Improved Filtration Performance of Continuous Alumina-fiber Reinforced Mullite Ceramics

Improved Filtration Performance of Continuous Alumina-fiber-reinforced Mullite Composites for Hot-Gas CleaningSatoshi Kitaoka w and Naoki KawashimaJapan Fine Ceramics Center,Nagoya 456-8587,JapanYoshinobu KomatsubaraMitsui Mining Material Co.,Omuta 836-0062,JapanAkira YamaguchiNagoya Institute of Technology,Nagoya 466-8555,JapanHisao SuzukiShizuoka University,Hamamatsu 432-8561,JapanThe effect of filtration layer morphology on filtration perform-ance at 673K was investigated for continuous alumina-fiber-reinforced mullite composite filters that capture fly ash on their outer surfaces.Two types of filtration layers were prepared on the outer surfaces:a mono-layer consisting of mullite agglom-erated particulates and a bi-layer,with mullite whiskers formed by a vapor-phase reaction and strongly adhered to the particu-lates exposed on the outer surfaces.For filters with mono-fil-tration layers,the filtration efficiency was improved slightly by increasing the filtration layer thickness.However,the maximum differential pressure increased during the filtration tests.Adhe-sion of whiskers to the outer surface decreased the maximum differential pressure to about one-third less than that with the mono-filtration layer.This low differential pressure remained constant throughout the duration of the test,with corresponding increases in filtration efficiency.I.IntroductionHOT -GAS particulate filters are key components for advanced coal-based power generation systems such as pressurized fluidized-bed combustion (PFBC)and integrated coal gasifica-tion combined cycle (IGCC).Filters should allow easy removal of dust cakes accumulated on the filter surfaces and excellent reversibility of the differential pressure during pulse cleaning.For this reason,adhesion of the dust cakes to the filters should be suppressed as far as possible.Adhesion is generally related to the formation of alkali-containing liquids in the cake and chem-ical reactions between the filter element and the fly ash.1Chem-ical reactions have a particularly strong effect on dust adhesion and make pulse cleaning difficult.To simplify the study of the interaction of fly ash with filters,we have considered a repre-sentative ash containing SiO 2and Al 2O 3(the principal constit-uents)and K 2O (one of the most reactive constituents,whichsignificantly affects the nature of the ash at high temperatures).As the SiO 2–Al 2O 3–K 2O phase diagram indicates,2mullite is one of the equilibrium phases that can form from ashes of simple composition (e.g.,SiO 2:Al 2O 3:K 2O 571:28:1wt%for Ash No.10,distributed by the Association of Powder Proc-ess Industry and Engineering,Japan;SiO 2:Al 2O 3:K 2O 563.8:35.8:0.4wt%for the fly ash that was collected from a hop-per at the PFBC demonstration plant z ).The phase diagram indicates that Al 2O 3reacts with fly ash 3and,therefore,direct contact should be avoided.Because of the poor thermal conductivity of unreinforced ce-ramic filters,however,large temperature gradients form between the inner and outer surfaces of the filters during excessive gas cleaning,or as the result of sudden ignition of char accumulated on the filtration surfaces or from a combination of the two rea-sons.Thermal stresses induced by the temperature gradient lead to catastrophic failure of the filters,4,5resulting in low reliability.To be used successfully,these filters require precise control of operating conditions and constant observation.Continuous-fiber-reinforced filters are expected to show ex-cellent damage tolerance,durability,and reliability.Oxide ma-terials are inherently stable in oxidizing environments such as the PFBC system and have more corrosion resistance to alkali species than the non-oxides SiC and Si 3N 4.6Because of the poor sinterability of mullite,if this material is used as the matrix of the fiber-reinforced filter,the composite must be sintered at high temperatures to increase the bonding strength among the mullite grains.Unfortunately,mullite fibers stable at high temperatures have not yet been produced,so that only high-purity a -Al 2O 3fiber (which has the highest thermal stability,1673K,of any oxide fiber)is available.However,if the Al 2O 3fibers are exposed on the filtration surfaces,the fibers react with the ash cakes,leading to an increased differential pressure of the filters.There-fore,the filtration surfaces of the filters should be fully covered with mullite.We have developed cylindrical composite filters that capture fly ash on their outer surfaces.3The filters consist of a mullite filtration layer and a substrate layer of continuous Al 2O 3-fiber-reinforced mullite matrix composite.The filters were fabricated by dipping a two-dimensional woven sheet of the fibers into a slurry that had been prepared from large mullite particles and an alkoxide-derived mullite precursor solution.The precursor coat-ed onto the large mullite particles formed very fine mullite45J ournalJ.Am.Ceram.Soc.,88[1]45–50(2005)DOI:10.1111/j.1551-2916.2004.00027.xRalf Riedel—contributing editorThis research was supported in part by the Center for Coal Utilization,Japan (CCUJ),under the Next-Generation Coal Utilization Technology Development Study (Environ-mentally Friendly Coal Combustion Technology),promoted by the New Energy and In-dustrial Technology Development Organization (NEDO),under contract from the Ministry of Economy,Trade,and Industry (METI).It has also been supported by Japan Coal Energy Center (JCOAL),under a contract from METI,as a part of regional R&D consortium projects.wAuthor to whom correspondence should be addressed.e-mail:kitaoka@jfcc.or.jpManuscript No.10492.Received August 28,2003;approved July 20,2004.zWakamatsu Pressurized Fluidized-Bed Combustion combined cycle demonstration plant of the Electric Power Development Co.,Japan.particulates during heating.These particulates accelerated the neck growth between the large particles,even at low tempera-tures.Therefore,the precursor allowed the sintering of large mullite particles without Al2O3fiber damage,resulting in in-creasedflexural strength for the composites while inhibiting de-lamination of the wovenfiber sheets.Although the morphology of thefiltration layers strongly af-fectsfiltration behavior,there is limited information on the op-timal morphology for improvingfiltration performance.In particular,if a layer consisting of particles with large aspect ra-tios,such as whiskers,is formed on thefilter surfaces,it can be expected to have a small open pore size and large open porosity, resulting in a decreased differential pressure and more efficient dust removal.In this case,the particles need to be stronglyfixed to thefilter surfaces to prevent them from falling out as a result of shock during gas cleaning.For the present study,the effect offiltration layer morphol-ogy on the performance of continuous Al2O3-fiber-reinforced mullite composites was investigated.Two types offiltration lay-ers were produced on thefiber-reinforced composites:a mono-layer consisting of mullite particulates and a bi-layer with mul-lite whiskers formed by vapor-phase reaction and strongly ad-hered to mullite particulates exposed on the outer surfaces.II.Experimental Procedure(1)MaterialsThe mullite precursor coated onto the large mullite particles ac-celerated sintering of the large particles,resulting in a lower sin-tering temperature without degradation of the Al2O3fibers.3 The precursor solution was prepared by a polymeric sol–gel process from aluminum and silicon alkoxides as follows:7First, tetraethoxysilane(TEOS)was added to dehydrated ethanol (TEOS concentration of10mol/m3).Distilled water and hydro-chloric acid were then dropped into the solution in molar ratios of2and0.1,respectively.The TEOS was partially hydroly-zed by reflux at343K for50h,producing what is hereafter called the silicon precursor.Next,aluminum isopropoxide (Al(OCH(CH3)2)3)at a concentration of10mol/m3was dis-solved in2-methyl-1-propanol at room temperature and ref-luxed at368K for24h.The solution obtained is referred to hereafter as the aluminum precursor.A mullite precursor solu-tion was synthesized by mixing the silicon precursor with the aluminum precursor(Al/Si molar ratio53)and stirring at room temperature for5h.The chemical copolymerization of partially hydrolyzed species resulted in high compositional homogeneity of the precursor solution.A slurry was prepared from the mullite precursor solution and mullite particles(KM101,mean diameter0.75m m;KCM Co.,Nagoya,Japan).The concentration of the mullite particles was25wt%.A two-dimensional woven sheet of a-Al2O3fibers (ALMAX,Al2O3purity of99.5wt%,mean diameter10m m; Mitsui Mining Material Co.,Tokyo,Japan)was vacuum infil-trated with the slurry.The infiltrated sheet was wound around an Al2O3cylindrical mold and then dried during rotation at100 rpm to prevent gravity segregation of the slurry in the sheet.The centrifugal force induced by the rotation moved part of the in-filtrated slurry to the outer surface of the sheet,resulting in the buildup of a mullite layer on the outer surface.Next,the sheet with the mold was sintered at1673K in air for2h.The cylin-drical composite obtained(outer diameter55mm,thickness 2.5mm,width150mm)is hereafter referred to as sample A.The thickness of the mullite layer on the outer surface of sample A was B5–10m m,which acted as afiltration layer.Thefiber frac-tion of the supporting layer was B30vol%.In case of dustfiltration at the exterior wall of the cylindrical composite,a decrease in pore size from the interior of the cyl-inder to its exterior,as well as a uniform size distribution of the exterior pores,was necessary for easy removal of the cake dur-ing back blowing.To achieve this,two types offilters were pro-duced using sample A as a substrate layer.For onefilter,a slurry consisting of15wt%mullite particles and the precursor solutionwas sprayed onto the outer surface of sample A,and thefilterwas sintered again at1673K,in air,for1h.Thisfilter is referredto as sample B.Anotherfilter was prepared by adhering mullite whiskersonto the outer surface of sample A by a vapor-phase reaction.This whisker-synthesizing method has been referred to in previ-ous studies.8,9First,TEOS and aluminum nitrate enneahydrateAl(NO3)3Á9H2O were dissolved in dehydrated ethanol.AlF3 powders then were added to the solution and fully mixed in.Theamount of AlF3was2wt%of the Al(NO3)3Á9H2O,and the Al/ Si molar ratio of the mixture was3.An ammonia solution wasadded to the mixture and precipitates were produced.The pre-cipitates were dried at573K for24h.A slurry consisting of thedried powder(15wt%),polyvinyl butyral(1wt%),and dehy-drated ethanol was prepared.The slurry was applied liberally to the outer surface of the sample,and the coated sample then was dried at room temperature.The cylindrical composite was placed in an airtight container andfired at1473K for2h.After firing,the surplus mullite whiskers that had formed on the outer surface were removed by ultrasonic washing in water. The edges of the remaining whiskers were strongly adhered to the mullite particles exposed on the outer surface,which was fully covered by the whiskers.Thisfilter is referred to as sample C.When the adhesion of the whiskers to the top surface was achieved at higher temperatures and for longer times than those for sample C,the whiskers on the surface grew,and neighboring whiskers combined,resulting in the formation of plate-like particles.(2)Measurement and AnalysisThefiltration surfaces and cross sections of thefilters were ob-served by scanning electron microscopy(SEM).For sample C, mullite whiskers on the surface were examined by transmission electron microscopy(TEM).Chemical analysis,with a spatial resolution of B10nm,was performed by energy-dispersive X-ray spectroscopy(EDS)and electron diffraction(ED).The po-rosity and distribution of the pores in the composite were meas-ured by mercury intrusion porosimetry.The permeability,k,of thefilters was determined from the relationship between theflow rate and the applied N2pressure, according to Darcy’s law,k¼Lu mD Pð1Þwhere D P is the differential pressure across afilter of thickness L.The superficial velocity,u,of the gas with viscosity,m,is as-sumed to be the same throughout thefilter.Thefiltration performance of the samples was evaluated us-ing an apparatus(Hosokawa Micron Corp.,Osaka,Japan)in accordance with the German VDI/DIN standard3926.10The cylindrical composites were cut into100-mm lengths,and both bases of the cylindrical samples were sealed by the holders in the apparatus.Dust-containing air was sucked inward through the samples,and the dust cakes that accumulated on the outer sur-faces were removed by cyclically blowing air outward.The dif-ferential pressure across the samples was monitored during the filtration tests.A sample offly ash,which was collected from a hopper at the PFBC demonstration plant,was used as the dust for thefiltration tests.Figure1shows the particle size distribu-tion of the dust,which has two peaks.The average diameter of the dust particles was7.4m m.Figure2shows an SEM micro-graph of the dust.Spherical particles,which were probably formed by solidification of melted dust,and agglomerated par-ticles were observed.Thefiltration performance using this dust was evaluated at673K,the upper-limit temperature of the ap-paratus,for a dust concentration of5g/m3,afiltration velocity of0.097m/s,a pulse blowing pressure of0.2MPa,a blowing interval of150s,and a total number of204blowing cycles. These tests were conducted as a preliminary assessment of the46Journal of the American Ceramic Society—Kitaoka et al.Vol.88,No.1performance of the filters,which in practice would be used at temperatures 41100K.The dust passing through the filter samples was captured with filter paper.The average dust con-centration in the clean gas after filtration was determined by di-viding the mass of the captured dust by the total amount (5.1m 3)of clean gas that passed through the filter paper.The filtra-tion efficiency,Z ,was calculated as follows:Z %ðÞ¼1ÀC =C i ðÞÂ100ð2Þwhere C i and C are the dust concentrations before and after fil-tration,respectively.III.Results and DiscussionFigure 3shows SEM micrographs of typical cross-sections and filtration layer surfaces of the samples.For sample A,the outer surface was covered with mullite particulates that had accumu-lated on the Al 2O 3fiber sheets.The thickness of the particulate layer was B 5–10m m.This layer consisted of agglomerated mul-lite particles with many pores (less than a few micrometers)among the agglomerates and acted as the filtration layer.Al-though regions among the fiber bundles were filled with the mullite particulates,there were many pores in the bundles.As shown in Fig.3,large gaps were observed among the bundles and agglomerated particulates.For sample B,the mullite par-ticulate layer,which was 20–30-m m thick,acted as the filtration layer.The filtration surface seemed less uneven,and the pores smaller,among the agglomerates of sample B than in sample A.In sample C,the whiskers were arranged on the particulate layer perpendicular to the outer surface.The thickness of the whisker layer was the same as the length (B 4m m)of the whiskers.The aspect ratio (length/width)of the whiskers was B 15.The filtra-tion surface of sample C was fully covered with whiskers.There were no large depressions,such as those of sample A,and the pores observed on the surface of sample C seemed to be smaller than those of the other samples.Incidentally,the microstructure of the substrate layer of sample C was the same as those of the other samples.The ED pattern of a single whisker corresponded to that of a single crystal of mullite.TEM–EDS analysis of the whiskers indicated an average Si/Al molar ratio of 0.34,similar to the stoichiometric composition (0.33)of mullite.No deteri-orated layers were observed at the interfaces between the mullite whiskers and the mullite particulate layers that had accumulated on the Al 2O 3fiber sheets.We propose a mechanism for the formation of whiskers per-pendicular to the outer surface based on thermodynamic argu-ments.Figure 4shows the equilibrium partial pressures of the gaseous species and stable regions of the condensed phases un-der a constant F (g)partial pressure of 1.3Â10À8Pa,calculated from the decomposition reaction of AlF 3(s)(AlF 3(s)-Al (1)13F (g))as a function of O 2partial pressures at 1473K for the Al–O–F–Si system.As shown in Fig.4,when AlF 3(s)particu-lates exist on the outer surface of the cylindrical composites,the AlF 3(s)reacts with oxygen to mainly produce AlOF 2(g)as follows:AlF 3ðs Þþ1=2O 2ðg Þ!AlOF 2ðg ÞþF ðg Þð3ÞConsequently,an oxygen potential gradient forms on the outer surface.Under O 2partial pressures of o 3.2Â10À24Pa,the partial pressure of AlF (g)is the highest of all the gaseous species in the system.The partial decomposition reaction of AlF 3(s)can be represented as follows:AlF 3ðs Þ!AlF ðg Þþ2F ðg Þð4ÞFor O 2partial pressures of 43.2Â10À24Pa,the partial pressure of the AlF decreases,and the AlF condenses into Al 2O 3.AlF ðg Þþ3=4O 2ðg Þ!1=2Al 2O 3ðs Þþ3F ðg Þð5ÞUnder O 2partial pressures of o 9.6Â10À19Pa,silicon species in the vicinity of the AlF 3(s)particulates change to Si (s),accompa-nied mostly by the production of SiF 4(g).When O 2partial pres-sures are 49.6Â10À19Pa,SiF 4(g)condenses into SiO 2as follows:SiF 4ðg ÞþO 2ðg Þ!SiO 2ðs Þþ4F ðg Þð6ÞThe Gibbs free energy for the formation of mullite from Al 2O 3and SiO 2at 1473K is À20.7kJ/mol.Therefore,when both Al 2O 3and SiO 2coexist as stable phases at higher O 2partial pressures,as shown in Fig.4,mullite may be produced accord-ing to6AlOF 2ðg Þþ2SiF 4ðg Þþ7=2O 2ðg Þ!3Al 2O 3Á2SiO 2ðs Þþ20F ðg Þð7Þ10−1100101102103Diameter, µm543210V o l u m e , %d av = 7.4 µmFig.1.Particle size distribution of the flyash.Fig.2.SEM micrograph of the fly ash.January 2005Filtration Performance of Mullite Composites for Hot-Gas Cleaning 47Because the oxygen partial pressure at convex points on the mullite particulate layer on the outer surface of the composite is higher than in the troughs,the convex points may serve as nu-cleation sites for the vapor-phase reaction,forming mullite is-lands.Since the tip of the island is exposed to a higher oxygen partial pressure than the edges,the islands will grow perpendic-ular to the outer surface to form whisker-shaped mullite crystals.Because the pore size distributions of the filtration layers alone are difficult to evaluate and separate from those of thesubstrate layers,pore size distributions were measured for the entire sample.The thickness ratios of the filtration layers to the samples were o 1%.Consequently,the measured distribu-tions correspond most closely to those of the substrate layers.The distributions of all the samples were very similar.Figure 5shows the pore size distribution of sample C,as an example.As shown,the distribution consisted mainly of three peaks.The pores 4100m m in Fig.5probably correspond to crevices be-tween the fiber bundles and/or sheets.The pores of micrometer size are probably gaps in the fiber bundles,and those measuring o 1m m correspond to the pores among the agglomerated par-ticulates and/or whiskers.Table I lists the properties of the samples.The average pore size and porosity of each sample were B 50m m and 50vol%,respectively.The sample permeabilities were 2Â10À12m 2toFig.3.SEM micrographs of typical cross sections and filtration layer surfaces in each sample.105− 5− 10− 15− 45− 35− 25− 15− 55()PLog − 40− 30− 20− 100()i P L o g Fig.4.Equilibrium partial pressures of gaseous species and stable re-gions of condensed phases under a constant F (g)partial pressure of1.3Â10À8Pa,which is calculated from the decomposition reaction of AlF 3(AlF 3-Al (1)13F (g)),as a function of O 2partial pressures at 1473K for the Al–O–F–Si system.All pressures in the figure are inpascals.10100.40.30.20.10.30.20.10C u m u l a t i v e i n t r u s i o n , m l /gL o g d i f f e r e n t i a l i n t r u s i o n , m l /gDiameter, µm0Fig.5.Pore size distribution in sample C.48Journal of the American Ceramic Society—Kitaoka et al.Vol.88,No.13Â10À12m 2.Therefore,the permeation behavior of the samples was almost the same as for the clean gas without dust.During the filtration performance tests,because the gas flow in the inward direction through the samples caused the dust to accumulate on the outer surfaces of the samples,the differential pressure across the samples gradually increased.When the dust cake on the outer surface was removed by pulse blowing,the differential pressure drastically decreased.Figure 6shows pro-files of the differential pressures for the samples during the tests.The horizontal time axis in Fig.6is too long to show the pres-sure loss signal during one filtration-blowing cycle of 150s.However,the time dependence of the maximum differential pressure is shown.To illustrate clearly the pressure loss signals after 6h,where the maximum differential pressures for the sam-ples were at steady state,magnified portions of each profile are shown in their sets.If the pressure difference suddenly increased and quickly reached a constant value,the channels in the filtra-tion layer immediately filled with the dust,blocking the flow of gas into the filter.On the other hand,if the pressure difference increased during filtration,the dust cake apparently grew with-out blocking the filtration layer.The maximum differential pressure of sample A (Fig.6(a))increased with time for B 1h,after which it remained constantat 3kPa.The pressure difference under steady-state conditions suddenly reached a constant value immediately after the dust-containing air had been sucked inward through the sample and remained constant until the air was blown outward.This result suggests that a large amount of residual ash particles had be-come trapped in the filtration layer when the air was blown outwards,thereby blocking the gas flow in the inward direction.The particles captured on the filtration layer surface probably could easily enter the pores formed among the agglomerated particles,so that they became completely surrounded by the pore walls.As shown in Fig 6(b),although the time dependence of the maximum differential pressure of sample B was similar to that of sample A when the thickness of the filtration layer was increased,the maximum differential pressure of sample B was larger than that of sample A.However,for sample C (Fig.6(c)),adhesion of the whiskers to the outer surface dramatically de-creased the maximum differential pressure to about one-third lower than those of the other two samples,and this value re-mained constant during the test.The pressure difference at steady state gradually increased as the dust-containing air was breathed,but it did not become saturated.This finding suggests that dust accumulated on the whisker surfaces could be removed easily by pulse cleaning.Table II lists the filtration properties of the samples.The fil-tration efficiencies of the samples were all very high and very similar.The amount of captured dust for the samples differed by 0.58mg/m 3,a small but significant amount.For sample A,the dust concentration in the clean gas after filtration was 1.37Â10À3g/m 3,and the filtration efficiency was 99.973%.The filtration efficiency of sample B was slightly better than that of sample A,but the increased filtration-layer thickness led to an increase in the maximum differential pressure during the filtration tests.The filtration efficiency of sample C was the highest among all the samples (Table II).The low maximum differential pressure and high filtration efficiency were achieved simultaneously by the addition of the whisker layer for the following reasons:First,the large po-rosity and small pore diameters of the whisker layer maintained high permeability.Second,because the whiskers were perpen-dicular to the outer surface,the captured dust was in point con-tact with the edges of the whiskers and,thus,easy to remove during pulse cleaning.The whiskers rarely broke away or failed during the tests,retaining the same morphology as before the testing.IV.ConclusionsThe filtration performance of alumina-fiber-reinforced mullite composite filters with respect to fly ash was evaluated at 673K.For filters with mullite particulate filtration layers,although the filtration efficiency was improved slightly by increasing the fil-tration layer thickness,the maximum differential pressure dur-ing the filtration tests also increased.Adhesion of whiskers to the top surface,with a thickness of only one grain,dramatically decreased the maximum differential pressure over that achieved with particulate filtration.A constant maximum differential pres-sure was maintained throughout the testing.The filtration effi-ciency of the whisker-containing filter was the highest of all three filters.The whiskers remained intact and did not fail during the tests.Table I.Properties of SamplesSample A BCOpen pore size (m m)525454Open porosity (vol%)544951Permeability (m 2) 2.92Â10À12 2.35Â10À12 2.75Â10À12280D i f f e r e n t i a l p r e s s u r e , k P aTime, h642806428D i f f e r e n t i a l p r e s s u r e , k P aD i f f e r e n t i a l p r e s s u r e , k P aFig.6.Characteristic pressure loss signals for (a)sample A,(b)sample B,and (c)sample C.Table II.Filtration Performance of SamplesSampleABCDust concentration after filtration (g/m 3)1.37Â10À3 1.18Â10À30.784Â10À3Filtration efficiency (%)99.97399.97699.984January 2005Filtration Performance of Mullite Composites for Hot-Gas Cleaning49References1P.Hancock,‘‘Cracking,Spalling and Failure of Ash Deposits in FiltrationSystems’’;pp.3–13in High Temperature Gas Cleaning,Vol.2.Edited by A Dittler, G Hemmer,and G Kasper.,G.Braunems GmbH,Karlsruhe,Germany,1999. 2E.F.Osborn and A.Muan.‘‘Phase Equilibrium Diagrams of Oxide Systems,’’Plate5,The American Ceramic Society and the Edward Orton,Jr.,Ceramic Foundation,1960.3S.Kitaoka,N.Kawashima,H.Muto,H.Suzuki,A.Yamaguchi,Y.Ta-dakuma,and S.Ikeda,‘‘Fabrication of a Continuous Alumina Fiber-Reinforced Mullite Composite Filter,’’Cer.Eng.Sci.Proc.,22[4]279–84(2001).4M. A.Alvin,T. E.Lippert,and J. ne,‘‘Assessment of PorousCeramic Materials for Hot Gas Filtration Applications,’’Ceram.Bull.,70[9] 1491–8(1991).5M.Davidson,X.Guan,H.Hendrix,and B.Shirley,‘‘Power Systems Devel-opment Facility:Filter Element Evaluation During Combustion Testing’’;pp.363–74in High Temperature Gas Cleaning,Vol.2.Edited by A Dittler,G Hemmer,and G Kasper.,G.Braunems GmbH,Karlsruhe,Germany,1999.6N.S.Jacobson,J.L.Smialek,and D.S.Fox,‘‘Molten Salt Corrosion of Ceramics’’;pp.205–22in Corrosion of Advanced Ceramics:Measurement and Modeling,Edited by K.G.Nickel.Kluwer Academic Publishers,Dordrecht, 1994.7H.Suzuki,M.Shimizu,H.Kamiya,M.Takahashi,and T.Ota,‘‘Preparation of Fine Mullite Powders with High Surface Area by Agglomeration Control of Alkoxide-Derived Precursor Sol,’’J.Soc.Powder Technol.,Japan,34,170–75 (1997).8K.Okada and N.Otsuka,‘‘Synthesis of Mullite Whiskers by Vapour-Phase Reaction,’’J.Mater.Sci.Let.,8,1052–4(1989).9K.Okada and N.Otsuka,‘‘Synthesis of Mullite Whiskers and Their Applica-tion in Composites,’’J.Am.Ceram.Soc.,74,2414–8(1991).10D.-I.P.G TM ng,‘‘Test Method for Cleanable Filters under Laboratory and Operational Conditions,’’Adv.Filtration Separation Technol.,13A,104–12(1999).&50Journal of the American Ceramic Society—Kitaoka et al.Vol.88,No.1。