species
in air must be present in the test environ-
ment for the high-temperature embrittle-
ment to occur.
In order to identify the causative gas-
eous species, oxygen and argon gas mix-
tures were used to create different envi-
ronments. Oxygen partial pressures of lo4,
lo3, and 10' Pa (lo-', lo-', and atm)
in argon were employed. The testing tem-
perature was fixed at 900°C. The gas mix-
ture was dried before entering the furnace
chamber to eliminate moisture (gas mix-
tures contained a maximum 5 ppm H20).
The furnace chamber was evacuated to
0.1 mPa before the introduction of the gas
mixture and heating. The gas was kept at
lo5 Pa (1 atm) to maintain constant partial
pressure of oxygen. The results of the
flexural testing are shown in Table 11.
A progressive degradation of ultimate
strength was observed with increasing par-
tial pressure of oxygen. The macroscopic
views of these specimens are shown in
Fig. 4. All specimens failed in a brittle
fashion from the tensile surfaces, except
the one tested in 10' Pa atm) Po2. It
failed in compression/shear, and its be-
havior was identical with that of the nonde-
graded specimens. These results clearly
indicate that the high-temperature embrit-
tlement is associated only with oxygen.
Since the degradation happens during load-
ing, the interaction of oxygen and com-
posite occurs quickly. Oxygen diffusion
through the composite from the environ-
ment must be fast, probably via the first
crack. The short duration of the embrittle-
ment process makes in situ characterization
very difficult. Further research is needed to
understand the degradation mechanism and
to find a remedy.
In summary, a simple and very effec-
tive method of testing ceramic composites
at high temperature in air in tension
was described. The present investigation
indicates that Sic-fiber-reinforced lithium
September 1985
aluminosilicate glass-ceramic-matrix
composites experience severe thermo-
mechanical degradation at 900” and 1000°C
in oxidizing atmospheres. Composite em-
brittlement occurs during loading and at a
very fast rate. Oxygen partial pressure of
103 Pa (lo-’ atm) was sufficient to cause
composite embrittlement at the testing
speed employed. This problem must be
better understood and remedied for these
composites to reach their full application
potential.
J . Am. Ceram. Soc., 68 [9] C-251-C-252 (1985)
Concentration Dependence of the Diffusion
Coefficient of Water in Si02Glass
MINORUTOMOZAWA*
Materials Engineering Department and Center for Glass Science and Technology, Rensselaer
Polytechnic Institute, Troy, New York 12181
The concentration dependence of water diffusion in SiOz glass was analyzed
assuming that both molecular water and hydroxyl water can diffuse and that they
are in equilibrium as interstitial and substitutional impurities. The equation ob-
tained appears to be consistent with experimental observations of water diffusion
in Si02 glass.
HE DIFFUSION of water in S i 0 2 glass
Tplays an important role in the chemical
and mechanical properties of the glass as
well as in the oxidation of silicon. Water
can exist in glass both as hydroxyl (-SOH),
and molecular water (HzO). Both types of
water are believed to be in equilibrium by
HzO+-Si-O-Si-~-SiOH+-SiOH (1)
with the equilibrium constant
K=[-SiOHI2/[H,O] (2)
where [-SiOHj and [H20j are the concen-
trations of hydroxyl and molecular water,
respectively. When the total water concen-
tration is low, most of the water is in the
-SOH state. Among the various suggested
mechanisms of water diffusion in glass, the
molecular water diffusion mechanism sug-
gested by Doremus‘ appears to be most
plausible, giving a reasonable value of the
activation energy and the concentration de-
pendent diffusion coefficient roughly con-
sistent with the experimental data. In this
mechanism the minority water component,
molecular water, diffuses and reacts with
the SiO, network by Eq. (1) to form im-
mobile hydroxyl. The reaction is assumed
to be so fast that local equilibrium is estab-
lished. This assumption leads to the effec-
tive diffusion constant D, of water given by
CONTRIBUTING
EDITOR--. C. BOYD
Received May 3, 1985; revised copy received
June 13, 1985; approved June 13, 1985.
*Member, the American Ceramic Society.
Communications of the American Ceramic Society
ACKNOWLEDGMENTS
We acknowledge the experimental assistance pro-
vided by Mr. Mark Rowe. We also thank Mrs. H. L.
Henrich for preparation of the manuscript.
REFERENCES
‘K. Prewo and J. Rrennan, “High-Strength Silicon
Carbide Fiber-Reinforced Glass-Matrix Composites,’’
J. Mater. Sci., 15 [2] 463-68 (1980).
’K. Prewo and J. Rrennan, “Silicon Carbide Yam
Reinforced Glass Matrix Composites,” J. Muter. Sci.,
17, 1201-206 (1982).
3J. Brennan and K . Prewo, “Silicon Carbide Fiber
Reinforced Glass-Ceramic Matrix Composites Exhib-
iting High Strength andToughness,” J . Muter. Sci., 17,
2371-83 (1982).
4T.Mah and H. A. Lipsitt, ”Silicon-Carbide Fiber-
D, =2DHZo[-SiOH]/K
=2DH,,[H,0]/[-SiOH] (3)
where DH20 is the diffusion coefficient of
molecular water. This model was success-
fully used to analyze the concentration
profile of water in Si02 glass’ as well as
the oxidation of s i l i c ~ n .The
~ , ~ overall be-
havior of water diffusion in SiO, glass can
therefore be explained satisfactorily by
this mechanism.
When closely examined, however,
some discrepancies exist between experi-
mental results and the theory based on this
mechanism. For example, according to this
model, the diffusion coefficient is linearly
proportional to the concentration of -SOH
or approximately to the total water concen-
tration and should extrapolate to zero for
SiOz glass with zero water content at all
temperatures. Consequently, the concen-
tration vs diffusion distance profile should
have the same characteristic shape at all
temperatures and concentrations. In reality
the diffusion coefficient extrapolates, for
glass with zero water content, to nonzero
values, which are dependent on tempera-
t ~ r e Correspondingly,
.~ different concen-
tration profiles are observed at different
temperatures. Figure 14-6 shows that the
concentration profile at a high temperature
is similar to the theoretical prediction by
Doremus, whereas the profile at low tem-
perature is closer to the one expected for
a constant diffusion coefficient. Further-
more, the shape of the concentration profile
at low temperature was found to be de-
C-25 1
Reinforced MAS Glass-Ceramic Composites”;
pp. 153-71 in Proceedings of Joint NASAlDOD Con-
ference, Cocoa Reach, FL, Jan. 20, 1983. NASA Conf.
Publ., No. 2291.
5K. Prewo, “Advanced Characterization of S i c
Fiber Reinforced Glass-Ceramic Matrix Composites,”
Interim Rept. R83-915939-1 on ONR Contract No.
N00014-81-C-0571, June 1983.
6T.Mah, “Fiber-Reinforced Ceramic-Matrix Com-
posites”; pp. 245-60 in Proceedings of Joint NASA/
DOD Conference, Cocoa Beach, FL, Jan. 19, 1984.
NASA Cons Publ., No. 2357.
’T. Mah, M. G. Mendiratta, A. P. Katz, R. Ruh,
and K. S. Mazdiyasni, “Room-Temperature Mechani-
cal Behavior of Fiber-Reinforced Ceramic-Matrix
Composites,” J. Am. Ceram. SOC. 68 [I] C-27-C-30
(1985). 0
pendent on the concentration. At low water
concentrations, a concentration profile is
similar to a complementary error function
profile expected for a constant diffusion co-
efficient, whereas at higher water concen-
trations, a concentration profile approaches
the one expected from Eq. (3) (cf. Fig. 2).4
One might argue that the equilibrium con-
stant K in Eq. ( 2 ) may not be a strict con-
stant; instead it may be a function of water
concentration and this can account for the
observed discrepancy. But in general, the
best approximation to the constant equi-
librium constant K is expected at low water
concentration. Yet at this low concentration
the greatest deviation from the concen-
tration profile expected from Eq. (3) is
observed. It is the purpose of the present
report to formulate a realistic model of
water diffusion in SiO, glass which can
account for the observed discrepancy.
In the molecular water diffusion
mechanism suggested by Doremus,’ it is
assumed that only the concentration gra-
dient of the molecular water causes the
diffusion flux of water. And the molecu-
lar water is assumed to diffuse by an inter-
stitial mechanism in a similar manner to
noble gases in glass until it reacts with
the Si-0-Si network.
It is suggested here that water can
diffuse through the motion of -SOH in
addition to the motion of molecular water.
This possibility was suggested earlier by
Roberts and R ~ b e r t s When
.~ water can
diffuse through the motion of -SOH also,
the concentration gradient of -SOH, in ad-
dition to the concentration gradient of mo-
lecular water, becomes important in
water diffusion. Studies on diffusion using
isotope “0 show that the oxygen in -SiOH
and H,O in glass can exchange easily. For
example, the study on the diffusion of
H2% into S O z glass showed7 that the dif-
fusion distance of water is much greater
than that of “0. A separate experiment on
the diffusion of water into S O 2 glass con-
taining “0 also indicated* that the network
oxygen can exchange with the oxygen from
the molecular water by
Hz’60+-Si-180-Si- + -SiI60H
+-Si’*OH+ H2180+-Si-’60-Si (4)
This observation suggests that the follow-
ing diffusion mechanism is possible.