3384

Ind. Eng. Chem. Res. 1996, 35, 3384-3398

Advanced Ceramic Materials for Use in High-Temperature

Particulate Removal Systems
M. A. Alvin
Westinghouse Science & Technology Center, 1310 Beulah Road, Pittsburgh, Pennsylvania 15235

First-generation monolithic porous ceramic filter materials have experienced thermal fatigue,
high-temperature creep, and a loss of material strength when operated for extended periods of
time in advanced coal-fired combustion and gasification systems. Fiber-reinforced and advanced
ceramic composites potentially provide a means to mitigate the degradation mechanisms
encountered by monolithic filter matrices and ultimately extend operating life. In this paper
we will review advancements which have recently been made during the development of the
second-generation porous ceramic filter materials and provide insight into the performance of
the filter elements during bench-scale qualification testing and operation in demonstration plant
test facilities.
Introduction
High-temperature, high-pressure particle control is
required to meet environmental and turbine equipment
specifications in advanced coal-fired power generation
systems. Direct application of high-performance, hightemperature particulate control ceramic filters is expected not only to be beneficial to the advanced fossil
fuel processing technology but also to select hightemperature industrial and waste incineration processes.
Porous ceramic filters are currently being designed
to meet performance, life, and cost constraints of the
process application in which they are used. For each
application, the porous ceramic filters must withstand
variation in the effluent gas stream chemistry, changes
in the composition and loadings of the entrained fines,
and oscillations in the effluent gas stream temperature
and possibly pressure, while maintaining high particulate removal efficiencies, high flow capacity, and relatively low pressure drop characteristics. During process
operation, the ceramic filters must also withstand
stresses resulting from mechanical clamping and support structures, as well as vibration and thermal
stresses imparted from pulse gas cleaning. The criteria
for successful use and operation of porous filters as a
viable advanced particulate removal concept therefore
requires not only thermal, chemical, and mechanical
stability of the ceramic materials but also long-term
structural durability of each filter element and high
reliability of integrated process design features.
Development and production of highly durable and
reliable filters are dependent not only on properties of
the material used in the construction of each filter
element but also on the relative stability of that material
under corrosive high-temperature, high-pressure environments which contain gas-phase sulfur, alkali, and
chloride phases [Alvin et al. (1991)]. In addition, during
production of the filter body, manufacturing and process
control must be exercised to produce a quality-assured
component that meets both design dimensional tolerances and compositional criteria.
With respect to filter materials, strength, thermal
shock resistance, thermal expansion, modulus of elasticity, fracture toughness, thermal conductivity, hardness,
density, and chemical reactivity are important criteria
in the selection of a viable, durable, filter material.

Telephone: 412-256-2066. Fax: 412-256-2121.

S0888-5885(96)00128-5 CCC: $12.00

Many of these material properties are available for

dense ceramics but have not been established for the
porous matrices which are currently being considered
for filter fabrication.
As shown in Table 1, oxides, mixed oxides, and nonoxide matrices are utilized to construct porous rigid
barrier, cross-flow, candle, or tube filter elements.
Materials that were initially considered for use in the
construction of the cross-flow filters included alumina/
mullite, cordierite, aluminosilicate foam, cordieritesilicon nitride, reaction-bonded silicon nitride, and
sintered silicon nitride. As technology moved from the
stacked-plate cross-flow filter design to the production
of monolithic units, alternate materials and processes
were utilized to fabricate the cross-flow filter concept.
Similarly, as technology progressed, utilization of the
cross-flow filter alumina/mullite matrix and processing
techniques were transitioned into the manufacture of
alumina/mullite candle filters. Currently, the alumina/
mullite matrix, clay-bonded silicon carbide materials are
used to commercially manufacture ceramic candle filters. Alternate materials which have been used to
fabricate candle filters include fireclay, aluminosilicate
fibers, alumina, chemically vapor infiltrated silicon
carbide (CVI-SiC), and filament wound matrices. Numerous advanced materials and processes are currently
being considered in the production of prototypic candle
filter elements.
First-Generation Filter Materials
The alumina/mullite (Al2O3/3Al2O32SiO2) matrix consists of mullite rods that are surrounded by a grain
boundary amorphous or glassy phase which contains
corundum (Al2O3) and anorthite (CaAl2Si2O8). Cordierite (Mg2Al4Si5O8) when properly fired contains very
little glass. The aluminosilicate foam is comprised
principally of glass fibers which are coated and held
together by a mullite wash.
Cordierite-silicon nitride (CSN) is initially formed
from a mixture of cordierite and silicon nitride (Si3N4).
Upon firing, a Mg-SiAlON and a glass phase are
expected to result. In this material, cordierite and the
resulting glassy grain boundary phase can be considered
as the binder which holds the silicon nitride grains
together. Reaction-bonded silicon nitride (RBSN) is
manufactured through a process in which silicon is
nitrided in the presence of iron, forming a microstructure in which the Si3N4 grains are directly bonded to
1996 American Chemical Society

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3385
Table 1. Porous Ceramic Filter Materials
cross flow filters
Coors alumina/mullite
GTE cordierite
GTE cordierite-silicon nitride
Foseco aluminosilicate foam
AiResearch sintered silicon nitride
AiResearch reaction-bonded silicon nitride
DuPont SiC-SiC composite
Blasch injection-molded alumina/SiC
Blasch injection-molded mullite-bonded alumina
CeraMem cordierite
bag filters
3M Nextel 312, 440, 550, 610, 720
tube filters
Asahi cordierite

each other. In contrast, sintered silicon nitride (SSN)

is manufactured via direct sintering of silicon nitride
grains with oxide additives. Unlike the porous alumina/
mullite, cordierite, CSN, and SSN filter materials,
RBSN does not contain grain boundary phases.
Commercially available clay-bonded silicon carbide
candle filters generally contain silicon carbide support
grains (500-1000 m) that are held together with an
aluminosilicate or clay binder or a crystallized glass
phase. A fibrous aluminosilicate mat, or a finer silicon
carbide grain layer (80 m), is applied as an outer
surface skin or membrane to provide a barrier to micron
and submicron particles in the process gas stream. In
the clay-bonded silicon carbide matrices, an oxide-based
binder phase bonds adjacent grains to each other.
During the processing of these materials, the potential
exists for additional grain boundary phases to form.
The oxide-based binder in the clay-bonded silicon
carbide filter materials contains relatively high concentrations (0.5-10%) of sodium, iron, potassium, calcium,
and/or titanium. Efforts have been focused on producing a second-generation clay-bonded silicon carbide
matrix which has increased strength, as well as enhanced creep and corrosion resistance during use at
either high-temperature and/or in-process gas streams
which contain gas-phase alkali species.
Alternately, binderless (sintered) silicon carbide materials have been used to fabricate candle filters. These
materials contain relatively low concentrations of iron,
nickel, and/or chromium within the grain boundary
phase(s).
Mixed oxides as clay-bonded aluminosilicates or fireclay materials have also been used in the production of
first-generation candle filters. These matrices contain
relatively high concentrations of sodium, potassium,
titanium, iron, and chlorine. Alternately, relatively
pure candle filter matrices have been formed from
R-alumina bonded with aluminosilicate phases. Similar
to the alumina/mullite matrix, aluminosilicate fibers
have been utilized to manufacture both cross-flow and
candle filters. In order to contain the fibers and
increase the strength of the fibrous aluminosilicate
candles, a toughened outer skin is applied over the
support structure. The toughened skin typically is a
blend of both fibers and a silica and alumina binder
phase.
Advanced Second-Generation Filter Materials
Recent advancements have led to the production of
filament wound and fiber-reinforced CVI-SiC or SiC-

candle filters
Schumacher Dia Schumalith F40
Schumacher Dia Schumalith FT20
Pall Vitropore 442T
Pall Vitropore 326
Coors alumina/mullite
3M CVI-SiC composite
DuPont SiC-SiC composite
DuPont filament-wound PRD-66
IF&P fibrosic
Textron nitride-bonded SiC
B&W alumina-based CFCC
Westinghouse/Techniweave sol-gel
Pall Aerotech iron aluminide
Kyocera zirconia/mullite
Kyocera mullite
Third Millennium silicon carbide
Kaiser Aerotech non-oxide CFCC

SiC composite candle filter elements. The CVI-SiC

composite filter produced by 3M consists of three
layerssan outer open-mesh confinement layer, a middle
filtration mat, and an inner triaxial braided fabric layer
which forms the structural support matrix of the filter
element. Within the confinement and filtration mat
layers, an 1-2 m layer of silicon carbide is deposited
which encapsulates Nextel 312 or alumina-based fibers.
Alternately, an 100 m layer of silicon carbide is
deposited along the Nextel 312 triaxial braid in the
support matrix.
The SiC-SiC composite filter produced by the DuPont
Lanxide Corp. consists of an 10-20 m silicon carbide
layer that is chemically vapor infiltrated (CVI) along
an 2-5 m interface coating which encapsulates 15
m Nicalon fibers. Fine-grain silicon carbide grit is
utilized to form an outer membrane along the surface
of the SiC-SiC composite filter matrix. Originally, the
DuPont SiC-SiC composite candle was constructed
using two plys of Nicalon felt. In order to more
efficiently utilize the CVI-SiC process, production of
the candle was modified to include the use of a singleply felt layer and a mesh screen support layer. As a
hybrid construction, an enhanced phase was added to
the interface coating in the mesh screen support, and
the fine-grain silicon carbide grit was utilized to form
the outer membrane. Several issues have been raised
as to the oxidative stability of the CVI-deposited silicon
carbide, as well as the stability of the interface coating,
and the Nextel or Nicalon fibers in either the 3M or
DuPont matrices during use in high-temperature, oxidizing process applications.
The oxide-based, filament-wound matrix produced by
the DuPont Lanxide Corp. consists of a layered cordierite, mullite, cristobalite, and corundum microstructure. An amorphous phase is also present in the asmanufactured filament-wound filter matrix. Due to the
differences in the thermal coefficient of expansion for
mullite, cordierite, and corundum which are present in
the matrix, a microcracked structure is formed after
high firing.
Critical issues in demonstrating successful long-term
viability of the porous ceramic filters are the chemical,
thermal, and mechanical stability of the various materials at process operating conditions. In the following
section, the response of the monolithic first-generation
filter materials during operation in the pressurized
fluidized-bed combustion (PFBC) environment at the
American Electric Power (AEP) Demonstration Plant
in Brilliant, OH, is presented. Several of the secondgeneration fiber-reinforced and filament-wound filter

3386 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Table 2. W-PFBC Hot Gas Filter Testing at AEP
PFBC Test Segment

operating time, h
operating temp, C
no. of filter elements
Schumacher F40
Schumacher FT20
Pall Vitropore 442T
Coors alumina/mullite
3M CVI-SiC composite
DuPont PRD-66

1
10/92-12/92

2
7/93-9/93

3
1/94-4/94

4
7/94-10/94

5
1/95-3/95

464
730-790

1295
620-790

1279
650-780

1705
660-760

1110
760-845

384

384

384

258
8
8
8
3
3

5
153
98
10
22

elements were also installed and tested in the Westinghouse Advanced Particulate Filtration (W-APF) system at AEP. The response of the advanced secondgeneration filter materials to PFBC operating conditions
will also be described. Subsequently, a discussion is
presented which reviews the qualification testing which
was conducted in the Westinghouse high-temperature,
high-pressure (HTHP) PFBC and bench-scale flowthrough facilities which explored the response of the
various second-generation filter materials to extended
simulated process and thermal transient conditions and
high-temperature oxidizing environments containing
gas-phase alkali. Many of the material response results
that were obtained during qualification testing in the
Westinghouse HTHP PFBC test facility have been
directly observed during field operation.
Field Testing
Testing in the W-APF system was initiated at the
AEP Demonstration Plant in October 1992. Five hightemperature test campaigns were conducted primarily
utilizing first-generation, monolithic, Schumacher Dia
Schumalith F40 clay-bonded silicon carbide candle
filters (Table 2). During conduct of the last two test
segments, advanced second-generation candle filters
were installed and operated in the W-APF system.
With the exception of ash bridging, which generally
failed the filter elements, the stability and long-term
viability of the aged or conditioned candle filter
materials were demonstrated during operation in the
PFBC gas environment at AEP. A summary of the
residual strength and microstructural and phase changes
which occurred during operation of the first- and secondgeneration candle filters follows [Alvin et al. (1995a,b),
Mudd et al. (1995)].
Schumacher Dia Schumalith F40
The Schumacher Dia Schumalith F40 filter material
consists of an aluminosilicate binder phase that encapsulates and bonds together silicon carbide grains, forming 40-50 m pores within a 15 mm thick support wall.
An 100 m thick, aluminosilicate fibrous membrane
is applied to the outer surface of the filter body, in order
to prevent fines penetration into and/or through the
porous ceramic filter wall. During test operations at
AEP, the integrity of the membrane was retained,
imparting complete barrier filtration characteristics to
the Schumacher Dia Schumalith 40 filter matrix. Typically 288-384, 1.5 m clay-bonded silicon carbide Schumacher Dia Schumalith F40 candle filters were installed
and operated in the W-APF in Test Segments 1-5.
Schumacher Dia Schumalith F40 surveillance candles successfully achieved 5855 h of operation in the
W-APF system at Tidd. The residual bulk strength of

Figure 1. Schumacher Dia Schumalith F40 strength profile.

the Schumacher Dia Schumalith F40 material was

determined via room temperature and process temperature testing of C-rings in compression and tension. As
shown in Figure 1, the process temperature bulk
strength of the Schumacher Dia Schumalith F40 material decreased during the initial 1000-2000 h of PFBC
operation. With continued operation, however, the
residual bulk strength of the Schumacher Dia Schumalith F40 material remained constant. The residual
or conditioned strength was considered to result from
the complete or nearly complete crystallization of the
binder phase that encapsulated the silicon carbide
grains, as well as the bond posts or ligament surfaces
(Figure 2).
Previous use of alternate clay-bonded silicon carbide
candle filters indicated that creep crack growth occurred
primarily at the base of the flange, and failure of the
elements resulted after 500-1000 h of operation in the
830 C circulating pressurized fluidized-bed combustion environment [Alvin et al. (1994)]. In order to
ascertain whether the Schumacher Dia Schumalith F40
filter material experienced creep during operation at
AEP, the overall length of the surveillance filter elements was measured after the various test campaigns
and compared to the initial, as-manufactured lengths.
After 5855 h of operation in the W-APF at AEP, two
1.5 m surveillance filters were observed to have elongated by 5-7 mm. Cracks as a result of high-temperature creep were not evident within the densified section
of the flange nor along the remainder of the filter body.

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3387

lag required to initiate high-temperature creep in the

Vitropore 442T clay-bonded silicon carbide candle filter
matrix.
Schumacher Dia Schumalith FT20
Eight improved, high-temperature, creep-resistant
Schumacher Dia Schumalith FT20 clay-bonded silicon
carbide candles were installed and operated in the
W-APF system in Test Segment 4 at AEP. The initial
bulk strength of the 10 mm wall Schumacher Dia
Schumalith FT20 material was comparable to the
higher strength, 15 mm wall, Schumacher Dia Schumalith F40 filters which had been manufactured in
1992.
As shown in Table 3, the residual bulk strength of
the Schumacher Dia Schumalith FT20 matrix after
1705 h of successful operation in the W-APF at AEP was
comparable to that of the initial bulk strength of the
filter matrix.
Coors Alumina/Mullite

Figure 2. Crystallization of the Schumacher Dia Schumalith F40

filter matrix after (a) 3038 and (b) 5855 h of operation in the
W-APF system at AEP.

Pall Vitropore 442T

The Pall Vitropore 442T material also consists of
silicon carbide grains that are bonded together via an
oxide-based binder to form the 10 mm filter support
wall. A finer grit silicon carbide layer is applied to the
outer surface of the filter element, forming the barrier
filters membrane.
During operation at AEP in Test Segment 4, eight
Pall Vitropore 442T clay-bonded silicon carbide candle
filters were installed and operated in the W-APF system.
Similar to the conditioning experienced by the Schumacher Dia Schumalith F40 filters, the Pall Vitropore
442T clay-bonded silicon carbide filters experienced a
loss of bulk material strength after 1705 h of operation
in the PFBC gas environment (Table 3).
Three of the Pall Vitropore 442T filter elements which
had been operated in the W-APF in Test Segment 4 were
reinstalled and acquired an additional 1110 h of operating life in Test Segment 5. Post-test inspection indicated that the 2815 h PFBC-exposed Pall Vitropore
442T filters elongated by 17-20 mm. Cracks as a result
of high-temperature creep were not evident directly
below the flange nor along the external surface of the
filter elements.
During operation in Test Segment 5, 150 Pall Vitropore 442T candle filter elements were installed in the
W-APF system. All candles remained intact after 1110
h of operation. Post-test characterization of 17 elements
indicated that the filters elongated by 0-4 mm. Since
all of the Pall Vitropore 442T filter elements utilized at
AEP had been identically manufactured, the difference
in elongation which resulted between the two test
segments was considered to reflect aging or the time

Eight Coors alumina/mullite P-100A-1 filters were

installed in the W-APF system during conduct of Test
Segment 4 at AEP. All Coors alumina/mullite filter
elements remained intact after 1705 h of operation in
the 660-760 C PFBC gas environment. The residual
bulk strength of the 1705 h, PFBC-exposed, Coors
alumina/mullite filter matrix was comparable to that
of the as-manufactured filter matrix (Table 3).
Three of the Test Segment 4 Coors alumina/mullite
filters were installed in the W-APF system prior to
initiating Test Segment 5. Post-test inspection indicated that all three of the Coors alumina/mullite filters
remained intact, achieving 2815 h of successful operation.
In addition, 95 newly manufactured Coors alumina/
mullite filters were installed in the W-APF system
during conduct of Test Segment 5 at AEP. After 1110
h of PFBC operation at temperatures of 760-845 C,
two filters failed, while the remaining Coors alumina/
mullite filters remained intact. One of the Coors
alumina/mullite filters failed directly below the flange,
while the second filter fractured at approximately 1000
mm below the flange.
Characterization of the residual bulk strength of the
1110 h, PFBC-exposed, Coors alumina/mullite filter
elements indicated that the Coors alumina/mullite
matrix lost strength during operation in the 760-845
C oxidizing environment. This implied a sensitivity
of the Coors alumina/mullite matrix to higher process
operating temperatures which may either induce or
accelerate phase changes or enhance crack propagation
properties along the i.d. surface and/or through the
10 mm filter wall. A further reduction in the residual
bulk strength along the i.d. surface and/or through the
filter wall was exhibited within the Coors alumina/
mullite filter matrix that experienced 2815 h of operation in the PFBC environment (Test Segments 4
and 5).
As shown in Figure 3, extensive mullitization resulted
along the pore cavity walls of the Coors alumina/mullite
filter matrix as both operating time and process temperature increased. As determined by qualitative X-ray
diffraction analysis, minor dissociation of mullite that
was present in the as-manufactured Coors filter material was initiated early during process operation. Simultaneously, a reduction and/or depletion of the as-

3388 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Table 3. Candle Filter Residual Bulk Matrix Strength
strength, psi
filter
matrix

candle filter
location

S153/317B
S504/322B
S436/321B
S193/318B
S065/314B
S106/317B
S324/319B
S109/317B
S215/378B
S447/322B
S455/322B
S523/322B
S418/321B
S422/322B
S228/318B

Test
Segment no.

C/M-15

R2-325
R5-325
R2-360

high-temp (732 C) C-ring testing

compression
tension

Schumacher Dia Schumalith F40s1991 Production Lot

1300 ( 213
1907 ( 111
1
464
1120 ( 123
1
464
1096 ( 116
1
464
1147 ( 119
1
464
940 ( 60
1
464
1180 ( 98
1
464
1083 ( 148
1438 ( 108
1
464
1140 ( 120
1424 ( 162
1, 2
1760
908 ( 72
1117 ( 91
1, 2
1760
794 ( 50
709 ( 71
1, 2
1760
793 ( 58
711 ( 89
1, 2
1760
793 ( 39
1016 ( 134
1, 2, 3
3038
720 ( 57
944 ( 172
1, 2, 3, 4
4734
870 ( 145
816 ( 112
1, 2, 3, 4, 5
5855
592 ( 56
763 ( 87

B/T-1
B/M-1
B/B-1
A/B-6
B/B-45
B/B-9
B/B-41
B/M-22
A/T-22
B/T-22
C/T-22
B/T-16
B/B-16
A/T-16

S199/315E
S039/312E

operating
time, h

room-temp C-ring testing

compression
tension

Schumacher Dia Schumalith FT20

2296 ( 261
2268 ( 167
1705
2283 ( 184
2370 ( 238

B/M-20
C/T-1

Pall Vitropore 442T

2857 ( 186
1705
2311 ( 231
1110
2569 ( 132

4
5

2574 ( 177
2034 ( 139
2277 ( 156

Coors Alumina/Mullite P-100A-1

2575 ( 182
2721 ( 415
1705
2475 ( 189
2903 ( 289
1110
2079 ( 140
2392 ( 130

DC-013
DC-003
DC-056

B/M-16
A/T-17

4
5

DC-068

A/T-1

1110

1958 ( 68

2544 ( 214

DC-002

B/M-17

4, 5

2815

2097 ( 119

2063 ( 178

1416 ( 127
1226 ( 116
1172 ( 134
1245 ( 108
1056 ( 131
1230 ( 127
1137 ( 101
1132 ( 112
1064 ( 72
1028 ( 94
989 ( 77
968 ( 96
890 ( 65
1031 ( 105
777 ( 57
891 ( 52a

2328 ( 228

1778 ( 246
1873 ( 174
1418 ( 122
973 ( 121
885 ( 54
1252 ( 241
1284 ( 199
1139 ( 160
1208 ( 99
966 ( 47a

3034 ( 148
3041 ( 238

2708 ( 360
3102 ( 272

3430 ( 221
2453 ( 187
2721 ( 138
2454 ( 214a

3029 ( 149
2138 ( 180
2201 ( 212
2115 ( 238a

3107 ( 276
2738 ( 161
2368 ( 93
2512 ( 107a
1800 ( 135
2079 ( 85a
2200 ( 141
2360 ( 116a

3353 ( 231
3291 ( 246
2636 ( 238
2717 ( 167a
2542 ( 196
2659 ( 109a
2146 ( 362
2287 ( 251a

strength, psi
filter
ID no.
D-99
D-132
D-237

candle
filter
location

B/M-7
B/M-8

Test
Segment no.

4
5

operating
time, h

1705
1110

room-temp
C-ring testing
compression
tension

high-temp (732 C)
C-ring testing
compression
tension

high-temp (843 C)
C-ring testing
compression
tension

DuPont PRD-66
1219 ( 162
1265 ( 188
1830 ( 238
1725 ( 320
1533 ( 202
1380 ( 188

1277 ( 178
1884 ( 142
1897 ( 256

NAb
NAb
1872 ( 230

1304 ( 327
1642 ( 401
1356 ( 104

NAb
NAb
1460 ( 197

diametral O-ring testing

filter
ID no.
43-1-2
43-1-6
45-18-02
a

candle filter
Test
operating
location
Segment no.
time, h

B/M-15
C/T-18

4
5

1705
1110

room-temp
strength, psi
composite triaxial braid

high-temp (732 C)
strength, psi
composite triaxial braid

high-temp (843 C)
strength, psi
composite triaxial braid

3M CVI-SiC Composite
1341 ( 254 14026 ( 2012 1060 ( 219 11012 ( 1795 NA
NA
1696 ( 195 18220 ( 1356 1429 ( 159 15599 ( 2246 NA
NA
2333 ( 415 18975 ( 3117 2225 ( 361 18001 ( 3745 1850 ( 299 16173 ( 2245

High-temperature strength testing conducted at 834 C. b NA: not available.

manufactured anorthite and glass phases resulted,

leading to the production of trace and minor concentrations of alumina, cordierite, and cristobalite within the
bulk matrix. Retention of the bulk matrix strength with
time is expected to depend on the rate of cristobalite
and cordierite grain growth, which could ultimately lead
to the formation of microcracks and loss of material
strength. Since phase conversion appears to have stabilized within the first 1100 h of PFBC operation, additional microcrack formation and/or loss of bulk material strength would be attributed to thermal fatigue (i.e.,
increased pulse cleaning intensity or duration) and/or
thermal shock of the Coors alumina/mullite filter matrix
during high-temperature PFBC process operation.

3M CVI-SiC Composite
The 3M CVI-SiC composite filter consists of three
layerssan outer open-mesh confinement layer, a middle
filtration mat, and an inner triaxial braided fabric layer
which forms the structural support matrix of the filter
element. Within the confinement and filtration mat
layers, an 1-2 m layer of silicon carbide is deposited
which encapsulates Nextel 312 or alumina-based fibers.
An 100 m layer of silicon carbide is deposited along
the Nextel 312 triaxial braid in the support matrix.
Three 3M CVI-SiC composite filters were initially
installed in the W-APF in Test Segment 4, and after
1015 h of operation, two of the elements failed at the

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3389

Figure 3. Crystallization of the Coors alumina/mullite filter

matrix after (a) 1110 h of PFBC operation in Test Segment 4 and
(b) 2815 h of PFBC operation in Test Segments 4 and 5.

base of their flange due to ash bridging. The third 3M

filter element, which remained intact, achieved 1705 h
of PFBC test operation prior to termination of Test
Segment 4 at AEP.
Post-test, room-temperature, gas flow resistance measurements of the intact PFBC-exposed 3M filter indicated that the pressure drop across the element was
220 in-wg at a gas face velocity of 10 ft/min. The roomtemperature pressure drop across the ash-coated 3M
filter was relatively high in comparison to alternate
filter elements which were exposed to similar PFBC test
conditions. The high-pressure drop across the 3M
PFBC-exposed candle filter was attributed to the adherence of the dust cake layer along and within the outer
confinement layer, as well as through the filtration mat,
and triaxial support braid.
Characterization of the 3M CVI-SiC matrix via
scanning electron microscopy/energy-dispersive X-ray
analysis (SEM/EDAX) indicated that minor changes in
the morphology of the filter matrix had occurred after
1705 h of operation in the 660-760 C W-APF system
at AEP. Due to deposition of the thin <1 m interface
coating, it was frequently difficult to discern whether
the interface coating had remained intact. A gap or
separation was considered to exist between the CVISiC coating and the underlying Nextel 312 fibers in
several areas of the PFBC-exposed 3M composite filter
matrix (Figure 4). In the high-temperature PFBC
environment, oxidation of the interface coating is expected to occur.
Post-test diametral compressive strength testing indicated that the strength of the PFBC-exposed 3M CVISiC composite matrix was greater than that of the asmanufactured filter matrix (Table 3). Due to the

Figure 4. Morphology of the 3M CVI-SiC composite matrix: (a)

as-manufactured outer CVI-SiC-coated Nextel fibers; (b)
CVI-SiC-coated fibers after 1705 h of operation in the W-APF at
AEP.

accumulation of fines within the PFBC-exposed matrix,

wedging of fines in between the SiC-coated fibers may
require a higher load to be applied to the matrix prior
to failure, thus generating what appeared to be a
strengthened composite matrix. Alternate mechanisms
which include phase changes within the matrix have
been considered to occur during process operation which
would provide additional bulk strength to the 3M CVISiC composite. During diametral compressive strength
testing, the relatively low load bearing, PFBC-exposed,
3M CVI-SiC matrix generally retained the graceful
fiber pull-out characteristics of the fracture-toughened,
as-manufactured matrix.
Testing was reinitiated at AEP in Test Segment 5,
during which time 10 newly manufactured, 3M CVISiC composite filters were installed in the W-APF
system. After 1110 h of filter operation at temperatures
ranging between 760 and 845 C, testing was completed. Post-test inspection of the filter cluster indicated that all 10 of the 3M CVI-SiC composite filter
elements generally remained intact. Along one of the
3M filter elements, a 25.4 mm diameter hole was
evident along the outer confinement layer. The location
of the removed confinement layer was 75 mm below
the flange.
Assuming that the strength of the as-manufactured
3M filters that were used for testing at Tidd in Test
Segment 5 was equivalent to the strength of the
as-manufactured filter elements that were used in Test
Segment 4, characterization of the 1110 h, PFBCexposed, 3M CVI-SiC composite filter elements indi-

3390 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996

cated that the strength of the ash-filled matrix once

again appeared to increase.
DuPont PRD-66 Filament-Wound Filters
The oxide-based, filament-wound, DuPont PRD-66
matrix consists of a layered cordierite (Mg2Al4Ai5O18),
mullite (3Al2O32SiO2), cristobalite (SiO2), and corundum (Al2O3) microstructure. An amorphous phase is
also present in the as-manufactured PRD-66 filter
matrix. Due to the differences in the thermal coefficient
of expansion for mullite, cordierite, and corundum which
are present in the PRD-66 matrix, a microcracked
structure is formed after high firing.
The diamond pattern weave of the polycrystalline
refractory oxide-based fibers forms the 7 mm structural support layer of the DuPont PRD-66 filter matrix.
A thin membrane layer is wrapped along the outer
surface of the support matrix, producing a light-weight,
bulk filter element.
During operation of the W-APF system in Test Segment 4, three DuPont PRD-66 filter elements were
installed in a middle array. After 1705 h of operation,
all three filters remained intact. During C-ring preparation of the PFBC-exposed, PRD-66 filter element,
magnesium sulfate crystallized along the outer surface
of the filter, as a result of leaching of the ash that was
contained within the filter i.d. bore and wall during wet
cutting with a diamond wheel. Post-test strength
characterization of the PRD-66 C-rings indicated that
the bulk strength of the ash-filled matrix tended to
increase after 1705 h of operation at AEP in Test
Segment 4.
Twenty-two DuPont PRD-66 candle filters were installed in a top array in the W-APF prior to operation
in Test Segment 5. After 232 h of operation, sections
of the PRD-66 matrix were identified in the ash hopper
discharge, indicating that failure had occurred. Testing
continued, and after 775 h of operation, additional
sections of the PRD-66 filter matrix were found in the
ash hopper discharge.
Testing was terminated after 1110 h of operation in
Test Segment 5. Only two of the DuPont PRD-66 filter
elements remained intact, four had suffered either
midbody fracture or failure at a location that was 3/4
below the flange, and sixteen filters had fractured at
the base of the flange. The outer surface of the intact
and fractured filters was generally ash free, particularly along the portion of the body that was adjacent to
the plenum support pipe (i.e., approximately midway
down the length of the filter element). Alternately, a
1-2 mm ash deposit remained along the outer surface
of the PRD-66 candles, near the bottom end cap. Divotlike formations resulted in lines which ran parallel
down both sides of the remaining intact and fractured
filter elements. Localized divoting was also observed
below the gasket sleeve which was installed around the
filter flange, as well as in alternate, isolated areas along
the filter body.
The mechanisms leading to divoting and midbody
failure of the DuPont PRD-66 filter elements in Test
Segment 5 are considered to be primarily related to
delamination areas which were present within the wall
of the filament-wound matrix (i.e., uneven winding and/
or localized drying or positioning of the elements during
manufacturing of the candles) and are not expected to
directly result from reactions of the process gas environment with the PRD-66 filter matrix. Post-test inspection indicated that fines were present within the 7 mm

PRD-66 filter wall. These may have resulted from

penetration of submicron fines through the PRD-66
outer membrane or were backpulsed into the matrix
after failure of an alternate candle. Ash, which has a
high thermal coefficient of expansion in comparison to
the ceramic, may have induced localized internal failure
within the wall during the numerous plant shutdown
and startup cycles in Test Segment 5, thus enhancing
the potential for divots to form. Midbody failure of the
element conceivably resulted once the wall had sufficiently weakened or thinned. Failure at the base of
the PRD-66 filter flange is attributed to the low load
bearing capability of the PRD-66 candle filters in
conjunction with the high thermal coefficient of expansion of the ash which surrounded the filter holder
mounts and which became wedged in between the
outer surface of the filter element and the metal holder.
Assuming that the strength of the PRD-66 candles
used in Test Segment 5 was equivalent to the strength
of the PRD-66 candles used in Test Segment 4, posttest strength characterization of the 1110 h, PFBC
exposed, PRD-66 filter matrix indicated that an increase
in strength appeared to have occurred along the ashfilled o.d. surface, while virtually no change in strength
was detected along the i.d. or pulse-cycled surface of the
filament-wound candle. Both as-manufactured and
field-tested candles exhibited brittle fracture characteristics during compressive or tensile strength testing.
In an attempt to discern whether changes occurred
within the DuPont PRD-66 filter material as a result
of operation at AEP, sections of the as-manufactured
and field-tested candles were removed and were subsequently subjected to SEM/EDAX analyses. As shown
in Figure 5, the PRD-66 filter matrix consists of numerous fibers which formed a filament yarn or bundle.
Each fiber is 10 m in diameter and is either sintered
or bonded to an adjacent fiber in the bundle or is
surrounded by a fine-grain, polycrystalline matrix which
assists in bonding adjacent fibers to each other. Frequently, voids are evident at the center of individual
fibers, as well as within the interior of the fiber bundle.
At various locations, the surface of the as-manufactured
fibers is mottled.
The composition of the fibers was predominantly
cordierite, with mullite, alumina, and/or cristobalite
present in the encapsulating polycrystalline layer which
formed after high fire of an alumina slurry along the
filament bundle. Infiltration of the slurry appeared to
primarily result along the outer periphery of the filament bundle, with limited deposition within the interior
of the bundle (i.e., between fibers). Incomplete infiltration, particularly at the center of the filament bundle
resulted in localized areas that were devoid of the
polycrystalline phase, and as a result, uncoated fibers
were frequently adjacent to each other.
Although difficulty was encountered during the characterization of the field-tested PRD-66 filters due to the
presence of fines throughout the wall, the morphology
of the material after 1110-1705 h of PFBC operation
closely resembled that of the as-manufactured filter
material. Frequently, however, where fibers were not
encapsulated with the polycrystalline phase, melting
and/or coalescence of adjacent fibers resulted.
Comment
Post-test inspection of the candles that were utilized
in Test Segment 5 indicated that ash had been carried
into the i.d. bore of many of the elements due to failure

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3391

Figure 5. Filament-wound oxide-based DuPont PRD-66 filter matrix: (a) cross-sectioned as-manufactured matrix; (b) association of the
fibers and the polycrystalline phase; (c) cross-sectioned fibers; (d) fiber outer surface.

of candles in a single array. As a result, ash fines

became tightly packed at the bottom of several of the
filter elements. After cooldown of the W-APF, cracks
were evident at 6-8 in. above the end cap of a Coors
alumina/mullite, a Schumacher Dia Schumalith F40,
and a 3M CVI-SiC composite filter which had been
internally plugged with ash fines. The difference in the
thermal coefficient of expansion of the ash in the
internal plug and the porous ceramic filter matrices was
considered to be responsible for crack initiation within
the various filter elements during the numerous plant
shutdown and startup cycles in Test Segment 5.
In contrast, however, failure was not observed in the
two remaining DuPont PRD-66 candles that were
completely or nearly completely filled with ash. This
does not eliminate the possibility that failure of the
other 20 PRD-66 filters did not, in part, result from ash
i.d. bore plugging. Table 4 summarizes the observed
candle filter failure mechanisms that were experienced
in the W-APF system during the entire five test campaigns at AEP.
HTHP PFBC Filter Qualification Testing
Prior to selection, installation, and operation in the
field, candle filter qualification testing is frequently

conducted at Westinghouse to assure that prototypic or

developmental filters demonstrate a >99.99% particle
collection efficiency during steady-state, simulated PFBC
operation. Similarly, the initial pressure drop across
the elements at process temperature, dust cake removal
efficiency, and the as-manufactured strength of the
flange and mechanical sealing/mounting of the element
within the filter holder are assessed. In addition,
Westinghouse qualifies the performance of the various
filters under extreme conditions as accelerated pulse
cycling which monitors the performance of the matrix
with respect to thermal fatigue, as well as exposure to
thermal transient events which simulate rapid startup
or shutdown ramps experienced during process upset
conditions.
In order to evaluate the performance of the advanced
second-generation filter elements, an array which included a DuPont PRD-66, a 3M CVI-SiC composite, and
a DuPont SiC-SiC composite filter was installed in the
Westinghouse HTHP PFBC test facility in Pittsburgh,
PA. Initially, the filter array was heated to temperatures of 843 C prior to initiating thermal transient
testing. A series of seven increasing-severity thermal
transients were delivered to the array which reduced
the outer surface temperature of the filter elements by

3392 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996
Table 4. Candle Filter Failure Mechanisms Experienced during Operation in the W-APF at AEP
filter element

max hours of operation

(test segment)

failure mechanism

Schumacher
Dia Schumalith F40

5855 (1, 2, 3, 4, 5)

ash bridging

Schumacher
Dia Schumalith FT20
Pall Vitropore 442T
Coors alumina/mullite
P-100A-1

1705 (4)

none experienced

2815 (4, 5)
2815 (4, 5)

none experienced
possible thermal
fatigue

3M CVI-SiC composite

1705 (4)

ash bridging

DuPont PRD-66

1705 (4)

divot formations;
midbody and
flange failure

Figure 6. Strengths of the as-manufactured, thermal transient,

and accelerated pulse cycled second-generation candle filters as
determined via C-rings tested in compression or tension.

6-110 C and 20-240 C within the first 5 and 60 s,

respectively, after transient initiation. Nine maximumseverity thermal transients followed. Subsequently, the
array was subjected to 10 accelerated pulse cleaning
cycles and a final maximum-severity thermal transient.
After 42 h of thermal transient testing, the filter array
was slow cooled and the elements were removed for
destructive characterization. As shown in Figure 6,
both the high-temperature compressive and tensile
strengths of the DuPont PRD-66 and 3M CVI-SiC
composite filter materials increased, while the hightemperature compressive and tensile strengths of the
DuPont SiC-SiC composite decreased. What is expected to have occurred is that crystallization and/or
oxidation was initiated in the DuPont PRD-66 and 3M
CVI-SiC composite filter materials, causing a strengthening of the matrices, similar to what had been identified during operation of the filter elements in the W-APF
system at AEP.
Characterization of the resulting morphology and
phase composition via SEM/EDAX and X-ray diffraction
analyses has not been completed for the thermal transient-tested DuPont PRD-66 and 3M CVI-SiC filter
materials. However, when the DuPont SiC-SiC composite was subjected to SEM/EDAX analyses, several
interesting changes were observed as a result of HTHP

comment
maximum elongation of 5-7 mm; reduction of strength;
conditioned strength of 800-1200 psi; binder crystallization;
bowing under load; failure at the base of the flange during
bridging; failure along end cap when i.d. filled with fines
improved high-temperature creep-resistant binder; no loss
of strength indicated
maximum elongation of 17-20 mm; reduction in bulk strength
midbody or failure at the base of the flange; end cap failure
when i.d. filled with fines; loss of strength experienced at
elevated temperatures
failure at the base of flange during ash bridging; failure along
end cap when i.d. filled with fines; removal of CVI-SiC
layer along end cap; apparent strength increase due to fines
filling matrix wall and/or phase crystallization/oxidation
failure occurred at the base of the flange when ash
encapsulated filter holder; body thinned where divots
formed; divots may be responsible for midbody fracture;
apparent strength increase due to fines filling matrix wall
and/or phase crystallization; minimal adherence of ash
along the outer filtration surface; minimal impact of
bridging and ash i.d. bore filling

PFBC thermal transient testing. These included removal of the interface layer that was originally deposited around the Nicalon fibers in the single-ply felt layer
of the DuPont SiC-SiC composite (Figure 7); evidence
of oxidation of the CVI-SiC outer surface; and removal
of the interface layer particularly along the outer fiber
bundle or tow, directly beneath the CVI-SiC encapsulating layer in the mesh support screen. Melting or
sintering of adjacent Nicalon fibers with the enhanced
phase that was added during the manufacture of the
mesh screen support layer was also evident. Retention
of the interface layer along the interior of the fiber
bundle or tow in the mesh screen support layer remains
to be determined.
Based on the resulting load versus deflection curves
that were generated during C-ring strength testing, the
fracture toughness of the DuPont SiC-SiC composite
appeared to be reduced after 42 h of thermal transient
testing. Loss of fracture toughness was primarily
attributed to removal of the interface layer in the singleply felt and mesh screen support layers and to the
sintering of the Nicalon fibers in the mesh screen
support layer.
In order to identify the viability and thermal fatigue
resistance of the advanced second-generation filters, a
second set of DuPont PRD-66, 3M CVI-SiC composite,
and DuPont SiC-SiC composite candles was installed
in the Westinghouse HTHP PFBC simulator and subjected to 3514 accelerated pulse cleaning cycles during
a period of 197 h. The temperature of the filter array
was maintained at 843 C while accelerated pulse
cycling was conducted.
As shown in Figure 8, the interface layer that initially
surrounded the Nicalon fibers in the single-ply felt layer
of the DuPont SiC-SiC filter matrix was removed after
197 h of accelerated pulse cycle testing. Similarly, crack
formations resulted along the outer periphery of the
Nicalon fibers. Typically, the cracks had rounded tips,
as well as segmented step-like characteristics.
Halo-like areas were readily evident along the
periphery of the Nicalon fibers in the single-ply felt.
These areas were generally enriched with oxygen and
effectively demarcated the location to which the cracks
penetrated. Bonding of the Nicalon fiber to the inner
surface of the CVI-SiC encapsulating shell often resulted near the crack formations.

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3393

Figure 7. Morphology of the DuPont SiC-SiC composite hybrid filter matrix: (a) as-manufactured single-ply felt layer; (b) as-manufactured
mesh screen support layer; (c) presence of the interface layer in the as-manufactured single-ply felt layer; (d) removal of the interface
layer in the single-ply felt layer after thermal transient testing.

Within the mesh screen support layer, thin CVI-SiC

bands which followed the contour of the Nicalon fibers,
the enhanced phase, and perhaps the interface layer
were evident. Near the periphery of the fiber bundle
or tow (i.e., adjacent to the CVI-SiC encapsulating
layer), as well as within the bundle, irregularly shaped
Nicalon fibers were evident. Melting of the fibers was
frequently observed, as well as mottling of the fiber
surface. These were considered to result as a response
or reaction of the Nicalon fibers with the enhanced
phase that was included in the mesh screen support
layer. Adjacent to the CVI-SiC encapsulating layer,
the melted fibers formed an interconnected network
which readily formed cracks during fast fracture. Void
formations that were observed in the fractured mesh
screen support layer may have resulted from fiber
pullout during sample preparation or alternately reflected removal of the interface phase during exposure
to simulated HTHP PFBC process operating conditions.
Notably the Nicalon fibers in the mesh screen support layer of the DuPont SiC-SiC composite filter
matrix did not exhibit crack formations along their
periphery.
Oxidation of the outer surface of the CVI-SiC encapsulating layer was again evident after 197 h of acceler-

ated pulse cycling in the Westinghouse HTHP PFBC

test facility.
As previously discussed, the reduced fracture toughness of the DuPont SiC-SiC filter matrix after 197 h
of accelerated pulse cycling was primarily attributed to
removal of the interface layer in the single-ply felt and
mesh screen support layers and to the sintering of the
Nicalon fibers in the mesh screen support layer.
After 800 h of steady-state, thermal transient and
accelerated pulsing of DuPont SiC-SiC candles in the
simulated HTHP PFBC process environment which
contained ash and an equivalent of 20 ppm gas-phase
sodium chloride at 1 atm pressure, further changes were
evident within the filter material [Alvin (1996)]. These
included an apparent swelling of the Nicalon fibers
within the single-ply felt layer. As a result, the fibers
fully, or nearly completely, filled the void in the CVISiC encapsulating shell. Bonding of the Nicalon fibers
to the inner wall of the CVI-SiC shell was also apparent
in the single-ply felt layer.
Corrosion Testing
Westinghouse has previously demonstrated the relative thermal/chemical stability of the oxide-based alu-

3394 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996

Figure 8. Morphology of the DuPont SiC-SiC composite hybrid filter material after 197 h of exposure to HTHP PFBC accelerated pulse
cycle conditions: (a) single-ply felt layer; (b) mesh screen support layer.
Table 5. Strength Characterization of the Steam/Air and Alkali/Steam/Air-Exposed 3M CVI-SiC Composite Minicandles
strength at 870 C, psi
C-rings tested in compression
filter ID no.

exposure

3M-453707
3M-453715
3M-453710
3M-453708
3M-453708

as-manufactured (15 mm)

as-manufactured (25.4 mm)
steam/air (15 mm)
alkali/steam/air (15 mm)
alkali/steam/air (25.4 mm)

time, h

400
400
400

temp, C

compressive load, lb

composite matrix

triaxial support braid

870
870
870

3.3 ( 0.36
4.0 ( 0.14
2.10 ( 0.62
1.7
2.85 ( 0.64

1739 ( 259
1183 ( 78
987 ( 250
717
762 ( 151

11784 ( 1476
8561 ( 463
8533 ( 1546
6298
5808 ( 695

mina/mullite filter matrix during exposure to hightemperature, oxidizing conditions which contain gasphase alkali and steam. In order to demonstrate the
stability of the oxide-based DuPont PRD-66 filter material, minicandles, which were manufactured with an
75 mm section of open filtration area, were exposed
for 400 h at 870 C to 5-7% steam/air and 20 ppm NaCl/
5-7% steam/air flow-through testing. During the 400
h exposure, the minicandles were subjected to pulse
cycling at 20 min intervals.
High-temperature, flow-through testing was similarly
conducted for the non-oxide-based, advanced, secondgeneration filters using minicandles of the 3M CVISiC composite matrix and 70 mm diameter 2-3 mm
thick disks of the DuPont SiC-SiC composite matrix.
Post-test characterization of each material included
C-ring compression testing at 870 C of the as-manufactured and flow-through tested minicandles and 4-point
bend, 1/4-point flexural strength testing at 870 C of the
as-manufactured and flow-through tested filter disks.
SEM/EDAX analyses have been completed for both the
3M CVI-SiC and DuPont SiC-SiC tested materials. A
discussion of these results follows.
3M CVI-SiC Composite Matrix
After exposure to either the high-temperature steam/
air or alkali/steam/air flow-through environment, the
3M minicandles appeared to be intact, retaining their
initial configuration. What was readily apparent, how-

ever, was the very brittle nature of the outer confinement layer along the alkali/steam/air-exposed 3M CVISiC composite material. In several locations, the fiber
bundles of the confinement layer weave were broken or
missing.
During preparation of the 3M CVI-SiC matrix for
strength testing, two as-manufactured minicandles were
cut into 15 or 25.4 mm sections for high-temperature
strength testing of C-rings in compression. These
sections were removed from the open filtration area of
the minicandle. Similarly, a 25.4 mm O-ring section
was removed from the reinforced filter matrix directly
below the flange. All three layers of the as-manufactured 3M CVI-SiC composite matrix remained bonded
together during preparation of either the 15 and 25.4
mm C-ring or 25.4 mm O-ring samples.
Similar C-ring and O-ring samples were removed
from the steam/air and alkali/steam/air-exposed 3M
CVI-SiC minicandles. During sample preparation, the
three composite layers of the steam/air-exposed 3M
CVI-SiC matrix remained bonded together, while debonding occurred along both the 15 and 25.4 mm C-ring
sections that were removed from the alkali/steam/airexposed 3M filter material.
What was also apparent for both the steam/air and
alkali/steam/air-exposed 3M minicandles was the extreme difficulty that was encountered during cutting of
the reinforced material directly below the flange. As
shown in Table 5, the applied compressive load required

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3395

Figure 9. Accelerated oxidation of the 3M CVI-SiC composite

filter material after 400 h of exposure at 870 C to the 20 ppm
NaCl/5-7% steam/air flow-through test environment.

to fail C-rings that were removed from the 3M CVISiC composite filtration area, and the resulting strength

decreased after 400 h of exposure in the 870 C steam/

air of alkali/steam/air flow-through environment. The
resulting load versus deflection curves that were generated during C-ring compressive strength testing tended
to indicate that a reduction in the fracture toughness
of the embrittled, 3M CVI-SiC composite matrix resulted after exposure at high temperature to the steam/
air and alkali/steam/air test conditions.
SEM/EDAX analyses of the steam/air and alkali/
steam/air-exposed 3M CVI-SiC filter matrices indicated
that the 1-2 m SiC layer which had been deposited
along the outer confinement layer and the aluminabased filtration mat fibers had generally been removed.
The 100 m CVI-SiC deposited layer which coated
the triaxial support braid generally was intact but was
enriched with oxygen (Figure 9). Due to the very thin
nature of the interface layer that initially coated the
Nextel and alumina-based fibers, resolution and identification of its presence in either the as-manufactured
or flow-through tested minicandles could not be made
using SEM/EDAX analytical techniques. As shown in

Figure 10. Morphology of the 3M CVI-SiC composite filter material: (a) as-manufactured material; (b) embrittled material after 400
h of exposure at 870 C to 20 ppm NaCl/5-7% steam/air; (c) melt formation of the Nextel and CVI-SiC matrix; (d) melt formation of the
alumina fibers and the CVI-SiC matrix.

3396 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996

Figure 11. Morphology of the cross-sectioned DuPont SiC-SiC

composite single-ply felt layer after 400 h of exposure at 870 C
to 5-7% steam/air flow-through test conditions.
Table 6. Strength Characterization of the Steam/Air and
Alkali/Steam/Air-Exposed DuPont PRD-66 Minicandles
C-rings testes in
compression at 870 C
filter
ID no.

exposure

time, temp,
h
C

D-282 as-manufactured
(15 mm)
D-289 steam/air
400
(15 mm)
D-288 alkali/steam/air
400
(15 mm)

load,
lb

strength,
psi

9.65 ( 0.72 1352 ( 135

870

11.88 ( 0.71 1541 ( 146

870

8.58 ( 1.70 1093 ( 265

Figure 10, exposure of the 3M CVI-SiC composite filter

matrix to the 20 ppm NaCl/5-7% steam/air environment led to the formation of a melt-like bond between
the Nextel or alumina fibers and the remaining encapsulating CVI-SiC shell. Aluminum- and silicon-rich
micron, nodular formations were evident along the
Nextel and alumina-based fibers in all three composite
layers of the steam/air-exposed 3M CVI-SiC composite
matrix.
DuPont PRD-66 Filament-Wound Matrix
After 400 h of exposure in the 870 C steam/air and
alkali/steam/air environment, the DuPont PRD-66 minicandles generally remained intact. However, along the
outer membrane surface of the filtration area of the
alkali/steam/air-exposed minicandle, numerous longitudinal cracks were evident. In addition, a 1 1.5 cm
section of the membrane and several of the underlying
structural support fibers had been removed along the
alkali/steam/air-exposed minicandle.
As shown in Table 6, the high-temperature, compressive strength of the filtration area in the brittle PRD66 filter matrix appeared to slightly increase after 400
h of exposure to the steam/air flow-through environ-

Figure 12. Morphology of the cross-sectioned DuPont SiC-SiC

single-ply felt layer after 400 h of exposure at 870 C to the 20
ppm NaCl/5-7% steam/air flow-through test environment: (a)
oxidation of the CVI-SiC surface; (b) depletion of the interface
layer and formation of the oxygen-enriched phase surrounding the
Nicalon fibers; (c) glazed membrane.

ment, while a slight decrease in strength was observed

after 400 h of exposure in the alkali/steam/air environment. To date, SEM/EDAX analyses have not been
performed on either the steam/air or alkali/steam/airexposed DuPont PRD-66 minicandles.
DuPont SiC-SiC Composite Matrix
As previously discussed, the oxidation of the CVISiC layer that is deposited along Nicalon fibers to form

Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996 3397
Table 7. 4-Point Bend, 1/4-Point Flexural Strength of the Steam/Air and Alkali/Steam/Air-Exposed DuPont SiC-SiC
Filter Matrixa
membrane tested in compression,b psi
filter matrix
as-manufactured
steam/air
alkali/steam/air
alkali/steam/air

25 C
13118 (
NAf
NAf
NAf

870 C

1697d

3128d

12068 (
5427 ( 749e
3953 ( 90e
7269 ( 557d

membrane tested in tension,c psi

25 C
7269 (
NAf
NAf
NAf

272d

870 C
4981 ( 1801d
4450 ( 541e
3092 ( 304e
2106 ( 454d

a Span: 1.57 in. Cross-head speed: 0.02 in./min. b Pulse-cycled surface strength (open-mesh screen). c Membrane-coated surface strength
(single-ply felt mat). d Bend bars were cut parallel to the open-mesh screen ribs. e Bend bars were cut on a diagonal to the open-mesh
screen ribs. f NA: not available.

Table 8. High-Temperature Filter Material Stability

filter matrix

thermal fatigue

thermal shock

accelerated pulse cycling

thermal transient testing

Coors alumina/mullite
P-100A-1

intact; reduced strength

susceptible

Schumacher
Dia Schumalith F40
Pall Vitropore 442T

intact; reduced strength

intact

intact; possible strength loss

intact

3M CVI-SiC composite intact; strength increaseb

DuPont PRD-66
DuPont SiC-SiC
composite

intact; strength
increaseb

flange failure; matrix

intact; strength
strengtheningb
increaseb
intact; strength decrease;
butt seam rupture;
loss of interface coating;
loss of interface;
crack formations along
melting/bonding of
periphery of fibers; bonding
fibers with enhanced
of fibers to CVI-SiC;
phase
melting of enhanced phase
along fibers; surface
oxidation

flow-through oxidation/corrosion
steam/air

alkali/steam/air

conversion of
conversion of amorphous
amorphous phase
phase to albite/anorthite;
to anorthite;
strength increase
strength increase
eutectic formation; plastic
NTa
deformation
NTa
eutectic formation; plastic
deformation expected
strength loss;
strength loss; accelerated
surface oxidation
oxidation; eutectic
formation; embrittlement
strength increase
strength loss; membrane
cracking and spalling
strength loss;
strength loss; accelerated
surface oxidation;
oxidation; eutectic
bonding of fibers
formation; embrittlement;
to CVI-SiC
removal of
interface layer

NT: not tested. b Morphology changes to be determined.

the dual single-ply felt or hybrid (i.e., single-ply felt/

mesh screen support layer) DuPont SiC-SiC composite
matrix has been shown to occur during exposure to
simulated HTHP PFBC process conditions. Due to the
thickness of the CVI-SiC deposit (i.e., 10-20 mm),
oxidation may represent a long-term degradation mechanism for the DuPont SiC-SiC composite filter matrix.
Alternately, removal of the interface layer, crack formations along the periphery of the Nicalon fibers, melting
of the fibers in the presence of the enhanced phase,
reduced fracture toughness, and a resulting, relatively
embrittled material are likely to directly impact the
short-term stability and viability of the DuPont SiCSiC composite filter matrix.
After 400 h of exposure of the DuPont SiC-SiC hybrid
matrix to the 870 C flow-through, steam/air environment, melting or bonding of the Nicalon fibers to the
inner surface of the CVI-SiC encapsulating shell
resulted in the single-ply felt layer (Figure 11). Generally, the interface layer that initially surrounded the
Nicalon fibers was considered to have been removed.
Oxidation of the CVI-SiC outer surface was also
observed (Figure 12). Oxidation of the CVI-SiC outer
surface was accelerated when the DuPont SiC-SiC
material was exposed for 400 h at 870 C to the 20 ppm
NaCl/5-7% steam/air flow-through test environment.
Depletion of the interface layer along the Nicalon
fibers in the single-ply felt was also evident after 400 h
of exposure at 870 C to the 20 ppm NaCl/5-7% steam/
air flow-through environment. In contrast to the steam/
air exposure, oxidation of the outer surface of the
Nicalon fibers in the single-ply felt occurred during the
alkali/steam/air exposure. Oxidation was expected to

have resulted either from outward diffusion of oxygen

from the as-manufactured Nicalon fibers or from reaction of the fiber surface with the flow-through gases.
Frequently, halo-like rings which were enriched with
oxygen resulted around the periphery of the Nicalon
fibers. Minor crack formations were evident in these
areas.
Retention of the enhanced phase that encapsulated
the Nicalon fibers in the bundle or tow in the mesh
screen support layer in the DuPont SiC-SiC hybrid
filter matrix resulted during exposure to either the
5-7% steam/air or 20 ppm NaCl/5-7% steam/air flowthrough test environment. In localized areas, retention
of the interface layer may be evident within the interior
of the bundle.
In contrast, the fine-grain SiC grit which formed the
membrane of the SiC-SiC composite filter matrix
tended to form a glassy phase during the high-temperature, alkali/steam/air flow-through exposure. The
formation of the glassy phase reduced gas flow permeability through the matrix. The glazed surface which
was enriched with sodium could conceivably serve as a
potential site for collection and adherence of ash fines
at process operating temperatures, ultimately causing
blinding of the filter element surface.
The residual strength of the hybrid DuPont SiC-SiC
matrix was determined at 870 C via 4-point bend, 1/4point flexural strength testing of bend bars that were
removed from the flow-through tested filter disks.
Approximately 60% of the as-manufactured strength
remained along the surface of the material that had
been subjected to pulse cleaning, while only 42% of the
as-manufactured strength remained along the mem-

3398 Ind. Eng. Chem. Res., Vol. 35, No. 10, 1996

brane-coated surface that sorbed alkali during the 400

h alkali/steam/air, flow-through test exposure (Table 7).
Based on the load versus deflection curves that were
generated during high-temperature, flexural strength
testing, the fracture toughness of the low fiber volume,
DuPont SiC-SiC filter matrix appeared to decrease
after 400 h of exposure in either the 870 C steam/air
or alkali/steam/air environment.
Summary and Conclusions
Table 8 summarizes the results which were obtained
during HTHP PFBC and corrosion testing at Westinghouse for both the monolithic and advanced secondgeneration porous ceramic filter materials. The failure
mechanisms of the oxide-based materials include thermal fatigue and/or thermal shock for the monolithic
matrices, while component construction and low load
bearing capabilities primarily impact the integrity of the
advanced second-generation filter materials. The nonoxide-based materials tend to undergo oxidation, particularly in the presence of gas-phase alkali, leading to
enhanced creep of the monoliths and embrittlement in
the advanced second-generation filter materials.
Generally, the monolithic materials tend to lose bulk
strength during initial exposure to accelerated pulse
cycling and thermal transient qualification testing,
similar to what has been experienced in the field. The
DuPont PRD-66 and 3M CVI-SiC composite materials,
however, tend to experience an increase in their residual
bulk strength during accelerated pulse cycling and
thermal transient qualification testing, again similar to
what has been observed during field testing.
Removal of silicon carbide along the outer confinement layer of the 3M CVI-SiC filter was observed not
only after 197 h of HTHP PFBC accelerated pulse
cycling but also after 1100 h of operation in the W-APF
system at AEP. Exposure time at temperature in the
oxidizing environment is expected to impact the rate of
removal and/or oxidation of the CVI-deposited SiC
matrix.
Although the DuPont SiC-SiC composite material
tends to embrittle under oxidizing conditions, as well
as lose bulk strength, the residual strength of the
DuPont matrix remains relatively high in comparison
to the conditioned strengths of the monolithic materials.
The residual strength of the 3M CVI-SiC composite
material also significantly exceeds the conditioned
monolithic filter material strength.
Although residual strengths of the non-oxide, fiberreinforced porous ceramic filter materials exceed those
of the monoliths, the load bearing capability of the
advanced second-generation filter materials is substantially lower than that of the monoliths. This may
impact the ability of the advanced filters to remain
intact if ash becomes lodged within the filter i.d. bore
or bridges between adjacent filter elements. Similarly,
the ability of the advanced filters to be easily and
successfully removed from the holder mounts, which

generally become encased with deposited ash, may also

make disassembly labor intensive.
Although several critical issues need to be addressed
prior to long-term use of the advanced second-generation filter materials, 5855 h of successful operation have
been demonstrated for the monolithic clay-bonded silicon carbide filter elements under 620-845 C PFBC
conditions. The viability of the monoliths to successfully
operate in an oxidizing environment where temperatures are g900 C may require the use of high-temperature creep-resistant materials. Similarly, active oxidation of the non-oxide matrices may result in the 500700 C gasification environment, and therefore the use
of oxide-based filters may be required. Many of these
issues will be addressed in the near future during
operation of the W-APF systems in the Foster Wheeler
pressurized circulating fluidized-bed combustion test
facility in Karhula, Finland, as well as at the Sierra
Pacific integrated gasification and combined cycle (IGCC)
test facility in Reno, NV, and at the Southern Company
Service combustion and gasification test facilities in
Wilsonville, AL.
Acknowledgment
The continued support and guidance of DOE/METC
are gratefully acknowledged.
Literature Cited
Alvin, M. A.; Lippert, T. E.; Lane, J. E. Assessment of Porous
Ceramic Materials for Hot Gas Filtration Applications. Am.
Ceram. Soc. Bull. 1991, 70 (9), 1491-1498.
Alvin, M. A.; Tressler, R. E.; Lippert, T. E.; Diaz, E. S.; Smeltzer,
E. E. Durability of Ceramic Filters. Paper presented at the CoalFired Power Systems 94sAdvances in IGCC and PFBC Review
Meeting, Morgantown, WV, June 21-23, 1994.
Alvin, M. A.; Lippert, T. E.; Diaz, E. S.; Smeltzer, E. E. Filter
Component Assessment. Paper presented at the Advanced CoalFired Power Systems 95 Review Meeting, Morgantown, WV,
June 27-29, 1995a.
Alvin, M. A.; Lippert, T. E.; Diaz, E. S.; Smeltzer, E. E. Thermal
and Chemical Stability of Ceramic Candle Filter. Paper presented at the Advanced Coal-Fired Power Systems 95 Review
Meeting, Morgantown, WV, June 27-29, 1995b.
Alvin, M. A.; Smeltzer, E. E.; Sanjana, Z. N.; Oberst, J. P.; Lippert,
T. E. Stability of the DuPont SiC-SiC Composite Filter Matrix
during Exposure to Simulated PFBC Conditions. Work performed at Westinghouse under DOE Cooperative Agreement No.
DE-FC02-92CE40994, Jan 10, 1996.
Mudd, M. J.; Hoffman, J. D.; Reinhart, W. P. Tidd Hot Gas Clean
Up Program. Final Report, American Electric Power Service
Corp., work performed under Contract No. DE-FC21-89MC26042,
July 1995.

Received for review March 5, 1996

Revised manuscript received May 29, 1996
Accepted June 6, 1996X
IE960128I

X Abstract published in Advance ACS Abstracts, August 15,

1996.