Thermodynamic Analysis of Alumina Refractory Corrosion by Sodium or Potassium Hydroxide in Glass Melting Furnaces

Transcription

1 Jurnal f The Electrchemical Sciety, B551-B /2002/ /B551/9/$7.00 The Electrchemical Sciety, Inc. Thermdynamic Analysis f Alumina Refractry Crrsin by Sdium r Ptassium Hydrxide in Glass Melting Furnaces Karl E. Spear a, * and Mark D. Allendrf b, **,z a Department f Materials Science and Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802, USA b Sandia Natinal Labratries, Cmbustin Research Facility, Livermre, Califrnia , USA B551 In this paper the high-temperature crrsin f Al 2 O 3 refractries by MOH g (M Na, K) fund in the cmbustin atmspheres f typical air- and xygen-fired glass-melting furnaces is examined using thermdynamic equilibrium calculatins. These hydrxide species are cnsidered t be the primary reactive alkali species since their partial pressures are significantly larger than thse f M g, the next mst abundant gas-phase alkali-cntaining species expected in typical furnace atmspheres. Thermchemical simulatins shw that crrsin f -alumina by NaOH g at typical furnace p NaOH g f arund 200 ppm under xy/fuel-fired cnditins is unlikely as lng as the refractry temperature exceeds 1564 K. Fr KOH g at 200 ppm, the temperature f the refractry must exceed 1515 K t avid crrsin. Under air-fired cnditins, p NaOH g is cnsiderably lwer ppm ; at50 ppm, crrsin is thermdynamically unfavrable at temperatures abve 1504 K. Fr KOH g at furnace levels f 50 ppm, temperatures must be abve 1458 K. The paper als presents a re-evaluatin f the thermdynamic and phase equilibrium prperties f the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 binary systems t develp accurate and self-cnsistent thermdynamic data. The data fr MAl 9 O 14 -alumina and M 2 Al 12 O 19 ( -alumina) are particularly critical since these phases are likely prducts f the crrsin f alumina refractries by MOH vaprs in glass melting furnaces The Electrchemical Sciety. DOI: / All rights reserved. Manuscript submitted March 1, 2002; revised manuscript received May 24, Available electrnically Octber 24, * Electrchemical Sciety Fellw. ** Electrchemical Sciety Active Member. z mdallen@sandia.gv The crrsin f refractry materials used t line the ceiling r crwn f cmbustin-heated glass-melting furnaces is exacerbated by the intrductin f xy/fuel melting technlgy in which air is replaced by xygen as the xidizer. The cncentratin f alkali hydrxides NaOH and/r KOH prduced by reactin f cmbustingenerated water vapr with alkaline xides in the glass melt in these furnaces is a factr f tw t fur times higher than in air-fired furnaces, due t the higher water-vapr cncentratin that results frm remving nitrgen frm the cmbustin gases. This can lead t unusually high crrsin rates fr lw-density silica refractries, which are ften used in air-fired furnaces. 1 Anecdtal reprts indicate that reductins in silica crwn lifetimes by as much as a factr f tw can ccur due t crrsin f the refractry. As a result, there is great interest in understanding the mechanisms fr this enhanced crrsin and fr using this infrmatin t design new, mre effective refractries. Alumina-based ceramics, such as bnded and fused cast aluminas -alumina (Al 2 O 3 ), -alumina a measured experimental cmpsitin f NaAl 9 O 14 ), 2 and a mixture f -alumina and -alumina, which we refer t as -alumina/ -alumina, fused cast aluminazircnia-silica AZS, and fused mullite (Al 6 Si 2 O 13 ) prvide alternatives t silica refractries. These materials are als subject t varying degrees f crrsin in xy/fuel envirnments, hwever. While silica refractries react with alkali vaprs t frm lwmelting sdium- r ptassium-silicate glassy liquids that can drip frm the crwn int the mlten glass, 3 mixing f these crrsin prducts with the glass melt is nt always a prblem since the chemical cmpsitins f the tw liquids are similar. In the case f alumina-cntaining refractries, hwever, frmatin f crrsin prducts that can drip, run, r spall int the glass melt can lead t defects in the final prduct. 4 Thus, there is a need t develp an understanding f the prcesses leading t crrsin f these alternative refractries in xy/fuel furnaces t determine hw furnace designs and/r perating cnditins can be adjusted t reduce r eliminate crrsin. Thermdynamic calculatins i.e., predictins f chemical cmpsitins based n minimizatin f the Gibbs free energy can play a useful rle by identifying energetically stable species and predicting the extent f crrsin and frmatin f prducts. The high temperatures at crwn surfaces 1850 K in sme lcatins 5 and lng reactin times f mnths t years indicate that equilibrium shuld be a valid apprximatin t the chemistry. T perfrm such calculatins, hwever, it is critical t develp internally cnsistent thermdynamic prperties fr all phases in these systems, including liquids, since liquid crrsin prducts can be a factr in refractry crrsin in glass furnaces. With a cmplete set f data fr each system, equilibrium calculatins can then be used t assess the ptential imprtance f all pssible prduct phases in the crrsin prcess. Crrsin f alumina by sdium has been f interest fr sme time due t its use in lw-pressure sdium discharge lamp envelpes, 6,7 in cmpsites, 8,9 and structural ceramics. Using varius qualitative and semiquantitative methds, the stability f alumina-cntaining refractries in glass-melting envirnments has been evaluated. 4 In general, -alumina and -alumina/ -alumina perfrm better than mst ther refractries in high-alkali envirnments. -alumina is als f interest as a sdium-in-cnducting slid electrlyte, and thus its thermdynamic stability is f imprtance t this applicatin as well. 10 Equilibrium calculatins used t predict the stability f alumina refractries are reprted in tw studies; 4,11 hwever, it is unclear if internally cnsistent thermdynamic data were used. The results suggest that bth -alumina and -alumina/ -alumina shuld be stable under typical xy-fuel melting cnditins and that alumina is mre resistant t alkali attack than silica. Earlier reprts f thermdynamic prperties such as heats f frmatin, entrpies, heat capacities, heats f transitin, and activities and phase diagrams fr bth the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 systems are summarized and critically evaluated by Erikssn et al. 12 The Na 2 O-Al 2 O 3 phase equilibria data were als thrughly evaluated by Rth. 13 Thermdynamic data, primarily heats f frmatin and M 2 O activities, fr crystalline phases in the M 2 O-Al 2 O 3 (M Na, K systems are reprted by several investigatrs. Specific reprts and reviews are given fr Na 2 O activity measurements in - alumina tw-phase cmpsitins, 2,10,14-16 and fr K 2 O activity measurements in similar cmpsitin regins. 14,17,18 The ptimized phase diagrams published by Erikssn et al. 12 fr the Na 2 O- and K 2 O-Al 2 O 3 systems d nt extend t M 2 O mle fractins less than 0.5, but these authrs have generated thermdynamic infrmatin fr the liquid phases in the alumina-rich part f

2 B552 Jurnal f The Electrchemical Sciety, B551-B the diagrams. These cmpsitins include the sdium- r ptassiumcntaining liquid prducts that can frm when either pure -alumina r -alumina reacts with alkali vaprs. Erikssn et al. 12 used quasichemical mdel parameters t represent nnideal interactins in these liquid phases. Hwever, these parameters are nt easily adapted t the mdified assciate species mdel we are using fr liquid slutin phases. 3 In this paper, we present a re-evaluatin f the thermdynamic prperties f the M 2 O-Al 2 O 3 binary systems (M Na r K t determine self-cnsistent sets f accurate thermdynamic infrmatin. The beta-aluminas, MAl 9 O 14 -alumina and M 2 Al 12 O 19 ( -alumina, are likely prducts f the crrsin f -alumina by MOH vaprs present in glass furnaces. Cnditins required fr reactins f MOH g with Al 2 O 3 refractries in representative air- and xygen-fired cmbustin atmspheres were determined. Our fcus is n reactins invlving MOH g since the hydrxide is the primary alkali-cntaining gaseus species; M g pressures are significantly smaller than the MOH g pressures in these furnace atmspheres. In additin t discussing the temperature dependence f MOH g and M g cncentratins, we present graphs fr bth alkali-cntaining systems, frm which a critical temperature fr crrsin can be determined fr any MOH g partial pressure in the furnace atmsphere between 0 and 1000 ppm. Due t the high liquidus temperatures f 2150 K in the - regins f bth the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 systems, the frmatin f lw-viscsity liquids is unlikely under cnditins typical f mst air- and xy-fuel glass melting furnaces. Equilibrium Mdeling Apprach T mdel alumina crrsin at equilibrium, ne needs t knw the partial pressure f H 2 O(g) (p H2 O), the activities f M 2 O(a M2 O) as a functin f temperature and cmpsitin, and the temperaturedependent equilibrium cnstants K eq fr all phase regins in the M 2 O-Al 2 O 3 systems. Thus, thermdynamic data such as heats f frmatin, entrpies, heat capacities, heats f transitin, and activities are needed fr all gaseus and cndensed species in the systems, including the crystalline and liquid phases that may be frmed by the crrsin prcess. The required infrmatin fr mst gas-phase species was btained frm the assessed SGTE database. 19 Hwever, the H f,298 value fr NaOH g in this database kj/ml is quite different frm that fund in the JANAF Thermchemical Tables kj/ml. 20 The crrespnding values f the entrpy (S 298 ) frm these tw surces are quite similar, hwever, be- ing and J/ml K, respectively. The H f,298 and S 298 values fr slid NaOH s frm the tw surces are als essentially the same. Thus, the 12 kj/ml difference in the H f,298 values fr NaOH g appears t be the nly majr uncertainty in the sdium hydrxide data used in ur calculatins. In the calculatins described here, we chse t use the JANAF value fr H f,298 (NaOH,g), fr reasns that are discussed in detail in a previus cmmunicatin. 3 Tabulated thermdynamic prperties fr liquid-phase crrsin prducts are lacking, s these values were determined alng with an assessment and ptimizatin f the values fr crystalline phases in bth the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 binary systems. The cmputer prgram ChemSage 21 was ur primary tl fr develping an assessed, internally cnsistent thermdynamic database fr these tw M 2 O-Al 2 O 3 systems and fr perfrming subsequent calculatins f the high-temperature crrsin reactins between their MOH g vaprs and Al 2 O 3 (s). The fllwing paragraphs describe ur analysis f available thermchemical infrmatin fr these systems and ur chices fr the values used in the equilibrium calculatins described later in this paper. In develping self-cnsistent sets f thermdynamic prperties fr each system, we mdeled the thermdynamic stabilities f M 2 O-cntaining liquid-xide slutins using a mdified assciate species apprach described previusly. 22 The negative free-energy Table I. Typical input cnditins fr equilibrium calculatins. P ttal Ä 1 bar, T Ä K C. n(ch 4 ) a n(o 2 ) a n(n 2 ) a n(m 2 O) a,c a refractry b Calculatin f p H2 O Air-fired Oxygen-fired Crrsin predictins Air-fired Oxygen-fired Air-fired Oxygen-fired a Mles. b Activity f either -Al 2 O 3 r -alumina was fixed at unity. c M Na r K. terms caused by nnideal mixing f end-member liquid cmpnents in a system are cntained in the frmatin energies f the liquid assciate species in this mdel. Fr the tw systems examined in this paper, nly ideal mixing f assciate liquid species was needed t mdel the tw respective liquid phases; n nnideal slutin parameters were needed. In bth M 2 O-Al 2 O 3 binary systems, we mdeled the liquid by using fur liquid assciate species f M 2 O(l), MAlO 2 (l), (1/3)M 2 Al 4 O 7 (l), and Al 2 O 3 (l). T prvide equal weighting f liquid species, the cmpsitin f each liquid assciate has a ttal f tw nn-xygen atms in its frmula. While these fur liquid species may nt exist as chemical entities that can be islated and characterized, their frmatin energies and the ideal slutins cmprised f them accurately represent the negative interactin energies that ccur between M 2 O and Al 2 O 3 in these tw alkali liquid xide slutins. The ternary Na 2 O-K 2 O-Al 2 O 3 is nt examined in this paper. The thermdynamic data fr the binary M 2 O-Al 2 O 3 systems were assessed and ptimized by perfrming a manual thermdynamic fitting f the binary equilibrium phase diagram fr the tw alkali-cntaining systems see Ref. 12, 13, 23, and the references cited therein fr the latest phase diagram data. Results were als cmpared with published thermdynamic activity values and assessments fr the systems see Ref. 2, 10, 12, This prcedure prvides a means f testing and generating a set f self-cnsistent thermdynamic infrmatin fr the systems, including data fr the xide liquid phases. 3,22 Using the resulting thermdynamic prperties, we calculated the equilibrium cmpsitins f the tw systems using input cnditins typical f air- and xy-fuel glass melting furnaces. Input parameters used t perfrm the equilibrium calculatins are given in Table I. These cnditins were chsen s that either crystalline - r -alumina is always present, as indicated in the last clumn f Table I, which ensures the existence f the chsen crystalline alumina refractry at equilibrium. All calculatins were perfrmed at a pressure f 1 bar. The majr species included in the varius calculatins are given in Table II, which als includes the surce fr the thermdynamic data in each case. Table III gives thermdynamic data fr imprtant species used in the calculatins in the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 systems. Results Na 2 O-Al 2 O 3 phase diagram and Na 2 O activities. The calculated phase diagram resulting frm ur assessed dataset is shwn in Fig. 1a fr the Na 2 O-Al 2 O 3 system. This diagram is generally in gd agreement with the assessed diagrams given by Rth 13 and Erikssn et al. 12 except fr a slight difference in the melting temperature fr NaAl 9 O 14, fr which there is little experimental infrmatin. The thermdynamic stability f the liquid is changed very little by the differences. In Fig. 1b, we als cmpare ur calculated Na 2 O activities fr the -Al 2 O 3 - -NaAl 9 O 14 tw-phase regin with the mre recent values reprted in the literature see Ref. 2, 10,

3 Jurnal f The Electrchemical Sciety, B551-B B553 Table II. Majr species used in calculatins and surces f thermdynamic data. Gas-phase a N 2 H 2 O CH 4 O 2 NO H 2 OH CO 2 NO 2 N 2 O H CO HCN NH 3 Na NaOH b K KOH Liquid phase c Na 2 O(l) NaAlO 2 (l) (1/3)Na 2 Al 4 O 7 (l) Al 2 O 3 (l) K 2 O(l) KAlO 2 (l) (1/3)K 2 Al 4 O 7 (l) Slid Phase d Na 2 O NaAlO 2 Na 2 Al 12 O 19 ( ) NaAl 9 O 14 ( ) Al 2 O 3 ( ) K 2 O KAlO 2 K 2 Al 12 O 19 ( ) KAl 9 O 14 ( ) Na 2 CO 3 K 2 CO 3 a A ttal f 75 gaseus species were initially used in these calculatins. The majr surce f these data was SGTE Ref. 19 except as nted. b JANAF Ref. 20. c The set f liquid assciate species was develped in the current studies. d Additinal fixed-cmpsitin nitride and carbide phases were initially used in these calculatins, but were always at very lw activities. The surce f liquid- and slid-phase thermdynamic data is ur current assessment f the type described in Ref. 22, using initial thermdynamics values frm the literature Ref. 2, 10, 12, , 15. Nafe, 10 Petric and Chatilln, 15 and Jacbs et al. 2 als shw graphs including ther reprted activity values alng with their wn respective measurements. Except fr the high-temperature mass spectrmetric vapr pressure measurements f Petric and Chatilln, all ther activity measurements were made using lwer temperature slid-state electrlyte galvanic cells. K 2 O-Al 2 O 3 phase diagram and K 2 O activities. Our calculated phase diagram fr the K 2 O-Al 2 O 3 system is shwn in Fig. 2a, and is als in gd agreement with the assessed diagram given by Erikssn et al., 12 except fr a slight difference in the melting temperature f KAl 9 O 14, fr which there is little experimental infrmatin. The thermdynamic stability f the liquid is changed very little by the differences. As discussed by Erikssn et al., 12 the very existence f the -alumina (K 2 Al 12 O 19 ) is uncertain, let alne its peritectid decmpsitin temperature. Erikssn et al. 12 shw the -phase melting by a peritectic reactin at 1920 C t give -alumina and liquid mle fractin X(Al 2 O 3 ) 0.75]. In cntrast, ur calculated diagram has a cngruently melting -phase at 1989 C and an -alumina/ -alumina eutectic at 1984 C, as shwn in Fig. 2a. We als cmpared ur calculated K 2 O activities fr the -Al 2 O 3 - -KAl 9 O 14 tw-phase regin with thse reprted in the literature see Ref. 14, 17, 18 ; the results are shwn in Fig. 2b. Pssible reasns fr the high values f Kumar and Kay 14 have been thrughly discussed by Kale and Jacb. 17 The reprted experimental studies in this system all invlved slid-state electrchemical cells. The phase diagrams in Fig. 1a and 2a clearly shw why aluminabased ceramics are mre resistant t crrsin by NaOH r KOH than silica. 3 In cntrast with silica crrsin, the frmatin f liquidphase prducts relatively rich in M 2 O can nly ccur at very high temperatures in the M 2 O-Al 2 O 3 systems. Such prducts are nt present until the temperature reaches 1857 K 1584 C fr the sdium case, and 2185 K 1912 C fr the ptassium case, where -alumina reacts t frm a liquid. In the Na 2 O-SiO 2 system, liquids can frm at temperatures as lw as 1090 K. 3 Pure -alumina in cntact with MOH g cannt frm a liquid until the temperature reaches 2158 K fr M Na, r 2257 K fr M K. Crrsin mdeling apprach. We nw discuss the chemical reactins invlved in the crrsin f - and -alumina by NaOH and KOH. Equilibrium calculatins by urselves and thers 3,24,25 Figure 1. a Calculated phase diagram fr the Na 2 O-Al 2 O 3 system. Nte that temperatures are given in C fr ease f cmparisn with published phase diagrams; mst temperatures in this article are given in K. Current temperatures are shwn, alng with previusly reprted values in parentheses; Ref. 12, Erikssn et al. b Calculated ln a(na 2 O) vs /T in the -alumina/ -alumina tw-phase regin in the Na 2 O-Al 2 O 3 system. shw that NaOH g is the mst abundant sdium-cntaining gas when the cmbustin atmsphere equilibrates with sdiumcntaining glass melts. This is als true fr KOH g in glass melts cntaining ptassium. Therefre, we assume that NaOH g is the key sdium-cntaining species participating in the crrsin f alumina, and that KOH g is the key ptassium-cntaining species invlved. The frmatin f these hydrxides can be described by the fllwing generic chemical reactin, where M Na r K M 2 O in glass melt H 2 O g, cmbustin gas 2MOH g 1

4 B554 Jurnal f The Electrchemical Sciety, B551-B Table III. Thermdynamic data fr imprtant species in Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 systems. Species H f,298 J/ml S 298 J/ml-K T (K) Cp a bt ct 2 d/t 2 J/ml-K a b10 3 c10 6 Gases Na 107, NaOH a 197, K 89, KOH 232, Liquid assciates Na 2 O(l) 370, (1023 K) 1, H trans H trans (1243 K) 11, NaAlO 2 (l) 1,000, Na 2 Al 4 O 7 (1/3) 1,234, K 2 O(l) 327, KAlO 2 (l) 1,040, K 2 Al 4 O 7 (1/3) 1,210, Al 2 O 3 (l) 1,564, Crystalline slids Na 2 O 417, ,982 b b H trans (1023 K) 1, H trans (1243 K) 11, NaAlO 2 1,111, ,119, Na 2 Al 12 O 19 10,685, ,702, NaAl 9 O 14 7,873, ,873, K 2 O 361, , KAlO 2 1,144, ,145, K 2 Al 12 O 19 10,740, ,788, KAl 9 O 14 7,900, ,916, Al 2 O 3 1,675, ,675, a Data frm Ref. 20; see text fr discussin f uncertainties in NaOH g data. b Values fr H f,298 and S 298 in italics are frm Erikssn et al. 19 d10 5 At the surface f the refractry, the MOH can then react t frm several prducts, as is evident frm the phase diagrams in Fig. 1a and 2a. The activity f cndensed M 2 O in the glass, a(m 2 O), is determined by the temperature and cmpsitin f the glass melt. The p H2 O value expected in the cmbustin atmsphere f an airfired furnace is fixed by the fuel/air rati at 0.18 bar in the calculatins reprted here, p H2 O was calculated frm the fuel/xidizer ratis given in Table I. Anecdtal reprts indicate that the resulting NaOH g pressure is t bar ppm fr typical sda-lime-silica glass melting. The crrespnding values fr

5 Jurnal f The Electrchemical Sciety, B551-B B555 Fr crrsin f alumina by NaOH g t ccur, the phase diagram in Fig. 1a shws that several reactins are pssible. First, if NaOH g reacts with -alumina, crystalline -alumina can frm at T 2158 K 1885 C; Reactin 2, belw. AtT 2158 K, the NaOH g wuld react t frm an alumina-rich liquid slutin Reactin 3, belw T 2158 K 2NaOH g 9Al 2 O 3 2NaAl 9 O 14 H 2 O g, cmbustin gas 2 T 2158 K 2NaOH g Na 2 O in Al 2 O 3 -rich liquid H 2 O g, cmbustin gas 3 Since the equilibrium value f p H2 O in a melting furnace is fixed by the fuel/xidizer rati, the p NaOH in equilibrium with bth -alumina refractry and a crrsin prduct Reactins 2 r 3 is fixed at ne value fr a given temperature and partial pressure f water. Similarly, fr Reactin 3 under these cnditins, the cmpsitin f the Al 2 O 3 -rich liquid crrsin prduct is als determined. If the cmbustin atmsphere has a larger p NaOH than the equilibrium value fr Reactin 2, then this reactin prceeds t the right and -alumina (NaAl 9 O 14 ) can frm. Likewise, if the cmbustin atmsphere has a larger p NaOH than the equilibrium value fr Reactin 3, then this reactin prceeds t the right and an Al 2 O 3 -rich liquid can frm. Secnd, NaOH g can react with -alumina. At temperatures abve 1857 K 1584 C; the eutectic pint between NaAlO 2 and -alumina and Al 2 O 3 mle fractins 0.68 X Al2 O 3 0.9, frmatin f liquid prducts can ccur, with the liquid being in lcal equilibrium with -alumina. NaOH g can als react with -alumina t frm -alumina at T 1716 K 1443 C; Reactin 4, r NaAlO 2 at temperatures between 1716 and 1857 K 1443 and 1584 C; Reactin 5. 4NaAl 9 O 14 2NaOH g 3Na 2 Al 12 O 19 H 2 O g 4 NaAl 9 O 14 8NaOH g 9NaAlO 2 4H 2 O g 5 Figure 2. a Calculated phase diagram fr the K 2 O-Al 2 O 3 system. The 1 bar gas phase that can ccur at temperatures abve 1500 C at the K 2 O bundary f this diagram is nt shwn. Nte that temperatures are given in C fr ease f cmparisn with published phase diagrams; mst temperatures in this article are given in K. Current temperatures are shwn, alng with previusly reprted values in parentheses; Ref. 12, Erikssn et al.* Fr explanatin f melting in the -alumina/ -alumina tw-phase regin, see Results sectin f text. b Calculated ln a(k 2 O) vs /T in the -alumina/ alumina tw-phase regin in the K 2 O-Al 2 O 3 system. xy-fired furnaces are 0.65 bar fr the fixed partial pressure f H 2 O(g) and t bar ppm fr NaOH g ; these results are cnsistent with NaOH measurements by Walsh et al. based n atmic-sdium absrptin spectrscpy. 26 Fr purpses f illustratin in the remainder f the paper, we use 50 ppm as a representative p NaOH fr air-fired furnaces and 200 ppm fr xy-fired furnaces. We nte, hwever, that NaOH g partial pressures are expected t vary frm furnace t furnace. Thus, the thermdynamic ptential fr crrsin in a specific cmbustin envirnment can be evaluated frm plts f the equilibrium NaOH g pressure given belw. Similar arguments als hld fr the case f ptassium in glass melting furnaces. The K 2 O-Al 2 O 3 phase diagram in Fig. 2a shws that if K 2 O reacts with -alumina, crystalline -alumina in this system can frm at T 2257 K 1984 C; Reactin 6, belw. AtT 2257 K, the K 2 O wuld react t frm an alumina-rich liquid slutin Reactin 7, belw T 2257 K 2KOH g 9Al 2 O 3 2KAl 9 O 14 H 2 O g, cmbustin gas T 2257 K 2KOH g K 2 O in Al 2 O 3 -rich liquid H 2 O g, cmbustin gas Reactins 6 and 7 have nly ne equilibrium value fr p KOH at a given temperature and partial pressure f water if -alumina is present at equilibrium. Fr Reactin 7 under these cnditins, the cmpsitin f the Al 2 O 3 -rich liquid is als determined. If the cmbustin atmsphere has a larger p KOH than the equilibrium value fr Reactin 6, then this reactin prceeds t the right and -alumina can frm. Similarly, if the cmbustin atmsphere has a larger p KOH than the equilibrium value fr Reactin 7, then this reactin prceeds t the right and an Al 2 O 3 -rich liquid can frm. 6 7

6 B556 Jurnal f The Electrchemical Sciety, B551-B At temperatures abve 2185 K 1912 C; the eutectic pint between KAlO 2 and -alumina and 0.75 X Al2 O 3 0.9, liquid frmatin can als ccur with the liquid being in lcal equilibrium with -alumina. Finally, KOH g can react with -alumina t frm -alumina at T 1435 K 1162 C; Reactin 8, rkalo 2 at temperatures between 1435 and 2185 K Reactin 9 4KAl 9 O 14 2KOH g 3K 2 Al 12 O 19 H 2 O g 8 KAl 9 O 14 8KOH g 9KAlO 2 4H 2 O g The MOH g pressures required fr Reactins 4, 5, 8, and 9 t ccur can als drive the fllwing reactin with the CO 2 (g) in the cmbustin envirnment 2MOH g CO 2 g M 2 CO 3 H 2 O g 10 The partial pressures f all the gaseus species in Reactin 10 are fixed by the temperature and respective cmpsitins f the input fuel/xidizer rati and the glass system being melted. Als, in the burning f natural gas by the reactin CH 4 g 2 O 2 g CO 2 g 2H 2 O g 11 9 it is seen that the H 2 O:CO 2 rati is 2:1, independent f whether the surce f xygen is air r pure xygen. Thus, the MOH g partial pressure in the furnace atmsphere is the variable that determines if Reactin 10 is thermdynamically favrable. Whether r nt the alkali carbnate cndensed phase can frm is determined by the equilibrium cnstant f Reactin 10, its cmpetitin with Reactins 4, 5, 8, and 9 and the partial pressures in the furnace atmsphere. Nte, hwever, that the temperatures required fr carbnate frmatin t ccur 1380 K are nt likely t be encuntered by refractries in glass-melting furnaces. Thus, the ptential fr carbnate frmatin, while required fr a cmplete thermdynamic treatment, shuld have n significant impact n alumina crrsin. Equilibrium predictins f alumina crrsin regimes. The cnditins in glass-melting furnace atmspheres that culd lead t crrsin f - r -alumina refractries can be determined by perfrming fixed-temperature equilibrium calculatins, using input cmpsitins and a range f temperatures representative f the furnace cmbustin atmsphere Table I. Selected results frm such calculatins fr the -alumina/ -alumina tw-phase regins f the Na 2 O-Al 2 O 3 and K 2 O-Al 2 O 3 systems are shwn in Fig. 3a and b, which display the partial pressures f MOH g and M g (p MOH and p M ). These are the partial pressures that exist when the gas is in equilibrium with a cndensed mixture cntaining bth - and -alumina phases. Similar plts are given in Fig. 4a and b fr the -alumina/m 2 O-rich phase. Several general bservatins fr bth the sdium and ptassium systems can be made frm the results in Fig. 3 and 4. First, in their -alumina/ -alumina tw-phase regins Fig. 3a and b, the partial pressure f MOH g always exceeds that f M g by a substantial factr under all cnditins, cnfirming that MOH g is indeed the equilibrium species that shuld be f cncern in crrsin. Secnd, the predicted equilibrium MOH g partial pressures are always abut a factr f tw higher in xy-fired furnace atmspheres than thse predicted fr air-fired furnaces, which, as discussed previusly, 3 is simply a result f the differences in H 2 O(g) partial pressures in these tw atmspheres fr example, at 1873 K xyfired p H2 O bar and air-fired p H2 O bar). Frm Reactin 1 r 2, it is seen that the gaseus rati MOH xy /MOH air H 2 O xy /H 2 O air 1/ / / Finally, Fig. 3a and b shw that the cncentratin f bth MOH and M gaseus species increases cntinuusly with temperature up t at least 2100 K. This is quite different frm the situatin that Figure 3. a NaOH g upper curves and Na g lwer curves partial pressures in equilibrium with the -alumina/ -alumina tw-phase regin as a functin f temperature. The hrizntal lines crrespnd t the example cases fr air-fired 50 ppm, lg P(bar) 4.30] and xy-fired 200 ppm, lg P(bar) 3.70] furnaces. b KOH g and K g partial pressures in equilibrium with the -alumina/ -alumina tw-phase regin as a functin f temperature. exists in the M 2 O-SiO 2 equilibria, in which there is a maximum in these cncentratins at 1873 K in the Na 2 O-SiO 2 system, 3,24 and 1800 K fr the K 2 O-SiO 2 system. 3 These temperature maxima are a result f the changing cmpsitin f the liquid crrsin prduct in equilibrium with crystalline SiO 2, which decreases in M 2 O cncentratin with increasing temperature until n M 2 O is present in the liquid at the melting pint f SiO 2 cristbalite. As is seen belw, the fact that p MOH and p M increase cntinuusly t very high tem-

7 Jurnal f The Electrchemical Sciety, B551-B B557 Figure 4. a NaOH g partial pressures in equilibrium with the -alumina/ Na-rich phase tw-phase system as a functin f temperature. See text fr further descriptin f Na-cntaining phases as a functin f temperature. The hrizntal lines crrespnd t p NaOH example cases fr air-fired 50 ppm, lg P(bar) 4.30] and xy-fired 200 ppm, lg P(bar) 3.70] furnaces. b KOH g partial pressures in equilibrium with the -alumina/k-rich phase tw-phase system as a functin f temperature. See text fr further descriptin f K-cntaining phases as a functin f temperature. peratures in the alumina systems has cnsequences fr the crrsin behavir f alumina refractries. Nte that the pints where the NaOH vapr-pressure curves intersect the hrizntal lines in Fig. 3a and 4a representing values f p NaOH in the cmbustin atmsphere indicate the lwest temperature at which the equilibrium vapr pressure exceeds that in the furnace. This leads t the cncept f a critical temperature see belw, abve which n crrsin can ccur. Thus, the predicted equilibrium NaOH g cncentratins in these figures represent a lwer limit fr furnace cncentratins that can result in crrsin. Figures 3a and b can be used t identify regins f chemical stability when alumina is expsed t cmbustin envirnments reprentative f glass furnaces. In the case f NaOH g, Fig. 3a indicates that -alumina shuld be stable in air-fired furnaces at T 1504 K and in xy-fired furnaces at T 1564 K, since their respective equilibrium p NaOH fr the - tw-phase system is higher than that fund in air- 50 ppm, r bar) and xy-fired 200 ppm, r bar) furnace atmspheres. Fr similar KOH g cncentratins in a furnace atmsphere, -alumina shuld als be stable at temperatures Fig. 3b T 1458 K fr air-fired furnaces and T 1515 K fr xy-fired furnaces. In cntrast, ceramics cntaining sdium -alumina shuld decmpse t -alumina and NaOH g at T 1504 K air-fired and T 1564 K xy-fired, since the equilibrium p NaOH exceeds that in the furnace; this phenmenn has been bserved. 4 Ceramics cntaining ptassium -alumina culd decmpse t -alumina and KOH g when the equilibrium p KOH exceeds the values in the furnace; fr furnace KOH g pressures f 50 ppm fr air-fired and 200 ppm fr xy-fired, the respective temperatures are T 1458 K air-fired and T 1515 K xy-fired. The -alumina/m 2 O-rich-phase equilibrium is smewhat mre cmplex than the -alumina/ -alumina equilibrium, since the M 2 O-rich prduct phases in equilibrium with -alumina depend n temperature. With increasing temperature, the pssible prduct phase changes frm -alumina t MAlO 2 (s) t liquid. Figures 4a and b shw the p MOH results fr tw-phase systems when MOH g reacts with -alumina. The nature f these prducts is indicated by the pssible chemical reactins discussed in the previus sectin. The M 2 O-rich prduct phase in equilibrium with -alumina in the Na 2 O-Al 2 O 3 system is Na 2 CO 3 (s) T 1220 K air-fired r T 1380 K xy-fired. The difference in these tw temperatures is the result f the difference in the assumed p NaOH in the respective cmbustin atmspheres fr air-fired and xy-fired systems 50 ppm vs. 200 ppm p NaOH ). At higher temperatures up t 1716 K 1443 C, the peritectid temperature, -alumina frms rather than Na 2 CO 3. At even higher temperatures, up t 1857 K 1584 C, the eutectic temperature, -alumina is unstable and NaAlO 2 frms. Finally, at temperatures between 1857 and 2158 K C, a sdium-rich cmpared t -alumina liquid phase frms in equilibrium with -alumina. As is shwn in Fig. 4a fr all temperatures 1260 K the NaOH g pressures required t prduce these sdium-rich prducts are much higher than thse present in a glass-melting furnace, s such prducts are nt predicted t frm under these cnditins. The M 2 O-rich prduct phases in equilibrium with -alumina in the K 2 O-Al 2 O 3 system d nt include K 2 CO 3 (s) fr the temperatures 1200 depicted in Fig. 4b. At temperatures abve 1200 K, ptassium -alumina frms rather than K 2 CO 3 up t 1435 K 1162 C, the peritectid temperature in bth air- and xyfired envirnments. At higher temperatures up t 2185 K 1912 C, the eutectic temperature, ptassium -alumina is unstable and KAlO 2 frms. Finally, at temperatures between 2185 and 2262 K C, a ptassium-cntaining liquid phase frms in equilibrium with -alumina. As is shwn in Fig. 4b fr all temperatures 1260, the KOH g pressures required t prduce these ptassium-rich prducts are 200 ppm lg P(bar) 3.70, the value estimated fr p NaOH in xy-fired glass-melting furnaces. Thus, if KOH g levels are cmparable t thse f NaOH g, K 2 O-rich crrsin prducts are nt predicted t frm in these furnace envirnments. Figure 4a indicates that the NaOH g cncentratin in equilibrium with the -alumina/na-rich prduct liquid, NaAlO 2, r -alumina tw-phase regin is much higher than that in equilibrium with the - alumina tw-phase regin by a factr f 36 at 1800 K fr bth air- and xy-fired cnditins; at 1200 K, a factr f 200 fr air-fired, and 102 fr xy-fired atmspheres. The difference

8 B558 Jurnal f The Electrchemical Sciety, B551-B between the p NaOH ratis at 1200 K fr air-fired and xy-fired cnditins is due t the reductin in p NaOH at lw temperatures caused by Na 2 CO 3 frmatin in the xy-fired cnditins. Figure 4a als shws that the equilibrium p NaOH is always higher than typical furnace cncentratins fr T 1250 K xy-fired and T 1200 K air-fired, as indicated by the hrizntal lines in the figure. Thus, the cnversin f -alumina t -alumina r any ther Na-rich prduct phase by NaOH g is nt thermdynamically feasible, and Reactins 4, 5, and 10 shuld nt ccur in typical glass melting furnaces. Similarly, Fig. 4b indicates that the KOH g cncentratin in equilibrium with the -alumina/k-rich prduct twphase system is significantly higher than that in equilibrium with the - alumina tw-phase system ranging frm a factr f 13 at 1800 K t a factr f 85 at 1200 K fr bth air- and xy-fired cnditins. Althugh t ur knwledge n KOH g furnace cncentratins have been reprted, Fig. 4b indicates that the equilibrium p KOH is always higher than furnace cncentratins f 200 ppm fr T 1240 K xy-fired, and f 50 ppm fr T 1200 K air-fired. Thus, the cnversin f -alumina t -alumina r any ther K-rich prduct phase by KOH g is prbably nt thermdynamically feasible either again assuming that KOH g cncentratins are cmparable t thse f NaOH g, and thus Reactins 8, 9, and 10 shuld nt ccur in glass melting furnaces. A final imprtant bservatin is that M 2 O-cntaining liquid alumina phases are nt thermdynamically favred under typical furnace perating cnditins, as they are when silica is used. Althugh the phase diagram fr the Na 2 O-Al 2 O 3 system Fig. 1a indicates the pssibility f a liquid phase in equilibrium with -alumina abve 1857 K 1584 C, the values f p NaOH in equilibrium with that twphase regin are always much higher than typically bserved under either air- r xy-fired cases, s that ne can safely cnclude that these liquids will never frm. The data in Fig. 3a and b can be repltted t shw temperature as a functin f MOH partial pressure t define regins in which crrsin is thermdynamically favred t ccur Fig. 5a and b. In general, crrsin ccurs i.e., - is cnverted t -alumina in the regin f cnditins belw each curve. Fr example, if the p NaOH in the cmbustin atmsphere is higher than that in equilibrium with - and -alumina, the bundary line in Fig. 5a, then crrsin can ccur. The curves in Fig. 5a and b thus define a critical temperature, (T critical ) abve which crrsin des nt ccur fr a given p MOH in the cmbustin atmsphere. This cncept was intrduced previusly by Faber and Verheijn 25 and expanded upn by us in subsequent publicatins. 3,24 Abve each curve, crrsin will nt ccur, since p MOH in the furnace atmsphere is belw the equilibrium value fr the - tw-phase regin. At NaOH partial pressures representative f xy-fuel furnaces 200 ppm, the value f T critical is 1564 K, Fig. 5a. Under air-fuel cnditins, where the NaOH partial pressure is much lwer 50 ppm, T critical is predicted t be 1504 K. Nte that, since ur predicted equilibrium p NaOH are expected t be lwer limits fr the NaOH g partial pressures needed t cause crrsin, ur predicted values f T critical are actually upper limits fr temperatures at which crrsin can ccur at a given p NaOH in the furnace atmsphere. Similar arguments can be made fr Fig. 5b and KOH partial pressures in the furnace atmsphere. The fact the curves in Fig. 5 increase cntinuusly with temperature indicates that in the K temperature range there will always be a temperature abve which crrsin will nt ccur. This cntrasts with crrsin f silica refractries, fr which, because f a maximum in the p MOH vs. T critical curves ccurring at 1800 K, there exist furnace cnditins in which crrsin is always thermdynamically favred. A similar maximum als ccurs in the M 2 O-Al 2 O 3 systems, but nly at high temperatures where -alumina can exist in equilibrium with a liquid phase, between 2158 and 2327 K fr the sdium system, and between 2257 and Figure 5. a Znes f alumina refractry crrsin by NaOH g as defined by T critical see text fr the -alumina/ -alumina tw-phase regin. The NaOH cncentratins are given in units f ppm parts per millin at 1 bar since this is the cmmn measure f mst glass manufacturers we use the cnversin f 1 ppm bar). b Znes f alumina refractry crrsin by KOH g as defined by T critical see text fr the -alumina/ alumina tw-phase regin. The KOH cncentratins are given in units f ppm parts per millin at 1 bar K fr the ptassium system. Therefre, a maximum in the p MOH vs. T critical curves fr the alumina systems wuld never be bserved under practical glass-melting cnditins. The results presented abve suggest that there may be several advantages t using high-purity -alumina as a crwn material. First, T critical (xy) f 1564 K fr 200 ppm NaOH g in the furnace atmsphere is cnsiderably belw the perating temperatures typically used fr silica crwns maximum cntinuus perating temperatures K, indicating that -alumina shuld be thermdynamically stable in this temperature regime. Secnd, replacement f silica with alumina may prvide additinal flexibility

9 Jurnal f The Electrchemical Sciety, B551-B B559 with regard t burner ptimizatin. Burner placement has been shwn t affect crwn temperatures; placing burners higher n the furnace wall which may reduce NaOH and KOH vlatilizatin rates increases crwn temperatures in sme lcatins. 5 Since the temperatures at which alumina either melts r frms liquid crrsin prducts are much higher than 1873 K Fig. 1a and b, it can tlerate placement f burners in lcatins clser t the crwn. Finally, the frmatin f liquid sdium silicates frm the crrsin f silica refractries is much mre deleterius t the refractry than the frmatin f -alumina frm the crrsin f -alumina. The mechanical prperties f crrded alumina refractries are evidently sufficient t inhibit structural damage. 4 Cnclusins The thermdynamic calculatins reprted here, using newly generated data fr sdium- and ptassium-cntaining alumina phases, indicate that alumina is quite stable with respect t crrsin by MOH g under the cnditins typical f glass-melting furnaces. Unlike silica, liquid-phase alkali aluminates are nt stable except at very high temperatures and in equilibrium with extremely high MOH g partial pressures, indicating that these liquid prducts will nt frm. Thus, prblems assciated with frmatin f liquid crrsin prducts, such as run ff f crrsin prducts int the melt and presumably, unacceptably high crrsin rates as is the case fr silica, shuld nt be a prblem fr crwns cnstructed f either r Na -alumina used in xy-fuel melting furnaces. The negative aspect f using alumina refractries instead f silica is that they are heavier, s that glass-melting furnaces require special cnstructin. They are als mre expensive than silica, raising the capital csts f such furnaces. Acknwledgments The authrs are grateful t the fllwing rganizatins fr their financial supprt f this prject: the U.S. Department f Energy DOE, Office f Industrial Technlgies, Glass Industry f the Future Team; American Air Liquide; BOC Gases; PPG Industries, Inc.; Praxair, Inc., Techneglas; Visten Autmtive Systems; and the DOE Envirnmental Management Science Prgram funded by the Office f Envirnmental Management s Office f Science and Technlgy, and administered jintly with the Office f Energy Research under cntract DE-AC05-00OR22725 with UT-Battelle, LLC. Sandia Natinal Labratries assisted in meeting the publicatin csts f this article. References 1. Crrsin f Materials by Mlten Glass, G. A. Pecrar, J. C. Marra, and J. T. Wenzel, Editrs, Vl. 78, American Ceramic Sciety, Westerville, OH K. T. Jacb, K. Swaminathan, and O. M. Sreedharan, Electrchim. Acta, 36, M. D. Allendrf and K. E. Spear, J. Electrchem. Sc., 148, B H. T. Gdard, L. H. Ktacska, J. F. Wsinski, S. M. Winder, A. Gupta, K. R. Selkregg, and S. Guld, Ceram. Eng. Sci. Prc., 18, K. T. Wu and H. Kbayashi, in Crrsin f Materials by Mlten Glass, Vl. 78, G. A. Pecrar, J. C. Marra, and J. T. Wenzel, Editrs, p. 205, American Ceramic Sciety, Westerville, OH J. A. M. van Hek, F. J. J. van L, and R. Metselaar, J. Am. Ceram. Sc., 75, and references therein. 7. G. Dewith, P. J. Vrugt, and A. J. C. Vandeven, J. Mater. Sci., 20, S. K. Sundaram, J. Y. Hsu, and R. F. Speyer, J. Am. Ceram. Sc., 78, S. K. Sundaram, J. Y. Hsu, and R. F. Speyer, J. Am. Ceram. Sc., 77, H. Näfe, J. Electrchem. Sc., 143, A. J. Faber and O. S. Verheijen, Ceram. Eng. Sci. Prc., 18, G. Erikssn, P. Wu, and A. D. Peltn, CALPHAD: Cmput. Cupling Phase Diagrams Thermchem., 17, R. S. Rth, Adv. Chem., 186, R. V. Kumar and D. A. R. Kay, Metall. Mater. Trans. B, 16, A. Petric and C. Chatilln, in High Temperature Materials Chemistry IX, K.E. Spear, Editr, PV 97-39, p. 718, The Electrchemical Sciety Prceedings Series, Penningtn, NJ M. Barsum, J. Mater. Sci., 25, G. M. Kale and K. T. Jacb, Metall. Mater. Trans. B, 20, M. Ith and Z. Kzuka, J. Am. Ceram. Sc., 71, C G. Erikssn and K. Hack, SGTE Pure Substance Database, 1996 Versin, prduced by the Scientific Grup Thermdata Eurpe and btained thrugh GTT Technlgies, Herzgnrath, Germany M. J. Chase Jr., J. Phys. Chem. Ref. Data, Mngraph 9, 1, NIST-JANAF Thermdynamic Tables, 4th ed ChemSage 4.1, GTT Technlgies, Herzgnrath, Germany K. E. Spear, T. M. Besmann, and E. C. Beahm, MRS Bull., 24, Phase Diagrams fr Ceramists, Vl. 1-12, The American Ceramic Sciety, Westerville, OH K. E. Spear and M. D. Allendrf, in High Temperature Crrsin and Materials Chemistry, The Per Kfstad Memrial Sympsium, M. McMallan, E. Opila, T. Manayama, and T. Narita, Editrs, PV 99-38, p. 439, The Electrchemical Sciety Prceedings Series, Penningtn, NJ A. J. Faber and I. O. S. Verheijen, Reprt NCNG-Prject: Reductin f Refractry Crrsin-Phase 1, TNO Institute f Applied Physics S. G. Buckley, P. M. Walsh, D. W. Hahn, and R. J. Gallagher, Ceram. Eng. Sci. Prc., 21,