`Tube Materials in Soda-Lime-Silicate Glass
`
`S. Kamakshi Sundamn,* Jen-Yan Hsu,* and Robert F. Speyer*
`School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0245
`
`The corrosion resistance of molybdenum, molybdenum
`disilicide, and a SiC,,,/AI,O, composite to molten soda-
`lime-silicate glass was studied. The ASTM-C621-84 cor-
`rosion test method was modified because of inherent
`inaccuracies in the method and Si attack of platinum cruci-
`bles. Specimen-glass interfacial regions were characterized
`using XRD, SEM, and EDS. After 48 h of exposure at
`1565"C, the half-down corrosion recessions of Mo, MoSi,,
`and SiC,,,/Al,O, were 0.11, 0.316, and 0.26 mm, respec-
`tively. Mo oxidized to form a MOO, surface scale which
`cracked, allowing glass seepage and further oxidation. Sili-
`con was leached out of MoSi, into the glass, leaving a Mo,Si,
`interface and particles of Mo near the interface. For the
`SiC,,,/AI,O, composite, bubbles observed at the interfacial
`regions formed from oxidation of Sic to form CO. Thermo-
`dynamic modeling corroborated
`these experimental
`observations.
`
`I. Introduction
`HE work discussed herein is part of a larger research thrust
`
`T to develop materials for immersed gas-fired radiant burner
`
`tubes for glass melters. In the evaluation of candidate materials,
`our research has adopted a parallel approach, separately evalu-
`ating molten glass corrosion, combustion gas corrosion, as well
`as using thermodynamic simulations for both processes. Gas
`corrosion investigations are reported elsewhere.'.' In the present
`paper, the molten glass corrosion behavior of candidate materi-
`als is presented.
`A selection of the candidate materials which were evaluated
`and the results of which are presented herein are the following:
`molybdenum, molybdenum disilicide, and SiC,,,/AI,O, com-
`posite (p refers to particulate). Molybdenum has been used as
`the electrode material for electric melting and boosting in glass
`melting tanks for many years.'., Molybdenum disilicide has
`demonstrated a self-protective mechanism of forming a protec-
`tive SiO, layer at the surface in an oxidizing atmosphere.' For
`temperatures below 600°C. oxidation of molybdenum disilicide
`forms solid MOO, and SiO, products on the surface. The forma-
`tion of MOO,,,, in microcracks extends them and ultimately the
`solid fragments into powder, referred to as MoSi, pest.6 In the
`case of SiC,,,/AI,O, composite, the combination of high ther-
`mal conductivity of SIC and the refractory properties of A1,0,
`were expected to make it potentially suitable for the combustion
`product side of the present application. The composite is fabri-
`cated by the Lanxide process:' directed oxidation of molten alu-
`minum infiltrating into a particulate Sic matrix. Fused-cast
`AZS (50.8 wt% alumina, 32.6 wt% zirconia, 14.9 wt% silica,
`plus impurity phases), used as molten glass container material
`in this investigation, is well known for its corrosion resistance
`
`to molten glass"o and is a common commercial glass tank
`refractory.
`Bockris and co-workers" used molybdenum crucibles to
`study the interaction of Mo with molten binary silicates and
`detected a 0.15 wt% molybdenum absorption as MOO, by mol-
`ten Na,O-SiO, at 1750°C after 3 h in a H,-N,
`atmosphere.
`Minute reaction between pure silica and Mo under vacuum at
`1600°C has also been observed." Soda-lime-silicate glass, free
`of iron and sulfur, has been shown to be only very slightly cor-
`rosive toward molybdenum in the temperature range 1OOO" to
`1300°C in an oxidizing atrno~phere.'~ The slight observed cor-
`rosion was due to reaction with dissolved oxygen, postulated
`(though not shown experimentally) to form MoO,.I4
`The oxygen solubility of a glass melt is not governed entirely
`by the oxygen partial pressure of the furnace atmosphere; bas-
`i ~ i t y , ' ~ e.g., the concentration of alkali, and the concentration
`and type of polyvalent species present in the melti6." are also
`controlling factors. Based on solubility measurements of CO,,,,
`in glass, Pearce18 showed that oxygen ion activity increased
`dramatically with increasing alkali content in sodium silicate
`glasses. Because of its open structure, however, oxygen is pre-
`dominantly absorbed molecularly rather than atomically.
`Oxidation of molybdenum by dissolved oxygen in pure
`soda-lime-silicate glass melts was minute as compared to
`melts containing 0.1 wt% Fe and 0.3 wt% SO,. On reacting
`with Na,SO,, MoS, formed by
`~Mo,,, + 2Na,SO,,,, = MoS,,,, + 2Na,MoO,,,,
`The addition of 0.10% Fe,O, only (no sulfur compounds) to the
`soda-lime-silicate melt did not show any appreciable corro-
`sion.', Ooka" studied the effect of the amount of As,O,,
`Na,SO,, and Sb,O, present in soda-lime-silicate glass on the
`weight loss of molybdenum at 1500°C for 50 h. As,O, additions
`were found to be highly corrosive to Mo, Na,SO, was found to
`be slightly corrosive (forming MoS,), and Sb,O, had no effect.
`Lead-containing glasses were highly corrosive to Mo. Pecoraro
`el uLZo observed that the extent of corrosion by the mechanism
`of direct elemental solubility was negligibly small; corrosion
`proceeded primarily through the oxidation of the metal caused
`by oxidizing agents, dissolved oxygen and water, or a melt con-
`stituent which could accept electrons from the metal. Literature
`studies do not clearly indicate the form of corrosion product nor
`were they able to accurately measure corrosion rates.
`Muan and S p e d ' evaluated thermodynamic calculations for
`the systems Mo-0, Mo-Si-O, and Na,O-CaO-SiO, and sug-
`gested MoSi, as a potentially suitable material for gas boosters
`immersed in glass melting tanks. No experimental evidence of
`the corrosion rate or mechanism has been reported. The corro-
`sion resistance of SiC,,,/AI,O, composite in molten glass has
`also not been reported. However, the corrosion resistance of
`pure alumina has been studied extensivelyzz-26 and shown to be
`less than optimum.
`
`1. Smialek+ontributing editor
`
`Manuscript No. 194696. Received April 12,1993; approved February 16.1994.
`Supported by the Gas Research Institute under Contract No. 5090-298-2073.
`'Member. American Ceramic Society.
`
`11. Experimental Procedure
`
`( I ) Corrosion Testing
`The ASTM standard method C621-84 for isothermal corro-
`sion resistance of refractories to molten glass was initially
`adopted for the present investigation." This method involves
`1613
`
`GE-1029.001
`
`
`
`Journal of the American Cerurnic Society-Sundaram et al.
`
`Vol. 77, No. 6
`
`1614
`partial immersion of a test specimen of rectangular or circular
`cross section in molten glass contained in a Pt crucible. The
`recession from original material dimensions at the glass line
`and at an immersed position are measured after high-tempera-
`ture exposure for various periods of time. Modifications of the
`standard were made based on the following encumbrances: (i)
`molybdenum oxidizes and volatilizes above the glass line at
`temperatures exceeding -600°C, and (ii) silicon dissolves into
`the melt from the MoSi, and SiC,,,/AI,O, test coupons, and in
`turn alloys with the platinum crucible to form low-temperature
`eutectic phases, destroying the crucible.
`The Pt crucible as specified in ASTM C62 1-84 was replaced
`by an AZS (UNICORSOI, Corhart Refractories Corp.) crucible
`with a containment volume of 3.81-cm diameter by 3.81-cm
`depth to contain molten glass. The choice of AZS was made
`after static corrosion testing of fusion-cast AZS, MgO-partially-
`stabilized zirconia (PSZ) (Coors Ceramics Co., Golden, CO),
`fusion-cast chromia (Carborundum Co., Monofrax Refractories
`Division, Falconer, NY), and bonded chromia (Corhart Refrac-
`tories Corp., Buchannon, WV) by following the ASTM
`C62 1-84 standard. The results of corrosion testing, presented in
`Figs. I(a) and (b), indicate that AZS was the least corroded
`material at both the glass-line and half-down regions.
`
`Chromia (F)
`
`The soda-lime-silicate glass (Owens-Illinois, Inc.) used in
`the present investigation is a common container-glass composi-
`tion. A chemical analysis of the as-received glass in weight per-
`cent shows 73.3% SiO,, 13.3% Na,O, 11% CaO, 1.5% AI,O,
`0.3% K 2 0 , 0.2% MgO, 0.2% SO,, 0.2% ZrO,, and 0.07%
`FezO,. Molybdenum (Johnson Matthey/AESAR) specimens
`were of 12.5-mm diameter by 12.5-mm length for the purpose
`of complete immersion. Chemical analysis for trace elements in
`as-received molybdenum showed less than 1 ppm of Al, Ca, Cr,
`Cu, Mg, Mn, Ni, Pb, Si, Sn, and Ti, less than 2 ppm of C and 0,
`and less than 14 ppm of Fe. Cylindrical specimens of MoSi,
`(Kanthal Super 33, Kanthal Corp.) and SiC,,,/AI,O, (Du Pont
`Lanxide Composites) of 12.5-mm diameter by 60-mm length
`were used. The as-received MoSi, contained 1.7 vol% Mo,Si,
`grains, generally in contact with amorphous aluminosilicate
`glassy pockets comprising 18.6 vol% of the specimen, as deter-
`mined by scanning electron microscopy image analysis.
`The test configuration is sketched in Fig. 2. The quantity of
`glass powder put into the AZS crucible was adjusted so that a
`molten glass level of 18.8 mm was maintained. The as-received
`glass was crushed to -40 mesh, and the AZS crucible with
`48.5 g of glass was placed inside an electrically heated furnace.
`The glass was heated to 1565°C and allowed to equilibrate at
`
`0.4
`
`T
`
`0
`
`10
`
`40
`
`50
`
`30
`20
`30
`20
`Time (hrs)
`Time (hrs)
`(a)
`(b)
`Fig. 1. Corrosion rate of refractory containment materials at (a) glass-line and (b) half-down regions. The error bars indicate the standard deviation
`in the measurement of specimen diameters before the corrosion tests. These out-of-roundness deviations represent the largest errors of all contribu-
`tions to error in the measurement.
`
`40
`
`50
`
`Fig. 2. Fully immersed static corrosion test (unit: mm).
`
`GE-1029.002
`
`
`
`June 1994
`
`Molten Glass Corrosion Resistance of Immersed Combustion-Heating Tube Materials
`
`1615
`
`i
`
`Fig. 3. Correlation of measured diameters 20 and 2b to the true diameter 2r of the corroded specimen, for a known blade thickness T. Left: case
`where blade cuts through the center. Right: case where blade misses the center.
`
`that temperature for 2 h to obtain bubble-free glass. Test speci-
`mens were then loaded into the crucible.
`The diameters of the pre-exposed specimens at the glass-line
`as well as the half-down positions were measured using digital
`vernier calipers (Mitutoyo Corp., Model CD-8") with a preci-
`sion of 20.01 mm. Each specimen was rotated and the mea-
`surement was performed at 10 points of contact at the two
`positions. The standard deviation of measured diameters ranged
`from 0.0052 to 0.0255 mm. The planeness and parallelness of
`the surfaces of the specimens were checked using a machinist's
`square. If the faces were not truly flat and perpendicular to the
`cylindrical axis, they were ground using 600-p,m' polishing
`paper and checked again until they were flat. In the case of
`molybdenum specimens, the center was marked using a lathe
`(Rivett Lathe and Grinders Inc., Boston, MA) and a self-center-
`ing collet before the corrosion test (precision: 2 0.05 1 mm).
`MoSi, and SiC,,,/AI,O, specimens were affixed to zir-
`con support wafers with zircon cement and allowed to dry
`overnight. They were preheated to 1OOO"C and maintained at
`that temperature for 10 min, to avoid specimen thermal shock
`when transferred to the equilibrated molten glass. Each molyb-
`denum specimen in a Pt wire basket was suspended by Pt wire
`into the molten glass so that the specimen was completely
`immersed. These specimens were quickly immersed in the mol-
`ten glass at the test temperature; they were not preheated so as
`to avoid atmospheric oxidation. The durations of corrosion tests
`were 12, 24, and 48 h at 1565°C in static air. The test tempera-
`ture of 1565°C corresponds to the hot-spot temperature in a
`commercial glass tank. One specimen was tested for each
`experimental condition. After exposure, the specimens were
`removed from the melt and allowed to cool in ambient.
`The corroded specimen rods were cut lengthwise along the
`rod axis using a diamond wafering blade (Buehler, Lake Bluff,
`IL) with a thickness of 0.34 2 0.1 mm ( 2 0.1 mm positioning
`precision). Several initial steps were taken to ensure that the
`wafering blade propagated through the center of the specimen:
`For Mo after exposure, the top Mo surface was ground until a
`flat surface was exposed, while for other specimens, the uncor-
`roded top surface was already exposed. The specimens were
`then mounted on this surface using thermal glue (Buehler). A
`cut running parallel to the axis of each specimen was made. The
`edges along this side face were checked as to whether they were
`parallel to each other using a digital vernier calipers. If they
`were not parallel, the side face was repeatedly ground with
`uneven pressure, and remeasured, until they were parallel; this
`indicated that the side face was parallel with the cylinder axis.
`For MoSi, and SiC,,,/AI,O, specimens, the center of the top
`surface was marked by the intersections of drawn diameters,
`
`while for Mo, the center was previously marked (see above).
`The specimens were then mounted on the side face, and another
`cut was made through the marked center, down the center axes
`of the specimens.
`The diameters at the glass-line and half-down positions on
`the cut surfaces were measured using the platform and micro-
`scope of a micro-hardness tester (Model HMV 2000, Shimadzu
`Co., Kyoto, Japan) where the precision of diameter measure-
`ment was 20.01 mm. An oil of refractive index of 1.51 was
`used to cause the glass to vanish from view so that the interface
`could be clearly seen through the microscope.
`The ASTM C621-84 procedure does not account for the
`thickness of the cutting blade when evaluating cylindrical spec-
`imens; it specifies only to determine the average of the mea-
`sured diameters of the specimen halves. Figure 3 illustrates the
`effect of blade thickness on the corrosion measurement. If 2u
`and 2b are the measured diameters, and 2r the genuine diameter
`of the specimen, then r can be calculated:
`. , ]
`
`[(" + ;; - UZf +
`
`r = ?
`
`where T is the thickness of the blade. The calculated propagated
`error by using the equation of r ranged from 0.0003 to 0.007
`mm. The computational procedure is elaborated upon else-
`where.2R Since the calculated errors corresponded to less than
`the y-axis dimension of data markers, error bars for radius
`recession plots are not shown.
`The reaction of the AZS container with molten glass was sep-
`arately evaluated. A 48.5-g sample of glass was placed in an
`AZS container and maintained at 1565°C. After equilibration, a
`small quantity of the melt was removed from the center of the
`exposed melt surface using a Pt boat, and quenched after 3.6,
`12,24, 36, and 48 h of reaction with the container. These sam-
`ples were then chemically analyzed for SO,, A1,0,, and ZrO,.
`(2) Characterization
`The glass-specimen interface was characterized using scan-
`ning electron microscopy (SEM) (Model 18 10, AMRAY, Bed-
`ford, MA). Cross sections of specimens were mounted in epoxy
`and the surfaces were ground smooth using 120-, 240-, 320-,
`400-, and 600-grit abrasive papers and then polished with dia-
`mond paste (30, 15,6, and l p,m). The polished samples were
`etched in 1% HF for -30 s, rinsed with distilled water, and then
`cleaned in an ultrasonic cleaner using distilled water, and dried.
`The etched specimens were coated with conductive Au-Pd
`alloy. Point and line scan elemental compositions were deter-
`mined using energy dispersive spectroscopy (EDS) (Model
`
`GE-1029.003
`
`
`
`1616
`
`Journal of the American Ceramic Society-Sundaram et al.
`
`Vol. 77, No. 6
`
`1.41
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`10
`
`40
`
`50
`
`30 30
`
`20 20
`
`30
`20
`Time (brs)
`Time (hrs)
`(a)
`(b)
`Fig. 4. Specimen recessions at (a) glass-line and (b) half-down regions when exposed to soda-lime-silicate glass at 1565°C.
`
`
`
`0 0
`
`
`
`10 10
`
`
`
`40 40
`
`
`
`50 50
`
`TS 16-501 8, Princeton Gamma Technology, Princeton, NJ) to
`establish the chemistry at and across the interfacial regions.
`A 48-h corrosion test was repeated using a specimen with a
`,emicircular cross section where the flat interface was for the
`convenience of XRD analysis. Diffraction patterns were taken
`on the glass-coated surface and then repeated after grinding the
`flat surface with 600-pm abrasive paper for 10 min and clean-
`ing. This process was continued until the interface was exposed
`and then penetrated. The specimen was analyzed using a com-
`puter-interfaced (Model VAX- I 1/730, Digital Equipment Co.,
`Northboro, MA) X-ray (CuKa) powder diffractometer (Model
`12045 X-ray diffraction unit, Philips Electronic Instruments
`Co., Mount Vernon, NY) with a I-s time constant and 0.02” step
`4ze over a 28 range of 20-80”. Using a computer search pro-
`gram, the X-ray diffraction patterns were matched to JCPDS
`data and corresponding phases were identified. In the phase
`identification procedure (D5000, Alfred University X-ray Lab-
`oratory, Alfred, NY), two types of searches, LIST and SUB-
`TRACT, were used. The LIST search performed a forced match
`of the input data with the entire JCPDS powder diffraction file
`and produced a list of potential matches. The SUBTRACT
`search performed a Hanawalt search and subtracted a matching
`compound that met a figure-of-merit criterion.
`(3) Thermodynamic Analysis
`Thermodynamic analysis of Mo-O, M-Si-0,
`Si-C-0, and
`A 1 4 systems were performed to predict solid-molten glass
`interfacial chemistry. The standard-state free energies of reac-
`tion, forming MOO,, MOO ?, ”.”’ Mo,Si, and Mo,Si,, and
`CaMoO,” were computed and plotted against temperature
`(Ellingham diagrams). Thermodynamic data were obtained
`from the Facility for the Analysis of Chemical Thermodynam-
`ics (F*A*C*T).+ These thermodynamic plots represent equilib-
`ria among solids and molecular oxygen in their standard states,
`that is, pure oxygen at 1 atm and pure solids.
`Since oxygen dissolves predominantly molecularly
`in
`glass,i9 the chemical potential of 0, in molten glass is equal to
`that in the furnace (electrically heated) atmosphere, where
`po. = 0.2. The activities of oxygen in the glass and in the
`atmosphere, however, are not equal, since soluble gas is gener-
`ally considered as a dilute solution which is non-Raoultian.
`Norton” has determined that the ratio of the concentration of
`molecular oxygen in the glass relative to that in the furnace
`atmosphere at 1078°C is 0.01. The “coefficient of solubility”
`does not change appreciably with temperature.”.” Thus, the
`Henrian activity of molecular oxygen in glass was taken to be
`2 X lo-’. and Richardson lines corresponding to this oxygen
`activity are drawn on the Ellingham diagrams.
`
`_ _ _ _ _ _ _ ~
`’Accessed through time-sharing with McGill University. Montreal. Canada
`
`111. Results and Discussion
`The chemical analyses of the glass (removed from the top-
`center of the melt) after interaction with AZS for varying times
`are shown in Table 1. No significant changes in these constit-
`uents were observed up to 12 h of exposure. Beyond 12 h, ALO,
`and ZrO, concentration gradually increased. After 48 h of expo-
`sure, the A120, content increased to 2.8%. and the ZrO, content
`to 0.15%. The effect of this compositional change on the speci-
`men corrosion rate is not known. Replenishing glass melt after
`every 12 h of exposure was considered and rejected since it
`changes the present static testing to a semidynamic corrosion
`test.2u
`Figures 1 (a) and (b) show the glass-line and half-down reces-
`sions, respectively, of potential containment materials. In the
`case of the glass-line corrosion, AZS was the least corroded
`material. In the case of the half-down region, initially fusion-
`cast chromia showed a corrosion rate less than that of AZS, but
`increased steeply beyond 12 h of exposure. Chromia also
`severely discolored the glass. The half-down corrosion rates of
`bonded chromia and PSZ could not be measured because of
`specimen cracking.
`Figures 4(a) and (b) show the glass-line and half-down reces-
`sions of cylindrical specimens (in turn contained in AZS cruci-
`bles), respectively, by molten glass. In the case of glass-line
`corrosion, SiC,,,/AI,O, showed a corrosion rate initially less
`than that of MoSiz but showed greater recession rates after 12 h
`of exposure, and was the most corroded specimen after 48 h of
`exposure. With regard to half-down corrosion, Mo was the least
`corroded material. MoSiz and SiC,,,/AI,O, showed similar cor-
`rosion behavior. The indicated downward trend for SiC,,,/AI,O,
`could not be corroborated by repeated testing because of speci-
`men availability.
`( I ) Molybdenum
`Figure 5 shows XRD patterns starting from the glass side,
`penetrating through the interface. On the glass side, only one
`crystalline peak is apparent, which matches with CaMoO,. This
`
`Table I. Chemical Analysis of Glass after Interactions with
`AZS Crucibles at 1565°C
`Composition ( ~ 1 % )
`A I J h
`I .5
`1.4
`1.4
`1.4
`2.6
`3.2
`4.3
`
`~~
`
`~
`
`Corrosion lime ( h )
`0
`3
`6
`12
`24
`36
`48
`
`SiO,
`73.3
`72.6
`72.8
`73.5
`73.3
`73.2
`73.1
`
`Zr02
`0
`0
`0
`0
`0.06
`0.1
`0.15
`
`GE-1029.004
`
`
`
`June 1994
`
`Molten Glass Corrosion Resistance of Immersed Combustion-Heating Tube Materials
`
`1617
`
`500 I
`
`1
`
`I
`
`I
`
`I
`
`I
`
`Y 20
`
`40
`
`60
`
`Two Theta
`
`80
`
`XRD patterns penetrating from the glass, through the interface. into the molybdenum bulk of a specimen after an exposure of48 h at 1565°C:
`Fig. 5.
`a. MOO,: c. CaMoO,; m. Mo; p. Pt: u. unidentified phase.
`
`peak gradually reduces, then increases in intensity with penetra-
`tion. Low-intensity peaks corresponding to this phase are still
`apparent within the interfacial region. The trace corresponding
`to deepest penetration indicates the presence of MOO, and Mo,
`with minor quantities of CaMoO,. The increase and decrease in
`the intensity CaMoO, peak are attributed to the discontinuous
`composition of the debns in the glass near the interfacial layer.
`A few small peaks (marked “u” in Fig. 5) could not be matched
`with any known phases in the JCPDS database. Pt XRD peaks
`correspond to the Pt wire used in suspending the specimen in
`the glass melt.
`Figure 6 shows the Mo-glass
`interfacial morphology. An
`interfacial reaction layer is apparent. From the backscattered
`image (right-hand side), atomic number contrast of the interfa-
`cial layer implies that it was composed of (on average) atomi-
`cally lighter elements. This was corroborated by the line EDS
`scan shown in Fig. 7 in which a Mo-depleted (relative to Mo
`bulk) interfacial region is shown. This corresponds to XRD
`results showing MOO? at penetration depths corresponding to
`the highest trace in Fig. 5. It is therefore clear that the interfacial
`layer is Moo2.
`
`The sharp increase in Si intensity in Fig. 7 near the Mo bulk/
`interfacial layer interface implies the penetration of glass into
`the interfacial region. It is interpreted that oxidation of molyb-
`denum caused local expansion and associated compressive
`stress. These layers thus cracked, allowing penetration of the
`molten glass.
`Our XRD results showed that no MOO,,,, or MoS2,\, was
`detected within the interfacial region. The latter is in spite of the
`fact that the glass used in this investigation contained 0.2 wt%
`of SO,. Figure 8 shows the Ellingham diagram of various
`Mo-O,-Ca equilibria. MOO, is clearly favored over MOO,
`since the equilibria line falls above the Richardson line at the
`testing temperature. This is in agreement with XRD results at
`the interfacial region. Only near an atmospheric activity of sol-
`uble molecular oxygen would MOO, be expected to form.
`XRD results identified the formation of CaMoO, at the
`interfacial region and floating out into the glass. From Fig. 8,
`reaction between formed MOO,, calcium oxide, and dissolved
`oxygen is more favorable than reaction between elemental
`molybdenum, calcium oxide, and dissolved oxygen. This is
`
`Molybdenum-glass interfacial region after 48 h of corrosion (B = bulk, I = interface). Left: secondary electron image. Right: backscat-
`Fig. 6.
`tered image.
`
`GE-1029.005
`
`
`
`1618
`
`Journal of the American Ceramic Society-Sundaram et al.
`
`Vol. 77. No. 6
`
`Line EDS of the molybdenum-glass
`
`interfacial region in
`
`Fig.7.
`Fig. 6 .
`
`consistent with XRD results showing the presence of both Mo
`and CaMoO,. If CaMoO, formation using elemental Mo had
`been more favorable, then the formation of MOO? would not
`have been expected.
`(2) Molybdenum Disilicide
`The material used in this investigation is a cermet principally
`containing MoSi, with a minor aluminosilicate amorphous
`phase, and a fused silica surface layer of - 10 pm, as shown in
`phase, regions of Mo,Si, in contact with the ahminosilicate
`Fig. 9. After pre-heat treatment at 1000°C for 10 min, the sur-
`face layer became somewhat thinner and discontinuous,
`implying the onset of minute pesting. The presence of this sur-
`face layer on all specimens is expected to offset the absolute
`values of recession by 0.01 mm, but not expected to affect the
`trend of recession of MoSi2. This thickness was subtracted from
`all calculated specimen recessions.
`
`Fig. 9. Secondary electron SEM micrograph of a cross section of the
`surface of as-received MoSi, showing a -7-km fused silica coating:
`s, silica-rich vitreous protective coating; m, MoSi,; f. Mo,Si,; a, alumi-
`nosilicate glass.
`
`Figure 10 shows successive XRD patterns starting from the
`glass, penetrating through the interface. On the glass side, sev-
`eral peaks were identified as a-quartz. This phase was not
`detected in the interfacial region. On approaching the interfacial
`region, first molybdenum was identified. The interfacial region
`was found to be a mixture of primarily Mo,Si, with minor quan-
`tities of Mo,Si and Mo,O?,.
`
`100 I
`
`I
`
`-700 '
`
`0
`
`I
`I
`2000
`1000
`Temperature (K)
`
`I
`3000
`
`Fig. 8.
`Ellingham diagram of oxidation equilibria. A Richardson line of 2 X 10.' represents the molecular oxygen activity in soda-lime-silicate
`glass. The dotted lines indicate extrapolations. The dot-dashed line indicates the testing temperature ( I838 K).
`
`GE-1029.006
`
`
`
`June 1994
`
`Molten Glass Corrosion Resistance of Immersed Combustion-Heating Tube Materials
`
`1619
`
`20
`
`40
`
`60
`
`80
`
`Two Theta
`
`Fig. 10. XRD patterns as a function of penetration, analyzing glass,
`through the interfacial region. to the molybdenum disilicide bulk: m.
`molybdenum; q, a-quartz; 1, MoSi,; 2, Mo,Si,; 3, Mo,02,; and u,
`unidentified phase.
`
`.,
`0
`
`1
`
`3
`2
`Energy (KeV)
`
`4
`
`5
`
`Fig. 11. EDS patterns of flat-faced MoSi, specimen corresponding to
`the XRD pattern second from the top in Fig. 10. The spectra marked
`“bulk” is from an as-received specimen.
`
`Molybdenum disilicide-glass interfacial region at (a) glass-line and (b) half-down regions after 48 h of corrosion: B, bulk; I, interface; and
`Fig. 12.
`G, glass. Left: secondary electron image. Right: backscattered image.
`
`GE-1029.007
`
`
`
`1620
`
`Journal of the American Ceramic Society-Sundaram et al.
`
`Vol. 77, No. 6
`
`(a)
`(b)
`Fig. 13. Line EDS of molybdenum disilicide-glass interfacial region at (a) glass-line, (b) half-down regions.
`
`100
`
`0
`
`-1 00
`
`-200
`
`.....2
`
`p
`a3 2 -300
`:$
`es
`2
`; -600
`P
`8
`
`-400
`
`-500
`
`-700
`
`-800
`
`-900
`0
`
`2000
`1000
`Temperature (K)
`
`3000
`
`Fig. 14. Ellingham diagram of compounds of interest in MoSi+xygen
`
`interaction.
`
`Figure 1 I shows an SEM EDS pattern of the outer interface
`region from which the XRD second from the highest in Fig. 10
`was taken. This figure shows appreciable molybdenum
`enhancement relative to silicon, corresponding to an XRD pat-
`tern showing predominantly Mo,Si,.
`interfacial fea-
`Figures 12(a) and (b) show the MoSi,-glass
`tures at the glass-line and half-down regions, respectively.
`From the backscattered images, the interfacial region was richer
`in high atomic weight elements (relative to the bulk). Since
`oxygen is lighter than both metals, the interfacial region is not
`an oxide such as MOO,, or SO,. Mo being heavier than Si, the
`interfacial region is interpreted to be Mo-rich. Further corrobo-
`ration is provided by EDS line scans (Figs. 13(a) and (b)). At
`the interface, the Si concentration decreased sharply. The fluc-
`tuations in Si and Mo concentrations in the MoSi, bulk were
`attributed to the amorphous aluminosilicate phase (dark second
`phase in Fig. 12). The interfacial layer had a thickness of -35
`pm at the half-down region, and -25 pm at the glass-line
`region. A small Mo EDS peak suggests a limited molybdenum
`debris stream into the glass.
`
`XRD, SEM, and EDS results provide irrefutable evidence
`that the mechanism of MoSi, corrosion is removal of silicon
`from MoSi, to the glass, leaving a silicon-deficient molybde-
`num silicide. This can be visualized as an oxidation reaction
`where oxygen was provided from dissolved molecular oxygen:
`SMoSi, + 70, = Mo,Si, + 7Si0,.
`Figure 14 is an Ellingham diagram for equilibria of interest
`among Mo, Si, and 0,. The formation of silica from MoSi, via
`the most oxygen conservative mechanism is favored. Thus sil-
`ica is formed by removing silicon from MoSi, to form Mo,Si,,
`rather than complete oxidation of the silicon component, the
`molybdenum component, or both. In addition, the most favored
`reaction is that most conservative in silicon removal from the
`compound; e.g., Mo,Si, forms in favor of Mo,Si or Mo. If no
`MoSi, was locally available, then it would be thermodynami-
`cally favorable for Mo,Si, to convert to Mo,Si, and by the same
`argument, for Mo,Si to convert to Mo.
`These predictions are consistent with our experimental
`results where at the interface, Mo,Si, forms along with a minor
`amount of Mo,Si. The Mo,Si, at the interface would not be
`
`GE-1029.008
`
`
`
`June 1994
`
`Molten Glass Corrosion Resistance of Immersed Combustion-Heating Tube Materials
`
`1621
`
`expected to convert to Mo,Si or Mo since MoSiz is locally
`available which has a greater affinity for dissolved oxygen. A
`volume change would be associated with silicon removal from
`MoSi,. Thus Mo,Si, surface layers may dislodge from the inter-
`face and float away as debris. No longer near regions of MoSi,,
`local dissolved oxygen converts Mo,Si, to Mo,Si, and Mo,Si to
`Mo. The latter was observed by XRD in regions in the glass
`extended away from the interface.
`After an appreciable silicon-deficient interfacial layer has
`formed on the specimen, the diffusion distance for silicon from
`layer, to the surface
`the MoSi, bulk, through the Mo,Si,
`becomes arduous, and the rate of further silica formation is
`decreased. This appears to be the justification for the decrease
`in corrosion rate for the half-down tests, and the apparent termi-
`nation of recession after I2 h for the glass-line tests. This flat-
`tening of the corrosion rate may not be found in a commercial
`glass tank where greater convective flow forces are expected,
`which may periodically break loose and sweep away portions of
`the interfacial layer.
`(3) SiC,,,IA~,O,
`Figures I S(a) and (b) show the SiC,,,/Al,O,-glass interfacial
`features at the glass-line and half-down regions, respectively.
`Bubbles formed within the glass, near the composite surface at
`both regions. Near the glass line, the bubbles are seen in contact
`
`with the bulk of the specimen. CO or CO, can be the product of
`reaction between S i c and oxygen.”
`Competing oxidation reactions are considered in Fig. 16. The
`Sic portion of the composite oxidizes most favorably to form
`silica and carbon monoxide. If the carbon monoxide diffuses
`away from the interface, it is favorable for it to react with dis-
`solved oxygen to form CO,. Thus, the observed bubbles are
`anticipated to be a mixture of the two gases, of unknown ratio.
`The presence of bubbles in contact with a refractory in mol-
`ten glass has been demonstrated to be a mechanism of highly
`accelerated corrosion referred to as upward or downward
`“drilling,”34 where convection driven by surface tension gradi-
`ents sweeps fresh glass to the interface and corrosion products
`away. Thus, fresh glass with a higher oxygen activity would
`replace more viscous silica-rich glass. This corresponds well to
`the observed high corrosion rates for these specimens.
`(4) Remarks
`It is interesting to note that the MOO, interfacial layer formed
`in Mo specimens (-IS km) was much less