ournal . An. Cera.soc,8292321-3l(1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors Monazite Coatings on Nextel 720TM Emmanuel Boakye, ,t Randall S. Hay, , and M. Dennis Petryt nC, Dayton, Ohio 45432; Air Force Research Laboratory, Materials Directorate, Wright-Patterson Air Force Base, Ohio 45433 Seven different aqueous or ethanolic precursors were used fiber-matrix interface. 28, 29 Ideally, fiber coatings should be to continuously coat monazite (LapO4) on Nextel 720TM smooth, the same thickness, and the correct composition. The fiber tows. Immiscible liquid displacement was used to coatings also should not degrade filament tensile strength. Un- minimize bridging of coating between filaments. Precursor fortunately, there is little information about the degree to which viscosities, densities, and concentrations were measured these ideal coating qualities must be approached to get accept and solids were characterized by DTA/TGA and X-ray able CMC mechanical performance Coatings were cured in-line at 1200%-14000C and charac. Current cmc fiber-matrix interfaces are either carbon 30,3 terized for thickness, microstructure and composition by or BN. 32,33 However, oxidation is a major limitation to car- optical microscopy, SEM, and TEM. Tensile strengths of bon 4 and BN3,35 interfaces, particularly in the presence o he coated fibers varied with the precursor used and were water, 35, 36 and"pest "oxidation is a severe problem at inter- ing thicknesses were typically -50-100 nm for precur- high tempera Po), that.as 25% to >50% lower than those of the as-received fiber The mediate temperatures. 7, 38 The e problems motivate research oating stoichiometry and coating thickness of a particular on oxide Cmcs with carbon and Bn replacements, such as Monazite is stable with many common oxides, and some nesses predicted by theory for 12 mm diameter monofila evidence suggests monazite-alumina interphase boundaries de- ments. There were significant thickness variations from flect cracks. 9- Similar results were found for YPO,(Xeno- filament to filament, but usually little variation in ce time)yttrium aluminum garnet interphase boundaries. Mona- sition or microstructure. Amorphous AlPOa layers forme zite has a coefficient of thermal expansion(CTE)of 9.6x from phosphate- rich precursors. Factors that could affect 106C, modulus of 130 GPa, 0 and density of 5.08 g/cm coating characteristics and tensile strength reduction were Hydrated LapO4 converts to monazite at -500oC and is called discusse rhabdophane. Nonstoichiometric monazite precursors can react with the fiber or matrix, and carbothermal reduction of phos . Introduction phate in monazite is also possible. 43 Precise control of mona- zite stoichiometry during fiber coating is, therefore, desirable nd some work has been done on cloths. ,/on thick. 3,46Ss 0 eled s g of plates has been extensively investigated and mod- 10 Dip coating of fiber monofilaments is also a com racking can be caused by shrinkage from rapid Coating fiber tows is also of interest. Most ceramic fiber sht loss, capillary forces, and sintering at high heating process I .8,48 or by residual stress from thermal expansion mis tows have 200-500 filaments that are 10-15 um in diameter match- Continuous fiber coating generally uses high heating each. 8-20 It is difficult to coat individual filaments uniformly rates, therefore, cracking from weight loss and shrinkage is of in a fiber tow, because coating cements the filaments to- particular concern. Micrometer-thick coatings require multiple gether, 21 and a thick crust often forms around the tow perim layers and can still crack if thermal expansion mismatch is eter. 22,23 Displacement of some coating liquid by an immis- large. Currently, there is no consensus for optimal coating cible liquid reduces, but does not eliminate, cementation and thickness; it can vary widely from one fiber-coating-matrix crusting 22-24 Immiscible liquid displacement is a common system to another. rocess in oil recovery, chemical engineering, and hydrol- Fiber coatings from seven different monazite liquid-phase precursors on Nextel 720TM(3M, St. Paul, MN) fibers are ogy. 8, 25 Simple geometries have been studied, but the process described. Tows of Nextel 720 consist of 420 alumina-mullite is still poorly understood, and there is little data on fiber tows filaments 12 um in diameter with a CtE of 6x 10 bAnd One application for fiber tow coating is ceramic fiber-matrix an elastic modulus of 260 GPa. 20 Overall goals are to(1)coat composites(CMCs). 26, 27 CMCs are flaw-tolerant, if the fiber coating promotes crack deflection and fiber pull out at the fibers to evaluate the monazite interface concept,(2)develop a rotocol to screen precursors for fiber coating and the criteria for evaluating coating quality, and (3)collect data to under- stand continuous tow coating by liquids. This paper is con- cerned with the last two points. The work is part of a general K. T. Faber--contributing editor effort to identify fiber coatings and coating methods for CMcs The methods and apparatus used for coating are described else- where, 122-24 1 including a demonstration with ethanolic pre lo. 190312 Received April 22, 1998; approved January 18, 1999 cursors e build off previous work on precipitation of rials Directorate, Wright monazite powder from aqueous and ethanolic solutions, Base under Contract No general methods for homogenous erican Ceramic Society rare-earth orthophosphate sols 53,55 The relationships between erson Air Force Base. OH. recursor and coating characteristics are reported and discussed 2321
Continuous Coating of Oxide Fiber Tows Using Liquid Precursors: Monazite Coatings on Nextel 720™ Emmanuel Boakye,*,† Randall S. Hay,*,‡ and M. Dennis Petry† UES, Inc., Dayton, Ohio 45432; and Air Force Research Laboratory, Materials Directorate, Wright-Patterson Air Force Base, Ohio 45433 Seven different aqueous or ethanolic precursors were used to continuously coat monazite (LaPO4) on Nextel 720™ fiber tows. Immiscible liquid displacement was used to minimize bridging of coating between filaments. Precursor viscosities, densities, and concentrations were measured, and solids were characterized by DTA/TGA and X-ray. Coatings were cured in-line at 1200°–1400°C and characterized for thickness, microstructure, and composition by optical microscopy, SEM, and TEM. Tensile strengths of the coated fibers varied with the precursor used and were 25% to >50% lower than those of the as-received fiber. The coating stoichiometry and coating thickness of a particular precursor did not correlate with tensile strength. Median coating thicknesses were typically ∼50–100 nm for precursors with 40–80 g/L monazite, much larger than thicknesses predicted by theory for 12 mm diameter monofilaments. There were significant thickness variations from filament to filament, but usually little variation in composition or microstructure. Amorphous AlPO4 layers formed from phosphate-rich precursors. Factors that could affect coating characteristics and tensile strength reduction were discussed. I. Introduction SMOOTH, uniform ceramic coatings can be made from liquidphase precursors by dip1,2 and spin coating3,4 on plates. Dip coating of plates has been extensively investigated and modeled.5–10 Dip coating of fiber monofilaments is also a common process,11–15 and some work has been done on cloths.16,17 Coating fiber tows is also of interest. Most ceramic fiber tows have 200–500 filaments that are 10–15 mm in diameter each.18–20 It is difficult to coat individual filaments uniformly in a fiber tow, because coating cements the filaments together,21 and a thick crust often forms around the tow perimeter.22,23 Displacement of some coating liquid by an immiscible liquid reduces, but does not eliminate, cementation and crusting.22–24 Immiscible liquid displacement is a common process in oil recovery, chemical engineering, and hydrology.8,25 Simple geometries have been studied, but the process is still poorly understood, and there is little data on fiber tows. One application for fiber tow coating is ceramic fiber–matrix composites (CMCs).26,27 CMCs are flaw-tolerant, if the fiber coating promotes crack deflection and fiber pull out at the fiber–matrix interface.28,29 Ideally, fiber coatings should be smooth, the same thickness, and the correct composition. The coatings also should not degrade filament tensile strength. Unfortunately, there is little information about the degree to which these ideal coating qualities must be approached to get acceptable CMC mechanical performance. Current CMC fiber–matrix interfaces are either carbon30,31 or BN.32,33 However, oxidation is a major limitation to carbon34 and BN33,35 interfaces, particularly in the presence of water,35,36 and “pest” oxidation is a severe problem at intermediate temperatures.37,38 These problems motivate research on oxide CMCs with carbon and BN replacements, such as monazite (LaPO4), that are stable with the matrix and fiber at high temperatures. Monazite is stable with many common oxides, and some evidence suggests monazite–alumina interphase boundaries deflect cracks.39–41 Similar results were found for YPO4 (Xenotime)/yttrium aluminum garnet interphase boundaries.42 Monazite has a coefficient of thermal expansion (CTE) of 9.6 × 10−6 °C−1 , modulus of 130 GPa,40 and density of 5.08 g/cm3 . Hydrated LaPO4 converts to monazite at ∼500°C and is called rhabdophane. Nonstoichiometric monazite precursors can react with the fiber or matrix, and carbothermal reduction of phosphate in monazite is also possible.43 Precise control of monazite stoichiometry during fiber coating is, therefore, desirable. The stoichiometry of oxides, such as LaPO4, can be easily controlled with sols8,44 or solutions,4,45 but coatings from such precursors usually crack if they are over ∼0.1–0.4 mm thick.2,46–48 Cracking can be caused by shrinkage from rapid weight loss, capillary forces, and sintering at high heating rates,2,8,48 or by residual stress from thermal expansion mismatch.49 Continuous fiber coating generally uses high heating rates; therefore, cracking from weight loss and shrinkage is of particular concern. Micrometer-thick coatings require multiple layers and can still crack if thermal expansion mismatch is large. Currently, there is no consensus for optimal coating thickness; it can vary widely from one fiber–coating–matrix system to another.26,50 Fiber coatings from seven different monazite liquid-phase precursors on Nextel 720™ (3M, St. Paul, MN) fibers are described. Tows of Nextel 720 consist of 420 alumina–mullite filaments 12 mm in diameter with a CTE of 6 × 10−6 °C−1 and an elastic modulus of 260 GPa.20 Overall goals are to (1) coat fibers to evaluate the monazite interface concept, (2) develop a protocol to screen precursors for fiber coating and the criteria for evaluating coating quality, and (3) collect data to understand continuous tow coating by liquids. This paper is concerned with the last two points. The work is part of a general effort to identify fiber coatings and coating methods for CMCs. The methods and apparatus used for coating are described elsewhere,11,22–24,51 including a demonstration with ethanolic precursors.52 We build off previous work on precipitation of monazite powder from aqueous and ethanolic solutions,53–58 general methods for homogenous precipitation,45 and work on rare-earth orthophosphate sols.53,55 The relationships between precursor and coating characteristics are reported and discussed K. T. Faber—contributing editor Manuscript No. 190312. Received April 22, 1998; approved January 18, 1999. Supported by Air Force Research Laboratory Materials Directorate, WrightPatterson Air Force Base under Contract No. F33615-91-C-5663. *Member, American Ceramic Society. † UES, Inc., Dayton, OH. ‡ Wright-Patterson Air Force Base, OH. J. Am. Ceram. Soc., 82 [9] 2321–31 (1999) Journal 2321
urnal of the American Ceramic Sociery-Boakye et al. Vol. 82. No 9 with use of existing coating models for flat plates and mono- Take Up Spool filaments. The strength and Weibull moduli of coated ceramie fibers are also very important for CMCs; 9,60 therefore, tensile trengths of coated filaments are also measured and discussed More detailed work on tensile strength appears in a subsequent IL. Experimental Procedures () Materials 3 The following chemicals were used: trimethyl phosphate, drated lanthanum nitrate, and diammonium hydrogen phos- Furnace hate(Aldrich Chemical Co., Milwaukee, WI), nitric acid ( Mallinckrodt Co., Phillipsburg, NJ); 1-octanol and hexadec- ane(Matheson, Coleman, and Bell, Houston, TX); phosphoric acid(Fisher Scientific Co., Pittsburgh, PA); Darvan 821-A (R. T. Vanderbilt Co., Norwalk, CT); and Duramax B-1043 (Rohm and Haas Co., Montgomeryville, PA). Water was pu- rified by deionization of distilled water with nanopure ultra- pure system(Model D4744, Barnstead/Thermolyne Corp Surge Tank Dubuque, IA)for all experiments 过 Gas iniet (2) Precursor Characterization Viscosities were measured in a programmable rheomete nmiscihl Coater (Model DV-Ill, Brookfield Engineering Labs, Stoughton, MA) at a shear rate of 1/300(unless noted otherwise ) Densities were measured with a 25 mL bottle. Differential thermal analy- sis(DTA)and thermogravimetric analysis(TGa)were done (Model STA-409, Netzsch, Exton, PA)at 10@C/min Powder samples for DTA and TGA analysis were formed by evapora- Fiber tow diffractometer(Mode/ rory diffraction(XRD)was done in a tion at 140C for 18h. x-r Unless stated otherwise, powder for XRD was heat-treated at 1200C for I h. Monazite(LaPO4): La3 PO, and monazite ( LaPO4): LaPs Oo ratios were estimated from the 20= 28.6 diffraction of LapO4, the 20= 29.2 diffraction of La PO, Recireulation and the 20 23.9 diffraction of LaP3 Oo. 6 Qualitative esti mates of monazite crystallite size were calculated using the Scherer formula. 62 The effect of temperature on crystallinity Fig. 1. Schematic diagram of fiber coater. was studied for precursors No. I and No. 2 after heat treatments at La: PO, ratios from 1: 1 to 1: 1. 8 were measured for precursor spooled; therefore, temperatures s1200.C were used for these precursors. Multiple coatings were applied with precursor No Precursor concentrations referred to monazite yield per unit 2. The different coating liquids and coating conditions are volume liquid and were determined by weighing the residue listed in Table I after evaporation of a fixed precursor volume in a tared alu- (4) Coating Characterization mina crucible followed by heat treatment at 1400oC for I h in Fiber coatings were characterized for uniformity, thickness, treatment at 1400 C was omitted to preserve the carbon with- with energy-dispersive spectroscopy(EDS)(Model 360FE, rhabdophane Tokyo, Japan). TEM specimens of coated fiber cross sections ( Fiber Coating vere prepared by a method described elsewhere. b3 Phase pres- A vertical coater, described elsewhere, was used(Fig ence was evaluated from EDS in scanning TEM(STEM)mode 1). 11 22-24, 51 To minimize filament bridging, an immiscible hy (40 nm spot size)and tron diffraction. Typical lly coating drocarbon was floated on the precursor to help displace excess thickness, grain size, and percentage of coverage were mea- precursor 22-24 The hydrocarbon layer was typically 2 cm sured from 20 to 50 filaments. The number of coating bridges thick. I-octanol was used for aqueous precursors, and hexa- per filament and prevalence of crust were qualitatively esti- decane was used for ethanolic precursors. 2 A nonionic surfac mated from at least 100 filaments. Tem was used for thick- tant(Triton X-100, Lab Chem, Inc, Pittsburgh, PA)was added nesses less than-05 um; SEM was used for larger thicknesses to some aqueous precursors to improve filament wetting and Filament tensile strengths were measured with 75 tests using 7)and pure monazite(No. 1-No. 5)coatings were heat-treated average l, auge length. Failure stress was calculated from nt diameter. Details of the method are discussed in-line in argon and air, respectively, at-11000-1400oC at 0.7-1.4 cm/s. The furnace hot zone was -8 cm in length, and total furnace length was 30 cm. The fibers and coatings were, lIL. Results and Discussion therefore, heated very rapidly and held at near-maximum tem perature for 10 s at most. The furnace core was alumina with a ()General platinum rhodium 60: 40 wire wound on the outside. for som TEM and SEM micrographs of coatings from different liquid aqueous precursors(No. 1, No. 5), temperatures >1200C de precursors are shown in Figs. 2-8. Bridging of the coating graded fiber strength so severely that the fibers could not be between filaments was observed for all precursors(Figs. 2-8)
with use of existing coating models for flat plates and monofilaments. The strength and Weibull moduli of coated ceramic fibers are also very important for CMCs;59,60 therefore, tensile strengths of coated filaments are also measured and discussed. More detailed work on tensile strength appears in a subsequent paper. II. Experimental Procedures (1) Materials The following chemicals were used: trimethyl phosphate, hydrated lanthanum nitrate, and diammonium hydrogen phosphate (Aldrich Chemical Co., Milwaukee, WI); nitric acid (Mallinckrodt Co., Phillipsburg, NJ); 1-octanol and hexadecane (Matheson, Coleman, and Bell, Houston, TX); phosphoric acid (Fisher Scientific Co., Pittsburgh, PA); Darvan 821-A (R. T. Vanderbilt Co., Norwalk, CT); and Duramax B-1043 (Rohm and Haas Co., Montgomeryville, PA). Water was purified by deionization of distilled water with nanopure ultrapure system (Model D4744, Barnstead/Thermolyne Corp., Dubuque, IA) for all experiments. (2) Precursor Characterization Viscosities were measured in a programmable rheometer (Model DV-III, Brookfield Engineering Labs, Stoughton, MA) at a shear rate of 1/300 (unless noted otherwise). Densities were measured with a 25 mL bottle. Differential thermal analysis (DTA) and thermogravimetric analysis (TGA) were done (Model STA-409, Netzsch, Exton, PA) at 10°C/min. Powder samples for DTA and TGA analysis were formed by evaporation at 140°C for 18 h. X-ray diffraction (XRD) was done in a diffractometer (Model Rotaflex, Rigaku Co., Tokyo, Japan). Unless stated otherwise, powder for XRD was heat-treated at 1200°C for 1 h. Monazite (LaPO4):La3PO7 and monazite (LaPO4):LaP3O9 ratios were estimated from the 2u 4 28.6° diffraction of LaPO4, the 2u 4 29.2° diffraction of La3PO7, and the 2u 4 23.9° diffraction of LaP3O9. 61 Qualitative estimates of monazite crystallite size were calculated using the Scherer formula.62 The effect of temperature on crystallinity was studied for precursors No. 1 and No. 2 after heat treatments of 140°, 800°, 900°, 1000°, and 1200°C for 1 h. Phase presence at La:PO4 ratios from 1:1 to 1:1.8 were measured for precursor No. 1. Precursor concentrations referred to monazite yield per unit volume liquid and were determined by weighing the residue after evaporation of a fixed precursor volume in a tared alumina crucible followed by heat treatment at 1400°C for 1 h in air. For precursors with organic binders (No. 6 and No. 7), heat treatment at 1400°C was omitted to preserve the carbon without reducing the phosphate; therefore, LaPO4 was present as rhabdophane. (3) Fiber Coating A vertical coater, described elsewhere, was used (Fig. 1).11,22–24,51 To minimize filament bridging, an immiscible hydrocarbon was floated on the precursor to help displace excess precursor.22–24 The hydrocarbon layer was typically 2 cm thick. 1-octanol was used for aqueous precursors, and hexadecane was used for ethanolic precursors.52 A nonionic surfactant (Triton X-100, Lab Chem, Inc., Pittsburgh, PA) was added to some aqueous precursors to improve filament wetting and enhance thin-film formation. The monazite–carbon (No. 6, No. 7) and pure monazite (No. 1–No. 5) coatings were heat-treated in-line in argon and air, respectively, at ∼1100°–1400°C at 0.7–1.4 cm/s. The furnace hot zone was ∼8 cm in length, and total furnace length was 30 cm. The fibers and coatings were, therefore, heated very rapidly and held at near-maximum temperature for 10 s at most. The furnace core was alumina with a platinum:rhodium 60:40 wire wound on the outside. For some aqueous precursors (No. 1, No. 5), temperatures >1200°C degraded fiber strength so severely that the fibers could not be spooled; therefore, temperatures #1200°C were used for these precursors. Multiple coatings were applied with precursor No. 2. The different coating liquids and coating conditions are listed in Table I. (4) Coating Characterization Fiber coatings were characterized for uniformity, thickness, microstructure, and composition by optical microscopy, SEM with energy-dispersive spectroscopy (EDS) (Model 360FE, Leica, Inc., Thorwood, NY) and TEM (Model 2000 FX, JEOL, Tokyo, Japan). TEM specimens of coated fiber cross sections were prepared by a method described elsewhere.63 Phase presence was evaluated from EDS in scanning TEM (STEM) mode (40 nm spot size) and electron diffraction. Typically coating thickness, grain size, and percentage of coverage were measured from 20 to 50 filaments. The number of coating bridges per filament and prevalence of crust were qualitatively estimated from at least 100 filaments. TEM was used for thicknesses less than ∼0.5 mm; SEM was used for larger thicknesses. Filament tensile strengths were measured with 75 tests using a 2.54 cm gauge length. Failure stress was calculated from average filament diameter. Details of the method are discussed elsewhere.64 III. Results and Discussion (1) General TEM and SEM micrographs of coatings from different liquid precursors are shown in Figs. 2–8. Bridging of the coating between filaments was observed for all precursors (Figs. 2–8). Fig. 1. Schematic diagram of fiber coater. 2322 Journal of the American Ceramic Society—Boakye et al. Vol. 82, No. 9
September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 2323 Table I. Coating Liquids and Conditions Precursor tmosphere coatings ILa-IP 1.020.011286 ILa -IP Air 0.0303.81 ILa-l 1.060.0613.91 1.060.0644.100.4428 1300 0.075 50 147 404.971.040.11414400.6 ILa-IP 404971.04 ILa -IP 4.9 0.84 8338911m12 La-rich 0.0875.500 ILa -IP 8044.101.0 1300 P-rich 3.1 0.147 P-rich 1.070.378 Crust and normal coating not distinguishable. fLow tensile strength-fiber could not be spooled During fiber spooling and handling, the bridges often broke ortional to the solution concentration(C) for C between 20 halfway between the filaments and became axial fins. Coating and 80 g/L by a constant(K)of -0.83 x 10-6(m/kg).As thickness tapered off away from fins. Attempts to quantify the discussed later, this is the expected result if precursor viscosity number of bridges per filament were unsuccessful, because m) does not change significantly with C. There was some compaction of filaments for SEM and TEM observation often bridging of coating between filaments(Fig. 2(b). Some coat crushed fins. However, some qualitative observations of bridge gs had large bubbles, possibly from gas evolved during dry bundance were made for different precursors. Despite the high ng(Figs. 2(d), (e), and(g). Relatively little coating crust was Coating crust was often present around the tow perimeter 40 nm (Table I). Monazite grain size averaged -25 nm for Crust was distinguishable from normal coating by a thickness coatings done at 1200@C Grain size was slightly smaller(17 10 times higher. Thickness histograms tended to best fit a nm)in thinner coatings and slightly larger(33 nm)for coating nal distribution(Fig 9)and, with crusting, tended to be made at 1300 C. Lower concentration solutions made bimodal. Crusts were broken during sample preparation but smoother coatin were still visible as thick areas. Thickness and abundance of In coatings deliberately made phosphorus-rich, there were crust could be estimated from the histograms, if the larger amorphous AlPOA layers of variable thickness between the thickness distribution was assigned to crust and the smaller to monazite and the fiber(Fig. 2(f)). Othe rwIse monazite was normal coating. The bimodal distributions were separated into only phase observed. Occasionally ly a thin AlPOA layer was best fit sum of two unimodal distributions and median values found in coatings made from 1: 1 solutions. Lanthanum-rich of the separate distributions were used for thicknesses of hases were not observed by tEM crusted and normal coating Coated fiber tensile strength (o)was comparatively low (2) Precursor No. I (Table I). Fibers coated at 1300 C could not be spooled when stoichiometric solutions were used. However, phosphate-rich (A) Sol/Solution Characterization: Lanthanum nitrate coatings could be spooled; tensile strength was 0.99 GPa for a and trimethyl phosphate were dissolved in 20% nitric acid 4s Solutions with La: pO ratios of 1: 1. 1: 1.12. 1: 1.20.1- 1.60 La: PO4 =1: 1.59 solution and 1. 12 GPa for a La PO4 =1: 1.82 solution. At 1200oC, all coatings could be spooled, and tensile 1:1.80,and 1:3 were made. The viscosity and density of the l: I strengths ranged from 0.94 for a 20 g/L solution to 0.97 GPa g/cmy, respectively, as concentration changed from 20 to 80 there was no dependence of tensile strength on 8 from 1. 1 to 1.22 mPas and 1.06 to 1.07 g/cm, respectively, (3) Precursor No. 2 the La: POa ratio changed from 1: 1 to 1: 3 (A) Sol/Solution Characterization: Colloidal rhabdo A 5%TGA weight loss between 100 to 180C was consis- phane(LaPOaxH2O) particles were formed in water from lan- tent with loss of -I mol of water/ (mol of monazite)(Fig. 10). thanum nitrate and diammonium hydrogen phosphate If this behavior was similar to that for other mixed nitrate phosphates, the product could have been lanthanum methyl La(NO3)3+(NH,)2HPO4 - hosphate, (Lao(H(CH3)( PO4))lH20). 45,65 A DTA exotherm LaPO4 xH,O+ 2NH NO3+ HNO (1) gested that this intermediate converted to monazite above The sols were peptized with 70% nitric acid to OM-0.48M. Sol 540°C(Fg.10) m had a strong dependence on acid concentration, with a maxi- After heat treatment at 140C, the powder was X-ray amor- mum of 6. 74 mPas at 0. 14/, 5.00 mPas at 0.07M, 3.31 mPas phous. After heat treatment at 800C, XRD was consistent at 0.0M, and 1.21 mPas at 0. 48M for 40 g/L sol concentrations ith pure monazite for the 1: I and 1: 1. 12 solutions. Mixtures Sol density(p)was 1.04 g/cm of monazite and lap,Oo formed from the 1: 1. 2 to 1- 1. 8 solu- A 3% TGA weight loss between 140%.C corresponded tions(Fig. 11). Crystallite size for the 1: I solutions was -32 to a loss of -0.55 mol of water/(mol of monazite )and was nm after 800%C heat treatment and -38 nm after 1200.C heat consistent with dehydration of rhabdophane(ApOA O), treatment where x=0.5(Fig. 10).54 58 A further 25% TGA weight loss (B) Fiber Coatings: Nearly f the filament sur- and sharp endotherm between -2600-380oC was attributed to faces were coated when 40 and L concentrations were decomposition of ammonium nitrate and nitric acid sed (Fig. 2). Coatings done g/L solution covered The powder was a mixture of monazite and nitrammite 95% of the surface. Med thickness(8)was pro- after heat treatment at 140.C and pure monazite after
During fiber spooling and handling, the bridges often broke halfway between the filaments and became axial fins. Coating thickness tapered off away from fins. Attempts to quantify the number of bridges per filament were unsuccessful, because compaction of filaments for SEM and TEM observation often crushed fins. However, some qualitative observations of bridge abundance were made for different precursors. Despite the high heating rates, thin coatings were not cracked. However, when the coating was crusted and >1 mm thick, it often cracked off. Coating crust was often present around the tow perimeter. Crust was distinguishable from normal coating by a thickness ∼10 times higher. Thickness histograms tended to best fit a lognormal distribution (Fig. 9) and, with crusting, tended to be bimodal. Crusts were broken during sample preparation but were still visible as thick areas. Thickness and abundance of crust could be estimated from the histograms, if the larger thickness distribution was assigned to crust and the smaller to normal coating. The bimodal distributions were separated into a best fit sum of two unimodal distributions, and median values of the separate distributions were used for thicknesses of crusted and normal coating. (2) Precursor No. 1 (A) Sol/Solution Characterization: Lanthanum nitrate and trimethyl phosphate were dissolved in 20% nitric acid.45,65 Solutions with La:PO4 ratios of 1:1, 1:1.12, 1:1.20, 1:1.60, 1:1.80, and 1:3 were made. The viscosity and density of the 1:1 mixture changed from 0.90 to 1.07 mPazs and 1.02 to 1.06 g/cm3 , respectively, as concentration changed from 20 to 80 g/L. The viscosity and density of an 80 g/L solution changed from 1.1 to 1.22 mPazs and 1.06 to 1.07 g/cm3 , respectively, as the La:PO4 ratio changed from 1:1 to 1:3. A 5% TGA weight loss between 100° to 180°C was consistent with loss of ∼1 mol of water/(mol of monazite) (Fig. 10). If this behavior was similar to that for other mixed nitrate phosphates, the product could have been lanthanum methyl phosphate, (LaO(H(CH3)(PO4))z1H2O).45,65 A DTA exotherm suggested that this intermediate converted to monazite above 540°C (Fig. 10).55–57 After heat treatment at 140°C, the powder was X-ray amorphous. After heat treatment at 800°C, XRD was consistent with pure monazite for the 1:1 and 1:1.12 solutions. Mixtures of monazite and LaP3O9 formed from the 1:1.2 to 1:1.8 solutions (Fig. 11).61 Crystallite size for the 1:1 solutions was ∼32 nm after 800°C heat treatment and ∼38 nm after 1200°C heat treatment. (B) Fiber Coatings: Nearly 100% of the filament surfaces were coated when 40 and 80 g/L concentrations were used (Fig. 2). Coatings done with 20 g/L solution covered ∼95% of the surface. Median coating thickness (d) was proportional to the solution concentration (C) for C between 20 and 80 g/L by a constant (K) of ∼0.83 × 10−6 (m4 /kg). As discussed later, this is the expected result if precursor viscosity (h) does not change significantly with C. There was some bridging of coating between filaments (Fig. 2(b)). Some coatings had large bubbles, possibly from gas evolved during drying (Figs. 2(d), (e), and (g)). Relatively little coating crust was present around the tow perimeter, and thickness uniformity was better than average (Fig. 9). Median crust thickness (dcrust) was 440 nm (Table I). Monazite grain size averaged ∼25 nm for coatings done at 1200°C. Grain size was slightly smaller (17 nm) in thinner coatings and slightly larger (33 nm) for coatings made at 1300°C. Lower concentration solutions made smoother coatings. In coatings deliberately made phosphorus-rich, there were amorphous AlPO4 layers of variable thickness between the monazite and the fiber (Fig. 2(f)). Otherwise monazite was the only phase observed. Occasionally a thin AlPO4 layer was found in coatings made from 1:1 solutions. Lanthanum-rich phases were not observed by TEM. Coated fiber tensile strength (s) was comparatively low (Table I). Fibers coated at 1300°C could not be spooled when stoichiometric solutions were used. However, phosphate-rich coatings could be spooled; tensile strength was 0.99 GPa for a La:PO4 4 1:1.59 solution and 1.12 GPa for a La:PO4 4 1:1.82 solution. At 1200°C, all coatings could be spooled, and tensile strengths ranged from 0.94 for a 20 g/L solution to 0.97 GPa for an 80 g/L solution. The proportionality of C with d implied there was no dependence of tensile strength on d. (3) Precursor No. 2 (A) Sol/Solution Characterization: Colloidal rhabdophane (LaPO4zxH2O) particles were formed in water from lanthanum nitrate and diammonium hydrogen phosphate: La(NO3)3 + (NH4)2HPO4 → LaPO4zxH2O + 2NH4NO3 + HNO3 (1) The sols were peptized with 70% nitric acid to 0M–0.48M. Sol h had a strong dependence on acid concentration, with a maximum of 6.74 mPazs at 0.14M, 5.00 mPazs at 0.07M, 3.31 mPazs at 0.0M, and 1.21 mPazs at 0.48M for 40 g/L sol concentrations. Sol density (r) was 1.04 g/cm3 . A 3% TGA weight loss between 140°–200°C corresponded to a loss of ∼0.55 mol of water/(mol of monazite) and was consistent with dehydration of rhabdophane (LaPO4zxH2O), where x ≈ 0.5 (Fig. 10).54,58 A further 25% TGA weight loss and sharp endotherm between ∼260°–380°C was attributed to decomposition of ammonium nitrate and nitric acid. The powder was a mixture of monazite and nitrammite53,61 after heat treatment at 140°C and pure monazite after heat Table I. Coating Liquids and Conditions Precursor Composition Atmosphere Number of coatings C (g/L) h (mPa?s) r (g/cm3 ) d (mm) h (mm) dcrust (mm) hcrust (mm) Temperature (°C) s (GPa) 1 1La − 1P Air 1 20 0.90 1.02 0.011 2.86 1200 0.94 1 1La − 1P Air 1 40 0.030 3.81 1200 0.96 1 1La − 1P Air 1 80 1.07 1.06 0.061 3.91 1200 0.97 1 1La − 1P Air 1 80 1.07 1.06 0.064 4.10 0.44 28 1300 ‡ 1 1La − 1.59P Air 1 80 1300 0.99 1 1La − 1.82P Air 1 80 1300 1.12 1 1La − 3P Air 1 80 1.22 1.07 1300 2 1La − 1P Air 1 40 4.97 1.04 0.075 9.50 0.77 98 1300 1.47 2 1La − 1P Air 2 40 4.97 1.04 0.114 14.40 0.62 79 1300 2 1La − 1P Air 5 40 4.97 1.04 0.170 21.50 0.81 102 1300 2 1La − 1P Air 8 40 4.97 1.04 1300 1.04 2 1La − 1P Air 10 40 4.97 1.04 0.280 35.50 0.80 101 1300 3 La-rich Air 1 50 1.39 0.84 0.055 5.60 0.39 40 1300 1.21 3 La-rich Argon 1 50 1.39 0.84 1300 1.24 4 1La − 1P Air 1 80 2.84 1.27 0.087 5.50 0.68 43 1300 1.07 5 1La − 1P Air 1 80 44.10 1.07 1300 ‡ 6 P-rich Argon 1 60 3.12 1.05 0.147 † † 1300 7 P-rich Argon 1 108 2.24 1.07 0.378 † † 1300 1.56 † Crust and normal coating not distinguishable. ‡Low tensile strength—fiber could not be spooled. September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 2323
2324 Journal of the American Ceramic Sociery-Boakye et al. Vol. 82. No 9 treatment at 600C(Fig. 11). Monazite crystallite size in- creased from -23 to -37 nm between 600-1000%C. but there was little change between I000°1200°C (B) Fiber Coatings. Coating coverage was nearly con- tinuous for 40-80 g/L sols with 0M-0 14M nitric acid, but sols ith 0.4M and 1.9M nitric acid had only 40%-50% coverage Addition of TX-100 surfactant did not improve coverage Monazite was the only crystalline phase in the coatings Granu- lar monazite formed at -1200oC. while vermicular monazite Sum formed at -1300C (Fig. 3). When coatings were extremel thin(30 La:P=1: 1.8.(g)SEM micrograph of coated fiber surface, 1200%C, 80 vol%. Average grain size was -200 nm, much larger than that gl of other coatings. Although X-ray analysis of the coating liquid
treatment at 600°C (Fig. 11). Monazite crystallite size increased from ∼23 to ∼37 nm between 600°–1000°C, but there was little change between 1000°–1200°C. (B) Fiber Coatings: Coating coverage was nearly continuous for 40–80 g/L sols with 0M–0.14M nitric acid, but sols with 0.4M and 1.9M nitric acid had only 40%–50% coverage. Addition of TX-100 surfactant did not improve coverage. Monazite was the only crystalline phase in the coatings. Granular monazite formed at ∼1200°C, while vermicular monazite formed at ∼1300°C (Fig. 3). When coatings were extremely thin (30 vol%. Average grain size was ∼200 nm, much larger than that of other coatings. Although X-ray analysis of the coating liquid Fig. 2. TEM micrographs of coatings from precursor No. 1: (a)–(d) 1200°C, 80 g/L; (e) 1300°C, 80 g/L, honeycombs; (f) 1300°C, 80 g/L, La:P 4 1:1.8. (g) SEM micrograph of coated fiber surface, 1200°C, 80 g/L. 2324 Journal of the American Ceramic Society—Boakye et al. Vol. 82, No. 9
200 100 monazite fiber 100nm 100m 400mm (d) 100 nm 300nm uIn (d) 100 (f) 500nm y (g) 10m 围己c则cmmp(2ma4(0(0) TEM micrographs of coatings from precursor No3 (c)I pass, (d) morphology, and cracking of coating), (f)SEM micrograph of coated coatings were done at 1300C with a 50 g/L solution fiber surface( Note vermi microstructure)
Fig. 3. TEM micrographs of coatings from precursor No. 2 made at 1300°C with a 40 g/L sol at 1.4 cm/s: (a) 1 pass, (b) 1 pass, agglomerate, (c) 1 pass, (d) 2 passes, (e) 10 passes (Note fins, irregular morphology, and cracking of coating), (f) SEM micrograph of coated fiber surface (Note vermicular microstructure). Fig. 4. (a)–(f) TEM micrographs of coatings from precursor No. 3 (Note bridges in (c)). (g) SEM micrograph of coated fiber surface. All coatings were done at 1300°C with a 50 g/L solution
232 Journal of the American Ceramic Society-Boakye et al. Vol. 82. No 9 (a 400nm nIIl 100nm S Fig. 5.(aHd)TEM micrographs of coatings from precursor No 4 (Note bridging and bubbles. )(e)SEM micrograph of coated fiber surface showing extensive bubbling of coating. All coatings were de (e) 250mm Fig. 7.(aHd) TEM m hs of coatings from precursor No, 6 (a) Vermicular coating, (b ble fin, (c)spotting, (d)agglomerat (e) SEM micrograph of fiber surface with vermicular micro- structure. All coatings were done at 1300C in argon with a 60 g/L sol. showed Lay PO, from phosphate deficiency LapO, was observed by tem in fiber coatings. Average o of coated fibers averaged 1.21 and 1.24 GPa for fibers coated at 1300%C in air and argon, respectively(Table 1) (5) Precursor No. 4 200mm (A) Sol/Solution Characterization: This is an aqueous So- lution supplied by P. E D. Morgan of the Rockwell Science Center(Thousand Oaks, CA)from phytic acid and lanthanum Fig. 6.(a)and (b) TEM micrographs of coatings from precursor No nitrate in water. At 80 g/L concentration, m was 2.84 mPas, 5. All coatings were done at 1300C with an 80 g/L solution and p was 1.27 g/cm. A large amount of gas evolved as the
showed La3PO7 from phosphate deficiency, only LaPO4 was observed by TEM in fiber coatings. Average s of coated fibers averaged 1.21 and 1.24 GPa for fibers coated at 1300°C in air and argon, respectively (Table I). (5) Precursor No. 4 (A) Sol/Solution Characterization: This is an aqueous solution supplied by P. E. D. Morgan of the Rockwell Science Center (Thousand Oaks, CA) from phytic acid and lanthanum nitrate in water. At 80 g/L concentration, h was 2.84 mPazs, and r was 1.27 g/cm3 . A large amount of gas evolved as the Fig. 6. (a) and (b) TEM micrographs of coatings from precursor No. 5. All coatings were done at 1300°C with an 80 g/L solution. Fig. 7. (a)–(d) TEM micrographs of coatings from precursor No. 6. (a) Vermicular coating, (b) possible fin, (c) spotting, (d) agglomerates, (e) SEM micrograph of coated fiber surface with vermicular microstructure. All coatings were done at 1300°C in argon with a 60 g/L sol. Fig. 5. (a)–(d) TEM micrographs of coatings from precursor No. 4. (Note bridging and bubbles.) (e) SEM micrograph of coated fiber surface showing extensive bubbling of coating. All coatings were done at 1300°C with an 80 g/L solution. 2326 Journal of the American Ceramic Society—Boakye et al. Vol. 82, No. 9
September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 200nm Histograms of coating thickness for precursors Nos. 1-7 utions are log- normal. some are bimodal. For bimodal distribu the larger thickness corresponds to crust and the smaller TGA 100 200mm DTA 500mm 800 Fig 8.(aHc) TEM micrographs of coatings from precursor No. 7. Temp (C fiber surface with vermicular microstructure. All coatings were done at Fig 10. DTA/TGA of LaPO4 from different precursors 1300C in argon with a 108 g/L sol viscous state, was probably responsible for the bubbles. Aver solution dried to a viscous liquid before precipitation. The gas age grain size was -50 nm. Average o was 1.07 GPa for fibers foamed the viscous liquid and the precipitate. Precipitate sput- coated at 1300.C (Table 1) tered out of the crucibles during heat treatment; therefore, ad (6) Precursor No. 5 curate DTA/TGA measurements were not possible (A) SolSolution Characterization: The sol was made by XRD after a 1200 C heat treatment was consistent with onazite(Fig. 11). Crystallite size was-40 nm (B) Fiber Coatings: For C =80 g/L, the median 8 of La(oH)3+ H3PO4- LaPO4 3H,O normal and crusted coatings was 87 and 680 nm, respectively a(Oh)3 powder was slowly added to a 20 wt% H3 POa solu- CTable I). The lognormal thickness histogram was bimodal tion and stirred for 10 h. The product was centrifugally sepa- ( Fig. 9). Coverage was continuous. Large bubbles similar to rated and washed with water three times. Larger particles were those found from precursor No. I were present, but these were sedimented for a short time. The resulting sol/slurry was white larger and more common(Fig. 5). These made the coating and opaque, and required agitation to keep the particles dis- morphology irregular, particularly when seen by surface SEM persed. m was 4.18 mPas at 40 g/L and 44 1 mPas(1/50 shear observation(Fig. 5(e)). The bubbles also made it very difficult rate)at 80 g/L, and p was 1.04 and 1.07 g/cm, respectively to identify bridging, because broken coating bridges and bro- DTA/TGA and XRD of the precipitate are available else- ken bubbles were similar in appearance. Both tended to break where b Crystallite size was -40 nm off when the fiber was handled, leaving fine powder behind. A (B) Fiber Coatings: The median 8 was-300 nm for C= thin(-10 nm) coating was usually still present underneath the 80 g/L(Table D), but coverage was poor, and the microstructure bubbles. Gas evolution from the precursor, while in a highly was not uniform. Four or five filaments were frequently ce
solution dried to a viscous liquid before precipitation. The gas foamed the viscous liquid and the precipitate. Precipitate sputtered out of the crucibles during heat treatment; therefore, accurate DTA/TGA measurements were not possible. XRD after a 1200°C heat treatment was consistent with pure monazite (Fig. 11). Crystallite size was ∼40 nm. (B) Fiber Coatings: For C 4 80 g/L, the median d of normal and crusted coatings was 87 and 680 nm, respectively (Table I). The lognormal thickness histogram was bimodal (Fig. 9). Coverage was continuous. Large bubbles similar to those found from precursor No. 1 were present, but these were larger and more common (Fig. 5). These made the coating morphology irregular, particularly when seen by surface SEM observation (Fig. 5(e)). The bubbles also made it very difficult to identify bridging, because broken coating bridges and broken bubbles were similar in appearance. Both tended to break off when the fiber was handled, leaving fine powder behind. A thin (∼10 nm) coating was usually still present underneath the bubbles. Gas evolution from the precursor, while in a highly viscous state, was probably responsible for the bubbles. Average grain size was ∼50 nm. Average s was 1.07 GPa for fibers coated at 1300°C (Table I). (6) Precursor No. 5 (A) Sol/Solution Characterization: The sol was made by the reaction:66 La(OH)3 + H3PO4 → LaPO4 + 3H2O (3) La(OH)3 powder was slowly added to a 20 wt% H3PO4 solution and stirred for 10 h. The product was centrifugally separated and washed with water three times. Larger particles were sedimented for a short time. The resulting sol/slurry was white and opaque, and required agitation to keep the particles dispersed. h was 4.18 mPazs at 40 g/L and 44.1 mPazs (1/50 shear rate) at 80 g/L, and r was 1.04 and 1.07 g/cm3 , respectively. DTA/TGA and XRD of the precipitate are available elsewhere.66 Crystallite size was ∼40 nm. (B) Fiber Coatings: The median d was ∼300 nm for C 4 80 g/L (Table I), but coverage was poor, and the microstructure was not uniform. Four or five filaments were frequently ceFig. 8. (a)–(c) TEM micrographs of coatings from precursor No. 7. (a) Typical rough coating (note gaps in coverage), (b) AlPO4 between coating and fiber, (c) smoother coating, (d) SEM micrograph of coated fiber surface with vermicular microstructure. All coatings were done at 1300°C in argon with a 108 g/L sol. Fig. 9. Histograms of coating thickness for precursors Nos. 1–7. Distributions are log-normal, some are bimodal. For bimodal distributions, the larger thickness corresponds to crust and the smaller to normal coating. Fig. 10. DTA/TGA of LaPO4 from different precursors. September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 2327
urnal of the American Ceramic Sociery-Boakye et al. Vol. 82. No 9 l000 crust =005 10u m=0s5 1200C LI #1(1: lumber of Coats (N) Fig. 12. Median coating s of crusted and normal coating vs number the bimodality was from crusting. Frequent bridging and coa lomeration were also responsible for thick areas 600c remore the histogram was not separated into crust and normal coatings. The median S for all coatings was 147 nm for a sol with a 60 g/L concentration. Both granular and dendritic coatings similar to those made with precursor No. 2 were found. but the thickness was less uniform. There were also areas coated with large(200 nm)spherical polycrystalline monazite agglomerates(Fig. 7(d)). Other areas were spotte 40 with 25-100 nm monazite grains, with bare or thin(10 nm) AlPOa- coated areas in between(Fi IPOa presence was consistent with the observed phosphate enrichment of the lighted peaks for No. 2. 140.C are nitrammite. Highlighted peaks for areas with spotty coating but still highly variable. Coated fiber No. 1, 1200 C, La: P= 1: 1.8, as well as for No 6 and No. 7 are tensile strength was not measured LaP3Oo. Highlighted peaks for No. 3 are La]PO7. All other peaks are (8) Precursor No. 7 consistent with monazite A Sol/Solution Characterisation: This was an aqueous ented together over a large area. Separation of the coating nium polymethacrylate(Darvan 821-A)was added to colloidal into crusted and normal areas was not possible, because so rhabdophane(Eq (4)at a 50: 50 carbon: monazite weight ratio many filaments were cemented together. Filaments were fre- At a 108 g/L concentration, n was 2.24 mPas, and p was 1.0 quently spotted with polycrystalline agglomerates of monazite g/cm3 Fig. 6). Average grain size was -60 nm. Fibers coated at 1200 and 1300%C could not be spooled, so tensile strength was A 2. 5% TGA weight loss between 140 and 200C was assumed to be very low and was not measured. This low consistent with loss of-05 mol of water/(mol of monazite)and strength might at least be partly related to the excessive ce- was also consistent with dehydration of rhabdophane(LaPO4 mentation of the tow xH,O), where x =0.5(Fig. 10)5458 Weight losses at 480% 800°,900°,andl000° C were presumably due to decomposi (7 Precursor No 6 tion and oxidation of daryan a (A) Sol/Solution Characterization: This was an XRD was consistent with monazite with slight ex rhabdophane-Duramax B-1043 sol. Colloidal rhal LaP3Oo(Fig. 11). Crystallite size was-40 nm particles formed by reaction of lanthanum nitrate and B) Fiber Coatings: The median 8 was 378 nm for a sol with 108 g/L concentration(Table 1). The thickness distribu- La(NO3)3+ H3PO4 - LaPO 3HNO (4) than for most other coatings(Fig 9). However, in many places The polyelectrolyte Duramax B-1043 was added to the rhab- the coatings were"ragged"(Fig. 8(a)), and the coating was dophane sol at a weight ratio of 32: 68. At a 57 g/Concentra- occasionally debonded but adjacent to the filament it had ob- tion, m was 3. 12 mPas with no nitric acid and 2.90 mPas with viously debonded from. The microstructural variation 0. 28M nitric acid; p was 1.04 and 1.05 g/cm, respectively with Duramax-B in precursor No 6 was absent. The DTA/TGA measurements were not successful because of were loosely cemented, slightly elongated monazite foaming. XRD was consistent with monazite with slight excess average size( Fig. 8). A thin layer (-10 nm) f Lap3Oo( Fig. 11). Crystallite size was-40 nn was also observed at some coating-filament interfaces, agai (B) Fiber Coatings: There was significant bimodality of consistent with the observed AlPOA enrichment of the precur- the thickness histograms( Fig 9). Some crusts(500-1000 sor(Fig 8(b). o was 1.56 GPa for fibers coated at 1300 C, the thick )around the tow perimeter were obvious, but only part of highest of any measured (Table D)
mented together over a large area. Separation of the coating into crusted and normal areas was not possible, because so many filaments were cemented together. Filaments were frequently spotted with polycrystalline agglomerates of monazite (Fig. 6). Average grain size was ∼60 nm. Fibers coated at 1200° and 1300°C could not be spooled, so tensile strength was assumed to be very low and was not measured. This low strength might at least be partly related to the excessive cementation of the tow. (7) Precursor No. 6 (A) Sol/Solution Characterization: This was an aqueous rhabdophane–Duramax B-1043 sol. Colloidal rhabdophane particles formed by reaction of lanthanum nitrate and phosphoric acid: La(NO3)3 + H3PO4 → LaPO4 + 3HNO3 (4) The polyelectrolyte Duramax B-1043 was added to the rhabdophane sol at a weight ratio of 32:68. At a 57 g/L concentration, h was 3.12 mPazs with no nitric acid and 2.90 mPazs with 0.28M nitric acid; r was 1.04 and 1.05 g/cm3 , respectively. DTA/TGA measurements were not successful because of foaming. XRD was consistent with monazite with slight excess of LaP3O9 (Fig. 11). Crystallite size was ∼40 nm. (B) Fiber Coatings: There was significant bimodality of the thickness histograms (Fig. 9). Some crusts (∼500–1000 nm thick) around the tow perimeter were obvious, but only part of the bimodality was from crusting. Frequent bridging and coating agglomeration were also responsible for thick areas (Fig. 7). Therefore, the histogram was not separated into crust and normal coatings. The median d for all coatings was 147 nm for a sol with a 60 g/L concentration. Both granular and dendritic coatings similar to those made with precursor No. 2 were found, but the thickness was less uniform. There were also areas coated with large (200 nm) spherical polycrystalline monazite agglomerates (Fig. 7(d)). Other areas were spotted with 25–100 nm monazite grains, with bare or thin (∼10 nm) AlPO4-coated areas in between (Fig. 7(c)). AlPO4 presence was consistent with the observed phosphate enrichment of the precursor. The monazite grain size was 28 nm outside of the areas with spotty coating but still highly variable. Coated fiber tensile strength was not measured. (8) Precursor No. 7 (A) Sol/Solution Characterization: This was an aqueous rhabdophane–Darvan 821-A sol. The polyelectrolyte ammonium polymethacrylate (Darvan 821-A) was added to colloidal rhabdophane (Eq. (4)) at a 50:50 carbon:monazite weight ratio. At a 108 g/L concentration, h was 2.24 mPazs, and r was 1.07 g/cm3 . A 2.5% TGA weight loss between 140° and 200°C was consistent with loss of ∼0.5 mol of water/(mol of monazite) and was also consistent with dehydration of rhabdophane (LaPO4z xH2O), where x ≈ 0.5 (Fig. 10).54,58 Weight losses at 480°, 800°, 900°, and 1000°C were presumably due to decomposition and oxidation of Darvan A. XRD was consistent with monazite with slight excess of LaP3O9 (Fig. 11). Crystallite size was ∼40 nm. (B) Fiber Coatings: The median d was 378 nm for a sol with 108 g/L concentration (Table I). The thickness distribution was strongly unimodal and significantly more uniform than for most other coatings (Fig. 9). However, in many places the coatings were “ragged” (Fig. 8(a)), and the coating was occasionally debonded but adjacent to the filament it had obviously debonded from. The microstructural variation present with Duramax-B in precursor No. 6 was absent. The coatings were loosely cemented, slightly elongated monazite grains of 38 nm average size (Fig. 8). A thin layer (∼10 nm) of AlPO4 was also observed at some coating–filament interfaces, again consistent with the observed AlPO4 enrichment of the precursor (Fig. 8(b)). s was 1.56 GPa for fibers coated at 1300°C, the highest of any measured (Table I). Fig. 11. XRD from LaPO4 derived from various precursors. Highlighted peaks for No. 2, 140°C are nitrammite. Highlighted peaks for No. 1, 1200°C, La:P 4 1:1.8, as well as for No. 6 and No. 7 are LaP3O9. Highlighted peaks for No. 3 are La3PO7. All other peaks are consistent with monazite. Fig. 12. Median coating d of crusted and normal coating vs number of coats. 2328 Journal of the American Ceramic Society—Boakye et al. Vol. 82, No. 9
September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 2329 (9) Coating Thickness Analysis (A General: During continuous fiber tow coating with immiscible liquid displacement, the precursor wets filaments and can then be displaced by floating immiscible liquid, if m and coating speed(U) are low enough, and the density differ- 击e·口 ence(Ap) between precursor and immiscible liquid is high untrusted enough(Fig. 1). The precursor must wet the fiber-immiscibl m=049 liquid interface; otherwise, it will all be displaced. 22 Wetting a cylindrical fiber surface is more difficult than a planar su face. 67 The precursor and immiscible liquid are then partially displaced by air (or argon) as the fiber tow leaves the immis-= cible liquid. Typically, some of the precursor spheroidize at the immiscible liquid-air interface and then drips back through the immiscible liquid. Ideally, the m is low enough so that all mrd附 but thin precursor layers wetting each filament are displaced in r=6 a few seconds, either by immiscible liquid alone or by immis- cible liquid and air. If there is little interaction between pre- cursor liquid menisci on adjacent filaments, the precursor thickness wetting each filament should be similar to that wet- ting monofilaments. The extent to which this ideal situation is approached is now examined B)Monofilaments: (a) Large Fibers: Dip coating has n/p12 been modeled for flat plates. The plate entrains a precursor layer of thickness(h) that depends on m, P, U, and surface Fig. 13. 8 of precursor required to deposit coatings Nos. 1-4 vs tension(y). For low U and n, i.e., Ca(capillary number)= precursor n; (a--)represent predictions from coating models(Eqs Un/y1 mm. If no r dependence is assumed (large fiber limit), hicknesses must be calibrated empirically particularly at high prediction by Eq(6)agrees reasonably with derust, although the oating speeds. 6, During drying and heat treatment, the pre- cursor evaporates to leave a coating with thickness of 8 0.5)(Fig. 13). This agre on n is off(0.75 instead of predicted ement is either coincidence or suggests hC/ps, where C is precursor concentration(mass/unit volume that an m increase during heat-up causes different behavior, as and ps the coating density. For fibers, this is modified for discussed elsewhere I cylindrical geometry (C) Multifilaments: Transverse impregnation of C2h+h)]2 cursor is assumed to be rapid for low m(1-5 mPas)I Therefore, the precursor meniscus height(Z)alor ertical fibers with equal spacing can be modeled as (b) Small Fibers: There are several different dip-coating models for fibers; 5, 69, 71 however, for low m, U, and r in our experiments, all models reduce to prgv3+d22-(m/2) h=1.33rC33 (8) where d is the spacing between filament surfaces. This ion is for static fibers; for moving fibers, the meniscus should (c) Data Interpretation: The h for precursors Nos. 1-4 be higher If Z exceeds the immiscible liquid height of -2 were calculated from 8 by Eq(7) for crusted and normal the immiscible liquid does not displace the precursor. a d of 7( coating. For crust, h had a m4 dependence; for normal coating, um is required for displacement by air in 2 cm, and a d of 240 this dependence was m /(Fig. 13). A m23dependence was um is required for Ap of 0. 1. The latter value requires the 420 predicted by Eq( 8), assuming similar y for precursors Nos filaments in the tow to spread out over a diameter of 5.5 1-4. The appropriate y values were not obvious. y of pure which, as discussed earlier, is too large. This suggests it is water at 100C was 0.060 J/m2. 72 y of 1-octanol saturated unlikely that the precursor is displaced in the immiscible liquid water was -0.030--0.035 J/m2 73 and the water-l-octanol Instead, both precursor and immiscible liquid are partially dis- terface energy was 0.010 J/m2. Solutes and atmosphere also placed by air and, afterward, may rearrange to minimize inter- affected y. For qualitative analysis, we assumed y was-0.030 energy J/m2 and acknowledged the difficulty in using surface tension For moving fibers, precursors that are too viscous are not io The h for precursors Nos 1-4 on Nextel 720 fiber tows was mPas, about an order of magnitude higher than the others. As displaced for kinematic reasons. Precursor No. 5 has m of 44 compared to that predicted for isolated monofilaments with Eq noted, filaments coated with this precursor are more frequently (8).h values from 50 nm(precursor No. 1)to 140 nm(precur- cemented together, probably due to lack of precursor displace- sor No. 2)were predicted for normal coating. These were ment because of high nearly 2 orders of magnitude than those of 4. 1 um for precursor menisci do not interact with each other precursor No. I and 9.5 um fo rsor No 2 calculated from and individual filaments should coat like separate monofila- 8 with Eq (7)(Fig. 13).Clea filaments did not behave ments. Conversely, if filaments are close-packed, precursor like isolated monofilaments coating. Much more pre menisci wick up the fiber tow by -2y/pr based on Eq (9). The
(9) Coating Thickness Analysis (A) General: During continuous fiber tow coating with immiscible liquid displacement, the precursor wets filaments and can then be displaced by floating immiscible liquid, if h and coating speed (U) are low enough, and the density difference (Dr) between precursor and immiscible liquid is high enough (Fig. 1). The precursor must wet the fiber–immiscible liquid interface; otherwise, it will all be displaced.22 Wetting a cylindrical fiber surface is more difficult than a planar surface.67 The precursor and immiscible liquid are then partially displaced by air (or argon) as the fiber tow leaves the immiscible liquid. Typically, some of the precursor spheroidizes at the immiscible liquid–air interface and then drips back through the immiscible liquid. Ideally, the h is low enough so that all but thin precursor layers wetting each filament are displaced in a few seconds, either by immiscible liquid alone or by immiscible liquid and air. If there is little interaction between precursor liquid menisci on adjacent filaments, the precursor thickness wetting each filament should be similar to that wetting monofilaments. The extent to which this ideal situation is approached is now examined. (B) Monofilaments: (a) Large Fibers: Dip coating has been modeled for flat plates. The plate entrains a precursor layer of thickness (h) that depends on h, r, U, and surface tension (g). For low U and h, i.e., Ca (capillary number) 4 Uh/g (g/rg) 1/2, where r is fiber radius.71 h, r, and g change as liquid evaporates. Therefore, Eqs. (5) and (6) can only be approximate; in practice, accurate film thicknesses must be calibrated empirically,1 particularly at high coating speeds.6,7 During drying and heat treatment, the precursor evaporates to leave a coating with thickness of d 4 hC/rs, where C is precursor concentration (mass/unit volume) and rs the coating density. For fibers, this is modified for cylindrical geometry: d = Fr 2 + C~2rh + h2 ! rs G 1/2 − r (7) (b) Small Fibers: There are several different dip-coating models for fibers;5,69,71 however, for low h, U, and r in our experiments, all models reduce to: h 4 1.33rCa 2/3 (8) (c) Data Interpretation: The h for precursors Nos. 1–4 were calculated from d by Eq. (7) for crusted and normal coating. For crust, h had a h3/4 dependence; for normal coating, this dependence was h1/2 (Fig. 13). A h2/3 dependence was predicted by Eq. (8), assuming similar g for precursors Nos. 1–4. The appropriate g values were not obvious. g of pure water at 100°C was 0.060 J/m2 . 72 g of 1-octanol saturated water was ∼0.030–0.035 J/m2 , 73 and the water–1-octanol interface energy was 0.010 J/m2 . Solutes and atmosphere also affected g. For qualitative analysis, we assumed g was ∼0.030 J/m2 and acknowledged the difficulty in using surface tension for a priori thickness prediction.1 The h for precursors Nos. 1–4 on Nextel 720 fiber tows was compared to that predicted for isolated monofilaments with Eq. (8). h values from 50 nm (precursor No. 1) to 140 nm (precursor No. 2) were predicted for normal coating. These were nearly 2 orders of magnitude lower than those of 4.1 mm for precursor No. 1 and 9.5 mm for precursor No. 2 calculated from d with Eq. (7) (Fig. 13). Clearly the filaments did not behave like isolated monofilaments during coating. Much more precursor was entrained on each filament in a tow than would be entrained on the same isolated filament. One interpretation of crusting is that the tow behaves like a monofilament for tow perimeter coating (crust). Close packing of 420 filaments 12 mm in diameter gives a 130 mm radius. If 9.5 mm of the precursor surrounds each filament, then r is 335 mm and Eq. (8) predicts the h of 7 mm. The observed h is 98 mm. Even if the tow spreads more than close packing would suggest, r of ∼5 mm is required to predict the observed dcrust of ∼0.4–0.8 mm (Table I). The tow is certainly not spread much >1 mm. If no r dependence is assumed (large fiber limit), prediction by Eq. (6) agrees reasonably with dcrust, although the exponential dependence on h is off (0.75 instead of predicted 0.5) (Fig. 13). This agreement is either coincidence or suggests that an h increase during heat-up causes different behavior, as discussed elsewhere.1 (C) Multifilaments: Transverse impregnation of the precursor is assumed to be rapid for low h (∼1–5 mPazs) liquids.74 Therefore, the precursor meniscus height (Z) along multiple vertical fibers with equal spacing can be modeled as:75 Z = g rrg p =3~1 + d/2r! 2 − ~p/2! (9) where d is the spacing between filament surfaces. This expression is for static fibers; for moving fibers, the meniscus should be higher. If Z exceeds the immiscible liquid height of ∼2 cm, the immiscible liquid does not displace the precursor. A d of 70 mm is required for displacement by air in 2 cm, and a d of 240 mm is required for Dr of 0.1. The latter value requires the 420 filaments in the tow to spread out over a diameter of 5.5 mm, which, as discussed earlier, is too large. This suggests it is unlikely that the precursor is displaced in the immiscible liquid. Instead, both precursor and immiscible liquid are partially displaced by air and, afterward, may rearrange to minimize interface energy. For moving fibers, precursors that are too viscous are not displaced for kinematic reasons. Precursor No. 5 has h of 44 mPazs, about an order of magnitude higher than the others. As noted, filaments coated with this precursor are more frequently cemented together, probably due to lack of precursor displacement because of high h. If d >> r, precursor menisci do not interact with each other, and individual filaments should coat like separate monofilaments. Conversely, if filaments are close-packed, precursor menisci wick up the fiber tow by ∼2g/rr based on Eq. (9). The Fig. 13. d of precursor required to deposit coatings Nos. 1–4 vs precursor h. (– – –) represent predictions from coating models (Eqs. (5), (6), and (8)). September 1999 Continuous Coating of Oxide Fiber Tows Using Liquid Precursors 2329
2330 Journal of the American Ceramic Society-Boakye et al. Vol. 82. No 9 precursor is not coating ce for thermodynamic and k Coating bridges between filaments and crust around the tow perimeter were more abundant for more viscous precursors or these coating experiments, the relevant distance The most viscous precursors cemented filaments together filaments is somewhere between these two extremes. The coat- Some precursors made coatings with bubbles that could have ing process itself affects filament packing, because the capil formed from gas evolved during drying. The rate of increase in lary force of a drying liquid may pack filaments, as it does with coating thickness decreased with the number of coats. Coatings particles. Quantitative comparison between coatings made thicker than I um tended to crack off, probably from CTE vith and without immiscible liquid and measurements of y and mismatch. Cracking in thinner coatings was uncommon, de- n under actual coating conditions may yield more insight. Rig- spite high heating and cooling rates rous analysis requires knowledge of filament misalignment ess viscous precursors made coatings with less 8 variation and packing during drying. The observed lack of filament ce- 8 fit a lognormal distribution that was bimodal when there was mentation suggests that some immiscible liquid displacement crust around the tow perimeter. Thicknesses were nearly 2 ccurs The assumption of equal spacing in Eq. (9)is optimistic orders of magnitude greater than those predicted for isolate The filaments are more likely filaments. so filaments in a tow must entrain much thicker sistent with observations of variable 8 related to different amounts uid layers than isolated filaments. 8 increased ith m bu of precursor displacement dependent on filament space data quality did not allow confirmation of the predicted m It might seem that spreading fiber tows would solve the dependence. A serious obstacle to modeling tow coating was coating problems. However, it is not enough to spread the tow lack of information on filament spacing during coating and in the precursor--it must remain spread as the precursor dries Existing to eading me nods, such as pneumatic devices Immiscible liquid displacement might have reduced bridging lectrostatic methods 78 or magnetic fields79 d crust, but it did not displace enough precursor to make tre not suitable for nonconducting fibers, 9 humid atmo- individual filaments coat like monofilaments. Most precursor spheres, or high temperatures. 76, 77 If tows are spread enough was probably displaced by immiscible liquid and air at the to make individual filaments behave like isolated monofila- immiscible liquid-air interface. Tow spreading eliminated the ments during coating, it is apparent from the r dependence ir need for immiscible displacement, but much more viscous or Eq(8)that much more concentrated or viscous precursors or concentrated precursors were necessary to make 50-100 nm much higher coating speeds are necessary to make 50-100 nm Coated fiber g was from 25% to >50% lower than those of inviscid, low concentration precursors suitable for coating un- s-received fiber o was not dependent on 8 or precursor stol- spread tows make coatings only a few nanometers thick if they chiometry but did depend on the precursor used, which sug are used on a spread-out tor gested that flaw growth was sensitive to the effect of precursor (0 Tensile Strength decomposition products on the chemical environment during Filament o reductions after coating were large, in some coating. Lack of a stoichiometry dependence might have been artifact of the short time at temperature during coating cases so large that the fibers could not be spooled. There were Longer times at high temperatures should react nonstoichio- cursors(Table 1). Strength reduction was independent of 8 for metric monazite with the fiber and possibly cause further precursor No. 1, and a comparison of all coatings showed no strength degradation tendency for o to decrease with 8. The thickest coating per pas The optimal precursor for coating fiber tows should was No. 7, and this was also the strongest. Multiply coated low viscosity but high concentration, show the correct wet fibers had thicker coatings and were weaker, but this could be ting relationships, and form smooth thin films without bubble due to sequential strength decrease per coating pass. For coat formation from degassing. The optimal precursor for coating an individual 12 um filament should have the same attributes ing No. 3, the strength reduction was the same in air and argon but much higher viscosity Minimal fiber degradation from coatI A variation in o with monazite stoichiometry was not obvi- recursor decomposition byproducts is essential. Harmful by ous. Deliberate phosphate enrichment of coating No. I intro products are not identified. Their isolation is the subject duced an AlPO4 layer between the fiber and monazite coating future research that actually made the fibers slightly stronger than fibers with toichiometric coatings(Table I). Fibers with coating No. 7 Acknowledgments: We thank S. Sambasivan of Northwestern Univer were the strongest measured and were also phosphate-rich precursor No, 4, Dave Wilson from 3M for the Nextel 720 fiber, and K. Keller Although no phases other than monazite were observed for No for a thorough manuscript review 3, the powder from the precursor was observed to be slightly lanthanum-rich by X-ray, but fibers coated with it were inter mediate in o and still significantly stronger than those coated vith other stoichiometric precursors, such as No. I and No. 4 References Thickness uniformity also seemed to have little effect on o B D. Fabes, B JJ. Zelinski, and D R. Uhlmann, " Sok-Gel-Derived Ce- The fibers saw 1200%1400C for at most several seconds mic Coatings", Pp. 224-84 in Ceramic Films and Coatings. Edited by J. B pletely reacted coating with the fiber could introduce a stoi- am.Soc.Bam2,69114l-4301990 chiometry dependence that was not observed here Films "MpET le and R.W. Schwartz, "Solution Deposition of Ferroelectric Thin Sensitivity of fiber strength to corrosion and the environment SB. M. Deryagin and S. M. Levi, Film Coating Theory, p. 190. Focal Press, is a common observation. 80, 81 Differences in strength degrada- tion between different monazite precursors are attributed to bL. Strawbridge and P. F. James, Thin Silica Films Prepared by Dip Coat- stress corrosion from different decomposition products from different precursor chemistries. Future work will concentrate on isolating the corrosive species Brinker and G. w. Scherer, SolGel Science: Ist ed, p 908. Academic Press, San Diego, CA, 1990. I. Summary and Conclusions Nextel 720TM fibers were continuously coated when precur- iel Dip Coating, "Thin Solid Films, 201, 97-108(1991). ors with >40 g/L monazite yield and <2 mPas m were used Res.Soc.Smp.Proe.,121,717-29(1988)
precursor is not displaced for thermodynamic and kinematic reasons, and the coating cements filaments together. Clearly, for these coating experiments, the relevant distance between filaments is somewhere between these two extremes. The coating process itself affects filament packing, because the capillary force of a drying liquid may pack filaments, as it does with particles.8 Quantitative comparison between coatings made with and without immiscible liquid and measurements of g and h under actual coating conditions may yield more insight. Rigorous analysis requires knowledge of filament misalignment and packing during drying. The observed lack of filament cementation suggests that some immiscible liquid displacement occurs. The assumption of equal spacing in Eq. (9) is optimistic. The filaments are more likely to pack erratically, which is consistent with observations of variable d related to different amounts of precursor displacement dependent on filament spacing. It might seem that spreading fiber tows would solve the coating problems. However, it is not enough to spread the tow in the precursor—it must remain spread as the precursor dries. Existing tow spreading methods, such as pneumatic devices,76 multiple rolls,77 electrostatic methods,78 or magnetic fields79 are not suitable for nonconducting fibers,79 humid atmospheres,78 or high temperatures.76,77 If tows are spread enough to make individual filaments behave like isolated monofilaments during coating, it is apparent from the r dependence in Eq. (8) that much more concentrated or viscous precursors or much higher coating speeds are necessary to make 50–100 nm thick coatings in one pass on 12 mm diameter fibers. The inviscid, low concentration precursors suitable for coating unspread tows make coatings only a few nanometers thick if they are used on a spread-out tow. (10) Tensile Strength Filament s reductions after coating were large, in some cases so large that the fibers could not be spooled. There were large differences in s between fibers coated with different precursors (Table I). Strength reduction was independent of d for precursor No. 1, and a comparison of all coatings showed no tendency for s to decrease with d. The thickest coating per pass was No. 7, and this was also the strongest. Multiply coated fibers had thicker coatings and were weaker, but this could be due to sequential strength decrease per coating pass. For coating No. 3, the strength reduction was the same in air and argon coating atmospheres. A variation in s with monazite stoichiometry was not obvious. Deliberate phosphate enrichment of coating No. 1 introduced an AlPO4 layer between the fiber and monazite coating that actually made the fibers slightly stronger than fibers with stoichiometric coatings (Table I). Fibers with coating No. 7 were the strongest measured and were also phosphate-rich. Although no phases other than monazite were observed for No. 3, the powder from the precursor was observed to be slightly lanthanum-rich by X-ray, but fibers coated with it were intermediate in s and still significantly stronger than those coated with other stoichiometric precursors, such as No. 1 and No. 4. Thickness uniformity also seemed to have little effect on s. The fibers saw 1200°–1400°C for at most several seconds. Nucleation of a solid-state reaction between the fiber and a lanthanum- or PO4-rich precursor might not occur during this brief period. However, heat treatment at longer times that completely reacted coating with the fiber could introduce a stoichiometry dependence that was not observed here. Sensitivity of fiber strength to corrosion and the environment is a common observation.80,81 Differences in strength degradation between different monazite precursors are attributed to stress corrosion from different decomposition products from different precursor chemistries. Future work will concentrate on isolating the corrosive species. IV. Summary and Conclusions Nextel 720™ fibers were continuously coated when precursors with >40 g/L monazite yield and 50% lower than those of as-received fiber. s was not dependent on d or precursor stoichiometry but did depend on the precursor used, which suggested that flaw growth was sensitive to the effect of precursor decomposition products on the chemical environment during coating. Lack of a stoichiometry dependence might have been an artifact of the short time at temperature during coating. Longer times at high temperatures should react nonstoichiometric monazite with the fiber and possibly cause further strength degradation. The optimal precursor for coating fiber tows should have low viscosity but high concentration, show the correct wetting relationships, and form smooth thin films without bubble formation from degassing. The optimal precursor for coating an individual 12 mm filament should have the same attributes but much higher viscosity. Minimal fiber degradation from precursor decomposition byproducts is essential. Harmful byproducts are not identified. Their isolation is the subject of future research. Acknowledgments: We thank S. Sambasivan of Northwestern University for precursor No. 3, P. E. D. Morgan of the Rockwell Science Center for precursor No. 4, Dave Wilson from 3M for the Nextel 720 fiber, and K. Keller for a thorough manuscript review. References 1 B. D. Fabes, B. J. J. Zelinski, and D. R. Uhlmann, “Sol–Gel-Derived Ceramic Coatings”; pp. 224–84 in Ceramic Films and Coatings. Edited by J. B. Wachtman and R. A. Haber. Noyes, Park Ridge, NJ, 1993. 2 F. F. Lange, “Chemical Solution Routes to Single-Crystal Thin Films,” Science (Washington, D.C.), 273, 903–909 (1996). 3 H. Floch and J.-J. Priotton, “Colloidal Sol–Gel Optical Coatings,” Am. Ceram. Soc. Bull., 69 [7] 1141–43 (1990). 4 B. A. Tuttle and R. W. Schwartz, “Solution Deposition of Ferroelectric Thin Films,” MRS Bull. [June] 49–54 (1996). 5 B. M. Deryagin and S. M. Levi, Film Coating Theory; p. 190. Focal Press, New York, 1964. 6 I. Strawbridge and P. F. James, “Thin Silica Films Prepared by Dip Coating,” J. Non-Cryst. Solids, 82, 366–72 (1986). 7 I. Strawbridge and P. F. James, “The Factors Affecting the Thickness of Sol–Gel Derived Silica Coatings Prepared by Dipping,” J. Non-Cryst. Solids, 86, 381–93 (1986). 8 C. J. Brinker and G. W. Scherer, Sol–Gel Science; 1st ed, p. 908. Academic Press, San Diego, CA, 1990. 9 C. J. Brinker, G. C. Frye, A. J. Hurd, and C. S. Ashley, “Fundamentals of Sol–Gel Dip Coating,” Thin Solid Films, 201, 97–108 (1991). 10L. E. Scriven, “Physics and Applications of Dip Coating and Spin Coating,” Mater. Res. Soc. Symp. Proc., 121, 717–29 (1988). 2330 Journal of the American Ceramic Society—Boakye et al. Vol. 82, No. 9