In-situ synchrotron microtomography reveals multiple reaction pathways during soda-lime glass synthesis E. Gouillart,1 M. J. Toplis,2 J. Grynberg,3 M.-H. Chopinet,3 E. Sondergard,3 L. Salvo,4 M. Su´ery,4 M. Di Michiel,5 and G. Varoquaux6 1Surface du Verre et Interfaces, UMR 125 CNRS/Saint-Gobain, 93303 Aubervilliers, France 2IRAP (UMR 5277, CNRS/University of Toulouse III), Observatoire Midi Pyr´en´es, 14, Ave. E. Belin, 31400, Toulouse, France. 3Surface du Verre et Interfaces, UMR 125 CNRS / Saint-Gobain, 93303 Aubervilliers, France 4SIMaP, UMR CNRS 5266, Grenoble INP, UJF, GPM2, BP 46, 38402 Saint-Martin d’H`eres Cedex, France 2 1 5ESRF, 156 rue des Martyrs, BP 220, 38043 Grenoble Cedex 9, France 0 6PARIETAL (INRIA Saclay - Ile de France), NeuroSpin CEA Saclay, Bat 145, 91191 Gif-sur-Yvette France 2 Ultrafast synchrotron microtomography has been used to study in-situ and in real n timetheinitialstagesofsilicateglassmeltformationfromcrystallinegranularrawma- a terials. Significant and unexpected rearrangements of grains occur below the nominal J eutectictemperature,andseveraldrasticallydifferentsolid-statereactionsareobserved 1 to take place at different types of intergranular contacts. These reactions have a pro- 1 found influence on the formation and the composition of the liquids produced, and control the formation of defects. ] t f o I. INTRODUCTION chotronRadiationFacility(ESRF).WeusedwhiteX-ray s radiationwithapeakphotonenergyof40keV.Thespa- . t tial resolution was 1.6µm. A mixture of Ronceveaux sil- a Manycommonplacematerialsaremanufacturedfroma m loose packing ofcoarse reactivegrains. Among such ma- ica sand, Solvayr sodium carbonate and Saint-Germain calcium carbonatewas poured into a 2-mm-diameter sil- - terials, window-glass production relies on synthesis from d amixtureofquartzsand,sodiumcarbonate,andcalcium ica crucible. Weight percentagesofthese three materials n were 64, 19 and 17 respectively, and eachmaterial had a carbonate. Industrialsynthesisofgood-qualitysoda-lime o glass is generally carried out at 1400−1500◦C, despite characteristic grain-size of: 160−200µm, 250−320µm, c and 80 − 100µm for sand, sodium carbonate and cal- [ the fact thattypicalcompositionsarecompletely molten at1050◦C.Suchhightemperaturesarerequiredtoelimi- ciumcarbonaterespectively(seeFig.1aandSupplemen- 1 tary Movie). Such grain sizes are consistent with those natedefects(gasbubblesandunmoltensandgrains),and v used by the glass making industry. The sample was first to homogenize the melt. Parameters such as the grain 9 heated from room temperature to 740◦C at a rate of 70 size of raw materials are known to influence the qual- 8 ◦ C/min, during which rapid thermal dilation prevented 3 ity of glasses [1–4]. However, little direct information is imaging. Abovethattemperature,thesamplewasheated 2 available concerning the complex interplay between the ◦ ◦ . geometryofthesystemandtherateofchemicalreactions at5 C/minupto1100 C.3-Dscansofthe samplewere 1 acquired every 2.6◦C, providing a direct picture of the 0 taking place during the earliest stages of melting. 2 In this respect, in-situ high-temperature tomographic evolution of the microstructure of the system, as illus- trated in Fig. 1(a-d) (see also Supplementary Movie). 1 X-ray imaging [5] is a potentially powerful technique as v: it provides the possibility to: a) quantitatively describe In order to obtain quantitative information, 3-D vol- umes have been processed and segmented (see Supple- i the distribution of grains and the nature of solid-solid X mentary Materials). All of the 175 grains in the ana- contactsinthe initialpile; b)identify whereandatwhat lyzed volume at room temperature have been identified r temperature reactions between grains occur; c) quantify a (seeFig.1a). Sodium carbonategrainsarecharacterized theextenttowhichthesereactionsoccur. Thistechnique by their large internal porosity (red grains in Fig. 1a). has been employed in the past to study the evolution of powder compacts during the sintering of glass beads [6] or metallic powders [7]. However, no tomographic data III. RESULTS AND DISCUSSION have been acquired concerning a high temperature sys- tem involving chemicalreactions between grainsand the irreversible formation of a liquid. Ourdatarevealtheimportanceofsolid-statereactions before the firstappearanceofmelts. Betweenroomtem- ◦ peratureand750 C,manysodiumcarbonategrainshave unexpectedlybrokenup(Fig.1b). Thisdramaticchange II. EXPERIMENTAL PROCEDURE inmorphologycontinueswithincreasingtemperature,as illustrated in Fig. 2 for the reaction between a sodium Anin-situ microtomographyexperimenthasbeenper- carbonate grain (in red), and two sand grains (in yellow formed on the ID15A beamline at the European Syn- andblue). Astemperatureincreases,theoriginalsodium 2 20◦C 760◦C topview 770(cid:0)C 790(cid:0) 795(cid:0) 800(cid:0) sideview (a) (b) FIG. 2. Top- and side-view of one sodium carbonate grain 860◦C 905◦C (red) and two sand grains (yellow and blue) at four different temperatures, that react together to form silicates (white). Forthesakeof clarity,neighboringgrains havenot been rep- resented. (c) (d) FIG. 3. (a-c) Evolution of two neighboring grains of sodium carbonate (large grain at the center) and calcium carbonate FIG. 1. (a) Slice through the reconstructed 3-D volume ◦ (small grain on the right) at room temperature, 760 C, and of the pile of raw materials at room temperature. The seg- ◦ 825 C. The calcium carbonate grain (delineated in yellow in mented volumes of sand, sodium carbonate and calcium car- (b) and indicated by the presence of additional facets) is in- bonate grains have been coloured in yellow, red and green corporated into its larger neighbor, suggesting the formation respectively. Thetop-rightquarterhasbeenleftuncoloredto ◦ of a double carbonate. At 825 C, the eutectic melting of showtheoriginalslice. (b-d)Slicethroughthesameplaneas ◦ ◦ ◦ sodium carbonate and the double carbonate produces a car- in (a), at 760 , 860 , and 905 C. Fragmented sodium car- bonate liquid, as shown by the absence of porosity on the bonate grains react with sand to produce porous crystalline right part of thegrain. silicates (panel b), and rare liquid bridges (white arrow in panel b). Furtherincrease in temperature leads to formation ofasignificantproportionofliquid(panelc),finallyreaching astate inwhich thethegranular packingistransformed into tioned by the proximity of sand and calcium carbonate a viscous melt with bubblesand grain inclusions (d). to sodium carbonate. Consideration of the statistics of contacttypesdemonstratesthatintheinitialpacking,28 sandgrains(outof108)and18calciumcarbonategrains carbonategrainshrinks,withtheappearanceofajagged (out of 56) did not have any contact with a sodium car- and fuzzy reaction layer (shown in white). This newly bonate grain. The volumes of 47 individual silica grains ◦ formed material sticks to either of the adjacent sand have been comparedat roomtemperature and at 760 C grainsandtheresidualcarbonategrainmoves,seemingly (Fig. 4a). These grains are a mixture of grains with and attracted to unreacted parts of the sand grains. A large without contacts with sodium carbonate. We find that fraction of the sand grains is rapidly covered in a loose a little over 30 of these grains show a volume increase of shell of the new high porosity material (Fig. 2). Consid- eration of the relevant phase diagrams (see Supplemen- 0.2 0.20 tary Materials) and data from the literature including 0.08 thermogravimetricexperiments [3], in-situ X-ray diffrac- RT 0.0 0.15 Nma2eCltOs3 tciroynst[a8l]liannedsiond-isuitmuNsilMicRate[9s](aNllas2uSgiOge3statnhda/tofroNrma2aStiio2nO5o)f 1V/V760−000...462 (a)# Na2CO3 neighbors >0 c00..1005 iosnrdestpyopnesoibflesofloidr-tshtiastefraregamcetinotnatoioccnu.rFsuirntvhoelrvminogret,haestewco- 0.80 1n0o Na2C2O03 neig3h0bors 40 0.00RT(7b6)0 80N0S−NS8240 880 grain index T(◦C) carbonates. Although only a minority fraction is con- FIG. 4. (a) Relative volume growth of sand grains between cerned, some grains of CaCO3 grains are found to react ◦ ◦ room temperature and 760 C, sorted by increasing volume with Na2CO3 at temperatures below 750 C to form a loss. Full(resp. hollow)circlescorrespondtosandgrainswith new compound(Fig.3a,b). Ithasbeensuggestedinthe (resp. without) sodium carbonate neighbors. (b) Evolution literature that formationof a Na-Ca double carbonate is of thefraction cof solid sodium carbonate and silicate phase possible [10], but no direct evidence for such a reaction withincreasing temperature. Two major dropsoccuraround had been found when silica is present. the eutectic temperatures between sodium metasilicate (NS) Inspection of the different evolutions of grains shows and sodium disilicate (NS2), and the melting temperature of thatthe occurrenceofsuchsolid-statereactionsiscondi- sodium carbonate. 3 8%±2% (dotted line in Fig. 4a) corresponding to that by a layer of fine-grained porous phase (Fig. 1b). This expected from the α to β transition of quartz. On the temperature corresponds to the equilibrium eutectic be- other hand, approximatelyone thirdof the originalsand tween sodium metasilicate and sodium disilicate. The grains have experienced a volume increase smaller than formation of this sodium-rich liquid is less favorable for 8%, or even a significant volume decrease. The grains glass melting than formation of the silica-sodium disili- concerned are found to be systematically surrounded by cate eutectic, which contains more silica, thus resulting the fine grainedphase produced from sodium carbonate. ingreaterconsumptionofcrystallinesilica. These obser- It is found that at least one sodium carbonate neigh- vations therefore provide strong indirect evidence in fa- boratroomtemperatureisanecessarybutnotsufficient vor of the low temperature formation of both crystalline condition for a sand grain to have decreased in size at Na2SiO3 and Na2Si2O5. An even more abrupt decrease ◦ 760 C. Only about one half of the original sand-sodium in the fraction of the fine grained porous phase is ob- ◦ carbonate contacts are concerned by a loss in volume served at 865 C (Fig. 4b), corresponding to the melting ◦ of the sand grain at 760 C, probably because some of of sodium carbonate. The proportion of porous solids the contacts are lost when sodium carbonate starts re- drops rapidly to zero afterwards. Also, visual inspection ◦ acting and moving, as illustrated in Fig. 2. Concern- revealsthatat825 C,theporosityofthefewsodiumcar- ing calcium carbonate, given the large fraction of sand bonategrainsthathavecementedtoneighboringcalcium and the reaction-induced elimination of sodium carbon- carbonate grains (Fig. 3b) is suddenly invaded by the ate (Fig. 2), the majority of calcium carbonate grains presence of a liquid (Fig. 3c). This temperature agrees are not in contact with sodium carbonate. Such grains wellwiththeincongruentmeltingtemperatureoftheNa- ◦ isolated from Na2CO3 are observed to simply lose their Ca double carbonate at 820 C [10, 13]. CO2, producing refractory grains of CaO (Fig. 1d), in Finally, our data offer insights into the generation of contrast to the reactive pathway shown in Fig. 3. defects. Quartzgrainsnotsurroundedbyporoussodium ◦ carbonateorsilicatesbelow820 Carefoundtohaveap- ◦ The appearance of the first melts appears to be deter- proximately the same size at 950 C as at room temper- ◦ mined by the solid-state reactions. Even at 760 C cer- ature. It is undoubtedly this population of grains that tain sand grains show textural evidence for the presence will remain as high temperature solid defects because of of a small amount of liquid (highlighted by the arrows slow local diffusion. In the same way, enhancing the for- in Fig. 1b), despite the fact that the lowest stable eu- mation of Na-Ca double carbonate will act to eliminate tectic in the Na2O−SiO2 system is at 790◦C [11, 12]. the generation of CaO. The presence of liquid at such low temperature may be In summary,these in-situ observationscombined with explained by the existence of a metastable eutectic be- quantitative image processing revealunprecedented pro- tweensodiummetasilicateandsilica(seeSupplementary cesses of glass melting. From a physical point of view, materials). Quantification of the amount of liquid is theimportanceofdirectobservationisexemplifiedbythe not possible directly, but may be estimated indirectly unexpectedanddramaticeffectofsolid-statereactionson from the evolution of the amount of texturally distinc- the spatial distribution of sodium carbonate, changes in tive sodium carbonate and associated secondary phases microstructure that in turn lead to acceleratedreactions (e.g. Fig. 1b). The proportion of these porous phases compared to a fixed geometry. From a chemical point of ◦ increases up to 820 C (Fig. 4b), interpreted to reflect view,manyexcursionsfromoverallthermodynamicequi- the conversion of sand grains to sodium silicates. From librium are observed. Short range packing arrangements ◦ 760 to 820 C, we observe in places a scarce produc- have a profound influence on local reaction pathways, tionofliquid betweensomeof the sandgrains;neverthe- as most eloquently illustrated by the divergent fates of less, the production of crystalline phases is much more differentcalciumcarbonategrains. Theoccurenceofdif- important than the production of liquid, as shown by ferent reactions at different places strongly encourages Fig. 4b. Despite the large fraction of silica in the fi- the use of a spatially-resolved technique such as tomo- nal glass composition (75%), we do not observe any sig- graphic imaging in order to study glass melting. Other ◦ nificant formation of melts around 790 C, the eutectic metastable phenomena include the generation of liquids temperature between silica and sodium disilicate. How- well below those predicted by the equilibrium phase di- ◦ ever, at 820 C there is a sharp drop in the proportion agram, and the observation of a eutectic transition in of porous phases, and clear appearance of a significant the Na2O-SiO2 phase diagram, which is not the closest amount of liquid around sand grains previously covered eutectic to the mean composition. [1] J. Potts, G. Brookover, and O. Burch, “Melting rate of andbatchgrainsizeonhomogeneityofasoda-lime-silica soda-lime glasses as influenced bygrain sizes of raw ma- glass,” JournaloftheAmericanCeramicSociety29,197 terials and additions of cullet,” Journal of theAmerican (1946). Ceramic Society 27, 225 (1944). [3] F.Wilburn,S.Metcalfe,andR.Warburton,“Differential [2] F.Tooley andR.Tiede, “Influenceofbatchmixingtime ThermalAnalysis,DifferentialThermogravimetric Anal- 4 ysis,andHighTemperatureMicroscopyofReactionsbe- tions using in situ x-ray diffraction. Part 2. Soda-lime- tween the Major Components of a Sheet Glass Batch,” silica glass batches,” Glass Technology 45, 167 (2004). Glass Technol 6, 107 (1965). [9] A.Jones,R.Winter,P.Florian,andD.Massiot,“Tracing [4] C.ShecklerandD.Dinger,“Effectofparticlesizedistri- the Reactive Melting of Glass-Forming Silicate Batches bution on the melting of soda-lime-silica glass,” Journal by In Situ 23Na NMR,” J. Phys. Chem. B 109, 4324 of theAmerican Ceramic Society 73, 24 (1990). (2005). [5] J. Baruchel, J. Buffiere, P. Cloetens, M. Di Michiel, E. [10] M.Chopinet,E.Gouillart, S.Papin,andM.Toplis, “In- Ferrie,W.Ludwig,E.Maire,andL.Salvo,“Advancesin fluence of limestone grain size on glass homogeneity,” synchrotron radiation microtomography,” Scripta Mate- GlassTechnology-EuropeanJournalofGlassScienceand rialia 55, 41 (2006). Technology Part A 51, 116 (2010). [6] D. Bernard, D. Gendron, J. Heintz, S. Bord`ere, and [11] F. Kracek, “The system sodium oxide-silica,” The Jour- J. Etourneau, “First direct 3D visualisation of mi- nal of Physical Chemistry 34, 1583 (1930). crostructural evolutions during sintering through X-ray [12] A.Zaitsev,N.Shelkova,N.Lyakishev,andB.Mogutnov, computed microtomography,” Acta Materialia 53, 121 “Thermodynamic properties and phase equilibria in the (2005). Na2O–SiO2 system,” Phys. Chem. Chem. Phys. 1, 1899 [7] O. Lame, D. Bellet, M. Di Michiel, and D. Bouvard, (1999). “Bulk observation of metal powder sintering by X-ray [13] A. Gadalla and M. Abadir, “Calcination of sodium- synchrotron microtomography,” Actamaterialia 52, 977 calcium carbonates in air,” Industrial & engineering (2004). chemistry fundamentals 23, 220 (1984). [8] M. Dolan and S. Misture, “Analysis of glass batch reac-