Journal of Non-Crystalline Solids 403 (2014) 38–46

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids

j ourna l homepage: www.e lsev ie r .com/ locate / jnoncryso l

Cellular glass–ceramics from a self foaming mixture of glass andbasalt scoria

M. Marangoni a,⁎, M. Secco b,c, M. Parisatto c,d, G. Artioli c,d, E. Bernardo a, P. Colombo a, H. Altlasi e,M. Binmajed e, M. Binhussain e

a Department of Industrial Engineering, University of Padova, Italyb Department of Civil, Environmental and Architectural Engineering (ICEA), University of Padova, Italyc Inter-Departmental Research Center for the Study of Cement Materials and Hydraulic Binders (CIRCe), University of Padova, Italyd Department of Geosciences, University of Padova, Italye National Center for Building Systems, KACST, 11442 Riyadh, Saudi Arabia

⁎ Corresponding author.E-mail address: mauro.marangoni@unipd.it (M. Maran

http://dx.doi.org/10.1016/j.jnoncrysol.2014.06.0160022-3093/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 15 April 2014Received in revised form 25 June 2014Available online 17 July 2014

Keywords:Foaming;Cellular glass–ceramic;Basalt;Iron oxidation state

Self foaming cellular glass–ceramics were obtained by sintering mixtures of a basalt scoria and soda lime culletfor 15 min at 1050 and 1100 °C. The effect of polyvalent ions (Fe3+/Fe2+) on porosity (from 53 to 86 vol.%)and crystallization was studied for different mixtures subjected to different thermal treatments. Due to therange of mechanical strength values (crushing strength from 2 to 50 MPa) and total porosity achieved,these porous glass–ceramics could be applied as building materials, as lightweight aggregate for concreteor as lightweight panels.

goni).

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Glass foams are appreciated for their lightness and capacity of ther-mal and acoustic insulation combinedwith goodmechanical properties,far superior than those of polymeric foams. Moreover, as they areincombustible and waterproof [1], they are increasingly considered foruse as lightweight filling or insulating materials in civil engineering.

The production of glass foams commonly relies on the viscous flowsintering of fine glass powders, which creates a pyroplastic mass inturn foamed by the action of specific powder additives (foamingagents),when heating at 850–1000 °C. Foaming occurs because of therelease of CO, CO2 or SO3 gases, generated from the decomposition oroxidation of the additives. Decomposition reactions typically derivefrom the presence of carbonates or sulfates,whereas oxidation reactionsare associated to the interaction of carbon-containing species (C, SiC)with oxygen, mainly derived from the atmosphere within the sinteringfurnace [1]. In all cases, the gases may represent an environmentalproblem (toxicity, greenhouse effect, etc.), in clear contradiction withthe “green character” of glass foams, associated to the use of manydifferent types of crushed recycled glass, including glasses derivingfrom the vitrification of inorganic waste [2–7].

This study focuses on the development of glass-based foams frommixtures of recycled soda-lime glass and iron-rich basalt scoria,

exploiting oxygen as foaming gas instead of greenhouse gases. Therelease of oxygen, caused by the reduction of ferric oxide into ferrousoxide, has been already reported as effective for the development ofexpanded materials [8,9]; the present work greatly extends previousinvestigations and provides evidence for the significant tunability ofthe approach, which leads to partially crystallized products with totalporosity and crushing strength varying in a wide range of values,depending on the balance between glass and scoria or on the sinteringconditions.

2. Experimental procedure

2.1. Starting materials composition and samples preparation

Table 1 reports the chemical compositions of the starting materials,inferred from X-ray fluorescence analysis (XRF, Philips PW2400,Eindhoven, The Netherlands). Both soda-lime glass cullet (C) and basaltscoria (B) were considered after ball milling (Fritsch Pulverisette 6,Idar-Oberstein, Germany) at 400 rpm for 30 min in an agate jar andsieving to a size below 90 μm. Both powders were subjected to differen-tial thermal analysis (DTA-TGA, DSC 404, Netzsch Gerätebau GmbH,Selb, Germany, 10 °C/min heating rate).

Cullet (C) and basalt scoria (B) powders were first dry mixed indifferent proportions, as reported in Table 1 (C:B equal to 4:6, 5:5 and6:4), and then added with 7 wt.% distilled water without any binder.For each formulation, powder was uniaxially cold pressed (OMCN 159,

Table 1Chemical analysis (oxide contents inwt.% andmol%) of thewaste glass cullet, basalt scoriaand mixtures C4B6, C5B5 and C6B4.

Cullet Basalt Scoria C4B6 C5B5 C6B4

wt.% mol% wt.% mol% wt.% wt.% wt.%

SiO2 71.7 71.2 47.1 52.9 57.3 59.8 62.3TiO2 0.1 0.1 1.9 1.6 1.2 1.0 0.8Al2O3 0.7 0.4 14.7 10.5 9.1 7.7 6.3Fe2O3 0.1 0.0 12.3 4.8 7.5 6.2 5.0MnO 0.0 0.0 0.2 0.2 0.1 0.1 0.1MgO 3.3 4.9 11.0 16.1 7.9 7.2 6.4CaO 10.1 10.7 7.7 9.2 8.7 8.9 9.2Na2O 13.2 12.7 3.2 3.7 7.3 8.3 9.3K2O 0.1 0.0 1.0 0.7 0.6 0.5 0.4P2O5 0.0 0.0 0.5 0.3 0.3 0.3 0.2SO3 0.22 0.2 0.03 0.0 0.1 0.1 0.1L.O.I. 0.5 0.9 – – –

39M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

hydraulic press, Villa di Serio, Italy) in a rectangular die (50mm×34mm)at 30 MPa. The resulting green bodies were cut into square pellets(10 mm × 10 mm), dried at 80 °C overnight. Only for the mixture C5B5,additional rectangular bars were prepared and dried overnight.

2.2. Oxidation tests and firing

The oxidation of basalt scoria powders was studied on pelletsobtained by uniaxial pressing (at 40 MPa) fine powders, with no glassaddition, in a cylindrical mold (diameter of 13 mm). The pellets werefired for 30 min at 4 different temperatures (1050–1200 °C) andquenched in air. The FeO content was determined after dissolution ofthe samples in a solution of sulphuric acid (32%) and hydrofluoric acid(13.3%), and subsequent titration with potassium permanganate;Fe2O3 was obtained by difference to the total content of iron oxidemeasured by XRF (Fe2O3 tot).

The firing of sampleswas performed in two distinct ways. In the firstcase, glass/scoria samples were treated at 1050-1100 °C by directinsertion in a muffle furnace, with a holding time of 15 min. At theend of the holding time, the samples were rapidly cooled, at approxi-mately 60 °C/min, below 900 °C (with the muffle switched off andmuffle door partially open), and then naturally cooled to room temper-ature (inside the muffle). In the second case, for a selected formulation,the samples were fired at 1100 °C for 15min operating with 3 differentheating rates (10 °C/min, 20 °C/min and 40 °C/min); the cooling wasperformed in the same conditions.

2.3. Characterization

The materials were subjected to X-ray powder diffraction (XRPD)using a Bragg–Brentano θ–2θ diffractometer equipped with a real timemultiple strip (RTMS) detector (PANalytical X'Pert PRO, Almelo, TheNetherlands), employing CuKα radiation (0.15418 nm) and workingat 40 kV and 40 mA. Data acquisition was performed by operating acontinuous scan from 3.01° [2θ] to 79.99° [2θ], with a virtual step scanof 0.02° [2θ]. The diffraction patterns were analyzed by means of theX'Pert HighScore Plus 3.0 software (PANalytical), using data fromthe PDF-4 database (International Centre for Diffraction Data — ICDD,Newtown Square, PA, USA). Mineralogical quantitative phase analysis(QPA), based on the Rietveld method [10], was performed using theTOPAS software (Bruker AXS, Karlsruhe, Germany). The contents ofcrystalline and amorphous phases were determined using thecombined Rietveld–RIR method [11]. The observed patterns weremodeled through a pseudo-Voigt function, fitting the backgroundby 14 Chebyshev polynomials. For each phase, the lattice parameters,Lorentzian crystal sizes and scale factors were refined and residualpreferred orientation effects were modeled with the March Dollasealgorithm [12].

Furthermore, the basalt scoria and the cellular glass–ceramics weremicrostructurally and microchemically characterized by ScanningElectron Microscopy coupled with energy dispersive X-ray spectroscopymicroanalysis (SEM-EDS), using an instrument equipped with aLaB6 cathode, a four quadrant solid state BSE detector for imagingand a LEAP + Si(Li) detector for microanalysis (CamScan MX2500,Waterbeach, UK; EDAX, Mahwah, NJ, USA). The analytical conditionswere: accelerating voltage: 20 kV; filament current: 1.80 A; emissioncurrent: 20 μA; aperture current: 300 nA; and working distance: 20–30 mm. Qualitative interpretation of spectra and semiquantitativechemical analyses were performed through SEM Quant Phizaf soft-ware (EDAX, Mahwah, NJ, USA).

Water absorption, WAB, was determined using the boiling method,according to the UNI EN ISO10545-3 standard protocol. The geometricor bulk density, ρb, was obtained by considering the mass to volumeratio for 10 selected cellular glass–ceramics; the apparent density, ρa,and density of the solid, or true density, ρt, were evaluated by meansof a gas pycnometer (Micromeritics AccuPyc 1330, Norcross, GA),employing glass–ceramic samples “as fired” or crushed into fine pow-der, respectively. The three density values were used to compute theamounts of open and closed porosity.

Some pellets were investigated using a high resolution X-raymicro-computed tomography scanner (Bruker microCT-Skyscan1172) operating at 66 kV and 149 μA. To scan the entire objectvolume, preserving at the same time a sufficient level of spatialresolution, the acquisitions of radiographs were carried out foreach pellet in two separate scans (upper and lower) that were thenautomatically connected by the instrument software. The nominalspatial resolution (pixel size) was 5.59 μm for all the investigatedsamples. A total number of 1800 radiographs per scan were acquiredover a 360° rotation (angular step 0.2°, exposure time ranging from930 to 975 ms). At each angular position, 8 frames were collectedand averaged together in order to improve the recorded signal-to-noise ratio. The reconstruction of cross-sectional slices from 2D X-ray projections was carried out using a modified FDK algorithm [13]for cone-beam geometry, implemented in the Skyscan NReconsoftware. Corrections for the beam hardening effect and ring artifacts(i.e. circular features in the slices caused by anomalous responsesfrom some pixels of the detector) were also applied during thereconstruction process in order to improve image quality [14,15].

Crushing and four-point bending (40 mm outer span, 20 mm innerspan) tests were performed on 10 samples using an Instron 1121 UTM(Instron, Danvers, MA, crosshead speed of 1 mm/min) on pellets(10 mm × 10 mm × 8 mm) and bars (4 mm × 5 mm × 50 mm) cutfrom tile samples, respectively. Before bending tests, the bars wereused for the determination of the elastic modulus, by means of thedynamic resonance method (GrindoSonic Mk5, Leuven, B).

3. Results and Discussion

3.1. Characterization of the raw materials

The DTA and TGA analysis of cullet, shown in Fig. 1, revealed anexothermic peak at about 300 °C and a slight mass loss attributed tothe decomposition of organic impurities present in the starting scraps.From the DTA curve (Fig. 1a), the Tg of glass cullet was determined tobe ~560 °C; the broad exothermic peak at 740 °C is attributed to thesintering of fine powders (since no crystallization occurs in soda limeglass powders upon heating).

As determined bymineralogical quantitative phase analysis (Rietveldrefinements), shown in Table 2, the basalt scoria mainly consisted ofplagioclase (andesine), olivine (forsterite), clinopyroxene (augite) andspinel (titano-magnetite). The refinement was performed utilizing anandesine structure with a Ca/Na molar ratio of 0.98 [16], a ferrousforsterite structure with a Mg/Fe molar ratio of 4.48 (unit cell volumeof 294.60 Å3) [17], a titanian augite structure [18] and a titanomagnetite

Fig. 1. DTA (a) and TGA (b) curves for basalt, cullet and C5B5 mixture.

40 M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

structure [19]. A significant amorphous fraction was also present (onethird of the total mass) in the as received raw material. The Fe2+/Fe3+

ratio was estimated to be 1.50; Fe2+ ions could be located in the M1and M2 sites of augite, in the octahedral site of ferrous forsterite and inthe tetrahedral and octahedral sites of spinel, while Fe3+ ions could belocated in the M1 site of augite and in the tetrahedral and octahedralsites of spinel. SEM microstructural analyses on scoria particles arereported in the Supplementary File.

As reported in Table 2, the basalt scoria had significant variations ofboth mineralogical and chemical characteristics after firing at differenttemperatures. The XRD patterns of basalt scoria heated at differenttemperatures are reported in the Supplementary File. The DTA analysisof the basalt scoria revealed two crystallization peaks at about 960 °Cand 1050 °C, while partial melting of the scoria started around1080 °C. Indeed, the thermal treatments at 1050 °C and 1100 °C causeda decrease in glass content, more marked at lower temperatures andcounterbalanced by a significant increase in plagioclase and hematiteand a less significant clinopyroxene increase. Such analytical evidencesindicate that partial melting phenomena occur in the glass matrix andcrystallization of Al-rich silicates – e.g. plagioclase over clinopyroxene– is favored due to the peraluminous nature of the material. Further-more, the oxidizing firing conditions favored oxidation processes ofthe Fe2+ ions in the glass matrix, with consequent crystallization ofhematite. Spinel and olivine were also interested by iron oxidation

Table 2Mineralogical quantitative phase analysis of the basalt scoria at room temperature and afterfirinpatterns according to the Rietveld method (Rwp, R-factor of the weighted profile for each refinment process). FeO concentration (wt.%), Fe2+/Fe3+ ratio and forsterite unit cell volume (Å3)

Room T 105

Rwp 4.22 4.34Amorphous wt.% 32.0 (0.7) 14.1Andesine wt.% 39.9 (0.3) 50.5Augite wt.% 10.7 (0.1) 12.5Forsterite wt.% 15.1 (0.1) 13.8Spinel wt.% 2.3 (0.1) 1.1 (Hematite wt.% – 7.2 (Maghemite wt.% – 0.8 (FeO wt.% 6.64 1.08Fe2+/Fe3+ 1.50 0.11Forsterite cell volume Å3 294.60 290

phenomena [20,21]: the spinel underwent partial conversion into Ti-rich maghemite [22], while olivine underwent a depletion of structuraliron clearly testified by a d-spacing variation of (hkl) planes. For thisreason, it was referred to a new structural model, consisting of olivine[23], with a Mg/Fe molar ratio of 9.53. The structural change caused aunit cell contraction, with consequent cracking of olivine crystals andprecipitation of extracted iron in the form of iron oxides inside themicrocracks, as seen by SEM-EDS analyses (Fig. 2). A greater unit cellcontraction is observable at 1100 °C (290.46 Å3 vs 290.92Å3), indicatinga higher degree of Fe oxidation and consequent extraction from theolivine crystal structure. Such analytical evidence is consistent withprevious analytical studies and can be related to the temperature-dependent thermodynamic stability of olivines [21]. The changes in Feoxidation state are characterized by a steep drop in the Fe2+/Fe3+

ratio to a value of 0.11.After firing at 1150 °C, a partial melting of silicate phases, in partic-

ular clinopyroxene, is observable, with formation of glass phase. Thedegree of Fe oxidation is still high, as testified by the hematite formationand by the total conversion of spinel into maghemite. The residualolivine is also interested by limited Fe oxidation, as shown by thelower unit cell contraction (290.77 Å3), still related to thermodynamicfactors [21]. Nevertheless, the overall degree of Fe oxidation is lowerthan the one observed for the lower heating temperatures, as testifiedby the higher Fe2+/Fe3+ ratio of 0.19. At 1200 °C, the degree of melting

g at increasing temperature (wt.%), obtained by full profilefitting of the experimental XRDed pattern is reported, numbers in brackets are estimated absolute errors from the refine-for each temperature is also reported.

0 °C 1100 °C 1150 °C 1200 °C

4.22 2.70 2.75(0.8) 21.9 (0.6) 35.8 (0.5) 78.6 (0.2)(0.4) 47.7 (0.3) 36.4 (0.2) 12.1 (0.1)(0.1) 11.5 (0.1) 4.4 (0.1) –

(0.1) 9.7 (0.1) 13.1 (0.1) 3.5 (0.1)0.1) 0.8 (0.1) – –

0.1) 7.5 (0.1) 8.5 (0.1) 1.3 (0.1)0.1) 0.9 (0.1) 1.8 (0.1) 4.5 (0.1)

1.11 1.75 1.890.11 0.19 0.21

.92 290.46 290.77 290.80

Fig. 2. SEM-EDS microanalyses on a relict olivine crystal after firing at 1050 °C. a) BEI,showing the effect of iron oxidation (cracking of the crystal and precipitation of ironoxides); b) EDSmicroanalysis of the olivine relict, with clear iron depletion; c) EDSmicro-analysis of iron oxide dendrites (Mg and Si are due to the interaction of the electron beamwith the surrounding olivine relict).

41M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

of the silicate phases is significantly accentuated, in particular forclinopyroxene, with a relevant formation of glass phase. Furthermore,the Fe oxidation process led to a preferential crystallization ofmaghemite over hematite, being the hematite amount lower withrespect to the other samples and the maghemite presence not fullyjustifiable with the spinel oxidation process. On the whole, the degreeof Fe oxidation is lower than the other samples, as testified by the higherFe2+/Fe3+ ratio of 0.21.

All the Fe oxidation phenomena are consistentwith the TGA analysisof the scoria (Fig. 1b), which revealed a mass increase starting above700 °C, with the exception of the treatment at 1200 °C; in fact, thechemical titration revealed a decrease of the Fe2+/Fe3+ ratio from1150 °C to 1200 °C. This discrepancy between the chemical analysisand the TGA is reputed to be apparent, since it could be simply due tothe fact that the TGA was recorded by applying a heating rate of10 °C/min whereas the titration was measured for samples directlyfired at selected temperatures. Moreover, as the diffusion coefficient ofoxygen in a basalt melt is less than 1.65 · 10−6 cm2/s [24], atmosphericoxidation is not effective after the formation of themelt unless the layerof the melt is very thin.

The prediction of the redox state of iron ions into themixture is evenmore complicated by the fact that the soda lime cullet softens at arelatively low temperature, thus limiting the surface area of the basaltwhich is exposed to the atmospheric oxygen and by the fact that bothbasalt and cullet have a different redox equilibrium. The redox reactionbetween polyvalent ions, such as iron, and oxygen may be written interms of the ionic species present in the system, as proposed by severalauthors [25–27]:

FeO 3−2nð Þþn ¼ Fe2þ þ 1

4O2 þ n−1

2

� �O2−

: ð1Þ

The equilibrium between the ferrous and ferric oxide complexes isregulated by the atmospheric oxygen O2 whereas according to Toopand Samis [28,29], a “free” oxygen anion (O2−) is bonded only tomodifier ions and is related to the bridging (O0) and nonbridging oxy-gen (O−) ions by:

2O− ¼ O2− þ O0: ð2Þ

It is worth mentioning that Eq. (2) is a shorthand notation for thevarious reactions that involve more complex entities with variousproportions of bridging and nonbridging oxygens.

3.2. Development of glass–ceramic foams

Considering the chemistry of the starting materials, shown inTable 1, the introduction of cullet reduces the content of network glassmodifier and the same occurs to the concentration of O2−. Furthermorea number of studies performed on glass melts confirmed that forachieving the thermodynamic equilibrium between polyvalent ionsand environmental oxygen, several hours or even days are required[25,30–35]. The thermal treatment applied to the selected mixtureswas far from the equilibrium, as when the B and C mixtures softened,oxygen gas was virtually cut off from the bulk of the melt. These twoeffects shifted the equilibrium to the right of Eq. (1) and providedoxygen evolution, in turn causing the foaming of glass/basalt mixtures,by reduction of Fe3+ to Fe2+, the TGA analysis reported in Fig. 1evidenced the mass loss which occurred to the mixture C5B5.

Themineralogical composition obtained by the Rietveld quantitativephase analysis of C5B5, fired at 1100 °C with different heating rates, isreported in Table 3. TheXRDpatterns of sample C5B5heated at differentrates are reported in the Supplementary file. Significant variations inthe mineralogical profile of the samples occurred with respect to themixture at room temperature. A slight increase in the content ofamorphous phase – markedly dominant over the crystalline fraction –

is always observable, while the most evident variation in the composi-tion of the crystalline fraction is the drastic decrease in plagioclaseamount, slightly correlated to the heating rate. The plagioclase decreaseis counterbalanced by a significant increase in clinopyroxene concentra-tion, caused by the introduction in themixture of a silicate-richmaterialproportional to the amount of scoria. The transition from basic tointermediate composition as regards silica content, from peraluminousto peralkaline conditions as regards alumina saturation, and the drastic

Table 3Mineralogical quantitative phase analysis of the C5B5 mixture at room temperature and after firing at 1100 °C with different heating ramps (wt.%), obtained by full profile fitting of theexperimental XRD patterns according to the Rietveldmethod (Rwp, R-factor of theweighted profile for each refined pattern is reported, numbers in brackets are estimated absolute errorsfrom the refinement process). Forsterite unit cell volume (Å3) for each thermal treatment is also reported.

Room T 1100 °C, 10 °C/min 1100 °C, 20 °C/min 1100 °C, 40 °C/min 1100 °C, DF

Rwp wt.% 3.25 3.13 3.27 3.19Amorphous wt.% 65.9 (0.4) 68.7 (0.2) 71.1 (0.2) 69. 9 (0.2) 68.3 (0.2)Andesine wt.% 19.9 (0.2) 0.7 (0.1) 0.3 (0.1) 0.2 (0.1) –

Augite wt.% 5.4 (0.1) 25.4 (0.1) 23.2 (0.1) 24.4 (0.1) 26.1 (0.1)Forsterite wt.% 7.6 (0.1) 3.5 (0.1) 3.8 (0.1) 4.1 (0.1) 4.0 (0.1)Spinel wt.% 1.2 (0.1) 0.3 (0.1) 0.3 (0.1) 0.2 (0.1) 0.3 (0.1)Hematite wt.% – 0.2 (0.1) – – –

Maghemite wt.% – 0.8 (0.1) 0.7 (0.1) 0.7 (0.1) 0.7 (0.1)Quartz wt.% – 0.4 (0.1) 0.6 (0.1) 0.5 (0.1) 0.6 (0.1)Forsterite cell volume Å3 294.60 291.36 291.96 291.89 292.42

42 M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

increase in calcium concentration, caused the dissolution ofalumina-rich phases – e.g. the plagioclase – and the precipitation ofcalcium-rich ones – e.g. the clinopyroxene – during the thermalprocess. Furthermore, the presence of low but clearly recognizablequartz amounts suggests a slight systematic silica oversaturation ofthe mixture.

As for the other Fe-bearing phases, a partial oxidation of spinel, withmaghemite formation, is always observable, associated with a decreasein olivine concentration. As for the thermal treatment of the scoriasamples, an oxidation process of ferrous iron in relict olivine is observ-able, testified by the unit cell contraction of the phase. The unit cellcontraction is always smaller with respect to the one in the pure scoriasample treated at the same temperature, with values around 291 Å3 forthe mixtures treated with heating ramps, and it is larger in the samplesthat were directly fired. These analytical evidences suggest a loweroxidation of iron with respect to the thermally treated scoria, alsoconfirmed by the detection of hematite only in the sample heated at10 °C/min. Furthermore, the directly fired sample apparently showslower degree of oxidation with respect to the other ones.

SEM-EDS analyses (Fig. 3) confirmed the analytical evidencesdeduced by XRD analyses. The foams were constituted by a dense

Fig. 3. SEM images of C5B5 foams fired at 1100 °C with a 10 °C/min ramp (a, b) and by directOxidized olivine (OI) and spinel crystals (Spl) are indicated.

glass matrix, with clinopyroxene, olivine and spinel primary crystals.Plagioclase primary crystals were almost totally absent, whereas diffusenewly-formed microcrystals of clinopyroxene were present in thegroundmass, often preferentially nucleated around relict phases. Asobserved for the thermally treated scoria samples, the olivine oxidationprocess is clearly observable, as testified by the presence of fracturedcrystals with precipitation of extracted iron in the form of iron oxidesinside the microcracks (Fig. 3b, d). Nevertheless, relict olivine crystalswere less stressed with respect to the one observed in the thermallytreated scoria samples, with lower occurrence of iron oxide dendrites:such evidence is in agreement with the lower degree of olivine oxida-tion observed by XRD analyses. Furthermore, the olivine crystals inthe directly fired sample were less oxidized with respect to the onesof the other foams (Fig. 3d), confirming the lower degree of reaction.

As reported in Table 4, bubble formation was strongly promoted bydirect firing, determining a total porosity of 86%, of which only 8% wasclosed. By decreasing the heating rate to 10 °C/min, the total porositydecreased to 53%, of which only 8% was open. By applying heating ratesof 20 and 40 °C/min, intermediate total porosities were achieved. Thespecimens produced by applying a progressive heating at 40 °C/min,showed a water absorption after boiling of about 0.3 wt.%, which

insertion (c, d). a, c) low magnification; b, d) high magnification of the highlighted areas.

Table 4Characterization data for foam samples fired at 1100 °C for 15 min at different heating rates (data between square brackets obtained from X-ray computedmicro-tomography; * = dataobtained from bars cut from panel samples).

Sample type C5B5 C4B6 C6B4

15 min at: (°C) 1100 1050 1050 1100 1050

Heating rate (°C/min) 10 20 40 DF DF DF DF DF

WAB (wt.%) b0.2 b0.2 0.3 N100 93 24 78 42

Density (g/cm3)Bulk [ρb] 1.27 ± 0.08 1.03 ± 0.05 1.05 ± 0.03 0.37 ± 0.02 0.53 ± 0.08 0.88 ± 0.09 0.51 ± 0.05 0.44 ± 0.01Apparent [ρa] 1.38 ± 0.01 1.15 ± 0.01 1.18 ± 0.01 1.70 ± 0.05 1.49 ± 0.07 1.80 ± 0.03 1.39 ± 0.04 0.79 ± 0.04True [ρt] 2.70 ± 0.01 2.69 ± 0.01 2.69 ± 0.01 2.69 ± 0.01 2.69 ± 0.01 2.68 ± 0.01 2.68 ± 0.01 2.62 ± 0.01

Porosity (%)Total porosity [TP] 53 ± 3 [44] 62 ± 2 [53] 61 ± 1 [53] 86 ± 1 [73] 80 ± 3 67 ± 3 81 ± 2 83 ± 1Open porosity [OP] 8 ± 6 [1] 10 ± 4 [3] 11 ± 3 [7] 78 ± 1 [72] 65 ± 6 51 ± 5 63 ± 4 44 ± 3Closed porosity [CP] 45 ± 6 [43] 52 ± 5 [50] 50 ± 3 [46] 8 ± 1 [1] 16 ± 6 16 ± 6 18 ± 4 39 ± 4

Elastic modulus (GPa)* 26.9 ± 0.5 19 ± 2 18.8 ± 0.6 – – – – –

Strength (MPa)Bending strength [σflex]* 17 ± 4 11 ± 1 13 ± 2 – – – – –

Crushing strength [σcr] 50 ± 10 22 ± 7 18 ± 6 2.5 ± 0.7 5 ± 2 15 ± 7 4 ± 1 2 ± 1

43M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

decreased below 0.2 wt.% for the samples treated with a heating rate of10 °C/min and 20 °C/min.

X-ray computed micro-tomography (X-μCT) measurements werecarried out on the C5B5 set of samples, in order to evaluate the effectof different heating rates on the pore space properties (open and closedporosity fractions, pore size distribution). The segmentation of the porespace from the solid matrix was obtained by selecting a common globalthreshold value for all the datasets. The optimal limit for the solid/voidthreshold was found in correspondence with the minimum of the grayvalue histogram, between the two main peaks corresponding to poresand solids. By means of a simple procedure based on the digital filling

Fig. 4.Results fromX-raymicrotomography: a, b, c, d) reconstructed cross-sections of selected srendering of a pellet (C5B5, 10 °C/min heating rate — a portion of the sample was virtually cut

of the pore space, both in 2D (i.e. slice by slice, as shown by Fig. 4a, b,c, d) and in 3D, it was possible to calculate the total porosity of thesamples and to evaluate the fractions of open and closed porosity. Theanalysis of the pore size distribution was carried out only on the closedporosity fraction of each sample; the results are reported in the histo-grams of Fig. 4e.

The results from X-μCT, as shown by Table 4, are in good agreementwith those from gas pycnometry. However, it is easy to notice that thevalues of total porosity measured with X-μCT are always significantlylower compared to those obtained with gas pycnometry. This can beexplained in terms of the limited spatial resolution of the technique,

amples from the C5B5 series; e) pore size distribution of the closed porosity fractions; f) 3Dout to display the interior).

Fig. 5. Cross-sections of selected samples from the C5B5 series: a) 10 °C/min heating rate; b) 20 °C/min heating rate; c) 10 °C/min heating rate; d) directly fired sample.

44 M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

whichmakes the fraction of pores smaller than a fewmicrons practical-ly not detectable. Sub-micron spatial resolutions could be obtained,though on significantly smaller samples, using synchrotron-based X-ray computed micro-tomography. A 3D rendering of one of the investi-gated pellet is shown in Fig. 4f. The actual appearance of glass–ceramicsamples is visible in Fig. 5, showing a C5B5-type samples heated at 10,20, 40 °C/min and directly fired.

The changes occurring in the C5B5 samples depending on theheating rates, stimulated further tests on the effect of composition andtemperature. The results of treatments at 1050 and 1100 °C, in termsof mineralogical composition, are summarized in Table 5, while densityand strength data are reported again in Table 4. The XRD patterns of allthe investigated glass ceramics are reported in the Supplementary File.A progressive decrease in plagioclase concentration with increasingtemperatures, counterbalanced by an increase in clinopyroxene, isconfirmed. The presence of the calcium silicate phase is related to theintermediate and peralkaline composition of the mixtures, highlyenriched in calcium, whereas its occurrence at low temperatures onlyis related to its thermal stability field. Traces of quartz are detectablein all the samples, indicating a slight silica oversaturation for all the

Table 5Mineralogical quantitative phase analysis of C4B6, C5B5 and C6B4mixtures at room temperaturXRD patterns according to the Rietveld method (Rwp, R-factor of the weighted profile for eachrefinement process). The forsterite unit cell volume (Å3) resulting from the refinement after e

C4B6, T room C4B6, 1050 °C C4B6, 1100 °C C

Rwp 3.15 3.32Amorphous wt.% 59.1 (0.4) 58.0 (0.3) 61.1 (0.3) 6Andesine wt.% 24.0 (0.2) 7.9 (0.1) 2.9 (0.1) 1Augite wt.% 6.4 (0.1) 24.4 (0.1) 27.8 (0.2) 5Forsterite wt.% 9.1 (0.1) 6.5 (0.1) 6.7 (0.1) 7Spinel wt.% 1.4 (0.1) 0.8 (0.1) 0.3 (0.1) 1Hematite wt.% – 0.5 (0.1) – –

Maghemite wt.% – 0.7 (0.1) 1.0 (0.1) –

Quartz wt.% – 0.3 (0.1) 0.2 (0.1) –

Wollastonite wt.% – 0.9 (0.1) – –

Forsterite cell volume Å3 294.60 291.57 292.17 2

mixtures. Furthermore, spinel and olivine underwent Fe oxidationprocesses in all the samples, with maghemite formation and olivineunit cell contraction due to iron extraction. As for the C5B5 mixturestreated with different heating ramps, the unit cell contraction wasalways lower with respect to the ones observed for the thermally treat-ed scoria samples, suggesting lower overall degrees of iron oxidation.No clear correlations between themeasured olivine unit cell contractionand temperature or composition variations were observable.

3.3. Mechanical properties and potential applications

As previously observed, directly fired samples show mostly openpores, whereas the samples fired using a defined heating rate featuredmostly closed cells. As illustrated by Fig. 4e, the samples with closedcells had also a practically continuous pore cell distribution, while theopen porosity was associated to a defined maximum pore size (a sortof threshold level) andwas generally muchmore uniform. These differ-ences could be ascribed to the mechanism of pore formation itself. Op-erating with direct firing, lots of pore nuclei formed almostsimultaneously; on the contrary, operating at defined heating rates,

e and afterfiring at 1050 °C and 1100 °C, obtained by full profilefitting of the experimentalrefined pattern is reported, numbers in brackets are estimated absolute errors from the

ach thermal treatment is also reported.

5B5, T room C5B5, 1050 °C C5B5, 1100 °C C6B4, T room C6B4, 1050 °C

3.20 3.19 3.125.9 (0.4) 68.3 (0.3) 68.3 (0.2) 72.7 (0.4) 74.1 (0.2)9.9 (0.2) 2.6 (0.1) – 16.0 (0.2) 0.6 (0.1).4 (0.1) 20.4 (0.1) 26.1 (0.1) 4.3 (0.1) 17.3 (0.1).6 (0.1) 5.9 (0.1) 4.0 (0.1) 6.1 (0.1) 4.9 (0.3).2 (0.1) 0.4 (0.1) 0.3 (0.1) 0.9 (0.1) 0.4 (0.1)

– – – –

0.7 (0.1) 0.7 (0.1) – 0.5 (0.1)0.3 (0.1) 0.6 (0.1) – 0.2 (0.1)1.4 (0.1) – – 2.0 (0.1)

94.60 292.79 292.42 294.60 291.04

Fig. 6. Ashby's plots of developed glass–ceramic foams compared with non-technical ceramics and glasses (a) compressive strength; b) flexural strength; c) Young's modulus); the linescorrespond to guidelines for lightweight design.

45M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

some pores could form and collapse during the heating period. Largerpores, for defined heating rates, likely incorporated smaller ones; thiseffect (cell coalescence) is well known to cause a reconstruction of thecell struts, which become increasingly thicker, and in turn containsecondary pores [3]. In Fig. 4c, as an example, some very large poresare present and surrounded by very thick and porous struts.

The presence of open porosity generally implies a degradation ofmechanical properties; thick and porous struts, however, makeclosed-cell foams practically as weak as open-celled foams [3]. As ageneral approach, it may be noted that a component with uniformlydistributed open cells is mechanically as efficient as one with closedcells, which are not homogenously distributed, and with thick struts.The Ashby's plot [36] in Fig. 6a may be seen as a confirmation of this

interpretation: the studied samples are nicely aligned in terms of specif-ic crushing strength (the line in the graph represents materials with thesame ratio between crushing strength and density, σc/ρ), except forC5B5-type samples fired at 10 °C/min (well above the line, meaningthat they are mechanically more efficient). Interestingly, the specificstrength of these glass–ceramic foams is higher or equal to that of vari-ants of lightweight concrete, suggesting their use as new lightweightaggregates in the building industry.

The plots of bending strength and elastic modulus for densersamples, in the form of panels and fired with a defined heating rate,also indicate the suitability of these materials for possible applicationsin the building industry; in fact, the indices for the design of lightweightpanels [36], i.e. the specific flexural strength index (σflex

1/2/ρ, Fig. 6b) and

46 M. Marangoni et al. / Journal of Non-Crystalline Solids 403 (2014) 38–46

the specific Young's modulus index (E1/3/ρ, Fig. 6c), well exceed thevalues for most non-technical ceramics, including traditional ceramics(brick and tiles), natural stones (granite) and concrete. Consideringthe low water absorption (not exceeding 0.3 wt.%), C5B5-based panelscould be applied in the so-called ventilated façades, i.e. a newgenerationof cladding components that can be applied on the surface of largebuildings, aimed to improve the thermal insulation, as an alternativeto conventional porous ceramics (fired at higher temperature andrequiring the addition of specific foaming agents to develop a homoge-nous cellular structure) [37,38], especially if coated with a glaze [39,40].

4. Conclusions

The addition of cullet to basalt scoria enabled the fabrication ofhighly porous glass–ceramic foams with good mechanical strength atrelatively low temperature and with fast firing. Total porosity rangedin awide range of values, depending on the composition and processingschedule adopted, and components with either closed cells and lowwater absorption values, or with open cells were fabricated. No foamingagents were necessary, and the self-foaming mechanismwas related tothe redox reactions occurring between different iron oxide species. Thedata here presented demonstrate that by following this approach it ispossible to produce porous glass–ceramics possessing high strength(suitable for instance for load-bearing applications) or possessing avery high total porosity (therefore more appropriate, for instance, forthermal management applications).

The choice of the most appropriate processing will depend on thetype of application pursued for the specific components produced, forinstance:

• C5B5-based light weight aggregates obtained by applying a heatingrate of 10 °C/min could be used to realize lightweight concreteconsidering that the specific strength of these glass–ceramic foamsis higher or equal to that of typical lightweight concrete;

• C5B5-based panels obtained by applying a heating rate of 40 °C/minor 20 °C/min could be applied in the so-called ventilated façadesconsidering their high specific flexural strength index, the limitedwater absorption and that porous materials may further improvethe thermal insulation.

Acknowledgments

The authors would like to express their gratitude to King AbdulazizCity for Science and Technology (KACST), Saudi Arabia, for thesupport and funding through Grant No. 32-639. Financial supportfor the X-ray computed micro-tomography laboratory of the Depart-ment of Geosciences at the University of Padua was provided byFondazione Cassa di Risparmio di Padova e Rovigo (CaRiPaRo).

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jnoncrysol.2014.06.016.

References

[1] G. Scarinci, G. Brusatin, E. Bernardo, Production technology of glass foams, in: M.Scheffler, P. Colombo (Eds.), Cellular Ceramics. Structure, Manufacturing, Propertiesand Applications, Wiley-VCH, Weinheim, Germany, 2005.

[2] W. Owen (Pittsburgh Plate Glass, USA), US Patent 2,310,457 (1943).[3] E. Bernardo, F. Albertini, Glass foams from dismantled cathode ray tubes, Ceram. Int.

32 (2006) 603–608.[4] F. Méar, P. Yot, M. Ribes, Effects of temperature, reaction time and reducing agent

content on the synthesis of macroporous foam glasses from waste funnel glasses,Mater. Lett. 60 (2006) 929–934.

[5] E. Bernardo, R. Cedro, M. Florean, S. Hreglich, Reutilization and stabilization ofwastes by the production of glass foams, Ceram. Int. 33 (2007) 963–968.

[6] H.R. Fernandes, D.U. Tulyaganov, J.M.F. Ferreira, Preparation and characterization offoams from sheet glass and fly ash using carbonates as foaming agents, Ceram. Int.35 (2009) 229–235.

[7] E. Bernardo, Micro- and macro-cellular sintered glass–ceramics from wastes, J. Eur.Ceram. Soc. 27 (2007) 2415–2422.

[8] F. Defrancesco, A. Tomasi, G.D. Soraru, Expanded materials from quartz-porphyritesands, J. Mater. Sci. 22 (1987) 2493–2496.

[9] I. Ponsot, E. Bernardo, Self glazed glass ceramic foams from metallurgical slag andrecycled glass, J. Clean. Prod. 59 (2013) 245–250.

[10] H.M. Rietveld, A profile refinement method for nuclear and magnetic structures, J.Appl. Crystallogr. 2 (1969) 65–71.

[11] A.F. Gualtieri, Accuracy of XRPD QPA using the combined Rietveld–RIR method, J.Appl. Crystallogr. 33 (2000) 267–278.

[12] W.A. Dollase, Correction of intensities for preferred orientation in powder diffrac-tometry: application of the March model, J. Appl. Crystallogr. 19 (1986) 267–272.

[13] L.A. Feldkamp, L.C. Davis, J.W. Kress, Practical cone-beam algorithm, J. Opt. Soc. Am.1 (1984) 612–619.

[14] M. Boin, A. Haibel, Compensation of ring artefacts in synchrotron tomographicimages, Opt. Express 14 (2006) 12071–12075.

[15] J. Sijbers, A. Postnov, Reduction of ring artefacts in high resolution micro-CTreconstructions, Phys. Med. Biol. 49 (2004) 247–253.

[16] I.D.R. Mackinnon, J.B. Parise, J.D.F. Gerald, Average structure of an An48 plagioclasefrom the Hogarth Ranges, Am. Mineral. 71 (1986) 1399–1408.

[17] G. Ottonello, F. Princivalle, A. Della Giusta, Temperature, composition and fO2 effectson intersite distribution of Mg and Fe2+ in olivines, Phys. Chem. Miner. 17 (1990)301–312.

[18] G.M. Crisci, P.F. Zanazzi, G.Molin, D. Pasqual, Clinopyroxene from Lipari: comparisonwith analogues from other Aeolian Islands, Italy, Can. Mineral. 36 (1998) 97–105.

[19] B.A. Wechsler, D.H. Lindsley, C.T. Prewitt, Crystal structure and cation distribution intitanomagnetites (Fe3 − xTixO4), Am. Mineral. 69 (1984) 754–770.

[20] B. Gillot, F. Jemmali, A. Rousset, Kinetics and mechanism of ferrous spinel oxidationstudied by electrical conductivity and thermogravimetry, J. Mater. Sci. 21 (1986)4436–4442.

[21] D. Świerczyński, C. Courson, L. Bedel, A. Kiennemann, S. Vilminot, Oxidationreduction behavior of iron-bearing olivines (FexMg1 − x)2SiO4 used as catalysts forbiomass gasification, Chem. Mater. 18 (2006) 897–905.

[22] S. Collyer, N.W. Grimes, D.J. Vaughan, G. Longworth, Studies of the crystal structureand crystal chemistry of titanomaghemite, Am. Mineral. 73 (1988) 153.

[23] Yu Shu Cheng, Effects of pressure on the crystal structure of olivine in harzburgitexenolith of basalt, Proc. Natl. Sci. Counc. Repub. China A Phys. Sci. Eng. 21 (1997)173–179.

[24] R.F. Wendlandt, Oxygen diffusion in basalt and andesite melts: experimental resultsand discussion of chemical versus tracer diffusion, Contrib. Mineral. Petrol. 108(1991) 463–471.

[25] S. Holmquist, Ionic formulation of redox equilibria in glass melts, J. Am. Ceram. Soc.49 (141) (1966) 228–229.

[26] D.S. Goldman, Oxidation equilibrium of iron in borosilicate glass, J. Am. Ceram. Soc.66 (1983) 205–209.

[27] M. Cable, A century of developments in glassmelting research, J. Am. Chem. Soc. 81(1998) 1083–1094.

[28] G.W. Toop, C.S. Samis, Some new ionic concepts of silicate slags, Can. Metall. Q. 2(1962) 129–152.

[29] B. Mysen, P. Richet, Silicate Glasses and Melts, 1st edition Elsevier, 2005. 179–183.[30] F. Irmann, A study of molten borates with the Cr(VI)—Cr(III) indicator, J. Am. Chem.

Soc. 74 (1952) 4767–4770.[31] W.D. Johnston, Oxidation-reduction equilibria in iron-containing glass, J. Am. Ceram.

Soc. 47 (1964) 198–201.[32] W.D. Johnston, Oxidation-reduction equilibria in molten Na2O·2SiO2 glass, J. Am.

Ceram. Soc. 48 (1965) 184–190.[33] W.D. Johnston, A. Chelko, Oxidation-reduction equilibria in molten Na2O·2SiO2

glass in contact with metallic copper and silver, J. Am. Ceram. Soc. 49 (1966)562–564.

[34] S. Banerjee, A. Paul, Thermodynamics of the system Cu–O and ruby formation inborate glass, J. Am. Ceram. Soc. 57 (1974) 286–290.

[35] A. Paul, Chemistry of Glasses, 2nd edition Chapman & Hall, 1990. 228–230.[36] M.F. Ashby, Materials Selection in Mechanical Design, third edition Butterworth

Heinemann, Oxford, 2005.[37] E. Bernardo, M. De Lazzari, P. Colombo, A. Saburit Llaudis, F. Javier Garcìa-Ten, Light-

weight porcelain stoneware by engineered CeO2 addition, Adv. Eng. Mater. 12(2010) 65–70.

[38] EU. patent application EP 2 752 395 A1 A porous glass ceramic composition andmethod for manufacturing the same.

[39] M.A. Binhussain, M. Marangoni, E. Bernardo, P. Colombo, Sintered and glazed glass–ceramics from natural and waste raw materials, Ceram. Int. 40 (2014) 3543–3551.

[40] EU. patent application EP 2 749 544 A1 A glaze composition, method formanufacturing the glaze composition and methods for glazing.