Influence of particle size on the crystallization process and the bioactive behavior of a bioactive...

Journal of Non-Crystalline Solids 357 (2011) 211–219

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids

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

Influence of particle size on the crystallization kinetics of glasses produced fromwaste materials

M. Erol ⁎, S. Küçükbayrak, A. Ersoy-MeriçboyuDepartment of Chemical Engineering, Chemical&Metallurgical Engineering Faculty, Istanbul, Technical University, Maslak 34469, Istanbul, Turkey

⁎ Corresponding author. Tel.: +90 212 285 3351; faxE-mail address: erolm@itu.edu.tr (M. Erol).

0022-3093/$ – see front matter © 2010 Elsevier B.V. Aldoi:10.1016/j.jnoncrysol.2010.09.027

a b s t r a c t

Article history:Received 4 August 2009Received in revised form 29 August 2010Available online 20 October 2010

Keywords:Glass;Crystallization;Glass-ceramic;Differential thermal analysis

The crystallization behavior and kinetics of glasses produced from coal fly ashes, redmud and silica fumewereinvestigated by using differential thermal analysis, X-ray diffraction and scanning electron microscopytechniques. The kinetic parameters of the glass-crystallization transformation were estimated under non-isothermal conditions applying three different equations, namely, Kissinger, Matusuta-Sakka and Ozawa.Non-isothermal differential thermal analysis curves were obtained using both coarse and fine glass samples.The crystallization activation energies of coarse glasses are in the range of 233–439 kJ/mol while theactivation energies of fine glasses change in the range of 369–450 kJ/mol. Avrami exponent, n, values of coarseglasses indicated the three-dimensional bulk crystallization. This result is in well agreement with the cross-sectional scanning electron microscopy investigations. The values of the n obtained experimentally are in therange of 1.24–1.36 for fine glasses which show the one-dimensional surface crystallization. The crystallizedphase of the glass-ceramic samples produced from waste glasses by applying the controlled heat treatmentprocess was identified as diopside by X-ray diffraction analysis.

: +90 212 285 2925.

l rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

An increasingly urgent problem for the near future of human kindis the recycling of industrial wastes. One of themajor wastes is coal flyash, which is produced in significant amounts in Turkey. Fly ash is amajor source for environmental pollution since it is in a fine powderform and its heavy metal content is very high. Because of increasinglystringent environmental regulations, coal fly ash is regarded ashazardous material in most countries. Therefore, it must be recycledor, at least, be deposited in standard landfill sites with minimum risk.Vitrification technology is one of the most promising solutions for theutilization of industrial wastes. It is feasible to use coal fly ash as a rawmaterial source to develop glass materials since it contains largeamounts SiO2 and Al2O3, which are main glass network formers. Theglass materials produced from industrial wastes can also be convertedinto glass-ceramic materials possessing outstanding mechanical,physical and chemical properties.

In recent years, the production of glass-ceramic materials made byrecycling industrial wastes such as, coal fly ash, iron blast furnaceslags, municipal incinerator fly ash and even red mud from aluminumproduction have been investigated by many researchers [1–6]. Glass-ceramic production is achieved through a two step process namelynucleation and crystallization stages. In the nucleation stage, smallnuclei are formedwithin the parent glass. After the formation of stable

nuclei, crystallization proceeds by growth of a new crystalline phase.Knowledge of nucleation and crystallization parameters is importantin the preparation of glass-ceramics with desired microstructure andproperties [7].

Thermal analysis is widely used in studying the crystallizationkinetics of glasses because of providing rapid and convenient means[8]. Isothermal and non-isothermal methods are applied for thermalanalysis. Most of the studies are connected with non-isothermalinvestigations of the nucleation, the crystal growth and the kinetics ofcrystallization [8,9]. Several equations have been proposed, attempt-ing to interpret the non-isothermal data. Most of these equations areavailable to analyze the DTA data and to determine the activationenergy for crystallization. These equations shed important light on theunderstanding the nature of the crystallization mechanisms of dif-ferent glass forming systems. Recently, there were reports on crys-tallization kinetics of glasses produced from industrial wastes[1–5,10,11]. The activation energy of the glasses produced from coalfly ash was calculated as 370 kJ/mol by Cioffi et al. [3]. Thecrystallization behavior of glasses made of mixture of coal ash andsoda lime glass cullet was investigated by Francis et al. [4]. Obtainedresults showed that the crystallization mechanism is diffusioncontrolled crystallization with a decreasing nucleation rate. Morerecently, Romero et al. [12] have investigated the effect of glassparticle size on the crystallization kinetics of an iron-rich glass from anickel leaching waste. It was reported that the activation energies ofthe glass were in the range of 349–423 kJ/mol and crystallization inthe glass occured by different mechanisms depending on the glassparticle size. It was also stated that coarse particles showed three-

212 M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

dimensional crystal growth controlled by diffusion while the fineparticles led to an interface reaction mechanism with two-dimen-sional growth of crystals.

In this study, the crystallization behavior of glasses produced frommixtures of coal fly ashes obtained from two different thermal powerplants, red mud and silica fume was investigated by using DTA. Forthis purpose, Kissinger [13], Matusita & Sakka [14,15] and Ozawa [16]equations were used to determine the crystallization mechanism andthe activation energy for crystallization. Particle size effect on thecrystallization kinetics of waste glasses was also studied.

2. Experimental procedure

2.1. Starting materials and glass preparation

The raw materials used in this study were coal fly ashes obtainedfrom the Çayırhan and Orhaneli thermal power plants in Turkey, redmud from aluminum production and silica fume. Chemical composi-tions of these wastes were given in Table 1 [6,17]. Five percent redmud was added to Çayırhan coal fly ash to produce glass samples,while 20% red mud and 20 % silica fume were added to Orhaneli coalfly ash to increase the SiO2 and Fe2O3 content of it.

Glass samples were prepared from themixedwastes. In each batch20 g of waste was melted in a platinum crucible for 2 h in anelectrically heated furnace at 1773 K. To ensure homogeneity, themelt was poured into water. The cast glasses were crushed, pulverizedand remelted at the same temperature for 3 h to remove the airbubbles from the melt. Following this procedure, the refinedmelt wascast in a preheated graphite mould (673 K) to form cylinders ofapproximately 0.8 to 1 cm in diameter and 1 to 4 cm in length. Thecylinders were cooled to room temperature. To remove thermalresidual stress, the cast glasses were annealed in a furnace at 873 K for2 h followed by slow cooling to room temperature. The annealingtemperature and timewere chosen on the basis of the results obtainedin a previous study [11].

CR and ORS codes were given to the glasses produced fromÇayırhan fly ash+red mud and Orhaneli fly ash+red mud+silicafume, respectively.

2.2. Differential thermal analysis and heat treatment

Differential thermal analysis (DTA) experiments were performedby heating 20 mg glass samples in a Pt-crucible and using Al2O3 as areference material in the temperature range between 293 and 1373 Kat the heating rates of 5, 10, 15 and 20 K/min. The glasses were groundto a coarse particle size (800–1000 μm) to be a representative of abulk sample and to a fine particle size (b180 μm) to be a repre-sentative of powder samples.

The variation of the crystallization peaks with different DTAheating rates can be used to estimate the activation energy forcrystallization and to determine the crystallization mechanism. Thecrystallization behavior of glasses was determined by using non-isothermalmethods. DTAwas performed on both coarse and fine glasssamples to determine the particle size effect on the crystallizationmechanism. As non-isothermal methods Kissinger, Matusita–Sakkaand Ozawa methods were used.

Table 1Chemical compositions of the waste materials [6,16].

Waste materials SiO2

(%)Al2O3

(%)CaO(%)

MgO(%)

Fe2O3

(%)Na2O(%)

K2O(%)

SO3

(%)LOI⁎

(%)

Çayirhan fly ash 41.53 17.77 12.52 4.46 9.93 2.57 2.43 8.03 0.68Orhaneli fly ash 32.83 13.34 30.35 4.51 5.61 2.15 1.37 5.85 3.67Silica fume 90.80 1.02 2.55 0.94 1.93 – – 0.50 1.57Red mud 10.40 28.50 3.90 7.70 35.1 3.80 1.60 4.65 3.90

⁎LOI: loss of ignition.

Nucleation and crystallization experiments were carried accordingto the results obtained in the other studies [18,19]. For this purpose,CR and ORS glasses were heated at a rate of 10 K/min to the nucleationtemperatures of 963 and 988 K, respectively. Maximum nucleationtemperature and time were determined in the previous studies[18,19]. CR and ORS glass samples were held at maximum nucleationtemperatures for 4 and 2 h, respectively to obtain fully nucleatedglasses. Following nucleation, the temperature was raised to thecrystallization temperatures of 1135 and 1188 K, for CR and ORSsamples, respectively. Samples were held at these temperatures for15 min and cooled in the furnace.

2.3. Microstructural characterizations

The characterization of the produced glass-ceramic samples wascarried out using both electron microscopy and X-ray diffraction(XRD) techniques. Scanning electron microscopy (SEM) investiga-tions were operated at 25 kV to observe the microstructure of theproduced samples. For the SEM investigations, optical mount speci-mens were prepared using standard metallographic techniquesfollowed by chemical etching them in an HF solution (5%) for1.5 min. The etched glass-ceramic samples were coated with carbonprior to examination.

X-ray diffraction was utilized to determine that the crystallinephases occurred in the glass-ceramic samples. The X-ray diffractioninvestigations were carried out using CuKα radiation at 40 kV and30 mA settings in the 2θ range from 10° to 80°. In all cases, sampleswhich were analyzed by X-ray diffraction were ground to fine powderform.

3. Results

3.1. Activation energy determination

The kinetic parameters of the glass-crystallization transformationwere estimated under non-isothermal conditions applying threedifferent equations, namely, Kissinger [13], Matusita–Sakka [14,15]and Ozawa [16]. Crystallization mechanisms and activation energiesfor crystallization were calculated in each method from the heatingrate dependence of the crystallization peak temperatures.

Apparent activation energies for crystallization may be deter-mined employing non-isothermalmethods to the crystallization peak.For example, in the Kissinger equation [13], the crystallization peaktemperature is monitored as a function of the heating rate; thefollowing relationship is then applied

lnαT2p

!=

−EckRTp

!+ cons tan t ð1Þ

where α is the heating rate, Tp is the crystallization peak temperature,R is the gas constant, and Eck is activation energy for crystallization,determined by the Kissinger equation. A plot of ln (α/Tp2) vs. 1/Tpshould be a straight line, from the slope of which Eck can bedetermined.

Matusita–Sakka [14,15] have stated that Eq. (1) is valid only whencrystal growth occurs on a fixed number of nuclei. Incorrect values forthe activation energy are obtained if a majority of the nuclei areformed during the DTA measurement, so that the number of nucleicontinuously varies with α. They have proposed a modified form ofthe Kissinger equation as given before (Eq. (1))

lnαn

T2p

!=

−mEcRTp

!+ cons tan t ð2Þ

Table 2Values of n andm for different crystallization mechanisms in the heating process [9,14].

Crystallization mechanism n m

Bulk crystallization with a constant number of nuclei (i.e.,the number of nuclei is independent of the heating rate)

Three-dimensional growth of crystals 3 3Two-dimensional growth of crystals 2 2One-dimensional growth of crystals 1 1Bulk crystallization with an increasing number of nuclei (i.e.,the number of nuclei is inversely proportional to the heating rate)

Three-dimensional growth of crystals 4 3Two-dimensional growth of crystals 3 2One-dimensional growth of crystals 2 1Surface crystallization 1 1

213M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

where Ec is the correct activation energy for crystallization, n is aconstant known as the Avrami parameter and m represents thedimensionality of crystal growth. When surface crystallizationpredominates, m=1 and when the crystallization is predominantlybulk, m=3 (from Table 2). The value of m is related to n as m=nwhen crystallization at different heating rates occurs on a fixed

1200800400

(a)

(b)

(c)

(d)

T (K)

1200800400

T (K)

Coarse

(a)

(b)

(c)

(d)

Fine

ΔT E

xoth

erm

icΔT

Exo

ther

mic

Fig. 1. DTAplots of the coarse andfineCRglasses scanned at theheating rates of: a) 5 K/min,b) 10 K/min, c) 15 K/min and d) 20 K/min.

number of nuclei (i.e., the number of nuclei is constant during DTAruns at different values of α), and m=n−1 when nucleation occursduring DTA and the number of nuclei in the glass is inverselyproportional to α.

In addition, when surface crystallization predominates, m=n=1and Eq. (2) essentially reduces to the Kissinger equation which willyield the correct value for the activation energy, i.e., Eck=Ec.

In the presence of bulk crystallization, Eck does not necessarily tobe equal to Ec. Rather, a close inspection of Eqs. (1) and (2) shows that

Ec = n=mð ÞEck−2 n−1ð Þ=mð ÞRTp ð3Þ

For most oxide glass systems, Ec≥20 RTp typically [20]. Therefore,the neglect of 2 ((n−1)/m)RTp≤2 RTp in Eq. (3) will result in an errorless than only 10 % in the value of Ec. This error is within the errorrange of the DTA experiment. Then we obtain

Ec≅ n=mð ÞEck ð4Þ

For m=n, i.e., when crystallization occurs on a fixed number ofnuclei, Eck=Ec. Thus, for predominantly surface crystallization or for

400 600 800 1000 1200 1400

(a)

(b)

(c)

(d)

T (K)

Coarse

400 800 1200

(a)

(b)

(c)

(d)

T (K)

Fine

ΔT E

xoth

erm

icΔT

Exo

ther

mic

Fig. 2. DTA plots of the coarse and fine ORS glasses scanned at the heating rates of:a) 5 K/min, b) 10 K/min, c) 15 K/min and d) 20 K/min.

Table 3DTA results of coarse and fine CR glass samples.

Heating rate(K/min)

Crystallization peaktemperatures forcoarse particles (K)

Crystallization peaktemperatures for fineparticles (K)

5 1157 108010 1187 109115 1205 110820 1222 1113

Table 4DTA results of coarse and fine ORS glass samples.

Heating rate(K/min)

Crystallization peak temperaturesfor coarse particles (K)

Crystallization peaktemperatures for fine particles(K)

5 1200 108210 1222 109715 1237 110820 1254 1110

8.0 8.2 8.4 8.6 8.8

-10

-8

-6

-4

-2

Coarse

ln(α

n /Tp2 )

-4 -1

214 M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

crystal growth that occurs on a fixed number of nuclei, the analysis ofDTA data by the Kissinger equation (Eq. (1)) yields the correct value ofEc. When the number of nuclei changes during the DTA measure-ments, either, Eq. (2) should be used or Eck determined from Eq. (1)should be multiplied by the term (n/m) to obtain the correctactivation energy [8].

From the exothermic peak, the Avrami parameter, n, can beobtained by using the modified Ozawa equation [16]:

jd ln − ln 1−xð Þð Þð Þd lnα

jT

= −n ð5Þ

where x is the volume fraction crystallized at a fixed temperature Twith the heating rate of α. x is the ratio of the partial area at a certaintemperature to the total area of a crystallization exotherm.

Non-isothermal DTA curves were obtained with selected heatingrates (5, 10, 15 and 20 K/min) using both coarse and fine glasssamples. Typical DTA graphs for coarse and fine samples of CR andORS glasses recorded at 4 different heating rates were shown in Figs. 1

1.6 2.0 2.4 2.8 3.2

-4

-2

0

2Coarse

Fine

ln(-

ln(1

-x))

lnα

Fig. 3. The Ozawa plots of the coarse and fine CR glasses.

and 2, respectively. The crystallization peak temperatures of CR andORS glasses of both coarse and fine particles measured by DTA weregiven in Tables 3 and 4, respectively. As can be seen in Figs. 1 and 2and Tables 3 and 4, crystallization peak temperatures of all glasssamples increased with the increase in heating rates. Tp for coarseparticles was significantly higher than that for fine particles and thisdifference in Tp increased consistently with increasing heating rate.

At a particular heating rate, the Tp for a glass depends on the totalconcentration of surface and bulk nuclei present in the glass anddecreases with increasing concentration of nuclei. Since the surfacearea increases with decreasing particle size, the concentration ofsurface nuclei is expected to be higher for fine particles. If a glass is notinitially saturated with internal nuclei during a DTA scan, theconcentration of it will be higher for slower heating rates, since theglass spends a longer time in the temperature range where nucleationcan occur. The effect of these internal nuclei on crystallization is morepronounced in coarse particles, as they provide larger effectivevolumes for internal nucleation. It is expected that at low heatingrates, all the different particle sizes are highly nucleated (surface orbulk). At higher heating rates, the concentration of nuclei in the coarseparticles is less than that in the fine particles so that crystallization for

8.9 9.0 9.1 9.2 9.3

-12.0

-11.5

-11.0

-10.5

-10.0

-9.5

ln(α

n /Tp2 )

1/Tp *10 (K )

1/Tp *10-4 (K-1)

Fine

Fig. 4. The Matusita–Sakka plots of the coarse and fine CR glasses.

215M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

coarse particles occurs at a higher temperature [21]. This is due to theparticle size effect on heat transfer. Temperature gradients developedbetween the surrounding air and the surface of the glass sample andinside the glass sample. Temperature gradients can arise when theglass samples' dimensions or the heating rate are relatively large.The temperature rises gradually from inside the glass sample to thesurface of it since the heat transfer coefficient on the surface of thesample is small. At a given heating rate, coarse particles have greaterheat transfer resistance, so it takes longer for the center of the particleto reach the furnace temperature and a higher observed crystalliza-tion temperature results. The concentration of nuclei in coarseparticles is expected to decrease with increasing heating rate (lesstime in the nucleation range), so the difference in nuclei concentra-tion between coarse and fine particles will increase with increasingheating rate. This will cause the difference in Tp between coarse andfine particles with increasing heating rate. This case has also beenobserved by several authors [22–24].

Assuming that the nucleation and growth processes had occurredsimultaneously in as-quenched samples during the DTA measure-ments, the data for as-quenched glasses were analyzed by the

8.0 8.2 8.4 8.6 8.8

8.9 9.0 9.1 9.2 9.3

ln(α

/Tp2 )

ln(α

/Tp2 )

1/Tp *10-4 (K-1)

1/Tp *10-4 (K-1)

-12.4

-12.0

-11.6

-11.2

-10.8

-12.4

-12.0

-11.6

-11.2

-10.8

Coarse

Fine

Fig. 5. The Kissinger plots of the coarse and fine CR glasses.

Matusita–Sakka equation (Eq. (2)) to determine Ec. Since Eq. (2)included the n and m, first of all their values were determined usingthe Ozawa equation (Eq. (5)). Plots of ln(−ln(1−x)) vs. lnα for CRglass of coarse and fine particles are shown in Fig. 3. x values weredetermined at the fixed temperatures of 1185 K and 1195 K for coarseand fine particles, respectively. The values of n determined from theslops of these plots are 3.42±0.21 for coarse particles (correlationcoefficient r=0.92) and 1.36±0.10 (r=0.98) for fine particles. Thesevalues indicate that bulk and surface crystallizations are dominant inthe coarse and fine particles, respectively. The m value for the coarseparticles should be equal to n−1, i.e., 2.42±0.21 from Table 2. Fineparticles have a larger effective surface area so that the number ofinternal nuclei formed during the DTA run could be neglected. Thismeans that crystal growth for glasses of fine particles should haveoccurred on a fixed number of nuclei during the DTA measurements.Therefore, we can assume that n is equal to m for the fine glasses.n=m=1.36±0.10 for CR glass of fine particles and this value is veryclose to 1 which means surface crystallization.

Using n, m and R (R value was taken as 8.3144 J/mol K) values,Matusita–Sakka plots, ln(αn/Tp2) vs. 1/Tp, for CR sample yield Ec values

-6

-4

-2

0

2

1.6 2.0 2.4 2.8 3.2

1.6 2.0 2.4 2.8 3.2

-2.0

-1.5

-1.0

-0.5

0.0

0.5

Coarse

ln(-

ln(1

-x))

lnα

lnα

Fine

ln(-

ln(1

-x))

Fig. 6. The Ozawa plots of the coarse and fine ORS glasses.

216 M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

as 350±11 (r=0.97) and 372±17 (r=0.91) kJ/mol for coarse andfine particles, respectively (Fig. 4). Also using the Kissinger equation(Eq. (1)), Eck values of as-quenched glasses of coarse and fine particlescan be determined from the slope of the plots ln(α/Tp2) vs. 1/Tp (Fig. 5).For coarse particle Eck is 233±9 kJ/mol (r=0.96) that is lower thanthe Ec values determined by using the Matusita–Sakka equation. Byusing Eq. (4), i.e., multiplying the Eck value for coarse particles by thefactor n/m (=3.42/2.42) yields the Ec value of 330±9 kJ/mol. For fineparticles Eck value is 369±14 kJ/mol (r=0.92) that is close to the Ecvalues determined from the Matusita–Sakka equation. For n=m, i.e.,when crystallization occurs on a fixed number of nuclei, Eck=Ec. Thiscase shows that the crystal growth for fine particles occurred on afixed number of nuclei during the DTA measurements.

The Ozawa plots of ln(−ln(1−x)) vs. lnα for the coarse and fineparticle sizes for ORS glass are linear according to Eq.(5) with n=3.68±0.18 (r=0.89) and n=1.24±0.11 (r=0.93), respectively (Fig. 6). xvalues were determined at the fixed temperatures of 1227 K and 1102 Kfor coarse andfineparticles, respectively.n=3.68±0.18 value for coarseparticles indicates the bulk crystallizationwhile n=1.24±0.11 value forfine particles shows the surface crystallization. For coarse particles m isequal to 2.68±0.18 whilem is equal to 1.24±0.11 for fine particles.

ln(α

n /Tp2 )

ln(α

n /Tp2 )

8.07.9 8.1 8.38.2 8.4

1/Tp *10-4 (K-1)

1/Tp *10-4 (K-1)

9.0 9.19.1 9.29.2 9.3

-8

-6

-4

-2

-12.0

-11.5

-11.0

-10.5

-10.0

Coarse

Fine

Fig. 7. The Matusita–Sakka plots of the coarse and fine ORS glasses.

By plotting ln(αn/Tp2) vs. 1/Tp, Ec values for coarse and fine particleswere estimated from the slope of straight lines as 439±12 (r=0.98)and 450±18 (r=0.94) kJ/mol, respectively (Fig. 7). The Kissingerequation was used to determine the Eck values of as-quenched ORSglasses. Plots of ln(α/Tp2) vs. 1/Tp, for coarse and fine particles areshown in Fig. 8 and the values of Eck obtained from the slopes of theplots are 305±10 (r=0.98) and 444±12 (r=0.95) kJ/mol, respec-tively. The activation energy of crystallization for fine glass obtainedfrom Kissinger equation is close to the Ec values of fine glass estimatedfrom theMatusita–Sakka equation. n value of fine ORS glass was found1.24±0.11 which indicates the surface crystallization. We know thatthe Kissinger equation (Eq. (1)) is valid only when the crystallizationoccurs on a fixed number of nuclei during the DTA runs or when thesurface crystallization mechanism is dominant in the glass. Therefore,the Eck value of fine ORS glass is more accurate than the Eck value ofcoarse ORS glass. If we multiply Eck value of coarse particles with n/m(=3.68/2.68) we obtained the Ec value as 414±10 kJ/mol by usingEq. (4).

Tables 5 and 6 show the Ec, n andm values of CR and ORS glasses ofcoarse and fine particles, respectively. As seen from Tables 5 and 6, nandm values are different for coarse and fine glasses. It is important to

ln(α

/Tp2 )

ln(α

/Tp2 )

8.07.9 8.1 8.38.2 8.4

1/Tp *10-4 (K-1)

1/Tp *10-4 (K-1)

9.0 9.19.1 9.29.2 9.3

Coarse

Fine

-12.8

-12.4

-12.0

-11.6

-11.2

-12.4

-12.0

-11.6

-11.2

-10.8

Fig. 8. The Kissinger plots of the coarse and fine ORS glasses.

Table 5Ec, n and m values of the coarse glasses.

Samplename

Activation energy (kJ/mol) n m

Matusita–Sakka method Kissinger method

CR 350±11 233±9 3.42±0.21 2.42±0.21ORS 439±12 305±10 3.68±0.18 2.68±0.18

20 40 60 80

Inte

nsity

(a.

u.)

Diopside

2θ

217M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

note that during the DTA run, surface and bulk crystallization mayproceed simultaneously but if there is a control of particle size in glass,the crystallization may be governed by particle size. The difference inactivation energies obtained in this study could result from a differentcrystallization mechanism, which is strongly influenced by theparticle size in glasses. As it was also observed in CR and ORS glasssamples a large Ec is associated with surface crystallization while asmall Ec is associated with bulk crystallization in the same glass.

Fig. 9. XRD pattern of the CR glass-ceramic sample.

3.2. XRD studies of the produced glass-ceramic samples

In order to identify the crystalline phase(s), X-ray scans werecarried out on both CR and ORS glass-ceramic samples. In the XRDscan of CR glass-ceramic sample, the d-valuesmatched the card valuesof the diopside (Ca(Mg,Al)(Si,Al)2O6) phase. As seen in the represen-tative XRD pattern of CR glass sample nucleated at 963 K for 4 h andcrystallized at 1135 K for 15 min (Fig. 9), all the diffraction peaks canbe indexed as arising from the reflection planes of the diopside-alumina phase.

Fig. 10 is the X-ray pattern of ORS glass sample nucleated at 988 Kfor 2 h and crystallized at 1188 K for 15 min. X-ray diffraction analysisrevealed that the main crystalline phase that occurred in ORS glass-ceramic sample was diopside (Ca(Mg,Al)(Si,Al)2O6). In ORS glass-ceramic sample, not only the coal fly ash was used, but also the redmud and silica fume were used to produce glass samples. With theaddition of red mud and silica fume, the chemical composition of ORSglass-ceramic sample was closed to the chemical composition of theCR glass-ceramic sample. Therefore, the diopside phase was alsoexpected in the ORS glass-ceramic sample.

tens

ity (

a.u.

)

Diopside

3.3. SEM studies of the produced glass-ceramic samples

Cross-sectional SEM investigations were carried out to character-ize the crystalline morphology in the bulk of the sample. Themicrostructural studies were performed on the bulk glass-ceramicsamples using SEM. Fig. 11 shows the cross-sectional SEMmicrographof the CR glass-ceramic sample. It can be seen from Fig. 11 that tinyequiaxed crystallites uniformly dispersed in the cross-sectional areaof the CR glass-ceramic sample with an average crystalline size of0.25 μm.

Fig. 12 is a representative SEM micrograph of the cross-sectionalarea of the ORS glass-ceramic sample. As can be seen from Fig. 12 thatboth locally oriented dendritic crystalline growth and a significantnumber of leaf-shaped crystals occurred in the ORS glass-ceramicsample. SEM investigations confirm that bulk crystallization is apredominant mechanism in both CR and ORS glass-ceramic samplesas predicted by the Ozawa equation.

Table 6Ec, n and m values of the fine glasses.

Samplename

Activation energy (kJ/mol) n m

Matusita–Sakka method Kissinger method

CR 372±17 369±14 1.36±0.10 1.36±0.10ORS 450±18 444±12 1.24±0.11 1.24±0.11

4. Discussions

In this study, the kinetic parameters of both coarse and fine glasssamples obtained from waste materials were estimated under non-isothermal conditions applying three different equations, namely,Kissinger [13], Matusita–Sakka [14,15] and Ozawa [16]. For CR andORS glass samples, the Eck values of coarse particles obtained fromKissinger equation are lower than the Ec values of coarse particlesdetermined from the Matusita–Sakka method. However, activationenergy values of crystallization for fine glasses estimated fromKissinger and Matusita–Sakka equations are close to each other. Thedifference observed in activation energy values for coarse glasses isdiscussible when the Kissinger equation is used for the estimation ofEc. Matusita–Sakka stated that the Kissinger equation can be used onlywhen crystallization occurs on a fixed number of nuclei as in the caseof fine glass samples. This result confirms that Ec value will be correctif the crystallization occurred on a fixed number of nuclei however,the analysis of crystallization data using the Kissinger equation, yieldsincorrect values for Ec if nucleation and crystallization occursimultaneously. This result further confirms the necessity of usingthe Matusita–Sakka equation in determining Ec if the glass is initiallyunsaturated with nuclei. The Kissinger equation can be used todetermine Ec only when the glass is fully nucleated with surface orbulk nuclei prior to crystallization so that crystal growth at a differentα occurs always on a fixed number of nuclei. All equations gave similarresults for the crystallization activation energy of fine glasses(allowing for experimental errors, Ec values of all equations areclose to each other).

20 40 60 80

In

2θ

Fig. 10. XRD pattern of the ORS glass-ceramic sample.

Fig. 11. Cross-sectional SEM micrograph of CR glass-ceramic sample.

218 M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

As seen from Tables 5 and 6, crystallization of fine glasses occurswith higher activation energy of crystallization than coarse glasses forall glass samples. This result may be understood that the surface strainof the grain boundaries of the glasses effects on the nucleation andgrowth of crystals inside the glass. The applied energy from DTA maybe exhausted in the part for the crystallization inside the glass and theother part for the surface strain from the grain boundaries of the glassparticles. Since the small sized glasses have large surface grain energycompared with the large sized glass particles, thus the more energy isneeded relatively for the crystallization in the small sized glass sample[23]. It can be clearly seen that activation energy is dependent ontemperature, i.e., the higher crystallization temperature is correlatedwith lower activation energy. In the case of surface nucleation, smallerparticle size with its relatively large specific surface area helps theoccurrence of crystallization and thus decreases the crystallizationtemperature. Consequently, Ec is larger for fine glass samples than thatobserved for coarse glasses.

The explanation for different n and m values obtained for coarseand fine glasses can be discussed in terms of glass particle size. Largeparticles have much less surface area and few nuclei are formed. Inthis case, the dominant process would be the growth of nuclei. Onheating above Tg, the nuclei are surrounded by liquid and their growthwould be three-dimensional. As seen from Table 5, n values of CR andORS coarse glasses indicated the three-dimensional bulk crystalliza-tion. The larger surface area of smaller particles will contain largernumbers of nuclei. Supposing that glass surface is completelycrystallized, crystal growth proceeds one-dimensionally from thesurface to the interior of the glass [25]. This statement can be clearly

Fig. 12. Cross-sectional SEM micrograph of ORS glass-ceramic sample.

seen from Table 6. The n values of fine CR and ORS glasses obtainedexperimentally are 1.36±0.10 and 1.24±0.11, respectively. n valuesof fine samples showed that the crystallization mechanism is a one-dimensional surface crystallization. The crystallization mode of a glasshas a practical importance in the usage of it and also in the fabricationof a glass-ceramic since the surface crystallization mode mayintroduce huge thermal expansion difference at the boundarybetween the glass phase and crystallized phase, building up hightensile stress. This high tensile stress at the interface mostly causestotal failure of the glass. On the other hand, in the case of the bulkcrystallization mode where the crystal growth occurs at the finelydistributed precursor nuclei in the glass, the huge thermal expansioncoefficient gradient across the whole body does not occur and theglass body is safe against the thermal failure [26]. Therefore, from thisperspective the bulk crystallization is desirable compared to thesurface crystallization. The effect of glass particle size on thecrystallization kinetics of an iron-rich glass from a nickel leachingwaste was investigated by Romero et al. [12]. Similar results for n andm values both coarse and fine particles were reported. It was statedthat the crystallization mode and the dimensionality of crystals arestrongly dependent of the glass particle size. In their study, coarseparticles showed three-dimensional crystal growth controlled bydiffusion while the fine particles led to an interface reactionmechanism with two-dimensional growth of crystals.

The studied glasses are in the SiO2–Al2O3–Fe2O3–CaO quartetsystem since the chemical compositions of mixed wastes are mainlycomposed of SiO2, Al2O3, Fe2O3 and CaO.With the solid-state reactionsand the rearrangement of these structures, crystalline phases occur inthe glassymatrix when the heat treatment process was applied on theglass samples. This must require breaking and reforming of the Si–O,Al–O, Fe–O and Ca–O bonds. The single bond strengths of the Si–O, Al–O, Fe–O and Ca–O bonds are 445, 423, 335 and 128 kJ/mol,respectively [27–29]. The activation energies of crystallization forORS and CR glasses are close to Si–O, Al–O and Fe–O bond strengths.This result is consistent with the mechanism requiring the breaking ofSi–O, Al–O and Fe–O bonds and also rules out their possibilities as areaction controlling kinetics in forming crystalline phases in the glassymatrix. The activation energy of crystallization for ORS glass is higherthan the Ec values of CR glass. It was observed that the increase inFe2O3 content in the glass compositions caused a decrease in thecrystallization activation energies of the glasses. Since the Fe2O3

content of CR glass is higher than that of ORS glass, Ec value of it islower than the Ec value of ORS glass. This result was also observed inthe other studies [30,31]. Romero et al. [12] also reported that highiron oxide content resulted to the high crystallization tendency ofglasses. Iron oxides lower glass viscosity and consequently increasethe crystal growth rate, which also lowers the crystallizationtemperatures. Despite that it is usually an intermediate glass networkion, the Fe3+ could act as a modifier of the glass structure, breakingthe Si–O bonds [30]. Therefore, the Ec value of ORS glass is higher thanthat of the CR glass sample.

Only a few researchers studied on crystallization kinetics of coal flyash based glasses. Activation energy values of crystallization forglasses produced from coal fly ashes were found 283 and 318 kJ/molin the previous studies [11,32] which are lower than the Ec valuesobtained in this study since the chemical compositions of the studiedglassy systems are different from each other. The calculated activationenergy value of CR glass is close to the value of 370 kJ/mol found byCioffi et al. [3] for a glass produced from coal fly ashwhile the Ec valuesof ORS glass are higher than that value. The activation energy valuesfor crystallization are lower for the incinerator fly ashes than for thecoal fly ash containing glasses, which are in the SiO2–Al2O3–Fe2O3–

CaO quartet glassy systems. The activation energy values ofcrystallization for glasses obtained from incinerator fly ashes are inthe range of 293–468 kJ/mol [2,10], while the Ec values of glassesproduced from industrial wastes (such as coal fly ash, glass cullet and

219M. Erol et al. / Journal of Non-Crystalline Solids 357 (2011) 211–219

soda-lime) are in the range of 534–545 kJ/mol [1,4]. Romero et al. [12]calculated that the activation energy of the glass produced from nickelleaching waste is in the 349–423 kJ/mol interval which is very close tothe Ec value of the CR glass obtained in this study.

5. Conclusions

From the experimental results the following conclusions can bedrawn:

(1) DTA results showed that the crystallization peak temperaturesincreased with the increase in particle size.

(2) The crystallization behavior of glass samples has beeninvestigated under non-isothermal conditions. Using theOzawa equation, the Avrami constants (n) were calculated as3.42±0.21 for coarse CR glasses and 3.68±0.18 for coarse ORSglasses, indicating that bulk nucleation occurs in both glassesby three-dimensional growth. n values of fine samples showedthat the crystallization mechanism is one-dimensional surfacecrystallization.

(3) The crystallization activation energies of coarse glasses are inthe range of 233–439 kJ/mol while the activation energies offine glasses change in the range of 369–450 kJ/mol. Theactivation energy values for crystallization increased with thedecrease in particle size of the waste glasses.

(4) XRD results revealed that the CR and ORS glass-ceramicsamples produced from the different mixed wastes whichhad similar chemical compositions have the same crystallinephase determined as the diopside phase.

(5) SEM investigations revealed that crystallites were uniformlydispersed in themicrostructure for both glass-ceramic samples.The cross-sectional SEM investigations on the bulk CR and ORSglass-ceramic samples indicated that this result is in wellagreement with those determined by the Ozawa equation.

References

[1] A. Alvarez-Mendez, L.C. Torres-Gonzalez,N. Alvarez, L.M. Torres-Martinez, J. Non-Cryst.Solids 329 (2003) 73–76.

[2] Y.J. Park, J. Heo, Ceram. Int. 28 (2002) 669–673.[3] R. Cioffi, P. Pernice, A. Aronne, G. Quattroni, J. Mater. Sci. 28 (1993) 6591–6594.[4] A.A. Francis, R.D. Rawlings, R. Sweeney, A.R. Boccacini, J. Non-Cryst. Solids 333

(2004) 187–193.[5] M.L. Öveçoğlu, B. Kuban, H. Özer, J. Eur. Ceram. Soc. 17 (1997) 957–962.[6] N. Karatepe, A. Ersoy-Mreiçboyu, S. Küçükbayrak, Environ. Technol. 20 (1999)

377–385.[7] Z. Strnad, Glass-ceramic Materials, Elsevier Science, Amsterdam, 1986.[8] X.J. Xu, C.S. Ray, D.E. Day, J. Am. Ceram. Soc. 74 (1991) 909–914.[9] H.C. Park, S.H. Lee, B.K. Ryu, M.M. Son, I. Yasui, J. Mater. Sci. 31 (1996) 4249–4253.

[10] M. Romero, R.D. Rawlings, J.M. Rincon, J. Eur. Ceram. Soc. 19 (1999) 2049–2058.[11] M. Erol, S. Küçükbayrak, A. Ersoy-Meriçboyu, M.L. Öveçoğlu, J. Eur. Ceram. Soc. 21

(2001) 2835–2841.[12] M. Romero, M. Kovacova, J.Ma Rincón, J. Mater. Sci. 43 (2008) 4135–4142.[13] H.E. Kissinger, Anal. Chem. 29 (1957) 1702–1706.[14] K. Matusita, S. Sakka, Y. Matsui, J. Mater. Sci. 10 (1975) 961–966.[15] K. Matusita, S. Sakka, J. Non-Cryst. Solids 38–39 (1980) 741–746.[16] T. Ozawa, Polymer 12 (1971) 150–158.[17] M. Erol, S. Küçükbayrak, A. Ersoy-Meriçboyu, T. Ulubaş, En. Conver. Manag. 46

(2005) 1319–1331.[18] M. Erol, Key Eng. Mater. 264–268 (2004) 1883–1886.[19] M. Erol, S. Küçükbayrak, A. Ersoy-Meriçboyu, Chem. Eng. J. 132 (2007) 335–343.[20] H. Yinnon, D.R. Uhlmann, J. Non-Cryst. Solids 54 (1983) 253.[21] C.S. Ray, W. Huang, D.E. Day, J. Am. Ceram. Soc. 74 (1991) 60–66.[22] W. Li, B.S. Mitchell, J. Non-Cryst. Solids 255 (1999) 199–207.[23] S.J. Kim, J.E. Kim, Y.H. Rim, Y.S. Yang, Solid State Commun. 131 (2004) 129–133.[24] C.S. Ray, D.E. Day, Thermo. Acta 280–281 (1996) 163–174.[25] B.J. Costa, M. Poulain, Y. Messaddeq, S.J.L. Ribeiro, J. Non-Cryst. Solids 273 (2000)

76–80.[26] Y.M. Sung, J.H. Sung, J. Mater. Sci. 33 (1998) 4733–4737.[27] J.H. Jean, Y.C. Fang, S.X. Dai, D.L. Wilcox, J. Am. Ceram. Soc. 84 (2001) 1354–1360.[28] M.J. Hyatt, N.P. Bansal, J. Mater. Sci. 31 (1996) 172–184.[29] J. Husband, F. Aguirre, P. Ferguson, R.B. Metz, J. Chem. Phys. 3 (1999) 1433–1437.[30] L. Barbieri, A. Corradi, I. Lancelotti, A.P.N. De Oliveira, O.E. Alarcon, J. Am. Ceram.

Soc. 85 (2002) 670–674.[31] J. Williamson, J. Tipple, P.S. Rogers, J. Iron. Steel Res. Int. 206 (1968) 898–903.[32] M. Erol, S. Küçükbayrak, A. Ersoy-Meriçboyu, M.L. Öveçoğlu, Key Eng. Mater.

264–268 (2004) 1931–1936.