High Temperature Corrosion of Ceramics PDF

HIGH TEMPERATURE
CORROSION OF CERAMICS
J.R. Blachere
F.S. Pettit
Department of Materials Science and Engineering
University of Pittsburgh
Pittsburgh. Pennsylvania
NOYES DATA CORPORATION

Park Ridge, New Jersey,U.S.A.
Copyright @ 1989 by Noyes Data Corporation
Library of Congress Catalog Card Number: 88-38242
ISBN: O-8155-1188-4
Printed in the United States
Published in the United States of America by

Noyes Data Corporation
Mill Road, Park Ridge, New Jersey 07656
10987654321
Library of Congress Cataloging-in-Publication Data
Blachere, J.R.
High temperature corrosion of ceramics / by J.R. Blachere and F.S.
Pettit.
P. cm.
Bibliography: p.
Includes index.
ISBN O-8155-1188-4 :
1. Ceramic materials--Corrosion. I. Pettit, F.S. (Frederick
S.), 1930- . II. Title.
TA455C43B57 1989
620.1’404217--dc19 88-38242
CIP
Other No yes Publications
CERAMIC RAW MATERIALS
Edited by
D.J. De Renzo
The broad-based ceramics Industry encom- provides chemical and physlcal property data
passes all types of glass; refractones; abrasives; for more than 1000 products supplied by 181
whitewares. such as porcelain and pottery, ceramic raw material suppliers in the U.S. and
structural clay materials; etc. Increasing use of Canada.
advanced ceramic materials in the automotive Part I is an alphabetical listing of 99 raw
and aerospace Industries, as well as in such material categories. Raw material suppliers are
diverse areas as electronics and medical devices, then included alphabetically under each cate-
IS expected to push the demand for raw materials gory, along with their product information. The
far beyond that associated with the traditional categories in Part II include additives; semi-
ceramics industry. processed materials, some of which are available
Prepared directly from manufacturers’ data as unfinished shapes or substrates; and materi-
sheets and tables at no cost to, nor influence als intended for specific end uses. Raw materials
from, the contributing companies, this book categories are:
Parr I Corundum Potassium Compounds Yttrium Oxide

Alumina Diatomaceous Earlh Pyrophyllite Zinc Oxide
Alumina Chrome Dolomite Duartz Zircon
Alumina Zirconia Feldspar Rutile Zirconates
Aluminum Nitride Ferrites Sand Zirconia
Aluminum Titanate Fireclay Silica Zirconium Boride
Andalusite Flint Clay Silicon Zirconium Nitride
Antimony Compounds Fluorspar Silicon Carbide
Ball Clays Fluxes Silicon Nitride
Barium Compounds Frits Sillimanite Part II
Bauxite Garnet Slags Binders
Beryllium Oxide Graphite Soapstone Ceramic Additives
Bismuth Compounds Iron Oxide Soda Ash Ceramic Adhesives,
Bone Ash Kaolin Sodium Silicate Potting Materials,
Borates Kyanite Sodium Sulfate and Putty
Borax Lanthanide Compounds Spars Ceramic Coatings
Boric Acid Lead Compounds Spine1 Ceramic Colors
Boron Lime Spodumene Ceramic Fibers
Boron Carbide Limestone Stannates and Whiskers
Boron Nitride Lithium Compounds Stannic Oxide Ceramic Materials
Calcium Compounds Magnesia Strontium Carbonate Ceramic Precursors
Carbon Magnesite Talc Dielectric Compositions
Celestite Manganese Compounds Titanates Dispersing Agents
Chlorite Mineral Mica Titanium Boride Electronic Ceramics
Chrome Ore Mullite Titanium Carbide Reagents
Chromium Oxide Nepheline Syenite Titanium Dioxide Glazes
Clays (Miscellaneous) Nickel Compounds Titanium Nitride Glaze Stains
Cobalt Compounds Ochre Ulexite Pie20 Compositions
-. .
Cofemamte Periclase Vermiculite Refractory Materials
Copper 8 Copper Oxide Petalite Whiting Sealing and Solder
Cordierite Phosphates Wollastonite Glasses
ISBN O-8155-1143-4 (1987) 8%” x 11” 900 pages

Foreword
This book describes high temperature corrosion of ceramics. The materials in-
vestigated in this particular study were silica, alumina, silicon nitride and silicon
carbide. In addition to the pure single crystals or CVD materials, typical engi-
neering materials of various purities were included in the study. The corrosion
conditions were ‘hot corrosion’ in which gaseous corrosion was enhanced by
Na2S04 deposits. Some gaseous corrosion and oxidation experiments were also
performed. The hot corrosion was studied at IOOO’C and at lower tempera-
tures, in the presence of pure oxygen and oxygen containing SO2 and SOs. The
changes in morphology of the surfaces were observed in the scanning electron
microscope. This instrument and the related x-ray microanalysis were the
major tools of research. A method of measurement of oxide thickness in the
electron microprobe was developed for the experiments on silicon nitride and
silicon carbide.
In the use of materials at elevated temperatures in harsh environments, it is ap-

parent that, in most instances, ceramics are the best choice to provide corrosion
resistance. While ceramics may be more corrosion resistant than metallic alloys
and polymers, ceramics can react with certain environments. The purpose of
this study, then, was to systematically investigate the corrosion of ceramics and
to develop a theory generally applicable to all ceramics.
A great number of ceramic materials are available for use in a variety of corrosive
environments. In order to develop a theory applicable to the corrosion of cer-
amics in general, it was necessary to investigate a variety of representative cer-
amic materials exposed to a number of different corrosive environments. In
order to keep the number of experiments at a reasonable number, the ceramic
materials used in this study were selected on the basis by which resistance to
corrosion was developed, in particular, by being immune to the environment
or by developing passivity. Furthermore, the environments used to produce
corrosion were selected based upon the likelihood of their being encountered in
practice and their severity.
V
vi Foreword
The information in the book is from High Temperature Corrosion of Ceramics,

prepared by J.R. Blachere and F.S. Pettit of the University of Pittsburgh for the
U.S. Department of Energy, December 1987.
The table of contents is organized in such a way as to serve as a subject index

and provides easy access to the information contained in the book.
Advanced composition and production methods developed by

Noyes Data Corporation are employed to bring this durably
bound book to you in a minimum of time. Special techniques
are used to close the gap between “manuscript” and “completed
book.” In order to keep the price of the book to a reasonable
level, it has been partially reproduced by photo-offset directly
from the original report and the cost saving passed on to the
reader. Due to this method of publishing, certain portions of
the book may be less legible than desired.
NOTICE
The materials in this book were prepared as an account

of work sponsored by the U.S. Department of Energy.
Neither the United States Government nor the Depart-
ment of Energy, nor any of their employees, nor any of
their contractors, sub-contractors, or their employees,
nor the Publisher, makes any warranty, express or im-
plied, or assumes any legal liability or responsibility for
the accuracy, completeness, or usefulness of any infor-
mation, apparatus, product or process disclosed or rep-
resents that its use would not infringe privately-owned
rights.
Final determination of the suitability of any informa-

tion or procedure for use contemplated by any user,
and the manner of that use, is the sole responsibility of
the user. The reader is warned that caution must always
be exercised when dealing with ceramics at high temper-
atures, and expert advice should be sought at all times
before implementation.
Contents and Subject Index
INTRODUCTION.........................................l
EXPERIMENTAL PROCEDURES. . . . . . . . . . .......... .6

Experimental .
Conditions. . .. . . . . . . . . . .......... .6
Special Techniques . .. . .. . . . . . . . . . . . . .......... .7
Measurement of Oxide Thickness . . . . . . .......... .7
Measurement of Contact Angles. . . . . . . . . .......... .9
Cross Sections and Related Techniques .......... .9
GASEOUS CORROSION ....................... ...... . . . . 11

Introduction. ............................. ...... . . . . 11
Gaseous Corrosion of Silica and Alumina. .......... ...... ... 11
Silica. ................................ ...... . . . 11
Alumina .............................. ...... . . 13
Gaseous Corrosion of Silicon Nitride and Silicon Carbide ...... . . . . 14
HOT CORROSION OF SILICA. . . . . . . . . . . .. . . . . . . .. . . . . . . .16
HOT CORROSION OF ALUMINA. . . . . . . . . . . . . . . .. . . . . . . . .21
HOT CORROSION OF SILICON CARBIDE AND SILICON NITRIDE. . . .26
REFERENCES..........................................30
APPENDIX A-GASEOUS CORROSION OF SILICA AND ALUMINA

IN SULFUR OXIDE ENVIRONMENTS . . . . . . . . . . . . . . . . . . . . . . . . 31
H. R. Kim, J. R. Blachere, F.S. Pettit
Introduction. .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Gaseous Corrosion of Ceramics . . . . . . . . . . . . . . . . . . . . . . . . . 33
vii
viii Contents and Subject Index
Experimental Procedure. .............. . . . . . . . . .35

Materials ....................... . . . . . . . . . . .35
Experiments. .................... . . . . . . . . . .35
Results and Discussion ................ . . . . . .37
Silica. ......................... . . . . . . . . .37
Alumina ....................... . . . . . . . .39
General Results. ................ . . . . . . .39
Thermodynamics of Sulfate Formation . . . . . . . . .41
Single Crystal .................. . . . . . . .43
Polycrystalline Aluminas. .......... . . . . . . .43
General Discussion. .................. . . . . . : :53
52
Conclusions ....................... . . . . . .
References ........................ . . . . . . . . .55
APPENDIX B-HOT CORROSION OF SILICA. ................. . .5B

M. G. Lawson, H. R. Kim, F.S. Pettit, J. R. Blachere
Introduction. .................................... . 58
Hot Corrosion. ................................... : :60
59
Experimental Procedure .............................
Materials ..................................... : :60
60
Gaseous Corrosion. ..............................
Hot Corrosion. ................................. . .60
Results and Discussion .............................. 63
Gaseous Corrosion. .............................. . .63
Wetting by Sulfates .............................. . .64
Hot Corrosion. ................................. . .66
Kinetics of Crystallization. ......................... : 176
72
General Discussion. ..............................
Discussion of Crystallization Kinetics .................. . .82
Hot Corrosion of Silica Formers. ..................... . .85
Conclusions ..................................... . .86
References. ..................................... . .87
APPENDIX C-HOT CORROSION OF ALUMINA . . . . . . . . . . . .89

M.G. Lawson, F.S. Pettit, J.R. Blachere
Introduction. . . . . . . . .
. . . .. . . .. . . . . .. . . . . . . .89
Experimental Procedure. . . . . . . . . . . . . .. . . . . . . .95
Materials . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . .95
Hot Corrosion Experiments. . . . . . . . . .. . . . . . . .97
Results and Discussion . . . . . . . . . . . . . . .. . . . . . . . .97
Wetting . . . . . . .. . . . . . . . . . . . .. . . . . . .97
Hot Corrosion of Al*Os .. . . . . . . . . .. . . . . . . 102
Results, Acidic Conditions . . . . . . . . . . . . . . . . . 107
Single Crystal Alumina. . . . . . . . . . . . . . . . 107
High Purity Polycrystalline Alumina. . . . . . . . . 109
Medium Purity Polycrystalline Alumina. . . . . . . . . 112
Low Purity Polycrystalline Alumina . . . . . . . . . . . 113
Results, Basic Conditions . . . . . . . . .. . . . . . . . 115
Contents and Subject Index ix
Single Crystal . . . . . . . . .. . . . .
. .. . . . 116
High Purity Polycrystalline . . .
Alumina . . 116
Medium Purity Polycrystalline Alumina. . . . 118
Low Purity Polycrystalline Alumina .. . . . . 119
Discussion of Results . . . . . . . . . . . . . . . . 122
References. . . . . . . .. . . .. . . . . . . .. . 136
APPENDIX D-HOT CORROSION OF SILICON NITRIDE AND

SILICON CARBIDE ............................... . . . . 137
J. R. Blachere, D. F. Klimovich, F.S. Pettit
Introduction. ................................ . . . . . 137
Experimental Procedure ......................... . . . 137
Results and Discussion .......................... . . . 138
Acidic Hot Corrosion of Silicon Nitride. ............ . . . 152
Discussion of the Hot Corrosion of Silicon Nitride. ..... . . 157
Acidic Hot Corrosion of Silicon Carbide ............ . . . . . 161
Model for the Oxidation and Hot Corrosion of Silicon
Carbide Under Acidic Conditions .............. . . . 168
Proposed Model .......................... . . . . . 169
Defect Structure and Stoichiometry of Silica ....... . . . . 170
Oxidation of Silicon Carbide .................. . . . . 172
Acidic Hot Corrosion. ...................... . . . 178
References. ................................. . . . . 182
APPENDIX E-PUBLICATIONS ............................. 184

References. ....................................... 185
Introduction
In the use of materials at elevated temperatures in harsh environments, it
is apparent that in most instances ceramics are the best choice to provide
corrosion resistance. While ceramics may be more corrosion resistant than
metallic alloys and polymers, ceramics can react with certain environments.
The purpose of this program was to systematically investigate the corrosion of
ceramics and to develop a theory generally applicable to all ceramics.
A great number of ceramic materials are available for use in a variety of
corrosive environments. In order to develop a theory applicable to the corrosion
of ceramics in general, it is necessary to investigate a variety of representative
ceramic materials exposed to a number of different corrosive environments. In
order to keep the number of experiments at a reasonable number, the ceramic
materials used in this study were selected on the basis by which resistance to
corrosion was developed, in particular, by being immune to the environment or
by developing passivity. Furthermore, the environments used to produce corrosion
were selected based upon the likelihood of their being encountered in practice
and their severity.
A material is considered to be immune to a particular environment when it
is in equilibrium with that environment. Total or complete immunity is rare in
practice but in some cases the amount of reaction required for equilibrium to be
achieved is very small, and consequently the changes in the properties of the
material are very small. For example, when Al303 is heated in oxygen at
elevated temperatures, the oxygen activity in the Al.303 may not be the same as
that in the gas, and oxygen will be incorporated into or removed from the Al303
depending upon the oxygen activities in the Al303 and in the gas. Upon obtaining
equilibrium between the Al303 and the gas, which may require extremely long
2 High Temperature Corrosion of Ceramics
times depending upon the temperature, the Al303 does not exhibit any significant
changes in mechanical properties. The concentrations of point defects in the
Al303 however will change and there could be significant changes in properties
such as electrical and ionic conductivities. Nevertheless, in terms of a corrosion
reaction the Al303 may be considered to be immune to this gaseous environment.
Alumina and silica are two ceramic materials which can be considered to be
immune to many environments which are extremely corrosive to metallic systems,
and therefore these two materials were studied in the present investigation. The
purity and structure of these ceramics can also affect their behavior in different
corrosive environments and therefore these two parameters were also examined
in the present investigation. Four types of crystalline Al303 were studied of
varying purity as defined in Table I. Only one type of silica was studied. As
described in Table I it was of relatively high purity and was vitreous.
Passivity to corrosive environments is achieved by the formation of a
reaction product barrier through which the reactants involved in the corrosion
process must diffuse. The development of passivity is the principal means by
which metallic alloys achieve resistance to corrosive environments, but passivity
is also an important mechanism in certain ceramic materials, depending upon the
ceramic and upon the environment causing corrosion. For example, Si3N4 and
Sic react with most environments encountered at elevated temperatures and
resistance to corrosion is achieved via the formation of a passive reaction product
barrier. Furthermore, Al303 can react with gaseous environments containing
sufficient SO3 to form sulfates and again the properties of the sulfate reaction
product play a significant role in the corrosion properties. In the present studies
Si3N4 and SIC specimens with purities as defined in Table I were therefore used.
TABLE I - MATERIALS UNDER STUDY
Designation Material Purity (Wt%) Supplier
S.C. Sapphire -99.99 Saphikon
998 poly A120; (a) 99.8(0.1MgO, O.llSiO2) Lucalox (G.E.)
995 poly A1203 (a) 99.6 - 99.5(0.17 M@ AlSi Mag 772 (3M)
0.17 Si02) or ADS 995 (Coors)
975 poly A1203 (a) 97.4 (0.75 MgO, 1.6SiO2 S-697 (Saxonburg)
0.1 Na20)
Silica fused silica 99.99 Corning 7940
SIN CVD CVD*Si3N4 99.99 Deposits & Composites
SiN H.P. H.P.** Si 3N4 93(6Y203Fe, Al, 02) Airesearch
Sic SC. s.c.*** (a6H S.C.) 99.99 W.J. Choyke
Sic CVD CVE (Sic) 99.98 W. J. Choyke
Sic H.P. H.P. (aSiC) 94(3Al203, 2.5W, Fe, 02) Norton NC-203
* Chemical Vapor Deposition *** Single Crystal

** Hot Pressed + Polycrystalline Alumina
When ceramic materials are used in practical applications, a variety of
corrosive environments can be encountered. Most of these environments will
contain oxygen but other reactants such as sulfur, nitrogen, carbon and chlorine
can also be present. Moreover, deposits such as metallic sulfates, carbonates or
chlorides may also accumulate upon exposed surfaces and substantially affect
the corrosion processes. In the case of studies concerned with the corrosion of
metallic alloys it has been useful to examine corrosion reactions in environments
of increasing severity extending from gaseous environments containing oxygen,
to mixed gases containing one or more reactants in addition to oxygen, and finally
considering the effects of deposits such as Na3S04, Na3C03 or NaCl. The
environments used in the present studied are identified in Table II. These
environments consisted of gases containing oxygen, and other reactants such as
sulfur, carbon and hydrogen at temperatures of 700, 1000 and 14OOOC. The
effects of deposits were studies by using Na3S04 deposits in O3-SO3-SO3 gas
mixtures at temperatures of 700 and 1000°C. These conditions were selected
because they are frequently present in many environments encountered in
practice. Furthermore, as established from studies using metallic alloys, the
principles established from studies using these selections should be generally
applicable to other corrosive systems.
In the following sections of this report the experimental procedures will be
described and then a summary of the results will be presented and discussed.
Some of these results have been presented in previous reports for this
program.(l) Other results are included in student theses and are also presented
in drafts of papers about to be submitted for publication. Consequently, some of
the results will not be repeated in this report but will be referred to by
references or by draft papers included in appendices in the present report.

introduction 5
Results obtained from studies using gaseous environments will be discussed first
and then results obtained from studies using deposits on silica, alumina, Sic and
Si3N4 will be presented in sequence.

Experimental Procedures
Experimental Conditions
The materials that were studied in this program and the environments that
were used to produce gaseous corrosion and hot corrosion attack have been
briefly discussed previously in this report (Tables I and II). The ceramics that
were studied (Table I) consisted of Al3O3, SiO3, Sic and Si3N4. A range of
purities were examined in the case of Al3O3, Sic and Si3N4. Single crystals of
SIC and Al303 were also studied and compared to polycrystalline specimens.
The gas compositions that were used are presented in Table II. Gas mixtures
consisting of O3-SO3-SO3 were used in both gaseous corrosion studies and the
hot corrosion studies. Hot corrosion experiments were also performed in pure
oxygen. The gaseous corrosion studies were performed at temperatures of 700,
1000 and 1400°C whereas the hot corrosion investigations used temperatures of
700 and 1000°C. When O3-SO3-SO3 mixtures were used at 700°C, this mixture
was passed over a platinum catalyst to ensure that equilibrium was achieved.
The catalyst was not required to achieve equilibrium at 1OOOOC.The flow of the
gas was 1 cm3/s.
The experimental procedures have been discussed in detail in previous
reports and are also discussed in Appendices A, 0, C and D. The procedure
usually consisted of exposing specimens in a horizontal tube furnace at a fixed
temperature to a flowing gas stream of fixed composition. In the case of the hot
corrosion studies the specimens were coated with deposits of Na3S04 for
6
Experimental Procedures 7
investigations at 1000°C, or a NaSS04-CoSO4 equimolar solution for studies
performed at 700C. These deposits were applied on specimens by spraying warm
specimens with an aqueous solution saturated with NaSS04 or the sulfate
mixture. Most specimens were diamond polished and they were ail cleaned
before the experiments.
The characterization techniques consisted of morphological studies in the
light microscope and particularly the Scanning Electron Microscope (SEM) with
microanalysis of salient features by EDS (Energy Dispersive X-ray Spectroscopy)
and WDS (Wavelength Dispersive X-ray Spectroscopy used particularly for light
elements). These techniques were supplemented by X-ray diffraction, weight
change measurements and in some cases surface analysis techniques (ESCA). A
large number of other techniques were used in special cases they are SIMS, ISS,
FTIR, Laser beam ellipsometry.
Furthermore, since after the corrosion tests some deposited salt often
remained on the sample, the Nag.504 was washed off to reveal the sample
surface. The samples were characterized before and after this washing.
Experience has shown that the examination of the sample before washing is
extremely fruitful, and the observed morphologies have taken many forms. The
washwater was analyzed as described in a previous report (l); this analysis is now
a routine semi-quantitative method which allows the identification of the
elements dissolved in the salt and the stoichiometry of the salt.
Special techniques
Measurement of oxide thickness
A method using WDS measurements in the electron microprobe (EPMA) was
adapted from that described by Yakovitz and Newbury to estimate the
thickness of coatings from the intensities of emitted characteristic X-rays. The

intensity of the oxygen Ka line is used in the present research to measure the
thickness of oxide layers generated on non-oxide materials. This intensity is
measured under constant conditions (say 1OkVaccelerating voltage and 2 x lo8
A beam current) for a scale and a bulk standard of pure fused SiO2. Both
samples are coated with 200A of carbon. The ratio k of their intensities
corrected for background is correlated to the thickness of the oxide layer.
The calibration curve for the relationship between the k ratio and the oxide
thickness was calculated with a computer program written for this research
following the semiempirical approach of Yakovitz and Newbury(2) to generate
the Q (pz) curve which is the intensity of X-ray generated at a weighted PZ depth
into the sample. This calculation includes many corrections and correlations
used to predict the characteristic X-ray intensity. It must be understood that
the data available for these corrections (such as absorption) for light elements is
relatively poor and that this calibration can only be approximate. Measurements
based on light element are usually not used quantitatively. However the method
developed in this research gives good reproducibility ? 2% for thickness
measurements on silica layers up to about 0.8pm - lum. It was published
recently(3) and more details will be found in Appendix E which contains a copy of
the article on this measurement. It must be emphasized that since the
characteristic X-ray intensity for oxygen is used to calculate the oxide
thickness, a stoichiometric silica (SiO2) was assumed in the calculations. This
may lead to significant error if the composition of the scale deviates markedly
from that assumed in the calculations. However the method is quite
reproducible and has a high spatial resolution.
The direct measurement of scale thicknesses on cross sections in the SEM
is probably not better than + 10% in accuracy considering magnification

Experimental Procedures 9
calibration and other sources of error under the best conditions. It is good down
to about lum and depends greatly on the thickness of the layer (constancy and
magnitude). In some cases the scales are difficult to separate from the substrates
in cross sections in the SEM.
Measurement of contact angles
The contact angles of the deposits have been measured in the SEM at room
temperature using a method previously described by Murr(4). The wetting angle
is particularly important since deposits often break into droplets decreasing the
area of interaction between the sample and the melt. After coating with carbon
(about ZOOA)the samples are placed in a tilted position in the SEM which allows
the location of the drop under the electron beam and then they are tilted further
to a vertical position to record the profile of the drop. The edges of the drops
are enlarged for the measurement. A number of drops are measured under those
conditions on the same sample. Experience with the measurements show that
their reproducibility is about f: 2O. Since this reproducibility is very good, it has
been possible to establish when two wetting behaviors occurred simultaneously
on the same sample.
Cross sections and related techniques
The study of cross sections is necessary for a number of measurements.
However the samples are often damaged (silica and silica formers) during exposures
and processing. Also, cross sections are often desired with salt on the sample, so
that sawing or polishing in wet media has been avoided to preserve the soluble
sodium compounds. Most cross sections were either dry-cut with a diamond saw
or fractured. Fractured specimens show structural features well as has been

established in many studies of the microstructure of ceramics. The multilayer
nature of the scale and sometimes its apparent layered growth is shown clearly
in cross sections. The cross sections can also be used directly for the measurement
of oxide layer thickness (~l~rn). In some cases salt drops have been crossed by
the fracture giving valuable information. However the lack of flat cross sections
has limited the use of the crystal spectrometers of the electron microprobe.
In many cases the deposit drops do not adhere welt to the sample surface
after cooling so that the underside of the drops and the area of the sample which
was under the drops can be examined and analyzed at least qualitatively.
Mechanical bursting of bubbles in surface layers and removal of drops are also
used to study the underlying microstructure prior to washing of the specimens.

Gaseous Corrosion
Introduction
The results from the gaseous corrosion studies will be discussed by considering
first the studies of silica and alumina, and then the studies performed using Sic
and Si3N4.
Gaseous Corrosion of Silica and Alumina
The gaseous corrosion of silica and alumina were performed at temperatures
of 700, 1000 and 14000C in a number of different gas environments which included
oxygen, 02-H20-H2, 02-CO2-CO and O2-SO2-SO3 gas mixtures (Table II). The
results obtained from these studies are discussed in detail in Reference 1 and
Appendix A. The specific conclusions developed from these studies are as follows:
Silica
(1) Devitrification of silica glass to cristobalite took place rapidly under all
atmospheres studied at 14OOOC. The rate of crystallization increases with
increasing temperature and time.
(2) The devitrified layer undergoes a displacive transformation with a large
volume change on cooling which causes cracking. Such a transformation in crystalline
SiO2 is important with regards to the use of crystalline SiO2 scales on Sic, Si3N4
and metallic alloys as protective barriers for high temperature applications.
(3) The silica is significantly affected when exposed to low oxygen pressure
at 1400°C. Silica weight losses occurred after exposure to either wet hydrogen
or a CO-CO2 gas mixture and are related to the decomposition of silica in the
low oxygen pressure and the reaction of either hydrogen or CO with silica to
form SiO vapor and either H20 or CO2 gas. Weight losses increase with increasing
temperature and decreasing oxygen pressure.
11
Gaseous Corrosion 13
(4) No significant reaction of silica in CO2 or 02 was observed at all
temperatures except for devitrification at 14OOOC.
(5) Silica is very resistant to attack by SO2-SO3-02 gas mixture under the
test conditions.
Alumina
(1) Alumina is resistant to attack by H2-H20-02 gas mixture but impurities
in the alumina materials such as SiO2 and Na20 can result in volatilization. The
volatilzation is favored by low oxygen pressures and high temperatures.
(2) Alumina is very resistant to attack by CO2 gas.
(3) Alumina is resistant to corrosion by SO2-SO3-02 gas mixtures. Some
reactions occurred especially at high SO3 pressures (e.g. 7~10~~ atm. SO3) and
low temperature (7OOOC).
(4) Corrosion of alumina in SO3 containing gas can occur even on the highest
purity alumina where reaction products with activities less than unity are formed.
It may be due to the formation of a solid solution of Al2(SO4)3 with Al203 or a
nonstoichiometric sulfate. The observed sulfur was identified as a ~+6 (as in
sulfate) by ESCA.
(5) Degradation of alumina in SO3 containing gas becomes more severe
when Mg or Ca containing impurities are present, and it increases as impurity
content increases. The formation of products occurs preferentailly along the
grain boundaries (i.e., on the impurity second phases at the grain boundaries).
(6) Corrosion of alumina in SO3 containing gas becomes more favorable at
higher SO3 pressures and lower temperature (e.g. at 7000 than 1000°C), since a
lower SO3 pressure is necessary to form sulfate at 700°C than at 1000°C. The
severity of corrosion increases with time.

Gaseous Corrosion of Silicon Nitride and Silicon Carbide
The gas induced corrosion of Si3N4 and SIC was studied only at 1400°C in
oxygen and at 1OOOoC in an 02 - SO2 - SO3 gas mixture with an initial SO2
pressure of 0.01 atm. These limited studies were performed to determine if
devitrification of the silica scales formed on these materials occurred at 1400°C,
and to provide baseline data for comparison with the hot corrosion studies at
1ooooc.
The studies in oxygen at 1400°C showed varying results depending upon the
form of the substrate material. In the case of the single crystal Sic
crystallization of the silica scales was observed. These scales which were about
1 urn thick after 12 hours of oxidation and cracked upon cooling to form star
patterns that were believed to radiate from nucleation centers indicating the
original nuclei of the crystallization of the amorphous layer into cristobalite.
The silica scales which formed on the polycrystalline samples were thicker in
most cases than those formed on the single crystals of Sic being about 10 urn
after 12 hours of oxidation. The hot pressed silicon carbide exhibited a glazed
surface and contained bubbles resulting from CO and CO2 evolution. The lack of
crystallization was attributed to impurities, particularly Al2O3, stabilizing the
glass structure. Some cristobalite was identified in these scales however by
XRD along with a large “glass” peak. The hot pressed silicon nitride also formed
a 10 urn thick scale which was composed of cristobalite and enstatite (MgSiO3).
The CVD silicon nitride, which was much purer than the other silicon nitrides
developed an extremely thin silica scale after 12 hours of oxidation. There was
virtually no change in surface morphology of the oxidized specimens compared to
specimens prior to oxidation and the weight change of specimens after 12 hours
of oxidation was below the detection limits of the techniques used in this program
Gaseous Corrosion 15
to measure weight change. No impurities were detected in these silica scales on
CVD silicon nitride which XRD analyses showed to be a mixture of cristobalite
and glass. The sintered silicon nitride developed a scale composed of glass
containing some cristobalite. Yttrium silicate crystals were observed to protrude
out above the surface of the silica scale. Impurities influenced the oxidation of
this material substantially.
The results obtained from the studies performed at 1400°C show that the
oxidation of SIC and SiSN4 is dependent upon the structure and composition of
the silica scales that are formed upon these materials. Glassy silica provides a
more protective reaction product barrier than crystalline silica, however, the
incorporation of impurities into the glassy silica can cause the protectiveness of
the glass structure to be decreased very substantially by promoting devitrification.
At 1000°C the pressure of SO2 and SO3 in the gas mixture along with oxygen
did not significantly affect the oxidation of pure silicon nitride or pure silicon
carbide compared to oxidation in pure oxygen. The major effect of SO2 and SOS
occurred when the specimens contained impurities. While the effects of impurities
were significant but not documented extensively, these effects were not as
substantial as those observed at 1400°C, since the impurities cannot concentrate
in the oxide scale at 1000°C and thereby affect the protective properties of the
glassy silica scales.

Hot Corrosion of Silica
The results obtained from the studies on the hot corrosion of silica are
presented and discussed in more detail in Appendix B of this report. In the
following the important results from these studies are briefly summarized.
Specimens of fused silica (Corning 7940) about 1 cm x 1 cm square and 1
mm thick were exposed to a variety of conditions known to cause the hot corrosion
of metallic alloys. The experimental procedures have been described in the
experimental section of this report. Results from four different sets of
experimental conditions will be discussed in this summary. These experiments
were performed at 700 and 1000°C. The deposits of Na3S04 which were applied
to the specimens’ surfaces were liquid at 1OOOoC. At this temperature two
different gas compositions were used. One consisted of an SO3-03 gas mixture
and the other was pure oxygen. The SO3 pressure in the gas mixture was 1.5 x
10e3 atm. When Na3S04 is exposed to gases containing SO3 the activity of
Na90 in the sulfate is inversely proportional to the SO3 pressure. If the activity
of Na30 in the sulfate is taken as a measure of the basicity of the Na3S04,
higher SO3 pressures established less basic, or more acidic melts, whereas the
gas containing only oxygen causes the more basic liquid sulfate to develop.
Similar sets of experiments were performed at 700°C however at this
temperature the SO3 pressures in the gas mixtures was 7 x 10m3atm.
Furthermore the deposit applied to the specimen surfaces was Na3S04 - 50 mole
percent CoSO4 since pure Na$304 melts at about 883OC and the sulfate mixture
is liquid at 700°C.
16
Hot Corrosion of Silica 17
The liquid sulfate deposits wetted the silica specimens in varying amounts
depending upon the experimental conditions. Time at temperature and the
thickness of the deposit also affected wetting. Generally the liquid deposits
wetted the silica more completely under basic conditions. At 1000°C under
basic conditions the salt wetted most of the coupon after 1 hour with a wetting
angle of 20. After 24 hours the wetting was continous. At this temperature and
under acidic conditions droplets of liquid were formed with wetting angles Of
24O, and 130 on large droplets (’ 0.2 mm dia), after 24 hours. Wetting was not as
complete at 7OOoC. In the case of basic conditions droplets with wetting angles
between 13 -240C were observed after 24 hours, whereas under acidic conditions
the angles ranged between 36 - 49O.
At 7000C under acidic conditions some limited localized attack of the
silica was observed under of the salt droplets. No evidence of devitrification of
the silica was observed. After water washing to remove the salt, small weight
losses (0.1 - 1 mg/cm2) were detected and small voids were evident in the silica
where salt droplets had been present prior to water washing. The voids were
more concentrated near the perimeter of the droplets. Analysis of these results
has been complicated by the decomposition of the CoSO4 in the liquid via the
reaction,
coso4 * coo + so3
since the SC3 pressure in the gas was lower than the equilibrium pressure
required to maintain the initial liquid deposit. The principal reaction of the
deposit with the silica should involve the Na20 component of liquid since
sulfates and sulfides of silicon are not stable under the experimental conditions
that were employed. Hence, a reaction of the following type seema plausible,
XSiO2 + YNa2S04 = YNaZO-XSi02 + ~~03

Inspection of this reaction shows that the formation of sodium silicate phases
becomes less favorable as the SO3 pressure is increased. It is believed that the
SOS pressure in the melt over most of the specimen surfaces is too high to
permit silicate formation. However, some dissolution of silica occurred in the
sulfate at localized regions under the drops, in particular along the sulfate -fused
silica interface which caused voids to be evident upon water washing. This
localized dissolution is believed to result from impurities in the silica which
affect its stability.
Under basic conditions at 700°C the complication from decomposition of
the CoSO4 component of the liquid was more severe since there was no SOS in
the gas phase. Cristobalite spherulites were observed beneath the salt drops.
Weight change measurements after water washing were not meaningful because
some cobalt oxide resulting from decomposition of CoSO4 remained upon the
specimen surfaces. Since dissolution of the fused silica was observed under the
acidic conditions, and this process is believed to involve NaSO, such a reaction
would also be expected in the more basic melt. The important result obtained at
7000 under basic conditions however is that crystallization of the fused silica
was observed beneath the salt droplets. Since crystallization was not observed
at 700°C under acidic conditions, the activity of NaSO in the liquid deposit must
be established at some specific level in order for crystallization to proceed.
At 1OOOoCunder acid conditions cristobalite was observed to form beneath
all of the droplets, but no devitrification of the fused silica was evident away
from the droplets. The weight losses of specimens under acidic conditions at
1000°C were less than at 700°C. This can be accounted for by less sodium
silicate being formed at the higher temperature.

Hot Corrosion of Silica 19
The most extensive degradation of the vitreous silica occurred under basic
conditions at 1OOOoC. A layered reaction product was formed over the total
surface of the fused silica specimens. Proceeding from the salt-specimen
interface the following sequence of phases was observed after very long
exposures: sodium silicate, tridymite, cristobalite, unaffected fused silica. The
thickness of the crystallized silica conformed to a parabolic rate law under
isothermal conditions. Under cyclic conditions this crystallization proceeded
more rapidly and conformed to a linear rate law since the crystalline products
spa&d from the specimen aa a result of thermally induced stresses. The
observed parabolic rate has been accounted for by assuming the crystallization
of the fused silica is caused by sodium from the liquid deposit. There is no
question that some sodium silicate was formed. Analyses of wash water,
however, shows that more water soluble corrosion products are formed early in
the corrosion process than after long reaction times when the crystalline phases
have been formed. Such results suggest that the reaction to form silicates is
dependent upon the composition of the liquid deposit and the structure of silica.
Also as the silicate becomes richer in silica it is less soluble in water.
As discussed previously in the section on gaseous corrosion, silica was not
significantly affected by any of the gas environments used in these studies. No
corrosion nor devitrification was detected at 700 or 1OOOOC. On the other hand
significant degradation of fused silica was observed in SOS-02 gas mixtures, and
in oxygen, when liquid sulfate deposits were present. This degradation occurred
by two different, but related, processes both of which were dependent upon the
activity of NaSO in the deposit. The process which caused the most severe
degradation was devitrification. This process increased as the Na70 activity in
the liquid was increased and it occurred at both 700 and 1000°C. It requires a
threshold Nag0 activity. It was especially severe when the specimens were
thermally cycled since the crystallized products spalled under the influence of
thermally induced stresses. The other process involved the formation of sodium
silicate as a reaction product on the surfaces of fused silica. It appears that this
latter process is less prevalent when the liquid is reacting with crystalline silica
than vitreous silica, nevertheless this reaction also increases as the activity of
Nag0 in the liquid is increased. This reaction also appears to be affected by
impurities in the fused silica when its driving force is low.

Hot Corrosion of Alumina
The results obtained from the studies on the hot corrosion of alumina are
presented and discussed in detail in Appendix C. In the following important
results are summarized.
The weight changes measured after hot corrosion of the aluminas were
small but not negligible. They were due to offsetting reactions such as the
solution of Al203 into the sulfate melt, the silica and silicate precipitation
mostly due to impurity phases and the precipitation of Co0 at 7OOoC in pure
oxygen. In general the weight changes appear greater than for gaseous
corrosion.
The sulfate tended to wet the aluminas partially at 7OOoC under acidic
conditions (~20~ after 24 hours). The wetting of the purer aluminas decreased as
a function of time apparently as some alumina dissolved into the sulfate and
reduced the affinity of the sulfate for the alumina substrates. At higher
temperature (lOOO°C) under similar conditions the wetting improved a little.
The wetting was better at 1000°C under basic conditions than under acidic
conditions as the impurities tended to increase the wetting tendency. The wetting
is an indication of the reaction tendency of the sulfate with the substrate. Good
wetting results in more contact area between the sulfate and the substrate which
also promotes the corrosion.
The single crystal alumina tended to react very little with the Na2S04 and
in line with the gaseous corrosion results more reaction occurred under acidic
conditions than under basic conditions. The single crystal has basal orientation.
The corrosion of the polycrystalline aluminas indicated that the corrosion of the
grains is a function of orientation, with greater attack of planes away from the
21
basal orientation. This result is expected for solution of a single crystal in a
melt.
Under acidic conditions, at 7OOoC sulfates of aluminum and magnesium
were formed and after long exposures globular silica was observed on all
polycrystailine materials. This shows that under the most acidic conditions the
silicate impurities are attacked by the melt generating sulfates and precipitating
silica. The required solubility gradient is set for continued dissolution of alumina
by acid fluxing and precipitation of silica. The precipitation of silica globules,
probably cristobalite, occurs in the sulfate melt away from the interface with
the substrate and is not related to the microstructure of the substrate. The
corrosion is concentrated near the grain boundaries at 700°C and occurs on a
wider scale at 1OOOoC. At the higher temperature the regions under the sulfate
melt are smooth with no marked preferential attack at the grain boundaries. At
1000°C magnesium sulfate and calcium sulfate are formed and alumina is
incorporated in sodium aluminum silicates.
Under basic conditions significant corrosion occurred at 700 and 1000°C.
At 1000°C sodium magnesium aluminum silicates and sodium calcium aluminum
silicates were formed. The salt on cooling contained Mg, Al and some Ca. The
attack of the alumina grains was limited under the more basic conditions overall
at 1000°C and significant intergranular corrosion was evident in the
micrographs.
The impurities played a major role on the corrosion behavior of the
polycrystalline aluminas, particularly at high temperature (lOOO°C), in pure
oxygen (basic conditions). While there was little evidence of basic fluxing of the
single crystal, the high Nag0 activities promote reaction with the silicates
present at the grain boundaries of the polycrystalline materials. Sodium
aluminum silicates grew from the melt with transport of silica and other oxides
Hot Corrosion of Alumina 23
along the grain boundaries. This is illustrated by the perfect decoration of the
grain boundaries of high purity polycrystalline alumina with crystals of this
silicate after long term cyclic exposures. Some magnesium was present in the
silicates formed near triple points. With the more impure aluminas more silicate
was formed and the crystals contained various alkaline earth elements and
potassium which were present as impurities in the materials. While this fluxed
growth of silicate crystals feeding from grain boundaries and impurity grains is
interesting, it plays a fundamental role in the corrosion of the alumina. It is
proposed that the impurity reactions which form sodium silicates lower the Na30
activity and raise the SO3 locally in the melt. This more acidic Na3S04 then can
dissolve the alumina grains by formation of sulfate in the melt by acidic fluxing.
The two reactions, the grain reaction and the grain boundary reaction will
proceed cooperatively, as the ion needed for one is produced by the other.
Therefore a fundamental mechanism is proposed for the hot corrosion of
polycrystalline alumina in which the impurities play a major role. As shown
above they have a major influence even on 99.8% purity alumina, which means on
most technical ceramics. This may apply also to alumina scales grown on
coatings on superalloys but in a general manner since they do not contain silica-
based impurities for which the following mechanism applies.
Under acidic conditions, the sulfate tends to dissolve the alumina even
though conditions are not favorable for the formation of aluminum sulfate at
unit activity. The alumina is dissolved by acid fluxing. The formation of
aluminum sulfate in the field of stability of alumina was already observed in the
gaseous corrosion experiments. In a sulfate melt a wide range of activities can
be established locally. Under basic conditions little or no attack of single crystal
alumina was observed although basic fluxing should be possible, particularly at
high temperature (more basic conditions). However this is not promoted in

presence of silica-based impurities which are present at the grain boundaries and
as second phases in the polycrystalline aluminas. The impurities modify the local
conditions so that intermediate activities favoring the attack of the silicates and
the dissolution of the alumina in the sulfate prevail under the acidic and the
basic conditions of this study. Under acidic initial conditions, set by the gaseous
environment, the sulfate tends to dissolve alumina and as this is done the
activity of Nag0 in the sulfate is increased. This promotes the attack of the
silicates which are dissolved in the sulfate thus decreasing the Nag0 activity.
The two reactions can proceed cooperatively. At 700°C the activity of Na70
was always too low to form sodium silicates and SiO7 is precipitated. At
1000°C, higher Nag0 activities are generated by the same mechanism and
sodium aluminosilicates are formed. Under basic initial conditions, the sulfate
melt tends to attack the acidic impurity phases, thus promoting the dissolution
of neighboring alumina by acid fluxing. The two reactions proceed cooperatively
as discussed earlier.
The mechanisms just presented explain that the polycrystalline aluminas
were attacked under both acidic and basic conditions of the experiments,
however the processes are extremely slow and there was no experimental
evidence of catastrophic attack even after 400 hours exposure. Under the
circumstances one may ask if any of the proposed mechanisms would lead to slow
continuous attack without replenishment of the salt although acidic conditions
provide some of the requirements. However in view of the rate of attack by
sodium sulfate it appears that in many industrial processes in which it could be a
factor the sulfate will be replenished before it might become saturated or
depleted. Higher temperatures might increase the rate of corrosion and basic
conditions might become predominant under usual (percent or less) sulfur

Hot Corrosion of Alumina 25
concentrations in the atmosphere, however the sulfate vapor pressure then will
limit the corrosion since sulfate deposits are no longer formed at higher
temperatures.
Hot Corrosion of Silicon Carbide
and Silicon Nitride
The results of recent experiments are presented and discussed in detail in
Appendix D. Important results are summarized below.
The hot corrosion of silicon nitride and silicon carbide has been studied in
presence of Na9SO4, under acidic, 1% SO9-balance oxygen initially and basic,
pure oxygen, conditions in the temperature range 900-1000°C. During basic hot
corrosion the salt wets completely the samples while during acidic corrosion it
breaks up in droplets. The hot corrosion increases the rate of oxidation and the
thickness of the oxide layers formed increases markedly from acidic (measured
between the drops) to basic hot corrosion and both are greater than for dry
oxidation. The oxide layers formed tend to be vitreous and devitrify rapidly
under the liquid sulfate. For the purer materials devitrification was sparse and
very limited in between the sulfate droplets under acidic conditions. In general,
clearly different behaviors are observed for the pure materials under the two
conditions. They are controlled by the activity of sodium oxide in the sulfate
melt near the interface with the substrate. In all cases the materials oxidize and
the oxide dissolved into the sulfate. For acidic conditions, the sulfate does not
wet the oxide, and a surface activity of sodium oxide in equilibrium with the
atmosphere is set up in between the sulfate droplets. The sodium diffuses into
the vitreous silica and modifies it.
26
Hot Corrosion of Silicon Carbide and Silicon Nitride 27
Under basic conditions, a thick product layer was formed and the NagSO4
was consumed slowly in the reaction. The ceramics oxidize at their surfaces and
the oxide dissolves into the sulfate melt. The Nag0 activity builds up at the
interface and a silicate layer is formed. As the silicate enriches in silica,
cristobalite is nucleated at the interface. After long exposures, the silicate
phase remains on top of the silica and the sulfate left is in small isolated drops
on top of the silicate. The melt contains a high concentration of silica initially
as indicated by the wash water analysis and it has also been well documented by
otherst5). The observations are consistent with the mechanism proposed by
Mayer and Riley(S) except that the reaction is much slower with Nags04 than
for NagCOS which they studied. Some protection is offered by this complex
product layer since greater degradation was observed after preoxidation of the
samples.
Under acidic conditions (1.5~10-~ atm SOS), poor wetting and little
reactivity with the salt were observed. However the oxidation was enhanced
even between the sulfate drops. This oxide growth between the drops was studied
in detail. The oxide formed was mostly vitreous. In these regions a very thin
layer rich in sodium is detected (~10 A thick) and sodium diffuses into the silica
formed. The silica layers formed under the sulfate droplets devitrified rapidly
into cristobalite by spherulitic crystallization or random globular formation.
They were thicker than the vitreous layer formed outside the drops except for
the C-side silicon carbide for which these thicknesses were similar. The
thicknesses of oxide formed under the drops tended to be similar for all three
surfaces (CVD silicon nitride, C-side silicon carbide and Si-side silicon carbide)
studied but they were still smaller than the product layers formed under basic
conditions. This is generally consistent with a mechanism determined by the
liquid phase and therefore independent of the nature of the substrate surface.
Indeed for the basic conditions and under the droplets for acidic conditions there
was no more differentiation in the behavior of C-side and Si-side silicon carbide.
A model was proposed for the formation of random cristobalite globules which
correlates with the results obtained for the hot corrosion of bulk silica. Selective
attack of the substrate occurs by transport through the liquid phase between the
spherulite fibrils.
The acidic hot corrosion which occurs between the sulfate droplets is
essentially enhanced oxidation. The kinetics of this hot corrosion of silicon
nitride are complicated apparently by the formation of oxynitride. Some evidence
of this formation was obtained in this research. It is not clear presently how this
thin layer formed during the oxidation slows down the oxidation so strongly. The
silicon carbide has different behavior for the carbon-side and for the silicon-side.
During the acidic hot corrosion they oxidize parabolically at different rates.
Kinetic data and apparent activation energies were obtained for the hot corrosion
of each side. In order to interpret these results, it was logical to modify a model
for the oxidation of silicon carbide. However, a satisfactory explanation of the
oxidation of silicon carbide had not been proposed.
Based on the oxidation results of others, a model was developed on the
premise that the parabolic oxidation is controlled by diffusion of oxygen through
the oxide layers to the substrate interface. Although vitreous silica is formed in
all cases, the different rates of oxygen transport are associated with different
oxygen deficient vitreous structures produced in the oxidation of silicon, silicon
carbide C-side and silicon carbide Si-side. Tentative mechanisms were proposed
for the oxidation of the two sides of silicon carbide in which the oxides are formed
Hot Corrosion of Silicon Carbide and Silicon Nitride 29
under different oxygen pressures and therefore are expected to have different
defect structures. The observed trends in the variation of the parabolic constants
for the acidic hot corrosion of these materials compared to their oxidation and
that of silicon are predicted qualitatively by these models. High apparent
activation energies such as those measured for hot corrosion are generally
consistent with the models although that for the silicon-side of silicon carbide
could not be estimated.
The behavior of engineering materials is strongly influenced by the
impurities they contain. MgO and Y3O3 tend to segregate toward the surface
during oxidation and hot corrosion. They lower the acidity of the sulfate under
acidic conditions so that wetting is improved and thicker oxide layers are formed
as the conditions are more basic than promoted by equilibrium with the
atmosphere. Conversely, the wetting is not as good under basic corrosion for
these materials as for the purer ones. The impurities set conditions intermediate
between those promoted by the equilibria with the atmospheres selected for the
basic and acidic corrosion experiments. Alumina impurities in the silicon carbide
do not segregate to the product layer formed on oxidation or hot corrosion. The
alumina entering the oxide layer appears to stabilize it against devitrification.
Great progress has been made in the understanding of the hot corrosion of
silicon nitride and silicon carbide. The corrosion varies with environmental
conditions and it was shown to be controlled by the activity of sodium. The
behavior observed correlates well with the results on the hot corrosion of silica.
In particular the major role played in the degradation of the scales by their
devitrification is emphasized by these results. It was shown also that
preferential attack occurred locally under acidic corrosion. Overall the hot
corrosion of these materials is well understood qualitatively. The atomic
mechanisms proposed for the oxidation and hot corrosion of silicon nitride and
silicon carbide suggest directions for further studies.
1. J.R. Blachere and P.S. Pettit “High Temperature Corrosion of Ceramics”
a) DOE Report ER45117-2, March 1988
b) DOE Report ER45117-1, June 1985
C) DOE Report ER10915-4, June 1984
2. a) H. Yakovitz and D.E. Newbury, SEM/1976/1, IIT Research Institute,

Chicago, IP, p. 151.
b) J.I. Goldstein et al.,References
Scanning Electron Microscopy and X-ray
Microanalysis, Plenum, No. 4, 1981, p. 354.
3. J.R. Blachere and D.F. Klimovich, J. Am. Ceram. Sot., 70 [ll] C324-C326
(1987).
4. L.E. Murr, Interfacial Phenomena in Metals and Alloys, Addison-Wesley,

Reading, Mass., 1976, p. 67.
5. N.S. Jacobson, J. Am. Ceram. Sot., 69 [l] 74-82 (1986).
6. M.I. Mayer and F.L. Riley, J. Mat. Sci., l3, (1978) 1319-1328.
30
Appendix A-Gaseous Corrosion of Silica
and Alumina in Sulfur Oxide Environments
H.R. Kim, J.R. Blachere and F.S. Pettit
Abstract:
A fused silica of high purity and alumina in single crystal and poly-
crystalline forms with different impurity levels xiere exposed to mixtures of
SOj- SO2~02 at 700°C and 1000°C. The silica and alumina are very resistant
to corrosion in these environments. No reaction was observed with the silica.
Sulfates formed on the aluminas particularly at low temperature and high SO3
pressures. The reaction occurred under less severe conditions and with greater
intensity with increasing impurity content, particularly Ca and Mg.
Supported by the U.S. Department of Energy under Agreement Number
DE-ACOZ-8 lER109 15-A000
Presented in part to the 1982 Fall meetings of American Ceramic
Society in Cambridge, Mass.
31
I. INTRODUCTION
Ceramics are considered actively for components of advanced power systems
operating at high temperatures. (1) This is because they have high refractori-
ness ( chemical stability and mechanical properties much improved thrcugh
extensive research over the past twenty years. They have been partially
successful in recent demonstration projects (*) and it may be anticipated that
the needed improvements in reliability of components will result from the
present emphasis on ceramic processing. Ceramics may be used at high tem-
peratures as monolithic parts rePlacing superalloys, as thermal barrier
coatings or as coatings protecting metallic alloys from environments which cause
them severe corrosion problems.
One should not assume however that ceramics are totally corrosion resistant.
Corrosion of refractories by oxide melts such as slags and glasses are costly
to the steel and glass industries. As a result research has been performed on
the corrosion of ceramics in deep melts, (3) but little is known about the
corrosion df these materials in gaseous atmospheres. The environments encountered
in advanced energy systems may contain corrosive products of combustion.
In addition, deposits of salts such as NaZSO4 mpy form on the components and
tend to enhance the gaseous corrosion. It has been shown recently that ceramics
such as AlZO3, Si3N4 and yttria-stabilized ZrO2 can react with molten deposits
of Na2S04 in SO3 gas.(b) The gaseous corrosion of ceramics has been studied
in this research (5,6) not only to establish the susceptibility of ceramics to
the corrosion by gaseous atmospheres representative of those encountered in
energy systems but also as the baseline for research on the hot corrosion of
crramics( ‘1 C i n presence of salt deposits). The corrosion of silica and alumina is
discussed here. These materials are of great importance since they perform
as protective scales on superalloys and SiOZ forms on silicon nitride and silicon
carbide in oxidizing environments.

Appendix A-Gaseous Corrosion 33
II. GASEOUS CORROSION OF CERAVICS
Gaseous corrosion of ceramics can occur by a variety of mechanisms
just as in the case of the gaseous corrosion of metals and alloys. Moreover,
the corrosion process are ultimately controlled by the nature of the products
that are formed as a result of reaction between the gas and the ceramic. In
metals and alloys reaction with the gas is almost always characterized by
metallic elements being oxidized via transfer of electron to reducible species
in the gas. The cations of the ceramics are already in oxidized states and the
following possible corrosion processes can be proposed. First, the cations
may be oxidized to higher states which in the case of oxygen as an oxidant
would result in oxygen being incorporated into the ceramic. Such a process
could result in the formation of new phases or merely changing of stoichrometry
of the phase being oxidized. This plays a major role with degradation of chrome-
based refractories with temperature and particularly atmosphere cyclings.
Second, the atoms or molecules of the gas may form more stable compounds with
the cations of the ceramic, for example silicon nitride reacts with oxygen to
form silicon dioxide and nitrogen. This type of reaction will depend on the
properties of the products formed upon the surface of the ceramic but effects
produced by the elements displaced by the oxygen can also be very important.
h third possibility involves the formation of volatile species such as the re-
duction of SiOz (s) to SiO (g) or the oxidation of CrpOj (s) to CrOj (g). Such
reactions probably would involve diffusion through gaseous boundary layers and
could proceed at rates fast compared to those of processes involving the form-
ation of condensed phases on the surface of the ceramics. Finally, molecules
in the gas such as SOj. CO2 or Hz0 may possess a significant affinity for the
cationic component or its oxide and react to form compounds such as sulfates
carbonates or hydrates.
The rates of such processes will be dependent upon transport through
corrosion products and rather complicated mechanisms may prevail.
The gaseous corrosion of ceramics is expected to be similar in some
respects to the corrosion of ceramics in melts which has been studied extensively
in the literature (gv 3l It is usually modelled as a dissolution process.
The kinetics involve the transport of reactants and products which often
determine the overall rate of the corrosion process. The corrosion reaction
rate depends on the relative corrosion rate of the various phases, the viscosity
of the liquid, the density gradient in the liquid, the concentration gradient
in the liquid, the porosity of the solid, the wettability of the solid phases
by the liquid, the boundary layer at the interface and the geometry of the system.
The grain size and the pore structure are major factors in controlling the rate
of dissolution. Fine grain materials dissolve faster than coarse grain materials.
The liquid may penetrate into the grain boundaries resulting in engulfment of the
(9,101
solid grains by the liquid .The influence of the microstructure on the
corrosion complicates greatly the interpretation of corrosion experiments so that
for fundamental studies Simple systems with simple geometries and microstructures
must be used. This is valid also for gaseous corrosion studies since except for
wetting effect, the grain size, the pore structure and the grain boundaries should
play a similar role as in liquid corrosion.
There is significant literature on the oxidation of silicon nitride and
silicon carbide (*l! the only literature on gaseous corrosion of simple oxides
appears to be on their reduction and volatilization particulary for silica.
Silica at high temperature under low oxygen pressures volatilizes forming
of SiO (g) by thermal decomposition and/or by reaction with hydrogen or CO (12*13!
For instance, significant weight losses were observed in wet hydrogen (51 at
1400oc. Alumina is much more resistant to conditions of this tvpe and tends
to lose impurities such as sodium at high temperatures. Significant losses of
alumina are only observed at very high temperatures, of the order of 19OOOC. (14)
Steam at low temperatures can form hydrates with silica resulting in degradation
of silicate refractories. (15) However, there are no reports on the Influence
of sulfur oxide gases on alumina and silica; this is discussed below.
III. EXPERINENTAL PROCEDURE
(1) Materials
They include four aluminas and a fused silica which were obtained from
commercial suppliers in relatively high densities and flat configurations.
A brief description is given in Table I. The aluminas were selected to investigate
the influence of microstructure and impurities on the corrosion. The single crystal
is a transparent substrate with (0001) orientation. The (L) material is
in the form of almost transparent washers; (S) and (M) are white substrates.
The more impure (S) alumina was included to show the gross effects of impurities.
The (S) and (MI aluminas have a much finer average grain size L-kin than (L)
I’ 1
alumina (19Urn). Their microstructure will be discussed later as needed.
(‘2) Experiments
Flat specimens ( z l~l.cm) were exposed to sulfur oxide eases at ml and lno~~c
in either as received, polished or relief polished conditions. Prior to exposure
they were cleaned thoroughly in acetone and alcohol and then heated in flowing
oxyeen atmosphere at IOOOoC for one hour. The ~02 and 02 mixtures were passed
over a platinized catalyst at the temperature of the experiments before contact
with the samples. The conditions were selected to have relevance to gas turbine
operations. The SO3 pressures were high enough in most instances to prevent the
decomposition of Na2Sfl4 in hot corrosion exueriments performed in the same
environments. The initial SO2 contents of the gases were “s~ally 1%. 0.1% and
TABLE I. MATERIALS UNDER STUDY
Designation Materials % Purity (Imp) Suppliers

3
Silica fused silica 2.20 99.89 (0.1 H20) Corning 7940(l)
8.C. sapphire 3.97 99.99 Saphikon(2)
L a-A1203* 3.97 99.79 (0.1 MSO. 0.11 Lucalox (3)

SiO2)
M u-A1203* 3.89 99.6 (0.17 t&O, AlStiS 772 (4)
a-A1203* 0.17 SiO2)
S 3.74 97.4 (0.75 ugo,
+l.6SiO2, O.lNa20 S-697 (5)
* Polycrystalline
(1) Corning Glass
(2) Tyco Laboratories
(3) General Electric, Lamp Div.
(4) 3M, Technical Ceramic Prod. Div.
(5) Saxonburg Ceramics. Inc.

0.01% with the balance 02 for a total pressure of one atmosphere and a flow race
of about lcm3/sec. The corresponding equilibrium SO3 pressures are shown in Table
II. The exposure times were usually one week but ware varied from one day to one
month. Some experiments were performed in sequential exposures to follow the
product development on the same area of the sample, examined also before exposure.
Since the materials are electrically non-conductive the samples were coated with
o
about 200A of carbon prior to any observation in the Scanning Electron Microscope
(SEM). This carbon was oxidized at 700°C in 02 before any additional exposure
of the samples in sequential experiments. It was checked by comparison of
duplicate samples exposed only once for the same total time that this procedure
did not interfere with the results of the sequential experiments.
The major investigative tools were the SM’s equipped with Energy
dispersive X-ray Spectrometers (EDS) for elemental analysis. X-ray diffraction
(XRD) was used to identify the products when they formed in sufficient quantities.
Some specimens were also analyzed by Electron Soectroscopy for Chemical Analysis
(ESCA) after exposure. The weight of the specimen was checked on an analytical
balance before and after the experiments.
IV. Results and Discussion
No significant weight change was measured in any of the experiments. It can
be shown, considering the size of the samples, that a weight change of D.Img/
cm2 of exposed surface corresponds to the formation of a continuous layer of
product about lum thick. The corrosion products were usually discrete and often
with little coverage of the surface, they could be detected using the SEM , EDS
and ESCA, but in no case were they equivalent to a continuous layer of the order
I urn thickness. These products are described in the following.
(1) Silica
Silica is very resistant to atta’ck by SD2- SO3 atmospheres at 700 or 1DDO’C.
No visible product formed and no sulfur was detected by EDS or ESCA after exposure
Tr\BLEII: Environmental Conditions in Experiments on

A1203 and Silica
Initial
Gas Pressure Pressure (Atm) at
(Atm) 7oooc 1ooooc
02 Sl. ‘Ll . -1.
SO2 10-2 2.9 x 10-3 8.7 x 1O-3
SO3 7.2 x 10-3 1.3 x 10-3
02 %I. Sl. %I.
SO2 10-3 2.9 x 10-4 8.7 x 10-4
SO3 7.2 x 1O-4 1.3 x 10-4
02 -1. Sl. %I.
SO2 10-4 2.9 x 10-5 8.7 x 10-5
SO3 7.2 x 10-5 1.3 x 10-5

up to one month. where is no evidence of sulfate formation and there is *o thermo-
dynamic data for silicon sulfate. The oxygen pressure of the experiments was too
high for the formation of silicon sulfide. No experiments in C3-SO3-So3
atmospheres were performed above 1OOOoC since Sulfate formation is less favorable
as the temperature increases. .No devitrificatioa of the silica glass was observed
after the exposures at 700 or 1OOOoC.
(2) Alumina
(a) General Results
The alumina samples were exposed to all three atmospheres at 700cC
and 1000°C in extensive experiments including sequential exposures to observe
the evolution of the products. The results are summarized in Table III. This
cable is based on three types of results: the observation of products in the
SEM. the detection of sulfur by EDS and the identification of sulfates
by ESCA. Since only very small amounts of products were formed in many cases and
the analytical techniques were close to their limit of detection, the formation
of products (P) is reported in the table only when it was indicated by at least
two techniques. If only one technique suRRested the presence of products, Su
was placed in the table.
It was established that the products were sulfates by the combined observations
of sulfur in the products by EDS and the identification of the valency of this
sulfur as S-6 as in sulfates by ESCA. Moreover, on the purer aluminas the
1
elements de:ec:cd in the products were only aluminum and sulfur. Extraction replicas
on exposed (L) samples showed also the same combination, thus leading to the
conclusion that aluminum sulfate was formed on the purer aluminas, Mg and Ca
sulfates were identified in a similar manner on the more impure samples.
Table111 clearly shows that there is an increasing tendency to form
sulfates with decreasing temperature, with increasing impurity contents and
with increasinR SO3 pressure in the atmosphere. These general trends are
TABLE III
Results of Alumina Corrosion Experiment in SO3 Gas at 700° add 1000°C for One Week
Tempe'rature SO3 pressure (atm.) Single Crystal Polycrystalline Aluminas

(OC) in approx. 1 atm 02 Alumina (L)99.8% M)99.5% S)97.4%"
700 7.2 x 10-3 P P P P
7.2 x IF4 P P P 1'
7.2 x IO-' N P P P
1.3 x 10-3 N P - P
1.3 x 10-4 N SU P P
1.3 x 10-5 N N SU SU
P: Sulfur containing products present

N: No sulfur containing products
su: Presence of sulfur suggested but not confirmed
* Purity
consistent with thermodynamic analysis.
(b) Thermodynamics of sulfate formation
Considering the reaction:
Al*03+ 3SO3(g)= Al*(S04)3
at equilibrium one has:

g_ a Al2 W4)3
a Al203 (pso3j3
and the SO3 pressure for the equilibrium between hl2(SO4) 3 can be calculated.
The stability diagrams of Figure 1 for A1303 in SO3-SO3 atmospheres were
generated bv considering the appropriate equilibrium reactions assuming
unit activity for the solid phases. The test conditions and the relevant
boundaries for the HgO- Mg SO4 and CaO- CaSO4 equilibria are shown on the
diagrams.
The oxide-sulfate equilibrium boundaries shift to lower SO3 pressures ac
lower temperatures. Furthermore, for a given initial mixture of 02- SO3 the
eqdlibfiumSO3 pressure increases with decreasing temperature. Therefore,
sulfates have a greater tendency to form at lov temperatures as shown by comparison
of Figures la and 16 and observed experimentally: It is also obvious from Figure 1
that the SO3 pressures required for sulfate formations increase from CaSO4
CO ?IgSO4 and finally co A12(SO4; 3 so thar under similar activity conditions
CaSC4 and YgSO4 will form more readily than Al?(SO4)3 this explains the increase
in product formation with increasing contents of CaO and YgO.
*Preliminary results obtained by Raman spectroscopy also indicated that
sulfates formed on S alumina (16)

a 0
cao C&m*
4- WN Me
log POY
-lo - A12(so4
Al203
-1a -
A12S3
-1(1 -10 4 0 I
log Psoa
b Ow
4-
log Por
-10 *
log PSOl
Figure 1. Stability diagrams for A1203 in SO@03 atmospheres at: (a)
700°C (b) 1OOOoC. The experimental conditions (+) and the
boundaries for CaO and MgO are also shown.
Cc) Single Crystal
Very little reaction of the single crystal alumina was observed during
exposure to any of the gas mixtures at 700°C and IOCWC. A few thin tetragons
about 0.5 pa on a side were detected after an exposure of one week to the highest
SO3 pressures of 7000C (Table III). For the same time of exposure to the inter-
mediate pressure of SC3 at 70C°C no products could be found but sulfur was
detected by EDS and sulfates were indicated by ESCA analysis. NO sulfur was
found after exposure to the lower SD3 pressure at 700°C nor for any of the
gas compositions at IOOOoC. These results show that the single crystal material
was affected by the environment when the experimental conditions were most
favorable for reaction. The stability diagrams of Figure 1 show that Al2(SC4)3
should not be formed in any of the experiments. These diagrams have been
constructed however for unit activity of the phases. Solid solutions (e.g.
Al2(SO4)3 at less than unit activity dissolved in Al2O3) or other phases not
considered in these diagrams may be formed.
(d) Polycrystalline Aluminas
The sintering aids added to the (L) material degrade its purity but it is
anticipated that the MgO isilithin its solubility limits in A120,.
No magnesium spine1 was found by optical metallography or the electron microprobe.
The role of Sin2 is less clear and it may have enhanced solubility in presence
of MgD.(L6) If its solubility is exceeded, it would be expected to segregate
to the grain boundaries, but this has not been observed. MgO apparently does not
segregate greatly at grain boundaries but Ca does even though its average concentrat_
ion in this lnrterial is very srr.alL. (17)

More products were evident on the (L) alumina compared to the single
crystal material. Products were evident after exposures of one week to all of
the three gas mixtures at 7OO’C. As shown in Figure 2 the amount of products
increased with increasing SO3 pressure. Corrosion products were also observed
at the higher SO3 pressure at 1OOOnC (Table III), but in less quantity than at
7oonc. Exposure to the intermediate SO3 pressure at 1OOW’C did not produce any
observable products but sulfates were identified on specimens by EDS and ESCA.
The products formed on the (L) alumina did vary for apparently, the
same exposure conditions. Two types of morphologies were detected which are
shown in Figure 2 and 3 for as received (L) altiina. One is small cuboids.
The other contains many cuboids but also other shapes are evident such as laths
and films. The prodlicts were concentrated on certain grain faces of the same
grain while other faces of these grains exhibited no products indicating a
strong influence of crystallographic orientation of the grain. This nay be
due in part to surface segregation of impurities but in no case could any
impurities such as Ca or Mg be detected on these surfaces or in the products.
?iore corrosion products formed on the aluminas
L. M, S than on the single crystal. These materials are all polycrystalline and of
greater impurity content than the single crystal. Products were formed on all
polycrystalline materials under all conditions at 700°C and at least under the
higher SO3 pressure at 1000%. In general. the products formed more readily
(Table III) and were larger and in greater quantity (Figure 4) with increasing
impurity content. The samples in Figure 4 were relief polished and exposed in
tha same axpari-nt (7 x 10-3atm.S03at 700°C for one week). It shows also that
the products form preferentially at grain boundaries.

Figure 2. (L) alumina surfaces (A) as received, B,C,D, exposed for one
week at 700oC to 7 x 10-4 atm. 503 and 7 x 10-3 atm. 503
respectively. More products form at higher P503 (5EM)
Figure 3. L alumina surfaces exposed to 7 x 10-3 atm. SQJ at 700oC

for one week. A shows various morphologies of reaction
products and 8 only a few discrete ones. (SEM).
Figure 4. Relief polished polycrystalline A1203 surfaces exposed to 7 x

10-3 atm. 803 at 700oC for one week (a) L , (b) M, (c)
another M-type (d) 8 aluminas. Products increase in size and
quantity with decreasing A1203 purity. They form
preferentially along grain boundaries. (8EM).
Two types of products were found on the (M) alumina, exposed to 7 ~10~~
atm SO3 for one week at 700°c. one was larger and was a magnesium sulfate
with no ca1ci.m in it and the other smaller in size was a CaSOh with little
or no magnesium.
The lowest purfty (S) alumina had the mosf product in both number and
size after exposure. EDS analysis of the products Indicated S, Al, Si, Mg,
Ca. They appear to be mainly magnesium sulfate, calcium sulfate and their
solid solutions. The larger products were mostly MgSDq which formed on impurity
magnesium silicate grains identified by EDS before exposure. The impurity
phase( s’) in (S) occur as large grains of the size of the alumina grains
and along grain boundaries. Some smaller products, mostly CaSO4, appeared to
be on the grain boundaries. (Fi_eure 51. As indicated by analysis of the large
grains of products and of products extracted by replica techniques. silicon
and aluminum are included in the products. It is believed that the silicon
is present as silica formed by reaction of SO3 with the silicate impurities
while Al is probably dissolved in the sulfates. In the (S) alumina, this
aluminum is expected to come mostly from the impurity phases Since talc and clay
are used as sintering aids.
The evolution of the corrosion products as a function of time is shown
in Figure 6, for the same area of a relief-polished sample (S) after a sequence
of exposures from three days to 30 days. The products nucleate on the impurity
phases which are darker on the initial micrograph due to a weak acomic number
contrast and the relief polishing. The products grow wirh coalescence and
coarsening occurring ddring the latter stages. The product marked by the arrow
is a calcium-magnesium sulfate of unknown aluminum content.

Figure 5. Relief-polished 5 alumina exposed to 7 x 10-3 atm. 503 at

700oC for one week. EDS spectra (1) whole micrograph, 2,3,4
areas indicated by numbers. (5EM).
Figure 6(a). Same area of relief-polished (S) alumina exposed to 7 x 10-3

atm. SQ3 at 700oC. (A) before exposure, (6) for 3 days, (C)
for 10 days, (D) for 30 days. (SEM).
~
tD...
QJ CJ
..:3
~'C
.-£
0.
00
.cVI QJ
0...
~ ~
/;;,~
0 ~
CJ ~ .
VI
E ~QJ
'C:3
0 QJ '6'
~ ~ 0.
QJ..~
..~QJ
~ E
..r/)0.
~0.U
QJ~r5
..~
~.,;<
o.~..
~:3°
UVl
~o~
l1Jo.o
<~~..QJ ..
'C0.~
QJ ~...
~U~
~ ~bD
E ~C::
r/) <.- ~
0...0
~~/;;,
:3
~
II
..
~
~
V. GENERAL DISCUSSION
‘Ihe results that have been obtained in the present studies show that
silica and alumina are considerably more resistant to sulfur-oxygen gas mixture
than metallic alloys. Nevertheless. reaction of such materials do occur when
the thermodynamic and kinetic conditions are appropriate. In the present studies
the experimental conditions were such that no reactions of the gas mixtures
with silica were possible. In the case of alumina, however, reactions did
occur and these reactions were significantly affected by the purity of the
specimens.
On the aluminas it appears that a layer of sulfate is first absorbed
on the sample building perhaps to thicknesses giester than a monolayer. The
reactions are possible by solid solutions of the sulfate probably
with alumina in the case of the single crystal and with the impurities in
the other materials. Calcium sulfate and magnesium sulfate form at lower PSO3
than aluminum sulfate so that their solid solutions maybe anticipated to form
at lower PSO3. The formation of intermediate compounds cannot be ruled out at
this point although no such sulfates are known with Ca and Mg. Intermediate
compounds occur in the sulfation of K20.3 A1203 in which above 300% a mixed
sulfdte K~u(Su4)3 is formed. (lg).Sulfates were produced on the impurity phases
and the grain boundaries as documented for the more impure aluminas; however,
the control of the sulfate formation by the impurities appears to play a major
role in all the polycrystalline aluminas studied. This explains at least in
part the prominent role of the grain boundaries in this corrosion. The
impurities may not only increase the driving force for sulfate formation but
also provide nuclei for the growth of sulfates. A major point is that when
reaction products involving the impurities are formed, due to their greater
thermodyhamic stability, reaction of the alumina itself becomes possible.

The observation that corrosion of alumina is dependent upon impurity

(19)
content is consistent with data in the literature for reaction of Sic and Si3N4
with oxygen where impurities such as Mg and Ca have been found to accumulate
in the oxidation products and thereby significantly modify the protective
properties of the scales that are formed on such materials.
In view of the results that have been obtained with the alumina samples,
deposits of Na7S04 can be expected to significantly influence the corrosion of
alumina (hot corrosion) because these products should have significant solubilities
in NazS.04. This effect will be especially pronounced when temperatures are
such that the Na7304 is liquid. Moreover, liquids can be expected to be formed
below the melting ooint of Na7SOh due to the formation of phases involving
solution of Na7S04 and impurity phases. This research is continuing with major
emphasis on the hot corrosion of ceramics.
IV CONCLUSIONS
(1) Silica and alumina are very resistant to corrosion by SO3 gases.
(2) No interaction was found between SO3 and fused silica.
(3) Small amounts of products were formed on the aluminas.
(4) The products identified as sulfates formed only under the more favorable
conditions on the high purity sapphire. They formed in greater quantities
and at lower PSC3 on the more impure polycrystalline materials.
(5) The general trends of the sulfate formation are in qualitative agreement
with predictions from thermodynamics which are that the corrosion of alumina
in SO3 gases becomesmore favorable at higher SO3 pressures and lower
temperatures. It is suggested that the aluminum sulfate formed on the
Single crystal is stabilized by solid solution with alumina.
(6) The severity of the corrosion increases with the impurity content.
(7) The severity of the corrosion increases with time.
(8) In the polycrystalline aluminas the impurities dominated the corrosion

by formation of various solid-solutions of CaSO;- MgS04 and Alp (So,)-, _
The corrosion products formed preferentially on impurity phases and at grain
boundaries.
Acknowledgements:
The authors thank the 3M and Saxonburg Ceramics Companies for donating
materials, B. Draskovich, M. Zedar and L. Fisher for contributions to the
preparation of the manuscript.

References:
1. Anon., " Reliability of Ceramics for Heat Engine Applications".
National Materials Advisory Board, Commission in Sociotechnical Systems
NMAB-357.
2. A.F McLean, "Ceramic Technology for Automotive Turbines" ~~11. AIII. &ram.
Sot. 61 (8) 861-71 (1982)
3. A.R. Cooper, "Kinetics of Refractory Corrosion" Ceram. Eng. Sci. Proc.
2. 1963-89 (1981)
4. R.H. Barkalow and F.S. Pettit, "Mechanisms of Hot Corrosion Attack
of Ceramic Coating Materials" Conf. on Advanced Materials for Alternate
Fuel Capable Directly Fired Heat Engines, Castine. Maine, Aug. 1979.
5. H.R. Kim u Gaseous Corrosion of Oxide Ceramics" M.S. Thesis
University of Pittsburgh, 1983.
6. H.R. Kim, J.R. Blachire and F.S. Pettit "Gaseous Corrosion of Ceramics",
paper 11411.164th Meeting Electrochemical Society, Washington, D.C. Oct. 1983.
7. B. Draskovich, J.R. Blachere and F.S. Pettit, "Hot Corrosion of Silicon
Nitride and Silicon Carbide", Basic Sci. Div. Fall Neeting Am. Gram.
sot., Columbus, Ohio, 1983
8.a W.D. Kingery, H.K. Bowen and D.R. Uhlmann, "Introduction to Ceramics",
2nd edition, Wiley, 1976.
b. A.R. Cooper, Jr and W.D. Kingery, "Dissolution in Ceramic Systems:
I Molecular diffusion. Natural Convection and Forced Convection Studies
of Sapphire Dissolution in Calcium Aluminum Silicate", J. Am. Ceram.
Sot. 41 (1) 37-43 (1964)

c. L. Reed and L.R. Barrett, "The Slagging of Refractories
II lhe Kinetics of Corrosion", Trans. Brit. CSUII. SOC. 63 ~~‘h
534 (1964)
9. M. Millard. H. Wuttig and H.U. Anderson, "Influence of Grain Boundaries
on Liquid Corrosion of Al203 and N&l" paper No. 76-SII-82, An. Gram.
sot. May 1982
10. K.K. Kappmeier, ItThe Importance of Microstructural Considerations in
the Performance of Steel Plant Refractories" in Ceramic Microstructures,
R.M. Fulrath and J.A. Pask eds. Wiley, 1968, p.22
lla..D. Cubicciotti and K.H. Lau "Kinetics of Oxidation of Hot-Pressed
Silicon Nitride Containing Magnesia", J. Am. Gram. Sot. 61 (11-12)
512-17 (1978)
b. S.C. Singhal "Oxidation Kinetics of Hot Pressed SIC".
J. Mat. Sci l_l, 1246-53 (1976)
12. R.A. Gardner, "The Kinetics of Silica Reduction in Hydrogen"
J. Solid State Cliem9 336-44 (1974)
13. M.S. Crowley, "Hydrogen-Silica Reactions in Refractories"
Bull. Am. Ceram. Sot. 46 (7) 679-82 (1967)
14. L.J. Trostel, Jr, "Stability of Alumina and girconia in Hydrogen",
Am. Ceram. Sot. Bull 46 (12) 950-52 (1965).
15. D.E. Day and F.S. Gac. "Stability of Refractory Castables in Steam-
N2 and Steam-CO Atmospheres at 199OC, "Bull Amer. Ceram. Sot. 5f!
(7 1 644-48 (1977)
16. A. Negelberg, private communication.
17. I.K. Lloyd and H.K. Bowen "Iron Tracer Diffusion in Aluminum Oxide"
J. Amer. Ceram. Sot. 64 (12) 744-47 (1981)

18s. W.C. Johnson "Magnesium Distributions at Grain Boundaries in Sintered
A1203 Containing Mg A1204 Precipitates," J. Am. Gram. Sot. 61 (5-6)
234-237 (1978).
b. P.E.C. Franken and A.P. Cehring 'Grain Boundaries Analysis of ?!gO-Doped
A1203", J. Nat. Sci., l6, 384-88 (1981)

19. N.G. Krlshnan and R.W. Bartlett "Kinetics of Sulfation of Reduced Alunite"
Environ. Sci. Tech, ,1,923-929 (1973)

Appendix B- Hot Corrosion of Silica
M.G. Lawson, H.R. Kim, F.S. Pettit and J.R. Blachere
INTRODUCTION
There is a great emphasis on higher working temperatures for heat engines
which is dependent on advanced materials. As progress has been made in the
science and technology of ceramics leading to greater reliability of ceramic
parts, they are beginning to be incorporated in car engines and other high
temperature-high stress device&). Since ceramics are becoming serious
candidates for these applications, understanding their corrosion resistance under
environmental conditions related to these uses becomes important. While
ceramic compounds, in particular oxides are quite stable at high temperatures in
air, the stability of ceramics under various conditions which might be
encountered, e.g. in a heat engine, must be established. At high temperatures a
material may not be stable in presence of some gases, i.e. metals in oxygen, and
the degradation of a material by reaction with gases is called gaseous corrosion.
In combustion systems, various compounds such as NagC03, Nags04 may be
generated in addition to combustion gases t2). Below some temperatures these
compounds condense in the engine and often enhance the gaseous corrosion in a
complex process named ‘hot corrosion’. This type of degradation is well known in
superalloys but has not been studied in detail in ceramics.
Silica forms as a protective scale at high temperatures in oxidizing
atmospheres on silicon carbide and silicon nitride and on some coatings on
superalloys. Vitreous silica is protective in clean oxidizing environments because
it is continuous with some ability to self heal at higher temperatures. However
above llllO-12000C it is less protective as it begins to crystallize@). Although
growing scales may differ in properties from the same materials in bulk, it was
To be submitted to the Journal of the American Ceramic Society
58
Appendix B-Hot Corrosion of Silica 59
believed that valuable information could be obtained from a study of hot
corrosion of silica as a part of a systematic investigation of the hot corrosion of
ceramics (4). This paper covers the study performed under conditions relevant to
gas turbines, with deposits of sodium sulfates.
HOT CORROSION
Hot corrosion is dependent upon temperature and the gas atmosphere since
they affect the stability of the deposit and its reactivity with high temperature
materials. If the deposit is in equilibrium with the atmosphere, equation (1)
applies:
Na2S04 (1) = SO3 (g) + Na20 (1) (1)
Kl = PSO3. a(Na20)
The activity of Na20 and the pressure of SO3 in the melt are fixed by the
temperature and PSO3 in the atmosphere as defined by Kl. For pure Na2S04, its
activity can be taken as unity, and as a(Na20) increases, PSO3 decreases and
viceversa. An Na20 activity of 2~1O-l~ was calculated from equation (1) for the
equilibrium between the melt and the S03-containing atmosphere selected for
the hot corrosion experiments at 1OOOoC. The corresponding activity at 7OOoC is
lower. In presence of pure oxygen, the activity of Na20 in the melt is not
defined by reaction (l), it is expected to increase to values of the order of 10-6
to 10s4 since SO3 and SO2 are evolved from the Na2S04 to the gas. These
values are high and such Na2S04 melts are considered basic. However
equilibrium of the deposit with the atmosphere may be established only at the
beginning of the hot corrosion process and acid or basic behavior depends on the
specific reactions that occur between the sulfate and the ceramics.
EXPERIMENTAL PROCEDURE
Materials
The experiments reported here were all performed on a high purity fused
silica (Corning 7940). The major impurities are 1OOOppmof ‘water’ and lo-100
ppm of chlorine. All other impurities were in the ppm or subppm range. The
samples were cut to about 1x1 cm in area and polished prior to exposure.
Gaseous Corrosion
The samples were exposed in tube furnaces at 700 and 1000°C. SO3-03 gas
mixtures flowed through a platinized catalyst in the furnace before passing over
the samples. The initial SO3 contents of the gas were usually 1, 0.1 or O.Ol%,
balance oxygen at a total pressure of 1 atmosphere. The corresponding equilibrium
PS03 and PSO3 are given in table I. The gas flow rate was 1 cm3/s. The exposure
time was usually one week but varied from one day to one month. Weight changes
were checked on an analytical balance. The major tool was an SEM equipped
with X-ray spectrometers (Energy Dispersive, EDS and crystal spectrometers
WDS) for elemental microanalysis. These experiments were supplemented by
ESCA. (X-ray photoelectron spectroscopy).
Hot Corrosion
The polished specimens were placed on a hot plate and sprayed with an
aqueous solution of Na3S04 to form a continuous layer of sulfate usually with a
loading of 5mg/cm2. The samples were placed in a boat lined with platinum foil
and exposed in tube furnaces for times of 1 to 495 hours. Flow conditions and
catalysts were as described for gaseous corrosion. The extreme conditions of

Table I
SO3 Pressures for Corrosion Experiments
Temp. (OC) Initial (SO2) (1) S03(atm.) (1)
700 1 7x10-3 (2)

0.1 7x10-4
0.01 7x10-5
1000 1 1.5x10-3 (2)

0.1 1.5x10-4
0.01 1.5x10-5
(1) 96 S02-Balance 02 - total pressure 1 atm.
(2) Environments for Hot corrosion experiments.

table I were selected since they correspond to relatively acidic and basic
conditions. They are the initial mixture of 1% S02-balance oxygen (acidic) and
pure oxygen (basic) at 700 and 1000°C. Since Nags04 is not molten at 700°C, an
equimolar mixture of CoSO4 and Na2SO4 was used at that temperature. Cycling
experiments were also performed with or without water washing and reapplying
the salt usually on a 45 hours cycle.
The weight of the coupons was checked on an analytical balance with a
sensitivity of 0.1 mg before salt application and after exposure once the salt was
washed off in 10 cm3 of high purity water per cm2 of exposed area. The washing
lasted 20 minutes in an ultra sonic cleaner. The washwater of some of the
samples was analyzed by EDS in the SEM using a procedure adapted previously
(5~6). A single drop of the washwater was deposited on a beryllium coupon and
dried. The deposit left on the coupon was analyzed and the spectrum was
processed semi-quantitatively* after subtraction of a water spectrum obtained
under the same conditions and scaled to the same background. The specimens
were examined in the light microscope and the SEM before and after removal of
the salt, however the SEM with its X-ray spectrometers for microanalysis were
the major tools of this investigation. When possible, crystalline products were
identified by X-ray diffraction with a diffractometer. The wetting angle of the
salt on the substrates was measured on photographs of the droplet profiles taken
in the SEM at room temperature. Contact angles have been measured previously
by this method(7). The crystallized thicknesses were measured on SEM
micrographs of cross sections of the coupons.
* SSQ software by Tracer-Northern

RESULTS AND DISCUSSION
Gaseous corrosion
The gaseous corrosion(*) was studied for comparison with the hot
corrosion(g). No weight changes were detected, no products were visible in the
SEM and no sulfur was detected by EDS or ESCA on the surface of the specimens
after exposures up to one month under the conditions given in table I. While the
smallest weight change that could be measured (0.1mg/cm2) would require a
uniform deposit of the order of 1 urn, the SEM with careful examination and
ESCA are much more sensitive. An estimate of the detection limits of discrete
products in the SEM is
10m7 to lo-* g/cm2. Under these assumptions* sulfur in the products seen in the
SEM could be detected by EDS. ESCA is a surface technique, not a
microanalysis technique, and its detection limits were of the order of lo-* to
10mg g/cm2.
No evidence of degradation or formation of solid products containing sulfur
was found in these experiments. This might have been expected since there are
no reports of the existence of a silicon sulfate, and no silicon sulfide was
expected to form under the experimental conditions since the oxygen pressure
was always around one atmosphere. The formation of silicon sulfide has been
discussed recently(lOl. No experiments were performed above 1000°C because
sulfate deposition becomes less favorable at higher temperatures. No
devitrification of the silica was observed after any of these exposures at 700 and
1OOOoCin many experiments for times as long as 720 hours. Others have
reported that SO2 has a fluxing action on fused silica.
* A 1% coverage with islands 20nm thick containing 20% sulfur was assumed for
the calculation of the limits of detection using SEM and EDS. In ESCA, 0.1% of
a sampling depth of 5nm was assumed.
Wetting by sulfates
The wetting angles of the salts used in the hot corrosion experiments varied
with the environments. They are given in table II. The general wetting morphology
developed in a few minutes at the temperature of the experiments. Some evolution
occured with time with the wetting increasing from 1 hour to 24 hours. In general
the sulfate wetted the silica better under basic conditions than acidic conditions
under which discrete small droplets were formed. After 24 hours under acidic
conditions at 1OOOoCthe wetting angle was 24O on the droplets and 13O on an
exceptionally large drop about 0.02cm across. Under basic conditions at 1000°C
the salt wetted most of the coupon after 1 hour, and a wetting angle of 2O was
measured; after 24 hours the wetting was continuous.
In some cases, the salt formed drops of several sizes and different wetting
angles were observed for different size-range (table II). The wetting angle
measurements were very reproducible for given size range. The angles reported
are called wetting angles and not contact angles since the latter terminology
may imply that equilibrium was reached. It is clear from the increased wetting
with time and the influence of atmosphere on the wetting that reactions are
occuring between the sulfate and the silica surface, even under acidic conditions.
The wetting tends to increase with the affinity between the liquid and the
substrate surface and the greater wetting of the silica by the salt under basic
conditions reflects the higher affinity of the basic salt for the acidic silica.
It was established by careful microscopy of the same areas of samples
which were cycled, with washing off the salt and reapplication between cycles,
that the droplets did not reform preferentially on previous droplet sites. This
suggested that no preferential local attack would occur under acidic conditions.
This conclusion was substantiated by a long term experiment reported below.

Table II
Wetting Angles of Na2 SO4 on Silica (O)
7oooc (1) 1ooooc (2)

lhr. 24hr. lhr. 24hr.
Acidic 52 36-49 t3) 13-24 t3)
Basic - 26 t4) 2 0
(1) (Nag Co) SO4
(2) Na2 SO4
(3) Range of angles with drop sizes - smaller

angles for larger drops.
(4) Salt phase separated.

Hot Corrosion
Under acidic conditions, the salt did not wet well the silica and it formed
droplets. At 700°C the weight changes were small, for example, after removing
the deposit by washing there were losses of 0.1 mg/cmz after 1 hour and 1.1
mg/cm2 after 24 hours. No evidence of devitrification was apparent. Some
localized corrosion occurred under the droplets, as shown in figure 1. It is
suggested that this ‘wormy’ void texture formed by dissolution of sodium silicate
in water. This corrosion was very limited as indicated by the small weight
losses. The attack occurred preferentially under the salt at the perimeter of the
drops suggesting an interaction with the atmosphere. No spalling occurred after
495 hours of cyclic exposure of the silica. The sample was washed and salt
reapplied (5 mg/cm2) between cycles. The substrate had a strong texture after
exposure and a region 33 urn thick contained sodium. This attack over the whole
surface of the coupons is consistent with the separate observation that the
droplets did not reform at the same locations after reapplication of the salt.
Under acidic conditions, at 1000°C the wetting had improved and the silica
was devitrifying under the drops. After 24 hours the cristobalite formed
characteristic patterns intermediate between spherulitic and globular (figure 2).
Devitrification occurred also under very small drops and no evidence of attack
was detected outside the drops. The weight loss of 0.3 mg/cm2, was smaller
than at 700°C. This can be explained by the greater stability of the cristobalite
compared to the vitreous silica which will result in the formation of less sodium
silicate to dissolve in the salt. The cristobalite layer which did not spall
probably offered some general protection to the vitreous silica but it was not
established if preferential attack occured at the boundaries between the
spherulate fibrils as reported for silicon carbide under similar conditions(l2).

Figure I. Hot corrosion of fused silica under acidic conditions,

24 hoursat 700oC. SEM of sulfate drop area after
washing off sulfate showing wormy voids formed
preferentially at edge of drop.
Figure 2. Fused silica exposed for 24 hours under acidic conditions

showing coarsened spherulites under sulfate drops (after
washing).
Under basic conditions at 7OOoCthe salt tended to decompose with
extensive formation of cobalt oxide which complicated the interpretation of the
data, particularly weight change. Under the salt drops there was limited
formation of cristobalite spherulites mixed with some random globules after 24
hour exposure and some localized cracking (figure 3) but there was no evidence
of spalling or extensive attack of the substrate.
The most extensive degradation of vitreous silica occurred under basic
conditions at 1OOOoC. After 1 hour the salt wetted the coupon almost
completely and a layer of cristobalite spherulites had formed under the salt. As
shown in figure 4, the fibrils of the spherulites had already broken down
significantly into arrays of globules. The spherulites had grown to impingement
with radii of the order of 30 pm, underlining the high velocity of the surface
crystallization. After 24 hours nothing was left of the spherulitic surface
morphology as tridymite had formed at the silica-salt interface. Cristobalite
separates the tridymite from the vitreous silica. After 212 hours the tridymite
near the surface had coarsened into laths about 15 pm wide and over 60 pm long.
The crystalline layer spalled extensively. The washwater analysis indicated that
significant silicon was water soluble with the sulfate after 1 hour, however this
was no longer found after 10 or 100 hours. The sulfur was never depleted as the
Na/S ratios in the water remained of the order of 2 (t 0.4). This suggests that
less reaction occurred after a continuous crystalline layer was formed and that
at the longer times sodium silicates with low solubility in the sulfate at 1OOOoC
formed between the crystalline silica and the sulfate.
It became clear, as the previous results were obtained, that the major
mode of degradation of vitreous silica under hot corrosion conditions was due to
the crystallization and associated spailing. Qualitatively the extent of the
degradation at constant time increased in the order:

Figure 3. Fused silica exposed under basic conditions at 700oC for 24

hours. Limited nucleation occured in small portions of drop
areas.
Appendix S-Hot Corrosion of Silica 71
Figure 4. Fused silica exposed to basic conditions at lOOOoC for 1 hour.

The fibrils of the spherulites have broken up into globular
arrays.
Acidic 7000 < Basic 700° < Acidic 1000° < Basic 1000°C
The degradation increases with increasing Nag0 activity in the salt and decreasing
equilibrium PSO3 under these conditions. Since the hot corrosion results
emphasized the degradation associated with the devitrification, quantitative
data was obtained on the crystallization.
Kinetics of Crystallization
The thickness of crystallized silica was measured on fused silica under
basic hot corrosion conditions at 1000°C for times up to 300 hours. The
measurements were made on the unspalled region of fractured specimens. The
isothermal kinetics are shown in figure 5. They are for the propagation of the
cristobalite-glass interface. The plot is parabolic and the incubation time is
under an hour. A long term experiment was performed with exposure cycles of
46 hours. In between cycles, the salt was washed off and new salt was reapplied
as described previously. Since this was done at room temperature, the samples
were thermally cycled. Extensive spalling of the crystalline layer occured and
the thickness of the remaining silica glass was measured where the fracture had
propagated at the interface. The plot of the thickness crystallized for 6 cycles
is given in figure 6. The plot is linear instead of parabolic and the total thickness
crystallized (about 1000 urn) is over twice that without cycling. In a separate
experiment of 100 hours total duration, a sample was cycled without reapplication
of salt by cooling to room temperature after 1 hour and 10 hours of exposure. It
had the same layered structure as the other samples after the long exposures,
cycled or uncycled, (figure 7). One can identify a thin sodium silicate layer, a
tridymite layer and a cristobalite layer over the glass. The crystallized layer
had a thickness of the order of 400 urn which is also about twice the thickness
I/2 l/2
TIME (Hours)
Figure 5. Kinetics of cr,stallization of silica glau during basic hot

corrosion at lODOW in Oxygen.
900
800
600
500
400
300
200
IOC
2 3 Y 5 6
0 I
CYCLES (48 Hour, Resa It)
Figure 6. Kinetics of crystallization of silica glass exposed to basic hot

corrosion at 1OOOoC in Oxygen, with 45 hours cycles and
reapplication of sulfate. Note linear kinetics.
+"..'0
01 ~ C
1n~0I
C tI ~
0 ~
.-In -
.-~
+" e
'0>.>.
Col'O .-
0-
tI ...
M-ln
tI In
.-~ 01
In~~-
.aoIC+"ba
.-~ ..
-tI -
C tI .-.c
.->.- +"
'0 tI .;n .c
~- e +"
In 01 .-
0 e ~ ~
0. .-
><..'0...
..~ 0 tI
-.Clnol
0I...~+"
tI C
.-.c ~ 0
-+"w
tI
1n~>.C ~ .-
'0
~~.Cln
In ~ E-o .-
~O .c
IntI
OO~:c
CO-~
O ba
c ~
... 0 ...
..+" .-
4i...0-
InUO~
1n0 eO
Inoln...
E o ~ .~
U °.C..
tI
~
GI
..
~
~
crystallized for the same time under isothermal conditions. The larger rates of
crystallization and the change in the type of kinetics under cycling show the
strong influence of the salt on the crystallization but the increased damage under
the cycling experiments is due to the stresses generated during the cycling and
not to the replenishing of the salt.
General Discussion
The wetting was greater under basic than under acidic conditions. The
wetting angle decreased in the sequence:
Acidic 700 > Basic 700 > Acidic 1000 > Basic 1000°C
The increased wetting at higher temperature (lOOO°C) under acidic conditions is
due to a decreased acidity of the salt at higher temperature as predicted from
equation (1). The negative free energy of a reaction between a substrate and a
liquid enters in the interfacial energy balance of a sessile drop so as to increase
the wetting tendency of the solid by the liquid(ll). Therefore the wetting
behavior of the salt and in particular the influence of the atmosphere on it
indicate a tendency towards basic fluxing according to a reaction of the form:
xSi02 + yNa20 = yNa2OxSiO2 (2)
I(2 = a&/ (aSiO2)x. (aNa2O)y
Assuming that the silicate activity asil=l, y=l and x=2, the equilibrium
constant at 1000°C gives aNa20=9.2xlO- 11. Therefore reaction 2 should proceed
to the right when aNa is greater than this value. The experimental conditions
for equilibrium with 1.5x lo-5 atm So2 which correspond to aNa20=2xlO-15
should therefore not be .sufficient to form the silicate. A similar analysis was
performed by Jacobson et. al for the corrosion of silicon carbide(13). However

the wetting behavior, the crystallization and the reaction of the sulfate with the
silica indicate that the equilibrium is shifted to much lower values of aNa so
that at least locally all conditions for this work might have been sufficiently
basic to permit for reaction 2 to proceed to the right. This can be explained at
least partially by aSiO2 > 1 since the glass is not an equilibrium phase and the
OH and Cl impurities decrease its stability. A discrete silicate phase was observed
under long term basic exposures at 1000°C. It is only expected to form under
strongly basic conditions where the silicate activity might get close to 1. In the
less basic experiments and early in the basic experiments the silicate formed is
dissolved in the sulfate and has activities much less than 1.
Some limited reaction occurs between the fused silica and the sodium sulfate,
even under the less basic conditions at 700°C (acidic 700°C in table I.) Under
these conditions, with cyclic replenishing of the salt, significant hot corrosion
accumulated over long term exposures. The quantitative interpretation of the
hot corrosion on the basis of reaction (2) was limited by the presence of cobalt in
the melt at 700°C and the overwhelming effect of the crystallization under all
conditions except the less basic (acidic 7OOOC). Neglecting these complications,
the major interactions between the salt and the silica occur at their interface.
They are the dissolution in the salt of the silicate formed via reaction (2) in the
salt and diffusion of sodium into the vitreous silica. Accordingly, under the more
basic conditions, the Si content of the washwater increases early during the
exposure but decreases later as the silica at the interface crystallizes; after a
long time a silicate layer is detected on top of the crystallized silica (after
washing). Under the less basic condition, this reaction is localized leading to the
observed ‘wormy’ texture. It is proposed that the reaction begins at impurity
clusters in the fused silica. The OH and Cl impurities are connected only to one
silicon and they may generate microregions with more open structures more
susceptible to attack under reaction (2). Since for the less basic condition (acidic
7OOoC), the driving force for the reaction is small, the influence of these defects
is emphasized and it will be suppressed by increases in concentration of the gaseous
products in the salt. As a result it occurs more readily near the edge of the salt
drops (figure 1) where the SO3 produced is released more easily to the atmosphere.
Under more basic conditions these defects may play a role, but it is not as obvious
in the surface morphology observed after corrosion. Others have reported no
interaction of Na2SOq with fused silica under acidic conditions(14). This was not
the case in any of the present experiments which included lower temperatures.
Some of that difference in behavior may be due to the influence of the
impurities discussed above. Silica scales formed in combustion environment are
expected to contain OH impurities, so that the samples studied here are
expected to be more representative of industrial applications.
Sodium diffuses rapidly into fused silica with reported diffusion
coefficients (DNa) of the order of 10s6 cm2/s at 1000°C(15). In type III silica
glass such as used in the present experiments, the OH’s may slow down the
sodium. In a recent discussion it was pointed out that impurity and structure
seem to affect DNa( 13). Sodium can be incorporated in the glass as a network
modifier(17) through the reaction:
Na20 + Si-0-Si = 2Si-O- + 2Na+ (3)
(CSiO)2. (CNa)2
g3 =_________________
Cst. aNa
in which an oxygen bridge (Si-0-Si) has been broken into two single bonded
oxygens to incorporate the oxygen into the network. The sodium ions, as
network modifiers, are located in holes of the structure near the single bonded
oxygens. The oxygen bridges most likely to be broken are the more strained in
the structure(l6), and their concentration is Cst. Although the sodium ions are
associated with single bonded oxygens they are mobile in the structure. These
considerations are important in the oxidation of silica formers and similar
concepts have been used by Schaeffer to discuss the influence of water on the
oxidation of silicon(lg). The penetration of sodium is expected to increase
rapidly with the activity of Nat0 in the melt as it increases the driving force
and probably the mobility of Na+ for the low concentrations of Na+. Diffusion of
sodium in silicate glass has been discussed in terms of a vacancy diffusion(20).
Initially it must enter the glass structure by reaction 3 which could be rate
controlling. If this reaction is slow it is likely that ion exchange between protons
of the hydroxyls present in the glass and sodium might play a role. However, the
situation is modified by the crystallization of the silica glass. Diffusion of
sodium through cristobalite is very difficult since it does not contain the
channels of vitreous silica or quartz. Diffusion is probably easier along the grain
boundaries.
Since it breaks up the network, reaction 3 is expected to promote the
crystallization of silica and the required rearrangement of the network to 6-
member rings. In this research, the silica did not crystallize under the lower
Na30 activity (acidic, 700°C) but it did at the same temperature under the more
basic condition. Qualitatively, the extent of the crystallization increased with
the Nag0 activity as expected from the previous discussion and reaction (3).
From reaction 3, the concentration of single bonded oxygens is proportional to
(aNaaO)*. The strong influence of sodium on the crystallization of silica is well
know in ceramics(21), here it has been shown that under basic hot corrosion
conditions at 1OOOoC the crystallization proceeds at a rate of the same order as
that observed above 1400°C on the same materials in oxygen or an enhancement

of many orders of magnitude due to the sodium. This is in line with previous
results on the crystallization of sodium silicate glasses(21). Much greater rates
of crystallization, of the order of 0.1 mm/min. have been observed at 1000°C in
soda silica glasses with Na20 contents of the order of 1%. This shows that only
very low concentrations of sodium, in the ppm range are necessary to obtain the
crystallization rates of the order of 0.01 -0.1 um/min. observed here.
Under the more basic conditions, at 1000°C, the greater Na20 activities
lead to dissolution in the salt as well as sodium penetration in the vitreous silica,
however this is short lived because of the onset of devitrification. The
devitrification and the resulting spalling as the specimens are cooled through the
a-g cristobalite and the a-B-y tridymite inversions as well as the differential
contraction of the crystalline phases on cooling are responsible for most of the
degradation of fused silica under basic conditions. The increased degradation
under cycling is due to the cracking and spalling of the crystalline scale and
penetration of the salt through the cracks. In line with this conclusion, little
degradation was observed with no significant increase under cycling for the less
basic condition at 7OOoC (acidic 7OOoC) for which the silica did not crystallize.
Since it was responsible for most of the observed degradation the crystallization
of silica under hot corrosion conditions will be discussed extensively.
The crystalline phase formed initially was always cristobalite which forms
instead of tridymite since a lower energy path is available because of the greater
similarity between the structures of high cristobalite and vitreous silica(221. The
cristobalite was globular in morphology (figure 2-4). Under the more basic
conditions cristobalite globules were aligned in radial arrays related to their
spherulitic formation. Under less basic conditions the spherulites did not
nucleate under all the salt drops or over the whole areas covered by the drops
and for less basic conditions yet (acidic 700°C) no crystallization was observed
even after nearly 500 hours. Under the intermediate conditions, cristobalite
globules formed randomly under the drops where no spherulites formed or in
between spherulites. This suggests that the crystallization was initiated by two
different mechanisms depending on the Na20 activity. In both cases the sulfate
reacts with the silica by basic fluxing as discussed earlier and shown in reaction
2. The relationship with SO3 pressure is obtained by combining reactions (1) and
(2). Under the greater Na20 activities the cristobalite nucleated at the sulfate-
glass interface and formed spherulites, with the well-known radial morphology of
their fibrils, however the previous results suggest that a threshold concentration
must be reached for the nucleation of the spherulites at the interface. These
fibrils coarsen and break down due to interfacial instability giving the radial
arrangements of globules. The spherulites grow in a thin surface layer in which
high Na20 activities promote the crystallization of silica by breaking down the
network, (reaction 3) and silica is transported rapidly through the liquid phase.
On the other hand under intermediate basic conditions, the surface
reaction of the sulfate with the silica is less extensive, the silicate ions dissolved
in the sulfate diffuse into the sulfate away from the interface on which no
crystallization is initiated. At the same time it is likely that the solubility of
silicon, probably as silicate, in the sulfate decreases as the activity of Na20 is
decreased as reported by Kim(22) in the same range of activities. Thus as the
reaction proceeds at the interface, gradients of Na20 activity and SO3 pressure
are set up across the salt. The solubility of the silicate decreases as it diffuses
away from the interface leading to supersaturation and homogeneous
precipitation of silica in the salt which generates the random globules of
cristobalite (figure 3).

Discussion of Crystallization Kinetics
The crystallization of silica which has rapid parabolic kinetics under the
more basic isothermal conditions is not controlled by the transport of reactants
or products through the crystalline layer, since the crystal has the same
composition as the glass. As already discussed qualitatively, the very large rates
of crystallization are associated with high Na30 activities. The sodium breaks
oxygen bridges according to reaction (3). It is proposed that the crystallization
begins when a threshold Na30 activity is present at the silica glass-sulfate
interface generating sufficient concentrations of oxygen bridges to allow the
local rearrangement of the network to the crystalline form. This could occur by
a mechanism similar to that proposed by Fratello et. a&18) for the high pressure
transformation of fused silica to quartz. The sodium in Si-0-Na groups would
play a role similar to that of hydrogen in SiOH bonds, although the ONa bonds
are more ionic than the OH bonds and as a result they were considered ionized in
reaction (3). Since it is less tightly bound to the oxygen and because of size
considerations, the Na+ is more mobile than the protons(24). The active defects
sre still the non-bridging oxygens which are very mobile in combination with
either thermally created single bonded oxygens, SiOH or other SiO- Na+ groups.
The single bonded oxygens can attack oxygen bridges by simultaneous bond
formation and breaking, and reshape the network into six member rings. After
nucleation the crystallization front propagates at rates many orders of
magnitude greater than in the pure system. From the present results this is due
to the sodium which catalyses the crystallization. It has to be present at the
cristobalite-glass interface, although it could not be detected with the electron
microprobe, except in special cases. Only very small quantities of sodium are
required for the previous mechanism which can rapidly propagate a crystal ledge
into the glass.

In such an interfacial reaction a sodium can rapidly rearrange many bonds and
the controlling step is the break up of a strained Si-0 bond(l@. Adapting this
kinetic model the rate of advance of the interface (u) should be a function of the
stoichiometry of the silica and directly proportional to the concentration of
sodium which take the place of the hydroxyls in the model. The hydroxyls have a
mobility 3 to 4 orders of magnitude smaller than Nalz41. It has been suggested
previously that most hydroxyls in synthetic silicas are strongly bound and nearly
sessile(z51. The sodiums are highly mobile although in small concentrations.
The concentration of sodium ions is not constant in the moving interface.
During the induction period some sodium diffuses into the glass. Soon this
diffusion is essentially stopped by the formation of a continuous cristobalite
layer, as discussed earlier. In the proposed model, the sodium at the
cristobalite-glass interface is supplied by this original diffusion through the glass
and essentially trapped once a significant crystalline layer has formed.
Neglecting any diffusion through the crystalline layer and the solubility in the
cristobalite, one can approximate the concentration of Na in the glass at the
interface with the cristobalite as a spike of constant area, representing the
sodium injected initially into the glass, which spreads as a function of time
(figure 8). This spread which is due to the sodium diffusion into the glass is
slowed by the rejection of the sodium ahead of the crystallization front. Based
on this model, the concentration of sodium CNa at the interface is expected to
decrease with time in line with the observed decrease of the rate of
crystallization. Quantitatively the constant sodium in the spike can be written
as:
cNa.(Dt)+ = s
idym.
n
%I,0
Figure 8. Proposed mechanism of crystallization of silica during basic

hot corrosion. (1) Diffusion of Na into glass, nucleation of
cristobalite; (2) A continuous layer of cristobalite forms
rejecting Na; (3) with cristobalite growth the Na
concentration at the cristobalite-glass interface decreases by
diffusion into the glass.
in which D is the diffusion coefficient of Na+ into the glass and t is the time.
And the rate of crystallization u is
u = dx/ dt = A CNa+ = A S / (Dt)*
which integrates to x = B tt / Dt and if D is
independent of time, x takes the form:
x=kt*
which predicts the observed parabolic behavior. In the previous discussion, x is
the thickness of the crystallized layer while S, A, B, and k are constants at
constant temperature.
One could still assume that the sodium at the interface is supplied mostly
be diffusion through the crystalline layer by grain boundary diffusion however
the cristobalite changes grain size by over an order of magnitude and it transforms
to tridymite at the salt interface side without significant departures from the
parabolic behavior. Furthermore Na is not consumed in the crystallization reaction,
it is rejected into the glass at the cristobalite-glass interface and since it
catalyzes the crystallization one cannot expect a falling rate of crystallization
such as the observed parabolic behavior. Therefore it is concluded that the
mechanism proposed above provides a better explanation.
Hot Corrosion of Silica Pormers
The present results are in general agreement with the protective properties
of silica scales on ceramics (silicon carbide and silicon nitride) and coatings on
superalloys. The silica was not affected by the SO3-SO3 containing environments
of this study, however sodium plays a dramatic role under hot corrosion conditions.
The vitreous silica has a tendency to react with Na2S04 by basic fluxing,
particularly at low PSO3, but more importantly it is very sensitive to
crystallization which promotes spalling through temperature cycling. Under
these conditions silica scales should be strongly affected and fundamentally not
afford good protection to silica formers. The validity of these conclusions will
depend on the purity of the silica formed, the nature of the impurities associated
with the sealest and the rate of growth of the scale since thick scales are more
sensitive to spalling. It is likely that OH impurities promote the hot corrosion
and crystallization. Although only indirect indications are presented here, it is
expected from the role of OH in glasses( 18p2@. This is important for high
temperature structural applications since water is a major product of
combustion.
The results suggest also that some silicate scales might be more desirable
than vitreous silica if they could be used at high temperatures without
devitrification.
CONCLUSIONS
- No evidence of gaseous corrosion was obtained in atmospheres containing

initially up to 1% SO2 and balance oxygen at 700 and 1000°C for times up to 720
hours.
- No devitrification was observed under these conditions and similar

experiments in pure oxygen.
- The wetting of the silica by the sulfate varied significantly with the PSO3
and the Nap0 activity. It increased with aNa20. It was greater under basic than
under acidic conditions at both temperatures. Under basic conditions at 1000°C
the wetting was complete.
- In all cases there was a tendency towards basic fluxing which increased
with the activity of Na20.
- Some limited corrosion was observed under acidic conditions at 7000C.

At 1000°C under the same conditions, the silica devitrified and little reaction
was observed between the salt and the silica.
- The most extensive degradation of vitreous silica occurs by crystallization

and the associated spalling during temperature cycling. In general it increased
with the activity of Na20 and was very severe under basic conditions (pure oxygen
atmosphere) at 1000°C.
- The sodium accelerates (catalyzes) dramatically the devitrification of

silica and the rate of crystallization at 1000°C under basic conditions is of the
same order as that observed by others at 1400°C in air (without sodium).
- The kinetics of crystallization at 1OOOoC under basic conditions were

parabolic with a short incubation time. Cycling increased the damage and changed
the kinetics to linear. The extra damage was associated with the strains due to
cycling not with additional salt applications.
- The parabolic behavior observed under isothermal conditions can be

explained with a model in which the crystallization is controlled by the sodium at
the crystal-glass interface which diffuses into the glass prior to crystallization.
- The need for the development of vitreous coatings for the protection of
ceramics and metallic alloys is stressed.
Acknowledgements
The authors gratefully acknowledge the support of Basic Energy Science Division
of the Department of Energy for their financial support.
REFERENCES
1. a. M. Taguchi, Adv. Ceram. Mat. 2 [4] 754-762 (1987).
b. Heat Engine Ceramics, Bull. Am. Ceram. Sec. 64 [Z] 268-294 (1985).
2. N. Birks and G.H. Meier, Introduction to High Temperature Oxidation of

Metals, Arnold 1984, Chap. 8.
3. P.J. Jorgensen et. al, J. Am. Ceram. Sot. 42 (12) 613-616 (1959).
4. J.R. Blachere and F.S. Pettit, High Temperature Corrosion of Ceramics,

DOE Basic Energy Sciences DEFG OZ-84-ER45117.
5. B. Draskovich, Hot Corrosion of Silicon Mitride and Silicon Carbide, M.S.

Thesis, University of Pittsburgh (1984).
6. J.I. Goldstein et. al, Scanning Electron Microscopy and X-ray Microanalysis,
Plenum, (1981), p. 378.
7. L.E. Murr, Interfacial Phenomena in Metals and Alloys, Addison-Wesley,

(1975), p. 69.
8. H.R. Kim, Gaseous Corrosion of Oxide Ceramics M.S. Thesis, University of

Pittsburgh (1983).
9. M.G. Lawson, Hot Corrosion of Silica and Alumina, M.S. Thesis, University
of Pittsburgh (1987).
10. 0.0. Vander Biest et. al, J. Am. Ceram. Sot. 70 [7] 456-59 (1987).
11. I.A. Aksay et. ai, J. Phys. Chem. 78 [12] 1178-1183 (1978).
12. J.R. Blachere et. al, Paper #77-BEG-86P, Basic Science Div. Am. Ceram.
Sot., New Orleans (1986).
13. N.S. Jacobson, Workshop on Corrosion in Ceramics, Pennstate, Nov. 1987.
14. N.S. Jacobson et. al, Workshop on Corrosion in Ceramics, Pennstate, Nov.
1987.
15. G.H. Prischat, J. Am. Ceram. Sot 5_1[9] 528-530 (1968).
16. G.H. Prischat and W. Beier, J. of Non-CrystaLline Solids 7l, 77-85 (1985).
17. W.D. Kingery et. al, introduction to Ceramics, Wiley (1976) p. 103.
16. V.J. FrateIlo et. al, J. Appl. Phys. 51 [12] 6160-64 (1980).
19 H.A. Schaeffer, J. of Non-Crystalline Solids, 38 and 39 (1980), 545-550.
20. Z. Boksay: “Mass Transport in Non-Crystalline Solids”, in The Physics of

Non-Crystalline Solids, G.H. Frischat, Ed. Trans. Tech., (1977) p. 428.
21. H. Rawson, Inorganic Glass-Forming Systems, Academic Press (1967), p. 53.
22. Reference 17, p. 314.
23. G.M. Kim, M.S. Thesis, University of Pittsburgh, (198-).
24. R.H. Doremus, Glass Science, p. 168.
25. R.W. Lee and D.L. Fry, Phys. Chem. Glasses, 3 19 (1966).
26. Reference 24, p. 229.

Appendix C-Hot Corrosion of Alumina
M.G. Lawson, F.S. Pettit, and J.R. Blachere
INTRODUCTION
Failures due to corrosion limit the selection of materials in countless
operations. The corrosion of ceramic materials is a significant factor limiting
the design of new systems for coal gasification, coal liquification, energy
conversion, thermal storage, and the battery storage of power. (l) Advances in
the science and technology of processing, fabricating, and testing of brittle
materials have resulted in ceramic parts with mechanical properties acceptable
in a broad range of energy related applications.
Severe material degradation is common in environments where the combustion
of fuel occurs. As superalloys and complex cooling systems have been developed
the operating temperatures of turbine engines has been raised to increase their
efficiency. Many of the raw materials used in superalloys are expensive and can
be obtained from a limited number of foreign sources. Low heat rejection diesel
and gas turbine engines are in development which allow a substantial decrease in
the size of power source, primarily due to the elimination of bulky cooling
systems.(2T3) The continuing development of more efficient and compact engines
as well as the reduction of cost and dependence on strategic materials is dependent
on advances in materials technology. In addition to being inherently refractory
and resistant to corrosion, most ceramic materials are available from domestic
sources and are much less expensive than the elements used in superalloys.
Although ceramics tend to be refractory, the corrosion of ceramics occurs
at an appreciable rate in many systems. This is reflected in cost of refractoris
used by the steel and glass industries. The research resulting from the
89
widespread use of ceramics as vessels to hold and direct the flow of molten metal
and glass has led to an understanding of the corrosion of ceramics in the presence
of a melt.
Alumina is an excellent bulk refractory which is used in impure and fairly
pure forms. It is also generated as a scale which is protective of coatings on
superalloys. Its corrosion resistance to gases at high temperatures, in presence
of deposits which may enhance the corrosion was studied in this research. Gaseous
corrosion of alumina was discussed in appendix A. The conditions of this study
are particularly relevant to turbine applications since the gases were generated
by SO2-02 mixtures and the deposits were Na2 S04.
In a turbine engine at 1000°C, gases have oxygen pressures of 0.2 atm or
greater and SO3 pressures of about 10e4 to lo- 5. f4) Deposits results from the
condensation of gases produced when burning fuels with sodium and sulfur impurities
in air-containing sodium. The composition of condensed Na2S04 at equilibrium
with gas mixtures will be governed by the decomposition reaction:
Na2S04 = Na20 + SO3 (11
K1 = [aNa [PSO3]
or
so42- = 02- + so3 (21
assuming that the salt is at unit activity. The Na20 and SO3 dissolved in the
melt are the basic and acidic species, respectively.
During hot corrosion the activity of Na20 and pressure of SO3 in the
deposit determine the type and extent of reaction. The composition of the
deposit in an engine can be determined by the gas phase or by the reaction of the
substrate material with components in the salt. If the activity of Na20 is high
and the pressure of SO3 is low or nil, solid oxides exhibit a basic solubilityin the
melt as in the reaction:

Appendix C-Hot Corrosion of Alumina 91
MO + 02- = M022- (3)
If the activity of Na20 is low and the pressure of SO3 is high, solid oxides exhibit
an acidic solubility in the melt as in the reaction:
MO = M2+ + 02- (4)
If the activity of Na20 (i.e. 02-) and pressure of SO3 in the deposit are within an
intermediate range then the oxide is stable and negligible solubility is observed.
More specifically, the degradation of an Al203 substrate by reaction with a
molten Na2S04 film may proceed by acidic or basic fluxing reactions:
3 Na2SO4(1) + Al203 = Al2(SO4)3 + Na20 (3)
Na2S04(1) + Al203 = s NaA102 + SO3(g) (6)
The regions where the hot corrosion of Al203 occurs by either mechanism are
shown on the stability diagrams in Figure l.@y6)
The solubilities of many oxides over a range of compositions of Na2S04
have been measured experimentally. Results for Al203 and SiO2 are shown in
Figure 2.(7-g) Plots of the solubilities of different oxides are displaced to the
left and right of each other depending on the relative acidity or basicity of those
oxides.
Alloys resistant to degradation at elevated temperatures are characterized
by the formation of continuous compact solid oxide scales with a low rate of
transport of metallic or oxidant species. In addition, the volatilization rate of
the oxides in the scale must be negligible. The extent of degradation by hot
corrosion is partially determined by the solubility of the oxide scale in the
molten salt.
Where the solubility of the oxide in the molten salt deposit is low or is not
a function of the acidity or basicity of the melt, corrosion may proceed at a
rapid rate until the liquid is saturated. When the solution reaction stops the
corrosion rate decreases.

LOG
P BASIC
O2
FLUXING
-10
AIO;
-20 -10 0
LOG &03
Figure l(a). :Stability diagram showing phases of alumina that can

be stable in Na2SO4 at 700°C, and defining regions
where acidic or basic fluxing are possible. Dashed lines
are SO4 isobars (atm).
0
t
I/ 1
-16
IO
/
/
/
/
/ -10.6,
/ IO /
/ / 1
/ /
/ /
/
/ -s ’
BASIC / IO ’ ACIDIC
LOG / /
FLUXING / / FLUXING
/ t
P / /
/
-
02
/ /
/ /
I
AIO; /
/
A13+
-a
/
A’203 c
AG3
1
-8 -Y 0 Y
LOG Ps,
3
Figure l(b). Stability diagram showing phases of alumina that can be

stable in NaZSO4 at 1000°C, and defining regions where
acidic or basic fluxing are possible. Dashed lines are sulfur
isobars. Very high SO3 pressures are required for acidic
fluxing. Refractory metal oxides are believed to make acidic
fluxing favorable at lower SO3 pressures as indicated by the
displaced boundary (arrows).(ll)
Figure 2. Solubilities of alumina in fused Na2S04 at 927OC.

In acidic or basic conditions sustained attack results if solution of oxide
and precipitation of the oxide in the salt away from the oxide-salt interface is by
a negative gradient of the solubility of the oxide across the salt film.(lO-lll The
gradient in the solubility of the oxide is established by the local variation in the
activity of Nat0 (i.e. 09-) across the film. The resultant solution and
precipitation of the oxide produces a porous scale. The growth of a non-
passivating oxide scale on a metallic engine component results in high corrosion
rates.
In acidic conditions sustained attack may arise when the acidic component
is derived from the gas phase.(5p12*131 Als o, acid fluxing can result when an
acidic component comes from the solid phase and reduces the Na90 activity in
the deposit. This causes the melt to become more acidic. Alloy induced acid
fluxing results from the solution of refractory metal components of superalloys
in Na$304. More specifically, the formation of tungstates, vanadates, and
molybdates reduces the oxide ion concentration in the melt.(121 In a similar
manner, the oxidation of the vanadium which is contained in many fuels can
promote acidic fluxing.
Materials
The materials chosen are single crystal alumina and three poiycrystalline
aluminas containing different levels of impurity and microstructures. They have
been discussed in previous reports. Their sources were given in Table I of the
first part of this report. The single crystal is 99.99% pure. The chemical
analysis of the polycrystalline aluminas is given in Table I.

TABLE I
Chemical Analysis of Polycrystalline
Aluminas* (units: wt%)
High PP** Med PP Low PP
Fe203 O.Ol^ O.Ol^ 0.02
CaO 0.01,. 0.02 0.07
MgO 0.10 0.17 0.75
Ti02 O.Ol^ O.Ol^ 0.02
SiO2 0.11 0,17 1.65
K20 O.OOl^ 0.001- O.OOl^
Na20 O.OOl^ 0.02 0.09
** High PP = High Purity Polycrystalline Alumina
Med PP = Medium Purity Polycrystalline Alumina
Low PP = Low Purity Polycrystalline Alumina
Undetected, Limits of detection

Hot Corrosion Experiments
All specimen had areas about 1 x 1 cm which were polished down to 1 urn
diamond. After cleaning, they were usually coated with 5 mg/cm2 of Na2S04
and exposed for various times from 1 to 100 hours. Some samples were also
subjected to cyclic exposures with 45 hours exposure for cycle, and washing and
reapplication of Na2SO4 between cycles. Two temperatures 700°C and 1000°C
and two atmospheres, pure 02 and initially 1% SO2-balance oxygen at a total
pressure of 1 atm, were used in the experiments. The pure 02 tended to give
basic conditions and the SC2 containing atmospheres set up acidic environments
with pSO3 = 1.5 x 10e3 atm at 1000°C and pSO3 = 7 x 10N3 atm after passage of
the initial mixture over a platinized catalyst at 700°C. Since Na2S04 melts at
883OC, the sulfate used at 700°C was an equimolar mixture of Na2S04 and CoSO4
which was molten at that temperature.
The changes in morphology and products focused on the samples were
characterized mostly with the SEM, and its X-ray spectroscopy attachment (EDS
and WDS) for microanalysis. After exposures and observations of the samples,
they were washed and the weight changes were measured to iO.l mg. The wash
water was analysed as required using a semiquantitative microprobe method.
When sufficient products were formed, they were identified by X-ray diffraction.
These procedures were described in detail in previous reports.
Wetting
The wetting angles for the 1 hour and 24 hours isothermal exposures were
measured and are given in Tables II, and III. In most cases the wetting morphology
TABLE II
WETTING DATA Al203
1 & 24 Hour Isothermal Exposures
ACIDIC
1 HR 7oooc 24 HR 1000°/24
Low PP MS04 MS04 Na2S04

BM(S) 140 190
LD18O RP=50°
SD 40°
Med PP MS04 MS04 Na2S04

300 220 130
RP=48O RP 2O NC
High PP MS04 MS04 Na2SO4

8O 200 120
sxtaI MS0 MS04 Na2S04

140 220 180
KEY: MS04 (Na2,CO)S04 CWContinuous

Wetting NCNearly Continuous
Wetting PSPhase Separated
Drop Wetting Angle (Cobalt
Oxide+SaIt) BMBimodal (Two
Wetting Angles) BM(S)Angles -
f(Drop Size) LDLarger Drop
Wetting Angle SDSmaller Drop
Wetting Angle BM(H)Angles =
f(Drop Habit) RFReaction
Product Angle
TABLE III
WETTING DATA Al203
BASIC
1 HR 1ooooc 24 HR 700°/24
Low PP Na2SO4 Na2SO4 MS04

30 BM(H1 PS
NC 40/21° =16O
Med PP Na2SO4 Na2S04 MS04

cw cw PS
=14o
High PP ;$2s04 Na2S04 MS04

40 PS
=14o
sxtsl ~o~so4 Na2S04 MS04

E” PS
200
KEY: MS04 WwWSO4

cw Continuous Wetting
NC Nearly Continuous Wetting
PS Phase Separated Drop Wetting
Angle (Cobalt Oxide+Salt)
BM Bimodal (Two Wetting Angles)
BM(S1 Angles - f(Drop Size)
LD Larger Drop Wetting Angle
SD Smaller Drop Wetting Angle
BM(H1 Angles = f(Drop Habit)
RP Reaction Product Angle
which distinguished each set of exposure conditions, atmosphere, and temperature,
was nearly developed minutes after the coupon reached the furnace temperatures
except in a few eases.
The phase separation of the (Na2,Co)SO4 deposit used in the 700°C exposures
only occurred in the oxygen atmosphere as predicted by thermodynamics. After
exposure the deposits were composed of Na2S04 and masses of equiaxed cobalt
oxide crystals. The cobalt oxide formed a discontinuous layer on the substrate
surface. The cobalt oxide was preferentially wetted by the molten salt which
was prevented from wetting the alumina substrate.
In some cases two distinct wetting angles were measured on some of the
coupons. Bimodal wetting morphologies resulted from the phase separation of
the salt or from the formation a reaction product in the deposit. In other cases
the difference in contact angle was a function of drop size as reported in Tables
II and III.
The wetting of all the aluminas exposed in acidic conditions was limited.
In acidic conditions the wetting angle observed on the two higher purity aluminas
(High PP and single crystal) increased between 1 and 24 hours of exposure at
700°C. The wetting angle observed on the two lower purity aluminas (LowPP
and MedPP) decreased between 1 and 24 hours of exposure at 700°C. No trend in
the wetting with composition is apparent. In general, the wetting angle of the
salt on the aluminas in acidic conditions is lower at 1OOOoC than at 700°C. The
reaction of the more impure aluminas with the molten salt results in the formation
of multiple reaction products and the spreading of the droplets. There is less
reaction on the two higher purity substrates and the wetting decreases as the
angle of contact approaches the equilibrium wetting angle.

In basic conditions the wetting angle of the salt on the alumina is lower at
1000°C than at 700°C. However, the crystallization of cobalt oxide from the
salt at 700°C affected the results. At both temperatures, the wetting of the
alumina by the salt is more extensive in basic conditions than acidic conditions.
In generai, a slight decrease in the contact angle occured between 1 hour and 24
hours in basic conditions at 1000°C. In basic conditions the wetting generally
increases with the impurity content of the substrate. The lower purity aluminas
were wetted much more extensively by the molten salt than were the higher
purity aluminas. This is attributed to the fact that the wetting behavior is
affected by the reaction of impurity phases with the basic component of the
melt, sodium oxide, which is promoted by an increase in temperature.
Photographs taken during long term exposures where partial wetting of the
coupons occurred did not provide substantial evidence of the repeated wetting of
specific areas of a coupon when salt was reapplied before each cycle. This is
important since localized attack could be very damaging.
In those experiments where the wetting was poor and the resultant contact
angle was high, relatively thick droplets were observed. When the deposit is
thick the effect of the atmosphere on the composition of the salt, especially in
the vicinity of the substrate, may be significantly less than when the salt wet the
coupon in a thin continuous layer. On coupons where a low wetting angle is observed,
the area for reaction at both the oxide-salt and salt-gas interfaces is increased
and the distance over which the oxidant must be transported inward from the
salt-gas interface is decreased. Hence, the corrosion rate may be enhanced
when the contact angle is low. John observed that an increase in salt film thickness
results in a decrease in corrosion rate, for exposures of alloys with significant
solubility in the molten salt.(141

Hot Corrosion of Al203
Even after long term exposure with thermal cycling and the reapplication
of the salt the total weight changes measured were very small, less than 1 mg/cm2,
even after 500 hours of cyclic exposure. The weight changes recorded for 1, 24,
405, and 495 hour experiments are listed in Tables IV,V, and VI.
A globular silica reaction product was observed on the washed polycrystalline
coupons from the exposures made in either atmosphere at 700°C. Well defined
crystalline silicate reaction products were observed on the washed polycrystalline
coupons from expures in either atmosphere at 1000°C. After exposure, the silica
or silicate reaction products were etched from some of the coupons using
concentrated hydrofluoric acid. The weight decreases are listed in Table VII.
The weight losses due to the etching of the coupons indicate that these reaction
products were present on the washed polycrystalline substrates in significant
quantities which were not unreasonably large compared to the impurity content
of the samples.
The weight changes must be affected by the smaller area of reaction when
the coupon surface was not wetted completely by the melt. In acidic conditions
at both temperatures all the materials lost weight. The formation of cobalt
oxide on the surfaces of the coupons exposed at 700°C in basic conditions increased
the weight of the samples. The weight changes recorded for any of the exposures,
particularly of the polycrystalline materials, are probably the result of two or
more contributions, possibly of opposite sign. Also, a combination of both gaseous
and hot corrosion occurred over fractions of the coupon surfaces during exposure.
Coupons of the same materials were exposed to gaseous corrosion in the same
atmospheres for 168 hours (appendix A) and no weight changes were detected
with a sensitivity of 0.1 mg.

TABLE IV
WT CHANGE ALUMINA [mg/cm2]

BASIC 1OoOoC 1 HR 24 HR
LOWPP 0 - 0.2
MedPP 0 + 0.1
HighPP - 0.5 - 1.1
sxtal - 0.1 - 0.4
BASIC 7OOoC
LOWPP + 0.4
MedPP + 0.1
HighPP 0
sxtal + 0.1
ACIDIC 7OOoC 1HR 24 HR
LOWPP - 0.1 - 0.2

MedPP - 0.8 - 0.4
HighPP - 0.5 - 0.2
sxtal - 0.1 0
ACIDIC 1OOOoC 24HR
LOWPP - 0.3
MedPP -0.3
HighPP - 1.1
sxtel - 0.3
LowPP = Saxonsburg
MedPP = 3M
HighPP = Lucallox
sxtal = Single crystal
TABLE V
495 HR Cyclic Exposure (45 HR Cycles + ResaIt)
ACIDIC 700°C
Material
Cycle LOWPP MedPP HighPP SXtal
-0.2 -0.3 -0.3 0

2 -0.1 -0.1 -0.1 0
3 0 0 0 0
4 0 0 0 0
5 0 -0.1 -0.1 0
6 -0.1 -0.1 0 -0.1
7 0 0 0 0
a 0 +0.1 0 0
9 -0.1 -0.1 0 0
10 -0.1 0 0 -0.1
11 0 0 0 0
-0.6 -0.6 -0.5 -0.2

TABLEVI
WTCHANGE ALUMINA[mg/cm2]
405 HR CyclicExposure(45
HR Cycles+ ResaIt)
BASIC 1000°C
Material
Cycle LOWPP MedPP HighPP SXtal
1 -0.2 0 0 0
2 +0.3 -0.1 0 0
3 0 +0.1 0 0
4 0 -0.1 0 0
5 0 0 0 0
6 0 0 0 0
7 0 0 0 -0.1
8 0 0 0 0
9 0 0 0 0
+0.1 -0.1 0 -0.1

TABLE VII
RESULTS HP ETCH
Exposed Samples Etched in HP
BASIC 1000°C 24 HR 405 HR
LOWPP 0 - 0.6
MedPP - 0.7 - 0.3
HighPP 0
ACIDIC 700°C 495 HR
LOWPP - 0.3
MedPP - 0.1
HighPP - 0.1
ACIDIC 1000°C 24 HR
LOWPP 0
MedPP - 1.0
The data given in Table VIII were obtained from the semiquantitative
processing of the EDS spectra taken from the wash water. This analysis was
performed on 1, 10, and 100 hour exposures in acidic conditions at 700°C.
Results, Acidic Conditions
The 700°C exposures in acidic conditions were emphasized because sulfate
formation is favored at the low temperature for a given SO2 pressure.
Single Crystal Alumina
At 700°C limited reaction was detected after 1 hour of exposure.
The wetting was spotty and a wetting angle of 14O was measured. After 24 hours
of exposure the wetting angle was 22 o. The EDS spectra of many drops below
100 microns in diameter indicated that limited solution of Al had occurred.
Negligible reaction was detected on the washed substrate after 100 hours of
exposure. Aluminum ions were consistently present in the wash water at 1, 10,
100 hours. A limited amount of Si was detected after 10 and 100 hours of exposure.
Since negliglble silicon was present in the single crystal, it is assumed that
limited contamination of the samples occurred and that the levels of Si detected
for the single crystal alumina are not significant. The alumina furnace tube in
which the samples were heated may have been a source of Si. After 495 hours of
cyclic exposure, networks of shallow depressions were observed over portions of
the coupon.
The coupon exposed at 1000°C for 24 hours exhibited a faint etching at the
perimeters of drop areas. A significant amount of Al was present in some of the
EDS spectra of the salt at the drop edges. The substrate was wetted by drops up
to several millimeters across with a wetting angle of 180.

TABLE VIII
Wash Water Analysis, Al203 Exposures

700°C Acidic Conditions
Material Element 1 Hour 10 Hour 100 Hour
LOWPP
Na 17.0 18.0 16.0
S 14.0 14.0 14.0
Si 0.99 0.96 1.4
Al 2.0 1.2 2.2
Mg 2.4 1.3 1.3
Ca 0.00 0.00 0.00
co 4.1 4.2 3.5
0** 60.0 60.0 61.0
MedPP
Na 18.0 19.0 16.0
S 13.0 11.0 13.0
Si 0.94 1.5 3.6
Al 2.2 1.5
Mb? 1.8 :::
Ci 0.00 0.00 I?“00
co 4.2 3.9 3.3
0 59.0 58.0 61.0
HighPP
Na 17.0 19.0 17.0
S 15.0 15.0 16.0
Si 0.46 0.00 0.40
Al 1.0 0.52 0.66
ME 1.4 1.0 0.43
Ca 0.00 0.00 0.00
co 3.0 3.9 3.9
0 62.0 61.0 62.0
sxtal
Na 19.0 18.0 16.0
S 15.0 15.0 16.0
Si 0.00 0.46 0.56
Al 0.60 1.3 0.16
I% 0.81 1.6 0.00
Ca 0.00 0.00 0.00
co 3.9 3.7 4.4
0 61.0 61.0 63.0
** by stoichiometry
High Purity Polycrystalline Alumina
At 700°C limited reaction was detected after 1 hour exposure. The
substrate was wetted by two large patches covering about half of the coupon
area. A wetting angle of 8Owas measured.
After 24 hours of exposure the wetting angle was approximately 20°. The
drops were patchy and typically under 1000 microns across. The solution of Al
was indicated by some of the EDS spectra of salt at the edges of drops. These
spectra also indicate the presence of significant amounts of Si. Washing the
substrate revealed the preferential solution of alumina grains and the presence
of a fine poorly defined silicate along the edges of the drop areas. A few
isolated patches of cobalt sulfate 10 to 20 microns across, containing Na and
significant amount of Si were observed.
The etching of alumina grains throughout the drop areas was evident after
100 hours of exposure. A thin discontinuous layer of silica on the edges of grains
in a drop area is shown in Figure 3 (top). Aluminum ions were consistently
present in the wash water after 1, 10, and 100 hour exposures.
After 495 hours of cyclic exposure small thin patches of silica covered
portions of the substrate and fractions of individual grains (Figure 3, bottom). In
these areas the etching and preferential solution of grains was apparent.
The coupon exposed at 1000°C for 24 hours was wetted by several discrete
drops about 2 millimeters across with a 12O wetting angle. Traces of Al and Si
were detected in EDS spectra of salt at the drop edges. Washing the substrate
revealed the presence of a thin poorly defined silica-containing layer with traces
of Na and Mg detected in some spectra. This discontinuous layer covers a large
fraction of the grains in the drop areas (Figure 4, top).

Figure 3. High purity polycrstalline alumina exposed in acidic

conditions at 700oC or 100 hours {top) and 495 hours
{bottom). Etched grains in drop areas and thin patches of
silica are shown on washed substrates.
Figure 4. High, medium, and low purity polycrystalline aluminas

(top, middle, and bottom, respectively) exposed in
acidic conditions at lOOOoCfor 24 hours. Silicate
reaction products on washed substrates are shown.
Medium Purity Polycrystalline Alumina
At 700°C after 1 hour of exposure a significant amount of Al and a
limited amount of Si were indicated by the EDS spectra of drops. Wetting was
spotty except for a single large patch of salt covering about a fifth of the coupon
area. The droplets were typically 10 to 200 microns across. A wetting angle of
30° was measured. A few isolated patches of cobalt sulfate 10 to 20 microns
across and containing Na and a significant amount of Si were observed.
After 24 hours exposure the wetting was spotty with some small patches.
Substantial solution of Al was indicated by the EDS spectra of salt in drops with
a measured wetting angle of 22O, and in droplets containing larger well defined
crystals richer in Al with a wetting angle of about 48O. Washing the substrate
revealed that fine poorly defined bands a few microns wide, containing Co, Si,
and traces of Mg and Ca, skirted large fractions of the perimeters of the large
drop areas. In the drop areas, some of the grain boundaries had been etched.
The etching of grains throughout the drop areas is evident after 100 hours
of exposure. Small globules rich in Si scattered on raised blocky areas 20 to 30
microns across were observed. Aluminum and Mg ions were consistenly present
in the wash water at 1, 10, and 100 hours in much greater concentrations than
for the single crystal or the high purity polycrystalline material. The
concentration of Mg detected after 10 hours of exposure was relatively large.
The concentration of Si ions in the wash water was also much greater, and increased
drastically between 10 and 100 hours.
After 495 hours of cyclic exposure patches of a continuous layer of well
defined globular silica covered portions of the substrate. The etching of grains
was apparent.
The coupon exposed at 1000°C for 24 hours was wetted nearly continuously
over some portions of the surface and by discrete droplets with a 13O wetting
angle in other areas. In both areas, substantial amounts of Ca, Mg, Si, Al, a
limited amount of Ba, and traces of K were present in EDS spectra of the salt.
Washing the substrate revealed the presence of numerous well defined sodium
aluminum silicate and sodium magnesium aluminum silicate crystals scattered
across the substrate (Figure 4, middle). Also, smaller sodium magnesium
aluminum silicate crystals had formed as platelets aligned in parcels on the
substrate. Some crystals were present in groups but in most cases they
protruded from grain boundaries where discontinuous etching had occurred.
Low Purity Polycrystalline Alumina
At 700°C after 1 hour of exposure the wetting was spotty with some
larger patches of salt over 500 microns across. Wetting angles of la0 for the
larger drops and 40° for the droplets were measured. A significant amount of Al
and a limited amount of Si were indicated by the EDS spectra of drops. Blocky
crystals about 20 microns across were visible below the salt in several droplets.
After 24 hours of exposure patches of salt nearly as large as 2000 microns
across wetted the substrate. Substantial solution of Al was indicated by the EDS
spectra of salt in drops with a measured wetting angle of 14O. Sulfate drops with
a wetting angle of about 500 contained well defined sulfate crystals rich in Al.
Some of the crystals contained Al and Mg as well as Si (Figure 5). Washing the
substrate revealed that a thin poorly defined layer as large as 100 microns wide
containing Co, Si, a significant amount of Mg, and traces of Ca skirted the
perimeters of the larger drop areas. In the drop areas a relatively even etching
between the grains had occurred. A few isolated patches of cobalt sulfate 10 to
30 microns across and containing Na and a significant amount of Si were
observed.
Figure 5. Low purity polycrystalline alumina exposed in acidic

conditions at 700oC for 24 hours. Salt crystals containing Al
and Mg are shown (top and bottom).
The etching of grains throughout the drop areas is evident after 10 hours of
isothermal exposure. Distributions of silica globules about 1 or 2 microns across
have formed on fractions of the drop areas. After 100 hours of exposure many of
the globules coalesced. Aluminum and Mg ions were consistently present in the
wash water at 1, 10, and 100 hours in much greater concentrations than for the
single crystal or high purity polycrystalline materials. The concentration of Si
ions in the wash waters was also much greater, but the increase observed
between 10 and 100 hours was not as great as that observed for the medium
purity polycrystalline material.
After 495 hours of cyclic exposure a thin discontinuous layer of silica
covered portions of the substrate. The etching of grains was apparent.
The coupon exposed at 1000°C for 24 hours was wetted by drops up to 3000
microns across with a measured wetting angle of 19O. Spotty wetting was also
observed, particularly in the areas outside the perimeters of the larger drops. In
both areas, substantial amounts of Ca, Mg, Si, Al, and traces of Ba and K were
present in EDS spectra of the salt. Washing the substrate revealed the selective
solution of grains in the drop areas. Well defined aluminum silicate containing
Na and Ca formed in limited quantities (Figure 4, bottom). Minor amounts of Mg
were present along with the Ca in a few of the crystals.
Results, Basic Conditions
The 1OOOoC exposures in basic conditions were emphasized because more
extensive reaction occurred at the higher temperature when preliminary short
term exposures were made in oxygen. At 700°C a large fraction of the molten
salt did not remain in contact with the substrate throughout the exposure due to
the formation of cobalt oxide.

Single Crystal
At 700°C phase separation of the salt occurred within 24 hours of
exposure. The wetting was patchy and irregular. A wetting angle of about 20°
was measured. The concentration of Al ions was limited to the EDS spectra of a
few small drops, where the molten salt had remained in contact with the substrate.
Cobalt oxide separated the sulfate from the substrate over most of the area of
the larger drops.
No reaction was detected after 1 hour of exposure at 1000°C. A wetting
angle of loo was measured and broad patchy wetting covered large areas of the
coupon. After 24 hours of exposure an So wetting angle was measured.
Some sodium aluminum silicate was observed after 405 hours of cyclic
testing at 1000°C. Negligible quantities of Si are present in the as-received
substrate material. The alumina furnace tube is a likely source of the Si.
High Purity Polycrystalline Alumina
exposure. The wetting was patchy and irregular. A wetting angle of about 14O
was measured. Traces of Si and limited concentrations of Al ions were indicated
by the EDS spectra of small drops where the molten salt had remained in contact
with the substrate. A network of globular silica associated with Co formed on
the substrate in the perimeter of the larger drop areas. From the morphology of
the washed substrate it appeared that preferential solution of grains and growth
of Si rich needles had occurred in the drop areas. The growth direction of the
needles was constant across the surface of each grain (Figure 6, top). A few
larger needles were observed at or near the triple points between grains and
contained significant amounts of Ca.

Figure 6. High purity polycrystalline alumina exposed in basic

conditions at 700oC for 24 hours (top) and 1000oC for
405 houl'S (bottom). Oriented silica rich needles (top)
and silicate reaction products at grain boundaries
(bottom) are shown on washed substrates.
After 1 hour of exposure at 1000°C a distribution of well defined
hemispherical silicates formed across the substrate surface in the drop areas.
Crystals of sodium aluminum silicate between 5 and 10 microns across protruded
from beneath the salt near the centers of some of the drop. A wetting angle of
So was measured. After 24 hours of exposure a 4O wetting angle was measured.
A distribution of globular silicate had formed in the drop areas. Some
anomalously large globules contained Ca and Mg. A consistent pattern of
formation of sodium aluminum silicate along grain boundaries and sodium
magnesium aluminum silicate at the triple points between grains was apparent on
the washed substrate after 405 hours of cyclic exposure (Figure 6, bottom). The
preferential solution of the surfaces of certain grains is apparent.
Medium Purity Polycrystalline Alumina
At 700°C, phase separation of the salt occurred within 24 hours of
was measured. The EDS spectra of small drops where the molten salt had
remained in contact with the substrate indicated that significant amounts of Ca,
Si, Al, and Mg were present in the salt. Washing the substrate revealed the
pronounced etching of the grains in drop areas accompanied by the deposition of
a poorly defined discontinuous layer of silica. A sparse network of globular silica
associated with Co formed on the substrate in the perimeter of the drop areas.
After 1 hour of exposure at 1OOOoCthe salt had formed a continuous film
over most of the coupon. Substantial amounts of Ca, Mg, Si, and Al were present
in EDS spectra of the salt. Washing the substrate revealed the presence of
numerous well defined magnesium aluminum silicate crystals scattered across
the substrate and a random distribution of vugs up to 20 microns across where

preferential solution of the substrate occurred. After 24 hours of exposure the
continuous wetting of the coupon resulted in a substrate surface which was
partially covered with poorly defined crystals. Some of the crystals, particularly
the one near vugs, contained substantial amounts of Si and Mg. Small patches of
a thin layer of silica were also observed on the substrate. On the unwashed
sample large peaks for Ca, Ba, Mg, Al, Si and traces of K were present in the
EDS spectra of the salt. Rosettes about 10 to 15 microns across were present in
some areas where substantial amounts of Be were present in the salt (Figure 7,
top). The pronounced etching of grains was apparent after 405 hours of cyclic
exposure. Numerous bunches of poorly defined sodium aluminum silicate crystals
were scattered over the substrate surface.
Low Purity Polycrystalline Alumina
was measured. The EDS spectra of small drops where the molten salt had
remained in contact with the substrate indicated that significant amounts of Ca,
Al, Mg and Si were present in the salt (Figure 8). Washing the substrate revealed
a dense network of globular silica associated with Co on the substrate over
portions of the drop areas. A significant amount of Mg was present in many of
the globules suggesting that they were a magnesium silicate.
After 1 hour of exposure at 1OOOoC the salt had formed a nearly continuous
film on over half of the coupon surface. A wetting angle of 3O was measured.
Substantial amounts of Ca, Ba, Mg, Si, Al, and K were present in EDS spectra of
the salt, both in the salt film and in small groups of well defined salt crystals
rich in impurity ions. Washing the substrate revealed a fine globular silicate
Figure 7. Medium purity polycrystalline alumina exposed in basic

conditions at lOOOoCfor 24 hours. A multi-phase
deposit with rosettes containing Ca, Ba, Al, Si, Mg, and
K (top) and a typical multi-phase deposit containing Ca,
Al, Si, Mg and K (bottom) are shown.
Figure 8. Low purity polycrystalline alumina exposed in basic

conditions at 700oC for 24 hours. Salt crystals
containing Ca, Mg, Al, and Si are shown (top and
bottom).
containing a significant amount of Mg. An intermittent distribution of larger
globules up to
over 2 microns across contained Ca. A random distribution of vugs typically
below 10 microns across where preferential solution of the substrate had
occurred was also observed.
After 24 hours of exposure a bimodal wetting morphology had developed.
A wetting angle of 4O was measured on drops as large as 4000 microns across. A
thin discontinuous film of salt with a network of holes and branches was left
behind by the dewetting of the molten salt. A 21° wetting angle was measured
on the edges of the film. On the unwashed sample a substantial concentration of
Al ions and traces of Ca, Si, and Mg were indicated by the EDS spectra of the
salt. On the washed sample well defined tabular crystals of sodium magnesium
aluminum silicate and blocky sodium aluminum silicate crystals containing Ca
were observed (Figure 9, top).
The etching of poorly defined grains was apparent after 405 hours of cyclic
exposure. Blocky crystals of sodium magnesium silicate were predominantly free
of Ca and had become less well defined than after 24 hours of isothermal
exposure (Figure 9, bottom). A phase rich in Al and Mg formed on some sites on
the substrate, probably magnesium aluminate spineL
Discussion of Results
Impurities increase the attack of the polycrystalline aluminas by the
molten sulfate. This effect is enhanced by their pronounced segregation, at the
grain boundaries and at the triple points between grains in impurity second
phases. In each condition of exposure the hot corrosion of the lower purity
polycrystalline materials was dominated by the reactions of the impurity silicate
phases with the melt.

Figure 9. Low purity polycrystalline alumina exposed in basic

conditions at 1000oC for 24 hours (top) and 405 hours
(bottom). Aluminum silicate crystals containing Ca left
and Na right (top) and Mg and Na (bottom) are shown on
washed substrates.
The data available on the solubility of alumina in Na2SO.t as a function of
the composition of the salt is not directly applicable when significant amounts of
Si are present in the melt. The formation of phases not predicted by thermodynamic
stability diagrams occurred at less than unity activity as in the previous work on
gaseous corrosion (Appendix A).
Activities of Na20 in the salt as great as 10-6 to 10e4 at temperatures
between 7OOoC and 1000°C in an oxygen atmosphere have been reported.
However, the actual activity of Na20 in the deposits during the exposures and
the profile of the activity across the thickness of the deposits is not known. In
the exposures of alumina the salt tended to form discrete droplets. Impurities
which result in more extensive wetting of the substrate increase the extent of
degradation by increasing the area of attack and promoting the formation of a
thinner molten salt film as discussed in the introduction. This is observed for the
polycrystalline aluminas, particularly under basic conditions.
At 400°C in acidic conditions all the coupons exposed were covered by
discrete droplets of salt. The data obtained from the analysis of the wash water
indicates that the atomic sodium to sulfur ratios were significantly below 2.0 for
all times of exposure (Table VIII). This is true for all four aluminas. Since no
insoluble products rich in Na were detected, it is likely that SO3 from the
atmosphere enriched the deposit as the composition of the melt shifted toward
being less acidic because aluminum sulfate and other sulfates formed in solution
in the melt.
The presence of Al in the wash waters and the weight losses recorded for
all exposures in these conditions are COnSiStent with the fOrtTIatiOn of A12(S0.+)3.
The wash water indicates that Mg was present as a soluble reaction product in
the salt, probably MgSO4, since it is stable at lower pressures of SO3 than
Al3(SO4)3. The well defined sulfate crystals in Figure 5, contain substantial
amounts of Al and significant amounts of Mg. An abundance of these crystals
was observed on the two least pure aluminas.
After 495 hour cyclic exposures silica was present on portions of all of the
substrates but the single crystal (Figure 10, top). The solution of Si rich impurity
phases at the grain boundaries of the two lower purity polycrystalline materials
resulted in a marked lack of connectivity of surface grains as in Figure 10 (bottom).
At 700°C in acidic conditions the principal reaction during the hot corrosion
of alumina is the formation of Al3(SO4)3, as was observed for gaseous corrosion
(Appendix A). A mechanism for the reactions which occurred during the long
term exposures of the less pure polycrystalline aluminas has been developed. It
is proposed that the acidic solution of alumina lowers the pressure of SO3 at the
substrate-salt interface below that at the salt-gas interface, resulting in a
negative solubility gradient for silica across the molten salt layer and satisfying
the Rapp-Gotto criteria for fluxing. (11) Impurity silicates are dissolved from the
grain boundary areas on the substrate. As the silicate ions are transported toward
the salt-gas interface they advance into a negative solubility gradient and the
precipitation of nearly pure SiO3 results. The coalesence, growth, and coarsening
of the SiO3 which is precipitating from the melt results in the formation of
globular silica patches on fractions of the substrate surface. Because the
precipitation of SiO3 in the melt occurs some distance away from the substrate-
salt interface the size and shape of the patches of globular silica were not dictated
by the microstructure of the substrate.
The morphology of the reaction products are shown in the schematic in
Figure 11. The solution of silicates present at the grain boundaries and as impurity
grains lowers the local Na30 activity of the melt in a manner analogous to the
Figure 10. Medium purity polycrystalline alumina exposed in acidic

conditions at 700°C for 495 hours. Globular silica (top) and
etched grains (bottom) are shown on washed substrates.
S/G
Distance
O/S = Oxide-salt
Interface
S/G = Salt-gas
Interface
Q = Impurity Silicate Phases

R = Globular Silica Reaction Product
S = Sulfate Deposit Containing (Al, Mg, Si)
A = Alumina Grains at Substrate Surface

Solution Reaction at A:
3NagSO4 + Al303 = Al3(SO4)3 + 3Na30
B = Intergranular Attack
Solution Reaction at B:
Silicate + Nag0 = Na3O*xSiOg (in sulfate)
C = Away from Oxide-Salt Interface where Melt is More Acidic
Precipitation Reaction at C:
Na3O.xSiOg (in sulfate) = xSiO3 + Nag0 (in sulfate)
(Between grains and impurity phases at substrate

surface Reaction A + Reaction B:
3NagSOq + Al303 + xSilicate = Al3(SO4)3 + 3NagO-xSi0g)
Figure 11. Schematic diagram of the morphology of reaction

products on polycrystalline aluminas exposed in acidic
conditions at 7OOoC.
way in which the solution of refractory oxides in molten Na3S04 lowers the
sodium oxide concentration of salt deposits during the alloy induced fluxing of
certain superalloys.(11~131 The formation of Al3(SO4)3 and MgSO4 occurs
concomitant to and in the vicinity of the localized solution of silicates, raising
the Nag0 activity and increasing the solubility of the silicate impurities in the
vicinity of the substrate surface. Intergranular corrosion results in a substantial
decrease in the connectivity of the grains of the substrate.
For the proposed model it is likely that the rate of reaction is controlled by
the inward transport of reactive species (i.e. SO3) from the salt-gas interface
across the thickness of the film or by the transport of soluble reaction products
away from the substrate-salt interface. This assumes that the transport of
reactant and product species proceeds more slowly than the dissolution of the
alumina grains or of impurity silicates at the substrate surface. The kinetics
could not be measured to check this statement, however, there is some evidence
that it is correct. As mentioned above, higher concentrations of Al were
detected in the drop edges than in the bulk of the salt on several of the samples
exposed in low temperature acidic conditions. The SO3 atmosphere has greater
effect on the chemistry of the salt at the drop edges because the distance for
transport of SO3 to interface or Na30 to surface is minimized.
At 1OOOoC in acidic conditions all of the coupons exposed were wetted by
discrete droplets of salt, except for the exposure of the medium purity
polycrystalline material where nearly continuous wetting was observed. Weight
losses were recorded for all the exposures in these conditions. After exposure,
well defined crystals with the form and chemistry of silicates were observed on
the washed polycrystalline substrates.

The silicates on the washed substrates of the polycrystalline materials
after 24 hours of exposure were shown in Figure 4. A fine poorly defined silicate
covers grains over large portions of the drop areas on the high impurity
substrate. Sodium aluminum silicate crystals containing substantial amounts of
Mg cover portions of the medium purity substrates. Sodium aluminum silicate
crystals containing substantial amounts of Ca cover portions of the low purity
substrate. In the drop areas on the washed substrates where the crystals are
present in substantial quantities, the substrates are etched more deeply than on
similar exposed at 700°C in acidic conditions. Within the etched drop areas the
intergranular corrosion which had been observed on the samples exposed at
700°C was absent (Figure 12).
The salt deposits on the two least pure materials contained significant
amounts of Al, Si, Mg, and Ca after exposure. The EDS spectra of most of the
smaller salt droplets do not contain Al or Si, so the Ca and Mg are in solution in
the salt as sulfates (Figure 13, top). Since the partial pressure of SO3 in the
atmosphere at 1000°C was less than at 7OOoC, the formation of Al2(SO4)3 would
not have been predicted by thermodynamics, but was observed for gaseous
corrosion (Appendix A). Under acidic conditions the formation of the CaS04
occurred at 1000°C but not at 700°C although it is stable at both temperatures.
This is believed to be the result of more severe degradation of impurity silicate
phases at the higher temperature. The uniform etching of the substrate, by the
formation of Al2(SO4)3, could result from local decreases in the basicity of the
melt at the oxide-salt interface in areas of copious silicate crystal growth.
The multi-phase salt deposit on the unwashed medium purity substrate
shown in Figure 13 (bottom) is characteristic of the two least pure materials.
The lighter blocky phase contains Ca, Si, Al, and Mg, and the remainder of the
Figure 12. Medium purity polycrystalline alumina exposed in

acidic conditions at 1OOOoC for 24 hours. Sodium
aluminum silicate reaction products and etching of
drop area are shown on washed substrates.
Figure 13. Low and medium purity polycrystalline aluminas (top

and bottom, respectively) exposed in acidic conditions
at 1000°C for 24 hours. Sulfate crystals containing
Ca, Si, Al, and Mg (bottom) are shown.
salt contains Al. The two phase salt droplet reaction product morphology must
be linked to the growth of aluminum silicates at surface heterogenieties in the
substrate.
As a result of the substrate microstructure unique reactions occurred
concomitantly over different areas of the substrate and produced local variations
in the composition of the melt. The formation of a network of well defined
aluminum silicate crystals occurred preferentially at the grain boundaries and
triple points of the microstructure. The results and the morphology of the
products described earlier and shown schematically in Figure 14 can be explained
as follows. As the alumina grains are dissolved by the sulfate, the local SO3
pressure decreases and the Nag0 activity increases accordingly near the alumina
grains. The increased Nap0 activity promotes the transport of Nag0 in the melt
to the grain boundaries and grains of silicate phases in the low purity aluminas,
where it reacts with these silicates, forming sodium silicates and alumina
silicates as well as calcium and magnesium sulfates. This maintains the acidity
of the sulfate melt thus promoting the dissolution of the alumina grains. As the
transport is rapid at 1000°C, the corrosion under the sulfate is quite uniform and
deeper than at ‘i’OO°C. At lower temperatures, 700°C under acidic conditions, a
similar mechanism operates but the transport is slower and the acid and basic
reactions cooperate mostly near the grain boundaries leading to strong
inter-granular attack (Figure 11).
The results of the exposures in basic conditions at 700°C are more difficult
to interpret due to the extensive formation of cobalt oxide crystals on the
substrate surface. Weight gains were recorded for the exposures.
Photomicrographs indicate that significant amounts of alumina were etched from
the substrates. The formation of nearly pure silica on the substrate did not
occur nearly as extensively as at 700°C in acidic conditions.

_ _----
r o 0-
,s
i
1
I
I0
I o 0 oI
I I
L-_-_-_-I
+ Transport of
Na20
-+ Transport of
so;
Q = Silicate Reaction Products
0 = Original Substrate Surface
Multi-phase Deposit
R = Bulk of Deposit
Sulfate Containing (Al, Mg, Ca, Si)
S = Other Phases
Complex Reaction Products Containing Silica
A = Alumina Grains at Substrate Surface

Solution Reaction at A:
3Na3SO4 + Al303 = Al3(SO4)3 + 3Na30
B = Intergranular Areas at Substrate Surface
Reaction at B:
Na3SO4 + Mg-, Ca-, Al- silicate =
NaSO*silicate + MgSO4 + CaSO4 + Al3604)3
Figure 14. Schematic diagram of the morphology of reaction

products on polycrystalline aluminas exposed in acidic
conditions at 1000°C.
The presence of CaS04 in the salt on the two lower purity aiuminas, which
was not detected on samples exposed in acidic conditions at the same
temperature (700°C), may be due to a more severe attack of the impurity
silicate in the oxygen atmosphere. The conditions in the oxygen atmosphere are
not conducive to the type of sustained fluxing which occurred at 700°C in acidic
conditions. Significant amounts of both Si and Al were dissolved in the salt.
Because of the depletion of the basic component of the melt due to the
formation of cobalt oxide, it is likely that the initial dissolution products
generated at the aluminum oxide-salt interface were silica and aluminum
sulfate.
In basic conditions at 1000°C the most extensive wetting of all four of the
aluminas occurred. The lower purity materials were wetted more extensively
emphasizing the influence of the impurities. Vugs developed at some of the
triple points between grains on the surfaces of the two least pure substrates.
Even after 405 hours of cyclic exposure negligible reaction was detected on
the single crystal substrate. Aluminum silicates formed on the substrates of the
polycrystalline materials. They were sodium aluminum silicates at grain
boundaries and sodium magnesium aluminum silicates at triple points on the high
purity substrate. Sodium calcium aluminum silicates and sodium magnesium
aluminum silicate formed on the lower purity substrate. A variety of multi-
phase deposit morphologies formed on the two least pure materials after
exposure at IOOOoC in oxygen. These deposits consisted of mixtures of sulfate
with various aluminum silicates of sodium, potassium and the rare earth
elements contained as impurities in the substrates.

The formation of NaA102 as the primary reaction product in the deposits
would be predicted on the basis of solubility data published by Rapp (Figure 2).
While this may be the case for the single crystal material, the degradation of the
polycrystalline materials cannot be analyzed without consideration of the
formation of aluminum silicates which occurred on the substrates and in the
molten salt. This is obvious when one considers the multiple phase morphology
of the deposits on the two least pure materials.
Relatively small weight gains and losses were recorded for the exposures at
1000°C in basic conditions. Intergranular corrosion resulted in a decrease in the
connectivity of the surface grains of the two lower purity substrates. The
weight changes and photomicrographs indicate that the formation of alumina
silicates and the solution of alumina were not as extensive as that which
occurred in acidic conditions at the same temperature. It is proposed that the
silicate impurities shift the activities in the melt towards values intermediate
between the acidic and basic conditions promoted by the atmosphere, thus
making the corrosion behaviors at 1000°C under both atmospheres similar.
However, the greater acidicity of the melt under acidic conditions increases the
dissolution of alumina. Under basic conditions the cooperative reactions initiate
with the attack of the silicates impurities by the basic sulfate forming various
sodium alumina silicates including those of the rare earths. This raises the pSO2
of the melt providing for acidic solution of the alumina grains. This sulfate
formation is limited and the silicate formation is more extensive than under
acidic conditions, leading mostly to intergranular corrosion.

REFERENCES
1. Wachtman, John B., et. al., “An Evaluation of Needs and Opportunities for
Fruitful Fundamental Research in Ceramics,” Ceramic Bulletin, Vol. 57,
No. 1 (1978), pp. 19-24.
2. Johnson, D.R., et.al., “Ceramic Technology for Advanced Heat Engines

Project,” Ceramic Bulletin, Vol. 64, No. 2 (1985), pp. 276-280.
3. T. Yonushonis, Cummins Engine Company, Private Communication.
4. Goebel, J.A. and F.S. Pettit, “Na2SO4 - Induced Accelerated Oxidation

(Hot Corrosion) of Nickel,” Metallurgical Transactions, Vol. 1 (1970), pp.
1943-1954.
5. Giggins, C.S. and F.S. Pettit, Hot Corrosion Degradation of Metals and
Alloys - A Unified Theory, Pratt and Whitney Aircraft, Report No. FR-
11545 (1979).
6. Goebel, J.A., P.S. Pettit and G.W. Goward, “Mechanisms for the Hot
Corrosion of Nickel-Base Alloys,” Metallurgical Transactions, Vol. 4 (1973),
pp. 261-278.
7. Jose, P.D., D.K. Gupta and R.A. Rapp, “Solubility of Alpha - Al203 in
Fused Na2S04 at lOOoK,” J. Electrochem. Sot., Vol. 132, No. 3 (1985), pp.
735-737.
8. Shi, D.Z. and R.A. Rapp, “The Solubility of SiO2 in Fused Na2S04 at
900°C,” J. Electrochem Sot., Vol. 133, No. 4 (1986), pp. 849-850.
9. Kim, G.M., “The Effect of Contaminants and Solubilities of SiO2 In Fused

Na2S04 at 1200°K,” Masters Thesis, University of Pittsburgh, 1982.
10. Rapp, R.A. and K.S. Goto, “The Hot Corrosion of Metals by Molten Salts,
“Symposium Fused Salts, Electrochemical Society Meeting (Pittsburgh,
1979).
11. Rapp, R.A. and K.S. Goto, Fused Salts, (J. Braunstein and J.R. Selman,
eds.), Electrochemical Society, 1978.
12. Birks, N. and G.H. Meier, Introduction to High Temperature Oxidation of

Metals (London: Edward Arnold, 1983), pp. 146-158.
13. Shores, D.A., “New Perspectives on Hot Corrosion Mechanisms,” H&$

Temperature Rapp, R.A. (ed.), (Houston: National Association
of Corrosion Eng., 1983), pp. 493-501.
14. John, R.C., “Corrosion of Metals by Liquid Na2C03,” Doctoral Thesis, Ohio
State University, 1979.
Appendix D-Hot Corrosion of Silicon Nitride
and Silicon Carbide
J.R. Blachere, D.F. Klimovich and F.S. Pettit
INTRODUCTION
Silicon nitride and silicon carbide are two ceramics materials considered
seriously for structural applications at high temperatures. They form a protective
silica scale in oxidizing atmosphere which is quite stable thermodynamically. In
highly corrosive environments such as those prevailing in incinerators or gas
turbines burning low grade fuels, SiOz may be attacked due to the pressure of
SOS, CO2 as well as oxides of metallic impurities (e.g. NagO, Na$O& It has
been shown in this program that Nap0 is the corrosive agent in the hot corrosion
of silica in contact with NagSO4 deposits (Appendix B). In particular the sodium
oxide promote devitrification and the crystalline layer formed tends to spa11
under temperature cycling. Using the results for silica as foundation, the
mechanisms and the extent of the hot corrosion of silicon nitride and silicon
carbide must be established.
The general procedures have been discussed in previous report&) and in
previous parts of this report. The materials are shown in Table I of the main
report. High purity materials (single crystal silicon carbide and CVD silicon
nitride) were studied in detail. Morphological studies of the corrosion were
performed on these materials and representative engineering materials usually
characterized by a significant level of impurities added during processing
particularly as sintering aids.
The two atmospheres used throughout the experiments were SOg-02
mixtures with a total pressure of 1 atmosphere. One contained 1% SO2 initially,
which generated a pressure of 1.5 x 1W3 atm of SO3 at 1000°C. The other was
pure oxygen. The gases flowed at the rate of 1 cm5/s. Usually the sodium
137
sulfate was applied with a surface loading of 5 mg/cmz on polished substrates.
The surface loading as well as the temperature were varied in the studies on the
purer materials. The times of exposure varied from 1 hour to 168 hours, with
standard times of 24 and 168 hours for all materials. The usual characterization
techniques were used as described earlier. They depended strongly on the
scanning electron microscope (SEM) with microanalysis of salient features by
EDS and WDS. The thickness measurements based on X-ray spectroscopy,
developed for this part of the research(2), where used for scale of thicknesses
under 1 pm. Above 1 urn they were measured on cross sections in the SEM.
These experiments were supplemented with a number of other techniques, in
particular X-ray diffraction, weight change measurements, and many types of
surface analysis (ESCA, SIMS, ISS). The methods will be discussed as needed in
the text of this appendix. The morphology was characterized after exposure
before and after washing off the soluble materials in water. Analysis of the
wash water was performed as described earlier.
The results of 24 and 168 hours exposures, in particular the product
morphologies, were described earlier for the various silicon nitrides and silicon
carbides(1~3~4). Scale thickness after 168 hours under gaseous, acidic and basic
conditions are compared in Fig. 1 and table I. For the purer materials (CVD and
single crystal) the thickness of oxide formed increased in order from gaseous
corrosion, which was essentially oxidation, acidic and then basic corrosion. This
trend is no longer clear for the more impure specimens whose behavior is
dominated by the impurities. Under basic conditions it must be emphasized that
the sulfate was not depleted on any samples which were not preoxidized before
Appendix D-Hot Corrosion of Silicon Nitride and Silicon Carbide 139
168 hour:
sic-c Sic-Si Si3 IV,
Figure 1. Thickness of layers formed for the oxidation, acidic and basic hot
corrosion of C-side and S-side single crystal silicon carbide and
CVD silicon nitride after 168 hours at 1000°C (measured between
sulfate drops for acidic corrosion). Note that oxide thicknesses
increase as OxA&.
Table I Thickness of oxide scales after 168 hours exposures (urn)
Gaseous Acidic Basic
SC SC - si 0.11 0.6
12-25
SC SC - c 0.5 1.1
CVD SC _- 1.2 (1.8)* 7.7
HP SC 0.78 0.61 -- 1.4 (24 hrs)
CVD SN 0.07** 0.3 (1.41) -6
HPSN 0.9 4.3 - 7.1 1.5 - 2.1
Sin SN 2.5 1.3 (2.1) 2.4 - 3.2
* Values in parenthesis are for spherulites under sulfate drops.

For acidic corrosion the values not parethesis were measured
between the droplets.
** 0.045 by WDS, 0.06 by ellipsometry, 0.09 SEM
SC SC -Si = single crystal Sic, silicon side
SC SC-C = single crystal Sic, carbon side
CVD SC = CVD silicon carbide
HP SC = Hot pressed silicon carbide
CVD SN = CVD silicon nitride
HP SN = Hot pressed silicon nitride
Sin SN = Sintered silicon nitride

exposure to the corrosion. Experiments on preoxidized specimens with thick
scales (>lO um)(3) and on bulk fused silica(s) lead to the essentially complete
consumption of the sulfate on these materials as indicated by EDS of the surface
before washing and by the semiquantitative wash water analysis (table II). This
shows the fundamental tendency for the reaction
Nags04 + SiOq = Na silicate + SO3 (1)
In many cases, no sulfate could be detected in the washwater and the silicate
overall stoichiometry was about NagSiO3. This is probably indicative of a
gradient in the silicate composition as the silicates in equilibrium with
crystalline silicas are richer in silica than this average composition. Impurities
in large amounts slowed this reaction as surface phases were formed on
preoxidation (Mg silicates, yttrium silicates)(1p3).
For the purer materials not preoxidized it is clear from these results that
while reaction (1) occurs under basic hot corrosion, it is the rate of oxidation
which controls the amount of silicate formed and the transport through the
silicate and sulfate layers decreases the rate of oxidation as the scale thickens.
Therefore the scale is protective. This is indicated since the preoxidized
samples not only formed more silicates which was water soluble but they
oxidized more as indicated by greater weight gains (after washing) than the same
materials not preoxidized. Since the preoxidation scales were formed at 14OOoC,
they had crystallized and cracked extensively during cooling before application
of the salt and were not protective during the hot corosion.
Under basic conditions the sulfate wets completely the sample surface and
reacts with the scale as it forms leading to very thick silicate layers even on the
purer silicon nitride (table I). This has been documented by Mayer and Riley(G)
who reported a transient acceleration of oxidation with injection of NagCOS.

Table II Wash Water Analysis From

Basic Hot Corrosion (168 Hours, 1000°C)
Material Si02 SO3 Na20

At% At% At%
SC SC 25 33 42
SC SC (PO)(l) 52 0 47
CVD SC 22 27 52
CVD SC (PO) 50 2 48
CVD SN 0 62 38
CVD SN (PO) 53 12 35
SN SN -_ -_ --
SN SN (PO) 59 13 27
SC SC = Silicon Carbide single crystal
CVD SC = CVD silicon carbide
CVD SN = CVD silicon nitride
Sin SN = Sintered silicon nitride
(1) PO = Preoxidized in 02 for 10 hours at 1400°C

With Na2S04 the reaction is milder(3~4~71 as all sulfate was not consumed in 168
hours. Also the extent of the reaction is affected by impurities and the purer
materials formed thicker scales except possibly for the hot pressed silicon
carbide which contained alumina as a sintering aid. The alumina showed no
preference for segregation in the scale either on oxidation or hot corrosion. The
aluminum present in the scale tended to stabilize the glassy silicate phase. As
discussed in a previous report, magnesium and yttrium tend to segregate into the
scale on oxidation. They complicate also the hot corrosion conditions. MgC in
particular forms magnesium silicates and tends to promote basic conditions on
the surface under acidic environmental conditions. In general the acidic
corrosion on the impure samples formed continuous silicate layers, these layers
were quite thick for hot pressed silicon nitride and sintered silicon nitride and
were similar in morphology to those observed under basic conditions.
Early in the research it became apparent that some hot corrosion occured
under acidic conditions in addition to the expected corrosion under basic
conditions. Data on pure materials for shorter times were needed in order to
understand the fundamental mechanisms of hot corrosion under acidic conditions
and establish if it could lead to significant degradation. Therefore the single
crystal silicon carbide and the CVD silicon nitride were exposed for 1 to 24 hours
under standard acidic conditions at 1OOOoC. In other experiments the
temperature and Na2S04 loading were varied as necessary.
Under acidic conditions the sodium sulfate does not wet the silicon’nitride
and silicon carbide completely. The wetting angle was of the order of 40° for
both materials. The drop size distribution was bimodal (fig. 2). There was a
variation in the values of the wetting angles measured but it was not consistent
with drop size, exposure time or nature of the sample. However the values
observed were greater than measured on fused silica under similar conditions.
The evolution of the deposit as a function of time is shown in figure 2 for single
crystal silicon carbide. The deposit which was sprayed as a layer around 120°C
breaks up as the sulfate melts and morphologies similar to those in fig. 2 were
observed after a few minutes of exposure at 1000°C. As shown in the figure the
smaller droplets tend to disappear with time, probably as a result of a coarsening
process. The large drops do not move on the surface and they are usually
surrounded after 10 to 24 hours by bands free of small droplets (fig. 2).
The kinetics of acidic corrosion for single crystal silicon carbide and CVD
silicon nitride are shown in figures 3-5 for times from 1 to 24 hours at 1000°C.
The plots of figures 3 and 4 were for thicknesses of vitreous oxide measured
between, and generally away from the sulfate droplets. All the thicknesses in
these two figures were obtained by the same method (X-ray spectroscopy(2)) in
order to minimize the influence of systematic errors in the interpretation. The
data for silicon carbide plots as good straight lines as function of the square root
of time suggesting a parabolic behavior. The kinetics for silicon nitride do not
show clearly this trend and are plotted as a function of time in figure 3a and as a
function of the square root of time in figure 3b. The lines for oxidation data
from previous workers and interpolated from previous experiments in this
research indicate that the hot corrosion, even under acidic conditions enhances
the formation of the oxide layers in all cases for the purer materials even
between the salt droplets.
The amount of sodium sulfate applied per unit area of surface was varied
from 0, 0.1 and 5 mg/cm2, the standard surface loading. While higher than that
for oxidation, the growth of the vitreous scale between the sulfate droplets is
independent of the surface loading as shown in fig. 4. The data for different
1000- -0
CVDSN /
/
2 /
J; /
v) .
z / Hot Corrosion
5 /
i
b .
/
500. /
---Q
/
/
/ Oxidation
168
TIME(HR)
Figure 3. Thickness of scale formed in the acidic hot corrosion of CVD

silicon nitride at 1000°C (5mg/cm2 of NaZSOq).
(a) linear plot
Appendix D-Hot Corrosion of Silicon Nitride and Silicon Carbide
147
/
/
/
/
/
//
/’
b
/O
d
01
I--_-_-l I I L
I.0 5.0 9.0 13.0
VT (h OUTS ) ‘L
Fimre 3. (b) parabolic plot measured between the drops of

Figure 3(a).
C-side
1.0 5.0
7rme
‘12 (/+%)
Figure 4. (a) Thickness of scale formed at 1OOOoC in the Acidic hot

corrosion of single crystal silicon carbide. Three different
loadings of NaZSO4 on the surface 5mg/cm2 (A), less than
0.1mg/cm2 (“) and 0.0mg/cm2 (0) fell on the same parabolic
plot. The data for silicon and carbon side fall on two
separate lines. (parabolic)
H(t)
-- --- -- -- -- A---
I
0 5.0
TIME “*H “*I
Figure 4. (b) Parabolic plots comparing of the data of figure 4(a) for single
crystal silicon carbide (solid lines) with the oxidation of
silicon (D + G )8 and single crystal silicon carbide (for
carbon side (“) and silicon side (A) (Harris) 22 at IOOOoC.
0
SPHERULITES
UNDER DROPS
O-SC(C)
22
A- SC(Si) ____I_
O-CVDSN
/
0
0
I3
A i-.---.
l:o ’ 510
l/2 I/2
TIME (HR )
Figwe 5. Thickness of spherulites (oxides) formed under the sulfate droplets

for the C-side and Si-side of single crystal silicon carbide and
CVD silicon nitride (Acidic hot corrosion at 1000°C). The solid
line corresponds to the data of figure 4 for the acidic hot
corrosion of C-side silicon carbide.
surface loadings fall on the same lines. No salt was applied initially on the
samples with 0 mg/cm2 loading but they were exposed with the other samples,
and they picked salt apparently through vapor transport. The independence of
the data on the surface loading is consistent with a constant salt activity
maintained through the vapor phase and a fixed pSOS (1.5 x 10e3 atm) setting a
constant aNa for the experiments at 1000°C according to the reaction
Na2S04 = Na20 + SO3
Kg = aNa20.pS03/aNa$304
No sulfate could be detected on the vitreous scale by SIMS and ISS, but sodium
was found by ISS in the surface layers of the scale suggesting the formation of a
silicate layer about 10 A thick at the surface of the scale. This conclusion and
the role of sodium in the acidic hot corrosion will be discussed in detail below,
but it underlines that the role of sodium is different in acidic corrosion from that
in basic corrosion.
The rate of oxide build up under the sulfate droplets formed during acidic
corrosion on the purer silicon nitride and silicon carbides is much greater than
that outside the droplets except for C-side SIC as shown in fig. 5. The silica
crystallizes rapidly under the salt forming cristobalite spherulites. The evolution
of the average thickness of this spherulitic material at the center of the
droplets, measured after washing, follows approximate parabolic behavior which
is close to that calculated from Deal and Grove%(*) data for oxidation of silicon
at that temperature. However this is fortuitous since this part of the scale is
fine crystalline material which grows from the melt.

Acidic Hot Corrosion of Silicon Nitride
The data of figure 3a, is fairly linear with an intercept with the vertical
axis at about 80 A. This suggests an initial oxide layer which is not consistent
with the present experiments. The samples were cleaned in HF prior to the hot
corrosion experiments so that an oxide layer close to 100 A is not expected
although some oxygen and water are always readsorbed, as established in the
electronic industry. The silicon carbide samples which were pretreated like the
silicon nitride do not have a significant y-intercept (fig. 4). The silicon carbide
data extrapolate essentially through the origin. The method of thickness
measurements by X-ray spectroscopy is also pushed to its limit with the very
thin oxide layers formed, but a systematic error in the measurements does not
appear responsible for this intercept. Another explanation is that oxynitride is
formed on oxidation as discussed below.
Laser ellipsometry measurements on the product layers formed on the
samples of table III were consistent with an index of refraction of 1.75 and the
thicknesses given in the table. This index of refraction is much higher than 1.46,
the one for vitreous silica. The corresponding thickness are larger than those of
figure 3 measured by X-ray spectrometry (Oku). IR measurements were not
conclusive since the spectrometer could not reach the major peak for the silicon
oxynitride. The method in general did not appear sensitive to the presence of
silicon oxynitride and was not pursued further. Tressler et al. recently reached
the same conclusion after extensive specular reflection FTIR spectroscopy@).
ESCA data is still expected. The thickness of the glassy product layers formed
on CVD silicon nitride were generally lower when measured by X-ray
spectrometry than measured on cross sections in the SEM. For the Sic single
crystal the thicknesses measured by X-ray were larger than measured in the
Table III
Scale Thickness of Selected Samples (A)
X-ray (4) Ellip (5)
A A
CVD SN (1) 450 605
CVD SN (2) 120 356
CVD SN (3) 130 326
(1) CVD silicon nitride exposed to gaseous corrosion 1% SO2-SO3

balance 02 for 168 hours at 1000°C.
(21 CVD silicon nitride exposed to acidic hot corrosion for 24 hours
at 91OOC.
(3) CVD silicon nitride exposed to acidic hot corrosion for 24 hours
at 955OC.
(4) X-ray spectrometry [2]
(51 Laser beam ellipsometry courtesy Terry O’Keefe Westinghouse

R & D.
SEM(2). The of thicknesses in the SEM on very thin layers such as those formed
on the silicon nitride are difficult and subject to errors but the trend in the
thickness measurements is consistent with an oxynitride layer formed on the
silicon nitride. This is in qualitative agreement with the results obtained by
ellipsometry. Further analysis of the ellipsometry data indicated that it was
consistent with a graded film of oxynitride with possibly an overlayer of silica
50-100 A thick ( in contact with the atomsphere). Raider et al.(lO) reported
graded layers of oxynitride apparently without overlayer of silica.
The present results are in line with an initial formation of silica which
would be rapid and controlled by the surface reaction:
Si3N4 + 3 02 = 3 Si02 + 2 N2 (3)
giving very steep linear kinetics for short times in figure 3. As a result of
reaction 3, the pressure of nitrogen rises and the oxygen potential decreases at
the interface favoring the formation of oxynitride which grows between the
silicon nitride and the vitreous silica as proposed previously(ll). Diffusion
through the vitreous oxynitride is expected to be more difficult than through the
silica(12) since it is more tightly bound. The transport of oxygen through this
layered scale becomes rate controlling at steady state giving parabolic
kinetics(l3). The sequence of the three mechanisms and the distortion of the
thickness data due to the formation of oxynitride instead of the silica assumed in
the thickness calculation from X-ray spectrometry are responsible for the linear
shape of the plot of figure 3a. The scale grows more slowly after 24 hours. The
thickness after 168 hours was measured as 1 pm by X-ray spectroscopy and 1.2
urn by imaging of cross section in the SEM. However data for longer times yet
would be required to establish if the growth had reached steady state and
became parabolic after one week exposure. Some data was obtained at lower
temperatures and shorter times in an attempt to determinethe activation energy
of the linear process but this was not justified because of the complexity of the
early kinetics.
The formation of silicon oxynitride during the oxidation of silicon nitride
has been proposed by previous workers but often not demonstrated convincingly
since they usually used one technique which could not be interpreted
unambiguously. Hench used IR spectroscopy (141, others a combination of ESCA
and depth profiling and concluded that an oxynitride was formed during the
oxidation of silicon nitride. Others report only silica as the product of this
oxidation(l@. In the electronic industry silicon oxynitride is prepared readily
and used as passivating layer. Comparison of the vapor deposited oxynitride and
material formed by thermal oxidation of silicon nitride deposits on silicon
indicated that graded oxynitride layers were formed by thermal oxidation(lO).
ESCA and index of refraction measurements were used to reach these
conclusions. Tressler et a1.(gy13) performed extensive characterization of scales
formed on oxidation of silicon and silicon nitride using many different methods
such as etch back rates, SIMS, FTIR, ellipsometry and concluded that a thin layer
of silicon oxynitride formed under the silica.
The steady state kinetics for a layered scale growing by transport through
the layers are controlled by the mobility of a specie through the scale. This was
treated in detail by Yurek et al.( 17) for the diffusion of metal ions through two
layers of oxide scale to oxidize at the gas-scale interface. In the present case
oxygen must diffuse through two layers, the silica and then the oxynitride in
order to oxidize the silicon nitride. Part of the oxidation occurs at the
oxynitride-silica interface and the other at the silica-silicon nitride interface. It
is assumed that oxygen transport is controlling(g~l*) as generally accepted for

dry oxidation of silicont lg). Since diffusion through the oxynitride is more
difficult than through silica, the oxynitride layer under steady state conditions
should be thinner than the oxide layer. Based on(17), it is expected that the
thickness of the layers under diffusion controlled steady state will be given by
y/x = Kp’oxy/ Kp’sil (4)
in which y, x, Kp’oxy, Kp’sil are the thicknesses and the rate constants for the
formation on silicon nitride of individual layers of oxynitride and silica,
respectively. Kp* for the total thickness y+x can be written
Kp* = (1 + Y/x)~ Kp’sil (5)
and if y is small Kp* = Kp’sil. Then the overall kinetics would be similar to those
of a silica layer forming directly on silicon nitride. This discussion assumes
distinct oxide layers. To a first approximation the formation of a graded layer
of vitreous oxynitride would not change this conclusion except that this layer
will spread over greater distance than a layer of specific composition. This is
expected from the previous kinetic arguments since the mobility of oxygen
through the oxynitride is expected to increase as the nitrogen content decreases.
The detailed shape of the nitrogen distribution would depend on the specific
concentration dependence of the diffusion coefficient through the layer. The
conclusion is consistent with the graded layer found by ellipsometry and the
large oxygen content of the scale found by X-rays. The index of refraction of
the scale appears high for the proposed graded structure of the scale, this may
be due to the fact that steady state was not achieved yet in the ellipsometry
samples. The conclusions of this analysis are in agreement with the extensive
work of Tressler et al.cg) on the oxidation of silicon nitride except that they
found a lower index of refraction to films grown under different conditions.
They find high actication energies for both the linear and parabolic constants of
a Deal and Grove analysis(*) for the oxidation of CVD silicon nitride between
1100 and 13OOOC. They conclude that the low rate of oxidation and
highactivation energies compared to silicon must be due to transport of oxygen
through the silicon oxynitride and it must be rate-controlling. The 110-120
kcal/mole for the parabolic activation energy seem to rule out the rate control
between oxygen molecular or ionic diffision through vitreous silica. While
oxygen transport, particularly molecular, through the oxynitride is expected to
be slower and with a higher activation energy than for fused silica the previous
analysis based on the formation of layered scale suggests that it cannot be rate
controlling.
Discussion of the Hot Corrosion of Silicon Nitride
The oxidation of silicon nitride is very slow. The larger oxidation rates(20)
observed in previous studies were due to impurities added for sintering as shown
in this research in the gaseous corrosion (and hot corrosion) of CVD, sintered and
hot pressed silicon nitride as shown in table I. The oxide layers formed on the
high purity silicon nitride (CVD) were thinner than observed on the high purity
single crystal or CVD silicon carbide under the same oxidation or hot corrosion
conditions. This may not establish that silicon nitride is intrinsically superior to
silicon carbide in this respect but it is certainly not inferior as stated in older
literature.
The oxidation of silicon nitride is much slower than that of silicon although
at 1000°C in both cases a vitreous oxide scale is formed and it appears oxygen
transport through the scale controls in both cases in that temperature range but
the detailed mechanisms are apparently different and not established for silicon
nitride. The general features of acidic corrosion already discussed are similar
for silicon nitride and silicon carbide and in particular the growth of the scale
was increased under acidic corrosion. The presence of Na20 on the surface of
the sample, supplied as Na2S04 (or Na2C03 etc...) modifies this oxidation. The
behavior of the Na20 depends on its activity in the sulfate melt as determined
by reaction 2. From the equilibrium constant Kp, the activity of sodium oxide is
then determined by the SO3 pressure as described in previous reports. Under
acidic conditions (here at 1000°C, a pSO3 of 1.5~10~~ atm was selected) the
aNa is small and the sodium sulfate reacts little with the silica formed by
oxidation of the silicon nitride or silicon carbide as indicated by the high values
of the wetting angles on bulk silica(5) and on the scales formed on the purer
specimens as reported above. Some penetration of Na20 into the silica scale is
expected, loosening the network by the reaction
Na20 + -Si-O-Si- = 2 -Si-O- + 2 Naf (6)
Kg = [Csio12.[CNa+12/ aNa20.Csiosi
* [Csio14/aNa20
This reaction should be applicable also to the oxygen bridges of the oxynitride.
As a result of this opening of the network molecular diffusion of oxygen and
nitrogen is increased through the vitreous scales, thus generating a higher rate of
oxidation. It is likely that the sodium ion diffuses to the reaction interface
where it is reduced as proposed for silicon, it was proposed also that it would
help in the silicon-oxide structural transition at the interface(21). The sodium
may affect also the stability of the oxynitride phase. In the present experiments
sodium has been suggested in the scale with the electron microprobe and ISS, but
the evidence is not conclusive. Only very small amounts of sodium need to be
dissolved in the network according to reaction 6 in order to increase interstitial
diffusion in silica glass as discussed later for SIC. Under acidic conditions the
rate of oxidation of CVD silicon nitride is increased between the droplets
probably by the mechanism discussed above. A very thin (slOA) layer rich in
sodium was found by ISS on the scale between the droplets, it did not contain any
sulfur. This adsorbed layer fed by the droplets by surface diffusion or vapor
transport would supply the sodium into the glass.
The silicon nitride under the sulfate droplets oxidizes more rapidly than
outside and crystallization starts early (in the first hour). After 24 hours of
exposure and washing off the salt the region which were under the droplets stood
out(l). The corresponding thicknesses in Table I were measured outside the
droplets. The crystal-glass interfaces provide high diffusivity paths normal to
the surface of the sample. Some liquid is present in the intercrystalline regions
in which sodium and other impurities concentrated during the crystallization.
Thus the crystallization provides fast transport paths for oxygen leading to
thicker scale under the sulfate drops and greater penetration of the scale under
the intercrystalline regions such as the interfilamentary regions of the
spherulites shown in figure 6.
Under acidic conditions for pure materials, vitreous scales are formed
which crystallize very slowly outside the droplets. Some crystallization occurs
in between the droplets but it is sparse and the growth is extremely slow as
judged by a few very small spherulites disperse in the glass even after 168 hours
exposure. The difference in growth rates is illustrated in figure 6 where
nucleation started under the edge of a sulfate drop. The growth away from the
drops appear slower yet than in this example. As discussed in detail in appendix
A, the crystallization of vitreous silica depends on the Oxygen ion activity (or
aNa20) in the structure, and the presence of defects in the network. The sodium
oxide reaction with the network (reaction 6) provides those defects and tends to
Acidic hot corrosion of CVD silicon nitride at lOOOoC. Under the

~6. sulfate drops (washed off in micrographs) the oxide crystallizes as
spherulites. After 24 hours the spherulites coarsen to globular
arrays.
occur preferentially at the interface between the salt and the sulfate where
aNa is highest. Under acidic conditions, in close proximity to the atmosphere
(outside the sulfate droplets) the gaseous potentials remain close to pSO3~1.5 x
1V3 atm and ~02~1 atm and the aNa is low. Under the sulfate droplets, the
oxygen and SO3 are more difficult to replenish as they are depleted by the
oxidation reaction at the silicon nitride-oxide or sulfate interfaces. Based on
equation (2) and
so3 = so2 + l/202 (7)
K7 = (p02)li2 .pSO2/pSO3
The activity of Na20 is raised under the sulfate droplets (it is higher than
outside) and the crystallization is promoted under the droplets.
Acidic Hot Corrosion Of Silicon Carbide
The morphologies of the samples of single crystal silicon carbide after
exposures were described previously (lp3) and for short times are generally
similar to those described above for the silicon nitrides (see fig. 2). They are
characterized by the formation of sulfate droplets separated by smooth vitreous
regions. The spherulitic growth of cristobalite occurs rapidly under the drops
and it is generally similar to the behavior observed on the silicon nitride (fig.6).
Preferential attack occurs also under regions between the cristobalite crystals as
shown in Figure 7 in which the silicon oxide has been etched away.
The kinetics of acidic hot corrosion for single crystal silicon carbide at
1OOOoCare plotted as a function of the square root of time in figure 4 and they
fall on two distinct lines both higher than the oxidation data of Harris(22)
sketched on the figure. The two lines have been associated with the carbon side
and the silicon side of the single crystal. In agreement with the oxidation results
of Harris and Tressler et al.(g) the carbon-side is the fast or thick side and the
silicon-side is the thin or slow side. Hot corrosion data was obtained also at
955OC and 910 oC. The data at 910°C are plotted in figures 8 and 9 and both
sides still show parabolic behavior at that temperature. Activation energies for
the growth of oxide scale in the temperature range 910-1000°C were determined
from the parabolic constants B* calculated from x2 = Bt in which x is the scale
thickness at time t. The thickness measurements were all performed by one
method (X-ray spectroscopy) in order to minimize the influence of systematic
errors. The plots of figures 10 and 11 give activation energies of 34 and 118
kcai/mole for the carbon-side and the silicon-side, respectively.
Silicon carbide is a polar crystal and has different basal surfaces (0001) and
(0001)(23) which can be differentiated by a number of methods based on different
surface morphologies after high temperature wet oxidation or attack by fused
salts(24). The present hot corrosion experiments resulted in milder attack but it
was clear after extensive experience that slight differences in morphologies
existed between the two surfaces which allowed their identification. Namely
after hot corrosion for long times, the silicon side between the droplets appeared
as-polished except for some interference colors on the thicker scales and the
carbon-side was rougher, duller in texture. Different oxidation behaviors have
been reported(g*22) for the two sides. This difference in behavior was no longer
apparent at high temperatures (1400-15OOoC). Similar basic hot corrosion
behaviors were found for the two sides in this research at 10000C(1~3~4(. Under
acidic hot corrosion, the vitreous scale between the droplets grows faster on the
*B was determined also by the Deal Grove formulation but it gave the same
results since the data plotted as functions of t1/2 falls on good straight lines
extrapolating through the origin (fig. 3,7,8).
~7. Acidic Hot CorroSion 0( silicon carbide single crystal (lOOOOC).

The substrate was attacked under the spherulites as shown after
disSOlution 0( oxide in HF.
31OC KINfllCS - C
5000
4000
3000
3
x
I?
5
0
5 2000
’ 000
0
0.00 2.00 4.00 6.00 a.00
SQRT(TIME(HR))
Figure 8. Kinetics of acidic hot corrosion of C-side single crystal silicon

carbide at 910°C (parabolic plot). Note reproducibility of data.
Appendix D-Hot Corrosion of Silicon Nitride and Silicon Carbide ‘I65
91OC KINETICS - SI
600
500
4oc
z
x
Y 300
$
I!
F
200
100
0
0 2.00 4.00 6.00 8.00
SQRT(TIME(HR))
Figure 9. Kinetics of acidic hot corrosion of Si-side single crystal silicon

carbide at 91OoC (parabolic plot).
ACTIVATION ENERGY - C
1400
i 3.80
13.60
13.40
1320
13.00
i 2.80
12.60 1
0. 78 0.80 0.82 0.84

1000/T
Figure 10. Arrhenius plot for the acidic hot corrosion of C-side single crystal
silicon carbide.
ACTIVATION ENERGY - St
13.0
12.0
1 1 .O
10.0
8.0
3.78 0.80 0.82 0.84
1000/-T
Figure 1. Arrhenius plot for the acidic hot corrosion Si-side single crystal
silicon carbide.
carbon side than on the silicon side as reported by Harris for oxidation. However
on figure 4 both sides show well defined parabolic behaviors while in Harris’
results the carbon-side was parabolic and the silicon side had linear kinetics over
the broad time scale of the experiments. For the hot corrosion the apparent
activation energy of the slow side (Si) is much larger than that of the fast
carbon-side.
Model for the Oxidation and Hot Corrosion of Silicon Carbide under Acidic
Conditions
In order to discuss the hot corrosion of silicon carbide under acidic
conditions in between the sulfate drops, it is important to understand its
oxidation. The formation of silica scales on silicon and on silica formers is
dependent on the transport of oxygen to the substrate for the oxidation to occur.
The molecular diffusion of oxygen through the scale might be rate controlling as
established for silicon dry oxidation. It has been proposed(g) that as the
temperature is increased contributions are made by the network oxygen
diffusion. Others have argued that the diffusion of CO produced in the oxidation
of silicon carbide is rate controlling and since the size of the two molecules is
about the same it is difficult to distinguish(z5). Considering recent
experiments(g) it is concluded that transport of oxygen is controlling the
oxidation of silicon carbide at low temperatures (below c 14OOoC). In earlier
work it was concluded that the diffusion of C-products was controlling at higher
temperatures(l*). They found two regimes in the oxidation of C-side silicon
carbide from 1200-1500°C, one with low activation energy (123kJ/mole) at low
temperatures and the other with high activation energy at high temperatures
(216kUmole) at 1 atm pressure of oxygen. However the oxidation of silicon-side
silicon carbide had only one regime associated with a high activation energy
(240kJ/mole in 1 atm of oxygen). The kinetics of the carbon-side oxidation
controlled by oxygen transport were modelled satisfactorily assuming that
several processes contribute to this transport; vacancy diffusion in the network
and molecular interstitial diffusion were considered specifically in the
analysis(g). The molecular diffusion dominates at low temperatures 1200-1350°C
while the vacancy diffusion dominates at high temperatures 1350-1500°C. In the
temperature range of the present experiments the oxidation of the carbon side is
similar to that of silicon, except slower with the same activation energy and
linear dependence of B on the oxygen pressure (gl. The molecular diffusion of
oxygen is dominating the kinetics. At low oxygen pressure a vacancy mechanism
appears to control the oxygen transport through the scale. The silicon side is
slower and has a high activation energy for the complete temperature range of
1200-1500°C(gl. Harris reported linear behavior at low temperatures. These
results must be considered in terms of the defect structure of the silica scale
and the surface reactions, then mechanisms will be proposed for the oxidation
and then for the hot corrosion.
Proposed Model
The following model is concerned with the lower temperature range
around 1000°C. It addresses:
il the different rates of oxidation and hot corrosion for Si-side

and C-side silicon carbide
ii) the enhancement of oxidation under acidic hot corrosion (the

role of Na20)
iii) the activation energies for hot corrosion
iv) the magnitude of B, the parabolic constant, relative to that

for silicon oxidation
It is based on:
0 the variation in stoichiometry of vitreous silica with its

formation under different ~02 at the reaction interfaces.
This variation in stoichiometry and the associated defect
structure modify the transport of oxygen through the scale
ii) the oxidation results of others summarized above
iii) the oxygen transport model of Tressler and Speartg) applied

to lower temperatures
iv) the surface analysis results of Muehloff et a1.(26)
v) our acidic hot corrosion results.
Defect Structure and Stoichiometry of Silica
Vitreous silica is not often considered non-stoichiometric in the glass
literature although glassy structures which are great solvents allow greater
variation in composition than the corresponding crystals. It is known that the
silica formed under oxygen deficient environments crystallizes more slowly than
similar silica formed under more oxidizing conditions(27). Silicon rich silica has
been prepared in the electronic industry. Fratello et a1.(28) discussed the
influence of OH impurities on the crystallization of silica on the basis of the
defects and stoichiometry of vitreous silica. While the defect structure of
vitreous materials is relatively controversial for strpctural defects and mass
transport, the electronic defects have been studied in great detai1(2gy30). In
general it is accepted that the SiOg glass structure is built with SiO4 tetrahedra
connected through oxygen bridges to form a continuous random network. Thus
structural defects may extend to any kind of orderin$31). The major defects in
silica have been reviewed by Motttao); they are dangling bonds in particular the
well known single bonded oxygen, but also 3 bonded silicon. They include also
oxygen vacancies in oxygen bridges as well as Si-Si bonds. The single bonded
oxygens tend to form in silicate glasses to accommodate extra oxygens
associated with network modifying ions in the structure. It is reasonable to
expect that oxygen vacancies and Si-Si bonds are favored under oxygen deficient
conditions. It is anticipated also that compensating defects can be generated
thermally. Reaction 8 generates 2 single bonded oxygens and an oxygen vacancy
2 Si-0-Si = Si- -Si + ZSi-0

nil = Vo + 2SiO (8)
K8 = [Vo] [SiO12 = Cv [SiO12
in which [V,] = C, is the concentration of oxygen vacancies Si- -Si in the
network. Under low oxygen pressure oxygen vacancies could form by the
equation so familiar for crystalline oxides
00 = Vo + I/2 02 + 2e
K = [e12 [Vo] [~02]~/~ = 4Cv3 [~02]l/~
where the vacancies are formed at the oxide-gas interface and the oxygen is
supplied directly by the atmosphere and the electron concentration [e] is
assumed to be supplied only by the reaction so that [e] = 2 Cv. It is more
appropriate here to consider that the oxygen molecules are dissolved into the
glass at the scale-gas interface with the concentration of dissolved oxygen
proportional to the oxygen pressure through Henry’s law. Reaction 9 occurs
inside the scale under pO2 lower than at the interface and the molecular oxygen
is that dissolved in that environment (concentration Cm) so that
Si-0-Si = Vo + l/2 02 dis + 2 Si- (9)
Kg = 4Cv3 Cm112
in which the electronic charges are associated with silicon dangling bonds and Kg
includes the dissolution of 02 gas into the glass. The formation of electrons in
equation 9 may have appeared unrealistic since silica is an insulator. However,
they exist commonly in vitreous silica in trapped form such as dangling bonds on
silicon shown in equation 9. It has been shown that thermally grown silica
contains positive and negative charge centers. Silica films support electronic
conduction during anodization(32).
The vitreous structure is built up on 5 and 6 member rings which are
randomly distributed on 3 dimensions. These rings provide tortuous paths for the
diffusion of gas molecules and impurity network-modifying ions smaller than the
size of the windows in the structure, a little over 3A(33). In the growth of a
scale the growth flux may introduce anisotropy to this structure by the
formation of more aligned channel in the flux direction(34). Larger channels less
than 50A in diameter have been postulated in order to explain silicon oxidation
and reported from observations in the TEM(35).
Oxidation of Silicon Carbide
The oxidation of silicon carbide at the reaction interface generates
very low oxygen pressures which cannot be calculated exactly except in presence
of free carbon, for the ternary equilibrium Sic-SiO2-C. The results of this
calculation which gives pO2 of the order of 10m30 atm at 1000°C can be
considered a lower bound for the oxidation of materials not containing excess
carbon. If active oxidation does not occur the silica formed at the reaction
interface is expected to be oxygen deficient with the formation of corresponding
defects: vacancies, 3-bonded silicons and possibly Si pairs. Thermal equilibrium
of the type of equation 8 will suppress the concentration of single bonded
oxygens
CJoxid = mNB/2x (10)
in which CJoxid is the sum of the oxidant fluxes which contribute to the
oxidation and N, B, x are the number of oxygen molecules incorporated in 1 cm3

of scale, the parabolic constant and the scale thickness, respectively. The
constant m is determined by the mass balance between oxidants and products(g).
Two types of contribution are made to the transport of oxygen, molecular in
which molecules of oxygens are transported through channels, in the vitreous
silica and network contribution which can occur through the various network
defects. Assuming that the major mode of network transport is by a vacancy
mechanism equation 10 becomes
DvCv + DmCm = mNb/2 (11)
in which Dv, Dm, Cv and Cm are respectively the vacancy and molecular
diffusivities and concentrations, respectively. Therefore the kinetics of
oxidation of silica formers may depend on the relative contribution of the two
transport paths which are determined by the stoichiometry of the scale at a
given temperature. Different oxygen pressures on formation of the scale will
result in different stoichiometries through equations 8 and 9. Under low oxygen
pressure the quantity of dissolved oxygen (Cm) will be depressed and on the basis
of equation 9 the concentration of oxygen vacancies (Cv) will be increased
tending to favor transport by a vacancy mechanism over transport by molecular
diffusion. This in line with the proposal of a gradual change from molecular
mass transport at high pressures to ionic mass transport at low pressures for the
thermal oxidation of silicon(g). Based on equations 8,9 and 11 one can explain
the oxidation and hot corrosion of silicon carbide in the low temperature range in
which oxygen transport is assumed to be rate controlling.
First let us consider the oxidation of silicon in 1 atm of dry oxygen,
molecular diffusion is dominating the transport@) and equation 11 becomes
simply DmCm = NB/2. The parabolic constant B is proportional to pO2 and it has
an activation energy Q = 28 kcal/mole which is similar to that measured for the

permeability of molecular oxygen through bulk vitreous silica(331. As discussed
later, if one assumes that the silica scale formed on C-side silicon carbide is
slightly oxygen deficient because it was formed under a low pO2 (Cm decreased),
then by equation 9, Cv increased. Although it is smaller than for the oxidation
of silicon, DmCm is still greater than DvCv. Molecular transport of oxygen
dominates but from equation 11, B is smaller than for silicon. This was observed
by Tressler et al.tgl, this oxidation has activation energy (27kcaVmole) and an
oxygen pressure dependence similar to that measured for silicon oxidation. On
the other hand if the scale on the silicon-side of silicon carbide was formed
under very low oxygen pressure, by the same reasoning Cm is greatly depressed
and Cv increased since Cv * Kg/(Cm) li6. So that now DvCv>>DmCm and the
network diffusion controls the oxidation resulting in a much lower B, a higher
activation energy (55-60 kcal/mole) and a low pO2 dependence(gl. Therefore
with the proposed model it is possible to derive a consistent picture for the
oxidation results obtained by others for silicon and silicon carbide. It postulates
different conditions of formation of the scales on silicon-side and carbon-side of
silicon carbide which are discussed below.
As discussed earlier the two sides of single crystal silicon carbide oxidize
at different rates for long periods of time at lower temperature below 1300° C.
This indicates that under those conditions the two sides maintain their specific
character as they oxidize. The two sides maintain their separate identity after
grinding and polishing as shown in the present experiments and after ion
bombardment (sputtering) to mix the surface. It was shown also by Muehloff et
a1.(261 that the two surfaces had different stabilities under low pressure
oxidation. Under ultra high vacuum the carbon side was observed to cover with
carbon above 900 K while the Si-side did not start to graphitize until 13000K. At
that temperature the graphitization of the carbon-side was extensive. The two
sides of silicon carbide maintained qualitatively their relative rates of oxidation
during the early oxidation for thicknesses of silica up to 20A, even after surface
mixing due to sputtering. This carbonization of the of the carbon side was
explained the evolution of silicon under the very low oxygen pressure at the site
of the reaction
Sic + l/2 02 = CO + Si (12)
Under 1 atm of oxygen, it is assumed that this difference in stability of the two
sides is maintained and that silicon still tends to volatilize from the C-side,
however both carbon and oxygen are oxidized in the reaction
SIC + 02 = SiO + CO (13)
In the pO2 range (s~O-~O atm) which is expected at the reacting interface SiO is
the prominant vapor specie in the Si-0 system(36). As shown in figure 12, it is
possible by this reaction to always maintain the order of the Si and carbon atoms
at the interface so that on the C-side carbon is always at the surface of the
silicon carbide as it oxidizes thus maintaining the C-side behavior. The SiO is
oxidized as it meets with higher oxygen pressure near the interface but not at
the interface and the silica forms as a rough porous scale and a rough interface
results from the active oxidation reaction. As the oxidation proceeds the SiO
oxidizes in the pores so that the scale porosity is filled. The scale is formed
under higher pO2 than is found at the interface and it will be more
stoichiometric.
On the Si-side, the silicons are at the surface and it is proposed that
oxygen is adsorbed first on the silicon (figure 13) and the carbon is oxidized by
penetration of oxygen through this oxide layer always maintaining silicon
between the oxide and the SIC in line with the Si character of that side. This in
1) SiC
Figure 12. Model for oxidation of C-side silicon carbide. A rough interface
is formed and the C-side is maintained at the interface during
oxidation. The SO2 is formed away from the interface.
Sic
Sic
- Atmosphere
3) I02 I
01
Sib
sic
Figure 13. Model for oxidation of Si-side silicon carbide. The silicon side
nature of the interface is maintained.
line with the results of Harris (linear kinetics of oxidation) as the CO would form
under the silica layer and desorption of CO from that position could control the
oxidation. At higher temperatures or in presence of sodium this desorption is no
longer rate controlling and parabolic kinetics are observed. The scale is formed
under very low oxygen pressure and it is expected to be very oxygen deficient as
discussed earlier. The scale formed under these conditions would be more
oxygen deficient than that formed on the C-side. In the oxidation of SiC the
network transport and the molecular transport add to supply the oxygen. This
interplay between the two types of transport processes is underlined as the
temperature of oxidation is increased and their relative contributions change for
the C-side and the Si-side. As the temperature is increased, transport by the
vacancy and other network mechanisms with high activation energies increases
rapidly and becomes a major contributor to the oxidation of the carbon side. The
oxidation of the silicon-side has been increasing also since it was dominated by
the network transport processes. As the temperature is increased the mobility
increases also in the SiC interfaces and they can no longer be differentiated by
the mechanisms of fig. 12 and 13 and around 1350oC the two sides oxidize at
about the same rate with the same activation energy as reported by Tressler(9).
Therefore the proposed model is qualitatively consistent with the data on the
oxidation of silicon and silicon carbide, now we shall apply it to our acidic hot
corrosion results.
Acidic Hot Corrosion
As discussed earlier acidic hot corrosion is oxidation enhanced by the
presence Na2Q at a low activity. Na2Q enters into the silica structure by
reaction 6 and generates single bonded oxygens thus modifying the defect
concentrations in the vitreous silica. Single bonded oxygens are oxygen excess
defects which lower the concentration of oxygen deficient defects through
reactions of the type
2 Si-0-Si = Si- -Si + 2 Si-O- (14)

K14 = [Vo] [Si-O12 = Cv [Si-012
Therefore as Na20 is introduced into the scale, it tends to increase the single
bonded oxygen concentration (eq. 6), this decreases the vacancy concentration
(eq. 14) which inturn increases the molecular oxygen concentration in the glass
(eq. S).This can be written as a single equation which is the algebraic sum of
equations 6, 14 and 9.
Na20 = 2 Na+ + l/2 02 gas + 2e- (15)
Kl6 = [Na+][02]1/2 [e12/aNa20
showing the Cm is proportional to aNa20*. Therefore in oxygen deficient scales
such as those formed on silicon carbide the sodium will decrease network
transport and increase molecular transport as it drives the scales towards more
stoichiometric compositions.
This result and the previous oxidation model will now be applied to the
acidic corrosion of C-side silicon carbide. According to the previous discussion,
the oxidation under 1 atmosphere dry oxygen at low temperatures (1OOOoC) is
controlled by molecular diffusion as DmCm >> DvCv. As shown in the previous
paragraph, in the hot corrosion Na20 increases Cm and decreases Cv, therefore
molecular transport is enhanced giving greater parabolic constant for the acidic
corrosion of C-side silicon carbide compared to that for oxidation under
*Equation 15 may be considered as the dissolution of sodium oxide in the glass in

an interstitial position (network modifier). Interpretation beyond the
relationship between aNa and [02dis] could be misleading since defects in the
glass structure are required to accommodate the Na+ ions and the electrons.
Therefore equations 6, 9 and 14 appear more representative of the reactions
occuring in the glass.
corresponding conditions. In our experiments the parabolic constant B was
increased by a factor of about 4 at 1000°C. It was about 3x lo-l4 cm2js which
is similar to B=3.2xlO-I4 cm2js of Deal and Grove for the oxidation of
silicon(*). This result is in line with the scales of silicon carbide being more
stoichiometric under acidic corrosion than those formed in dry oxygen oxidation
and with lower oxygen pressures expected in the formation of scales by oxidation
of silicon carbidethan in the oxidation of silicon. From the previous analysis the
activation energy for acidic hot corrosion is expected to be that of the
permation of molecular oxygen through vitreous silica, the same as for the
oxidation of silicon carbide and silicon. Therefore it should be 26-28 kcal/mole.
The 34 kcal/mole measured is close but significantly higher because of the
nature of the experiment. The parabolic constants plotted in figures 11 and 12
were determined at constant sulfur content in the atmosphere and not constant
Na20 activity. The initial mixture of 1% S02-balance oxygen at one atmosphere
total pressure was used at all three temperatures. After passage over the
catalyst it generated decreasing pSO3 for increasing temperatures (equation 7).
From equilibrium relation 2 the activity of Na20 increases with decreasing pSO3
and therefore with increasing temperature. Since B increases with aNa20, an
apparent activation energy greater than the 26-28 kcal/mole for molecular
diffusion of oxygen and the oxidation of silicon and C-side SIC will be measured.
When the generation of 02 molecules by reaction 15 will be dominating the
activation energy for diffusion will be increased since B will be proportional to
DmCm and both terms will depend on temperature. It can be seen from reaction
15 that Cm o (K aNa20P in which both K and aNa have exponential
temperature dependences and n is a fractional exponent such as 2/9 obtained
from the defect equations considered in this report. However the enthalpy of
reaction 15 and in general the operating defect equations are not known so that
it is doubtful that an exact temperature dependence can be derived. It is clear
that the activation energy for hot corrosion should be significantly higher than
that for oxidation. The difference between the measured apparent activation
energy (34 kcal/mole) and that for molecular diffusion (27 kcal/mole) may not be
representative because the activity of sodium may be too low to completely
dominate the transport. Simple considerations suggest that the two mechanisms
of generation of the diffusing oxygen molecules were of the same order of
magnitude in the present experimental conditions. Experiments with higher
Na30 activities produced by initial gas mixtures such as 0.1% SOZ-balance
oxygen would answer this question.
For the acidic hot corrosion of Si-side silicon carbide, the same general
reasoning is applicable, however the parabolic oxidation was expected to be
dominated by network transport (vacancy mechanism) with a low parabolic
constant B. As for the C-side, the Na30 introduced by the hot corrosion
decreases the network contribution (lowers Cv) and increases the molecular
contribution (raises Cm) according to equations 6, 9, and 14 or 15. In the
reactions considered Cm increases rapidly as Cv decreases since Cm = k/Cv6 in
equation 9 and Dm >> Dv by about 4 orders of magnitude. Therefore the change
in defect concentration affect dramatically the relative contributions to the
transport of oxygen although it is not possible to state if and at what
temperature the molecular flux became dominant. B is increased dramatically
as observed experimentally. It is increased by a factor of about 16 compared to
B extrapolated from the oxidation data of Tressler et al.tgl. The large apparent
activation energy is due to the same effect as for the carbon-side, the Na30
activity increases with temperature as discussed above. A quantitative
correlation between the results for C-side and Si-side acidic hot corrosion is
being considered.
REFERENCES
1. J.R. Blachere and P.S. Pettit “High Temperature Corrosion of Ceramics”

(a) DOE Report ER45117-2, March 1986
(b) DOE Report ER45117-1, June 1985
(c) DOE Report ER10915-4, June 1984
2. J.R. Blachere and D.F. Klimovich, *J. Am. Ceram. Sot. 70 [ll] C324-C326
(1987) paper appended in Appendix E.
3. B.S. Draskovich, MA thesis, University of Pittsburgh, 1985.
4, D.F. Klimovich, M.S. thesis, University of Pittsburgh, in preparation.
5. M.G. Lawson, M.S. thesis, University of Pittsburgh, 1987, (Appendix B).
6. M.I. Mayer and F.L. Riley, J. Mat. Sci., 13, (1978) p. 1319-1328.
7. (a) N.S. Jacobson, J. Am. Ceram. Sot., e, [l] 74-82 (1986).

(b) N.S. Jacobson and J.L. Smailek, J. Am. Ceram. Sot., a [8] (1985), p_
432-39.
8. B.E. Deal and A.S. Grove, J. Appl. Phys., 36 (1965), p. 3770.
9. R.E. Tressler and K.E. Spear, GRI Report GRI-87-0088, 1987.
10. S.I. Raider et. al., J. Electrochem. Sot., 123 [4] (1976), p. 560.
11. S.C. Singhal, Ceramurgia International, 2 [3] 123-130 (1976).
12. R.H. Doremus, Glass Science, Wiley 1973, p. 121.
13. R.E. Tressler and K.E. Spear, GRI Report, GRI-86/0066, 1985.
14. L.L. Hench et. al., Ceram. Eng. and Sci. Proc., 3 [9] (1982), p. 587-595.
16. I. Franz and W. Laugheinrich
17. G.J. Yurek et. al, Oxid. Metals, 4 265 (1974).
18. J.A. Costello and R.E. Tressler, J. Am. Ceram. Sot., 64, (1981) p. 327-331 .
19. F.P. Fehlner, Low Temperature Oxidation, Wiley 1986, p. 211.
20. D.C. Larsen et. al, Ceramic Materials for Advanced Heat Engines, Noyes
1985, p. 234.
91
-*. Reference 19
22. (a) R.C.A. Harris and R.L. Call, in Silicon Carbide 1973, R.C. Marshall
et. al., eds., University of South Carolina Press, 1974, p. 329.
(b) R.C.A. Harris, J. Amer. Ceram. Sot., 68, 1975, p. 7-9.
23. J.W. Faust Jr., in Silicon Carbide, A High Temperature Semiconductor,

J.R. O’Connor and J. Smiltens, (Pergamon, 1960).
24. Ibid, pv 403.
25. D.M. Mieskowski et. al., J. Am. Ceram. sot., 67, (1984), C17-Cl8
26. (a) L. Muehloff et.al., J. Appl. Phys., fl [7] (1986), p. 2558-2563.

(b) Ibid, fl [8] (1986), p. 2842-2853.
27. H. Rawson, Inorganic Glass Forming Systems, Academic Press (1967), p. 53.
26.. V.J. Fratello et. al., J. Appl. Phys., 51 [12] (1980), p. 6160-6164.
29. D.L. Griscom, MRS Bulletin, June 16/ August 15, 1987.
30. N.F. Mott, Adv. Phys., 3, 1977, p. 363.
31. Reference 19 p. 67
34. A.G. Revesz and H.A. Schaeffer, J. Electrochem. Sot., 129, 1982, p0 357.
36. U.S. Department of Energy, Thermochemical Stability Diagrams for

Condensed Phases and Volatility Diagrams - DOE/FE/13547-01, May 1980.
Appendix E-Publications
Many presentations have been made on this research. The most recent ones
are:
- Hot Corrosion of Non Oxide Ceramics [l]
- High Temperature Corrosion of Silicon Nitride and Silicon Carbide [2]
The research is being written up for publication. These publications are:
- Oxide Thickness Measurements in the Electron Probe Microanalyzer [3]
* Hot Corrosion of Silica (Appendix B) [4]
- Hot Corrosion of Alumina (Appendix C) [5]
- Hot Corrosion of Silicon Nitride and Silicon Carbide [6]
- Hot Corrosion of High Purity Silicon Carbide and Silicon Nitride [7]
- Corrosion of Ceramics [8]
The first paper is appended. Appendix B is ready for submission. The others
need further work. The material in Appendix C will be submitted next. Reprints will
be supplied as soon as they become available.
184
Appendix E-Publications 185
REFERENCES
1. J.R. Blachere, D.F. Klimovich and F.S. Pettit, “Hot Corrosion of Non Oxide
Ceramics,” paper # , Fail meeting, Basic Science Div. Am. Ceram. Sot.,
Nov. 5, 1966.
2. J.R. Blachere, D.F. Klimovich and F.S. Pettit, “High Temperature Corrosion
of Silicon Nitride and Silicon Carbide,” Invited paper, Workshop on Corrosion
of Ceramics, Penn State, Nov. 12-13, 1967.
3. J.R. Blachere and D.F. Klimovich, “Oxide Thickness Measurement in the

Electron Probe Microanalyzer,” J. Am. Ceram., 70 [ll], C324-C326 (1967).
4. M.G. Lawson et al., “Hot Corrosion of Silica,” ready for submission to J.

Am. Ceram. Sot.
5. M.G. Lawson et al., “Hot Corrosion of Alumina,” in preparation for J. Am.

Ceram Sot.
6. B.S. Draskovich et al., “Hot Corrosion of Silicon Nitride and Silicon Carbide,”
in preparation.
7. D.F. Klimovich et al., “Hot Corrosion of High Purity of Silicpn Nitride and
Silicon Carbide,” in preparation.
a. F.S. Pettit and J.R. Blachere, “Corrosion of Ceramics,” in preparation.

,” wh,ch the/(X) funcoons prwdc the cor-

reel,““, for dx rbaorpo”” of Ihe chrrac-
Oxide Thickness Measurement in the Electron ,cr,s,,c

delccwr.
X-ray, along
,, 15 dcfmcd
[he” prthr
ar I,,/,,)
1” ,hc
csc + I”
Probe Microanalyzer wh,ch s/p I, the mar, abwrpr,“”

event for the hnc measured I; the titm. p
cocffi-
J. K. BLACHERE* AND D. E KL~MOV~CH* is the denrny “f the tilm. and $ IS the

L3ke”ff angle.
cepamle”~ “f Mawnrtr Sacncr and E.ng,nrennp. V”lw,#ty “f P,,,,burgh. P,,t,burph.
The ex~en~~vc. but slmplc. ealcuta-
Pcnniytunlr ,526,
,mns are dr,cr,bcd ,n dclrd ,n Ref, ?
and 3. They rcpon ~“4 rcwl,, tar ,mgte-
An X-ray method/or rhe meosuremenfs of the thickness of wpporred rhmjilms in
elemcnl tilmr For muloclemcnl lilms. It 15
Ihe micromobr is modified for silica tilms on silicon nrrride and silicon carbide.
powble I” obtam buth Ihc tilm rhrcknos
Thr inctvhrrirs o/rhe o.&rn Ka linrbre measured on bulk SIO- und on rhe/i/m.
nod Ihe c”mpo,mon by mcrrur,ng r( ~1,“s
The derivation of the colibrarion curve giving the rhicknru of the film from rhe for all co,noonems of Ihe film. In txxh
rario of rhesr inrensiries is ourlined. The method has been used /or srl~ca Jilnts CBYI, all Acrclc”lc”lrl effccr, (atomic
rhwwr rhrr~~ I um wirh a lawral resolurion of a few micromrrurs. number. rbrorpo”“. nnd lluoreicencct arc
a\rumrd nept~glblc il, mlfhr be cxpccwd I”
a first ~ppruxmu,~“” fur lhm f,lm,
0.
X,DE thxkness mcasuremen,s arc The m~enuty-thickness rclaoonsh~p is nor
nctded for mnny applications such as slmplc smcc the Xaya generdlcd are a
Ihe rczareh on rhc oxidation and hot corro- function of depth I” the sample. and they
SL”” of uhcon “node and sdxon carbblde. are abrorbed rccordmg Lo the prrh of their
The prcw”“, method was adrpled Lo
Many methods hzve been used LOassess the cx~l from rhc samolc
meas~remcnta of the ,hxknos of s,hca
ev”luo”n of Ihe oxidaoon process; bow- The mrcroan~lyrir of dun tilmr has
films on sd,c”n carb,de md s,hc”” mmde
ever. they arc often mdlrect or desuuctive. bee” revxwcd by Guldslem.’ In order 1”
wng lhc oxygen Ko hnc t,, appllulu””
Fur ~“rtimcc. wghc changes arc the nc, calculate rhe inw&ry of the X-rays gener-
rewll of simuluncous rcacu”“~ whxh g,ve nled in [hc aample. Yakowtz and Newbury’
LO light elcmcms tatamlc oumbcr Z< tI)
req&cd scvcr.d mod,f,cau”n, ,,ncc the
“lfvxing comnbuuona. More dwcct mca- “rooaed a” emomeal aooruach bdscd on
h”rrmcnts arc oficn dependent on etch- ii&g Ihe X-ray’depfh oi
~roducrion ewe
low-energy X-ray, p&wed
absortxd and Ihe ab,“rprw”
,,re wungly
c”rrec,,“n 13
mg or frxurc of the specimens Other +(pzt to rhe combmmoo of a parabola and
w_mlarge I” use Ihe approruna~c/(~) ral-
methud> dppbcrble to rhm films ax very an ripmrnt,~l. The mdss rhicknrss p: IS
dceur.w bu! do nor have spatial re)oIulion. ,hc pr”duct “f ,he d,,ww from Ihc rurfacc
“es. The mwn,mc,,~ crlsut,md for Eq CI)
The ctecrron pr”tw microanalyzer IEPMA) : by the drnrlry p “f Ihc sample m Ihrt
xe d,ffercm trum Ihe mc.,rured m,c”,meS
1. becau,e of ~brvp,,“” .l’he X-rry, SC
otfcrr a nundcslruct~vc means of mcaur- lhlchne,,. Cornoared t” Ihnr for a bulk a!“-
grncrad A! var,““, dcplh, I” ,hr san~ptc
img ttlc Thickness of suppancd films wnh a &ard of Ihe co&wlion of Ihe tilm. the
and I” calcul~!c I,. lhe,r m,cn,,,y must be
high lalcral resoluuon which we have uxd @(pi) ~“rvc for the lhm tilm IS tnmcrred at
mtcgraad over dw depth of prwtucuon:
for rihca films a” silicon nilnde and sili- [he mas, tblckncs, pz, of dw lilm and it
con carb,dc. mctudo P mo,hf~ed b;rckscattering y!eld of
The menrny of characteristz X-rays, the cIcc~r”“s wh,ch accoum, fur the sub-
yeneraed by a” ctec~ron barn for an ele- ,,ra,c c”nlr~bu!wn The mwn~~ty of the
ment conlrmcd m a thin titm and not in the charnclerwc X-rays emmcd frum the film
bubruale. 15 related 1” the thickness of the II compared to thrl en,,rled by a bulk For un4 ahsorou”” [he ~“lal mtcn,W
thm film and c”mp”w~“n of the sample. ,tand.ud of Ihe film ma!eriA The in!may
r.11,” I, of these I~ncr. which can be
mr~wed. IS pr”pon~onrt 1” tbr rat” of
Ihc mlcn&llcs of lhc X-raya gencrafrd
1” the lilm l/L) 1” [hat generated I” a bulk
standard /,:
Appendix E-Publications 187
r
November\ 1987
h,k
#(PZ)
00
IO loo 1000 I‘XOO

Thickness ,nm,
Parabolic
h
kh and 0.1425 were calcul.Acd rer~ctwely for
bulk SIO,. S,,N,. md SIC The>e value)
were “blamed for a pnmrry clec~ron barn
ener~v of IO keV bv m~erwlrtwn of the
elem;;l,al drta of He;“r,ch’;nd rddmon of
mars-weighted elcmcmal c”clf~c,en,~.
Therefore, the fdm and Ihe wb\,rrtcs .~e so
and iir IS calculr,sd from the prcdlcled in which simdar dxa, .I c”“a,a”, brckscartermp cc&
mca,ured ,“,cn,,,,eh as fic,en,. Iha, of the ondc. ua, urcd for dll
film thxknesse$ rouhmg I” an enor es,,-
mr,ed I” br <I% of ,bc film dnckne\ses
for xhca glass on s,bcon nitride ,ubw.oes
X[-@z-t&f (p:-h++ j] and slightly over 1% for sd~con crrbrde
wbsorlo. For other subsmale-tilm comb,-
“a,,““,. when ,h,r a,aump,,“” 1s no, \rhd.
The inlegrrl for Ihe parabolic and ex-
in which If 15 Ihe X-WV m,e”s,,v measured a film b;rcksca,ten”p cwflicwn, “wt be
ponenoal r.eg,on 8” a lhxkness p ,uch as
from Ihe f~irn md /Z ;s rhu lot& mca- calculawd as a funcoon “1 p:, d.,. I. aod h
pi,>p:>l.Sh 1s
wed fur the bulk aodard. The mwgration are different for lilm .md wb,trr,e.”
for /: ,s over ,hr mas lh,ckness of [he film Cabbra”“” curve, of 1, II film ihlck-
and Ihrl for Ihe bulk mtenwy is over Ihe ncss were calculr~ed u,,h a compu,er u,mg
uhole depth of X-ray producd”” p:, g,ven Eq. (4) and Ihe mclhod oullmcd above.
by Ihe X-ray praluc!,“” range of Hemnch.’ -(pz,- IShI S,l,ca glr,s u!,h .I den,,,y of 2 2 g/cm’
Equroon (4) ha, ken wed rucce,rfully by
x Q (8) was ured for the lilm I” ,dI ma rage.
?+K(p-_.-I 5hl
KelM,. ma,, Ih,ckne,,. and ab,“rp,,on COICUII-
Ttw mwosmc, of Eq (4) were integrated wnh u”“,. The cahbrat,“” cunc 1, ,houn m
I” clozd form u~,“e Ihe “arabolic-ex- Fig. 2 for sdx”” carbldc md ,dxo” m-
ponenwdexpresrm-for &p;) The ” Q=exP[P&&-X j] mde ,ubwa,es.
arrumrd 4(p:c) cwve 1s sketched I” Fig. I.
1
For [he prrabohc ,eg,on
1.Shx The inteosmes of Ihe “x)pc” Ku line
lip.-J=h~‘(~:-h)‘(a-k)+k (5) were measured 1” a ,ca”“,“~ elcc~“”
m,croscope* fmcd wlh IWO ud\slmglh
and for the rrponenwl rrg,o” I;=I, (for pi = I.Sh)
disperwe spec,r”mctcn equ~ptwd for hgh,
The panme,er, C&Bk. and h are a func- eleme”, .mrly~r. All wmple, uere coaled
don of [he backscattering clewon yield. w,lb &a~, 20 “m of carbon. Duphcnle
Using values for the bulk matenal, the I”- mmsurcment~ were made ,,muluncously
x exp(w ) (6)
te”,,,y for lhr bulk standard I, crlculaed: wth ,hr IWO spec,r”me,ers ill a” .,ccelcrd-
lion voltage of IO kV and a beam curre”,
,“.=I, ifor pr=pz,)
of 20 rA. The ra,a”s kr tie calculaled from
in which Ihe parameters 6, k. and h are cab Ihe measured intms,l~es corrected for back-
culzued followmg Keuwr’ and Goldstem.’ However, for thm films Ihe elrcrron background. The oxide Ihicknes,es CA” be read
The mlegral for Ihe parabolic reg,““. for a xa,,er yield mcludes a conrnbution from from Fig. 2: ntedd. a am~ple m,erp&o”n
dwknes, ,x<l 5h. IP the subwale. Approx~ma,~“na of Ihe da, program and E.q. (4) were used LOgel them
of Cusslet were u,rd prev~ou,ly”’ dnd more exacdy. The meawreme”lS are very
wed m lh,s case Some var,~,,““s on Ihe reproducible. and a standard error of 1% or
electron ,rrnrm,ra~“n cwffiwn,’ of the less IP c;rlcula!ed’ for Le values corre,pond-
,ubauele were also med. The aweme val- me 10 th,cknesxs >IW nm for a se, 01
(7) “es of Ihe backsca,ter cwfticlen, of the lob-s co,m,~ on pea, md background. for
film ,h”uld be thrl of the ,ub,rrate for a film and srandsrd. I, i, ~mprowd by lhr
film mfmwly lhm md lbsl 01 Ihe bulk film comblnarlon of several of these mea-
mr,er,;ll for a IhIck film.’ I” tb,, case. suremen,,. They were uwlly repcrIed at
bilckscaltrr cocffic~ent?, of 0.1316. 0 I37 I, ,hr ,ame we and it, sevrral Icat~ons of
Communicdonr of rhe American Ceramic Societv November 1987
Table I. Oxide Thickness Comparison* better SUN+ Oxan Eq (I) for tits La\k. As
Ilwkncncu shown in Rg. 2. httle addmonal lhrckness
(PI, can bc measured w,th dur higher electron
SImplc Typr WDS SEM energy beau% the sofr X-ray, produced
I SIC.SC 0.32 0.21 deep in tbe sample by hght elements such
a, “xygro are almost completely absorbed
: SIC, SC
SIC. 0.31
0.20 0.30
0.13 Tb,s rc~ults m Ihe shallow ,lopc of the
4 Sic. SC 0.17 0.11 curve for duckcr tilms and Ihe prormuty of
rhe two curves for low film ducknose,.
2 SI,N,
SI,N.. CVD 0.10
0.05 0.09
0.12 Imermedmte electron beam energies such
.sc=nng!x clyrul. WDS..X.R) mlcmuulvrir. ud sEM=lnugw “lr- as 20 kV did not provide signlftcanl
yzuuu. impmvemcn~~
The X-ray melhod ,u,r dexnbed has

similar momholoev a” Ihe surface Of the by as much as 30 to 50%. However. tbe bee” used consi~~rntly I” our rebesrch’ for
sample. It I; esu&red that tix rcprcduc~- repmducibday of the X-ray method 1s ex- the measoremeot of StOl films wrh thick-
bdtty of Ihe thicknesses measured by this cellent. and Ihe meahurements are srraight- nesses from IO “m I” 1.0 pm with a re-
procedure was of Ihe order of II. forward. Reuter’ clawned an accuracy of producibdtry of a few percent. Thicker
+10% for the thicknesses of lhin films of films were imaged m cros) secuon m the
REIWLTS AND Discuss& Al. Cu. N,. and Au on various substrates SEM. a procedure whtch is earbe, and more
The Ihicknenes of glassy regions of measured by a slmdill. method Application accurare than dzscussed earher for the than-
“xzde layers formed by oxidation and hot of the X-ray method I” hgh! elements I”- ncr films. The X-ray method I, non-
corrouon on GIlcon carbide and silicon ni- creases the uncenrmry smce Ihe physical do,,uc,,ve and ha a lswal rraoluuon oi a
tnde al IooO”C were measured bv lhls quanuties, such as roass absorpoon coeffi- few mrcrolnel~r,; II ;Ill”WS a correlauon of
mothal and by unagrng of cross sec& in cientb. used in Ihe caIcuIa11ons are not as tb~kneas rorasureznen~ with surface fca-
tbe SEM. The rcsulls are gwe” I” Table 1. well estabhshed. Also. X-ray absorplnn [ores such a, spheruhlrs or glassy regtom.
The mraruremen~s by SEM Imaging play, a mqor role for Lhe longer wave- Other techniques such as chemical ;maly-
proved difficult for films under I gm duck. ,eng,hs. Con~parison wnh Ihe SEM mra- sir. opr~cal mwferometry. ad elhp5”m-
The magnX~ar~“n was cahbraled wilh stao- sure,,,en,s ruaeerts that the accuracy euy. whrle roorc accur.~~c. do no! offer lhls
dud latex spheres. The samples were frac- ,sbe,rer dun 30-g. The shape of the callbra- lateral re,o1u~~“” whtch 1s highly dewable
tured and brhdv elched I” HBF. to reveal t,“” curve of F,g. 2 s”gge,t, Ihat thick- in roatmal5 sc~cnce.
the oxide idye;. They had been carbon nessts inIhc rang~l0nmlo0.8
to I pm can
coated pnor 1” fracrure to observe tie mor- be measured: these values correspond ap
phology of the wface. and I, was found “roxlmatelv to the linear part of tbe curve.
that dos coating helped mantain Ihe sur- ihe lower iimir
is se1 by dir
counring statis-
Iace edge of the oxide lilm during etching. Lies. For the thicker lilms, the sens~liwly
The samples were recoated to avold charg- decreass as the curve of Fig. 2 levels off.
mg I” Ihe electron beam and the coating It was clamxd’ lhrl the ongmal method
dwk”ea,es were large compared lo some \yas reasonably accumte up 1” 306 of dx
of the “ride layers of Table 1. Under tbr bulk inter~cf~on volume, correspondmg in
c”“d#i”ns of Ihe measurements. il is ex- this car to <0.5 ,.un. However. the more
pxred lhar Ihe resolution of the SEM did extensive absorprion correction performed
not reach 10 em. The experimenrai diffi- here should allow the range of the “lea-
culties in the SEM imagmg of these thin surements to be extended
Inyen resulled in large experimenlal scaLLer Increasing the pnmuy electron energy
and poor accuracy. II is estimated that the should also increase Ihe film thicknesses
uncexiaintv of tbe thicknesses measured dl- which can be measured by generaling X-
recrly I” Ihe SEM could be as high as 40 to rays deeper in!” the sample. This WPI
50% for the dunner oxide films. As shown shown’ for metal films with Eq. (I). A
I” Table I, rhe LW” methods gwc tie same calibruion cuwe for 30.kV prunary elec-
order of magmwde but their results differ moos was generated wrh Eq. (4). which 16

High Temperature Corrosion of Ceramics PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

High Temperature Corrosion of Ceramics PDF

Hochgeladen von

Copyright:

Verfügbare Formate

HIGH TEMPERATURE

NOYES DATA CORPORATION

Published in the United States of America by

Library of Congress Cataloging-in-Publication Data

CERAMIC RAW MATERIALS

Parr I Corundum Potassium Compounds Yttrium Oxide

ISBN O-8155-1143-4 (1987) 8%” x 11” 900 pages

In the use of materials at elevated temperatures in harsh environments, it is ap-

The information in the book is from High Temperature Corrosion of Ceramics,

The table of contents is organized in such a way as to serve as a subject index

Advanced composition and production methods developed by

The materials in this book were prepared as an account

Final determination of the suitability of any informa-

EXPERIMENTAL PROCEDURES. . . . . . . . . . .......... .6

GASEOUS CORROSION ....................... ...... . . . . 11

HOT CORROSION OF SILICA. . . . . . . . . . . .. . . . . . . .. . . . . . . .16

HOT CORROSION OF ALUMINA. . . . . . . . . . . . . . . .. . . . . . . . .21

HOT CORROSION OF SILICON CARBIDE AND SILICON NITRIDE. . . .26

APPENDIX A-GASEOUS CORROSION OF SILICA AND ALUMINA

Experimental Procedure. .............. . . . . . . . . .35

APPENDIX B-HOT CORROSION OF SILICA. ................. . .5B

APPENDIX C-HOT CORROSION OF ALUMINA . . . . . . . . . . . .89

APPENDIX D-HOT CORROSION OF SILICON NITRIDE AND

APPENDIX E-PUBLICATIONS ............................. 184

corrosion resistance. While ceramics may be more corrosion resistant than

The purpose of this program was to systematically investigate the corrosion of

ceramics and to develop a theory generally applicable to all ceramics.

A great number of ceramic materials are available for use in a variety of

corrosive environments. In order to develop a theory applicable to the corrosion

of ceramics in general, it is necessary to investigate a variety of representative

ceramic materials exposed to a number of different corrosive environments. In

order to keep the number of experiments at a reasonable number, the ceramic

corrosion was developed, in particular, by being immune to the environment or

by developing passivity. Furthermore, the environments used to produce corrosion

and their severity.

A material is considered to be immune to a particular environment when it

is in equilibrium with that environment. Total or complete immunity is rare in

changes in mechanical properties. The concentrations of point defects in the

such as electrical and ionic conductivities. Nevertheless, in terms of a corrosion

reaction the Al303 may be considered to be immune to this gaseous environment.

immune to many environments which are extremely corrosive to metallic systems,

in the present investigation. Four types of crystalline Al303 were studied of

described in Table I it was of relatively high purity and was vitreous.

Passivity to corrosive environments is achieved by the formation of a

process must diffuse. The development of passivity is the principal means by

which metallic alloys achieve resistance to corrosive environments, but passivity

is also an important mechanism in certain ceramic materials, depending upon the

Sic react with most environments encountered at elevated temperatures and

resistance to corrosion is achieved via the formation of a passive reaction product

barrier. Furthermore, Al303 can react with gaseous environments containing

Designation Material Purity (Wt%) Supplier

S.C. Sapphire -99.99 Saphikon

998 poly A120; (a) 99.8(0.1MgO, O.llSiO2) Lucalox (G.E.)

Silica fused silica 99.99 Corning 7940

SIN CVD CVD*Si3N4 99.99 Deposits & Composites

SiN H.P. H.P.** Si 3N4 93(6Y203Fe, Al, 02) Airesearch

Sic SC. s.c.*** (a6H S.C.) 99.99 W.J. Choyke

Sic CVD CVE (Sic) 99.98 W. J. Choyke

* Chemical Vapor Deposition *** Single Crystal

When ceramic materials are used in practical applications, a variety of

corrosive environments can be encountered. Most of these environments will

can also be present. Moreover, deposits such as metallic sulfates, carbonates or

metallic alloys it has been useful to examine corrosion reactions in environments