Contents lists available at ScienceDirect

Journal of the European Ceramic Society

journal homepage: www.elsevier.com/locate/jeurceramsoc

Original Article

Effect of the phase transformation on fracture behaviour of fused silicarefractories

Yajie Daia,b,⁎, Yucheng Yina,b, Xiaofeng Xua, Shengli Jinc, Yawei Lia,b, Harald Harmuthc

a The State Key Laboratory of Refractories and Metallurgy, Wuhan University of Science and Technology, Wuhan 430081, ChinabNational-Provincial Joint Engineering Research Center of High Temperature Materials and Lining Technology, Wuhan 430081, Chinac Chair of Ceramics, Montanuniversitaet, Peter-Tunner Strasse 5, Leoben 8700, Austria

A R T I C L E I N F O

Keywords:Fused silica refractoryLow-high-temperatureCristobaliteMicrocrackingFracture process zoneDigital image correlation

A B S T R A C T

In this paper, the influence of phase transformation on the properties and fracture behaviour of fused silicarefractory was investigated. The virgin fused silica refractory is amorphous, and possible failure is attributed tothe propagation of a single crack in the structure. Due to the crystallization and phase transformation of low-/high- temperature cristobalite subpolymorphs occurring during the heat treatment, microcracks are formedespecially in the matrix and at the grain boundary. This microcracking enables the development of sizablefracture process zone, which is responsible for the increase of specific fracture energy even with the decrease ofstrength. Therefore, the heat-treated specimens exhibit lower brittleness and higher strain tolerance beforefailure compared with the virgin fused silica refractory. All of these properties represent a better thermal shockresistance. Furthermore, microcracking causes a characteristic temperature dependence of Young’s Modulus dueto phase transformation and partial crack closure at increased temperatures.

1. Introduction

Silica refractory is a special category of non-basic refractory pro-ducts containing more than 93wt% silica [1]. Silica is rather uniqueand complex because of its diverse polymorphs and subpolymorphs[2,3]. Thereinto, cristobalite and tridymite are the mineral silica phasesin silica refractories. Due to the low- and high- temperature sub-polymorphs of these two phases and the induced microstructurechange, silica refractories often exhibit temperature dependent elasticand inelastic properties [4,5]. The thermal expansion coefficient ofcristobalite, which becomes relatively small or even slightly negative attemperature higher than 800 °C, is predominant for silica refractories[6]. It was also proved by impulse excitation investigation that thedamage accumulation is more crucial for silica materials heated tolower temperatures [7]. Nevertheless, its good performance at hightemperature, remarkable refractoriness and good chemical resistance tocertain acidic components enable silica refractories specially used forconstructing coke ovens, glass melter roof, nonferrous metallurgicalfurnaces etc. [8,9].

In service, they undergo high temperature, thermal shock, me-chanical stress, corrosion and erosion. The thermo-mechanical stressinduced failure is one of the predominant factors limiting longevity ofrefractory linings. The compressive creep behaviour and the cyclic

fatigue of silica refractories were investigated [10–12]. Two principaltypes of silica refractories are commercially available for the users: oneis the conventional silica refractory produced from quartz raw materialsand mainly consisting of cristobalite and tridymite; the other is thefused-silica based refractory [5,10,13]. The latter one is amorphous andshows smaller thermal expansion compared to the crystalline silicarefractories at least in the low temperature region, which is believed tocause better thermal shock resistance. Although the interest in its ap-plication has grown, the fracture behaviour of fused silica refractorieswas not studied yet. The crystallization of cristobalite from amorphoussilica will gradually take place during their service lifetime. The crys-tallization as well as the further possible phase transformation cause acritical effect on the thermo-mechanical performance of fused silicarefractories. The influence of sintering temperature and additives (i.e.zircon and alumina) on the crystallization of fused silica based ceramicshave been investigated [14,15]. Furthermore, the phase transformationof low- /high- temperature cristobalite was simulated and its auxeticbehaviour was also studied to extend the understanding of this crys-talline silica [16,17]. The strength of fused silica refractories reducesdue to the devitrification of amorphous fused silica to cristobalite, andis attributed to microcracking of the cristobalite caused by the phasetransformation between the high-temperature cubic cristobalite andlow-temperature tetragonal cristobalite (about 2.8%–5% volume

https://doi.org/10.1016/j.jeurceramsoc.2018.08.040Received 5 June 2018; Received in revised form 21 August 2018; Accepted 24 August 2018

⁎ Corresponding author at: The State Key Laboratory of Refractories and Metallurgy, Wuhan University of Science and Technology, Wuhan 430081, China.E-mail address: yajie.dai@wust.edu.cn (Y. Dai).

Journal of the European Ceramic Society 38 (2018) 5601–5609

Available online 25 August 20180955-2219/ © 2018 Elsevier Ltd. All rights reserved.

T

contraction in the temperature range 200–250 °C during cooling)[18,19]. The strength of fused silica is dependent on cristobalite content[20]. Besides, by acoustic emission technology and fatigue test, An-dreev et al. [10] have proven that the presence of initial microcracks insilica refractories reduces the brittleness and limits the severe damage.

In the present paper, the temperature dependent Young’s Modulusand dilatation as well as the fracture behaviour of fused silica re-fractories is investigated. The main purpose is to receive insight into thetemperature dependent fracture properties and to understand the in-fluence of crystallization and phase transformation on the fracture be-haviour of fused silica refractories.

2. Experimental

2.1. Materials

The fused silica refractory investigated in this work is a commercialproduct from an industrial producer. The received fused silica re-fractory is unfired and consist of 98.63 wt% SiO2 and trace amount ofimpurities (mainly Al2O3, CaO, Fe2O3 and TiO2). The fused silica re-fractory was heated up to 1400 °C with a heating rate of 3 °C/min andheld at the target temperature for 0.5, 2, 5 and 10 h, respectively. Thecooling rate was also preset to 3 °C/min. However, the cooling ratebelow 300 °C slowed down due to the heat capacity of the furnace andprolonged the dwell time at low temperature range. Normally, theworking temperature of fused silica refractory is approximately1100–1200 °C. In order to imitate the crystallization degree in the firstseveral weeks of service, a higher temperature was adopted to accel-erate the crystallization of the fused silica. The virgin fused silica re-fractory specimens were named as 0H, which was amorphous as shownin Fig. 1. The other specimens were successively named 0.5H, 2H, 5Hand 10H, according to their dwell time at 1400 °C. The crystallization offused silica to cristobalite took place during the aforementioned heattreatment and the crystallization degree increased with the dwell time.

The selected physical properties of each specimen are shown inTable 1. The bulk density and apparent porosity were measured viaArchimedes method. The bulk density reached a minimum value forspecimen 2H, which has the highest apparent porosity. The theoreticdensity of low-cristobalite and glass phase are 2.32 g/cm3 and 2.20 g/cm3, respectively [2]. As seen from Table 1, the true density of virginfused silica refractory was 2.27 g/cm3 and increased due to the crys-tallization of fused silica. For the specimen 2H, 5H and 10H, the truedensity is rather close to the theoretic density of low-cristobalite.Meanwhile, the Young’s Modulus at room temperature decreased sig-nificantly because of the devitrification of fused silica refractory. Less

than 10% of the original stiffness remains after the heat treatment.

2.2. Investigation methods

The chemical and physical characterization of the studied materialswas determined by established measurement techniques. The phasecomposition was investigated using X-ray powder diffractometer(X’pert Pro, PANalytical, Netherland), set to copper radiation (CuKα,λ=1.5418 Å) at 40 kV/40mA. The micrographs were obtained byscanning electron microscope (Nova 400 Nano-SEM, FEI Company,USA). True density measurements were carried out using automaticTrue Density Analyzer (AccuPyc II 1340, Micromeritics InstrumentCorporation, US) according to GB/T 5071-2013 (Chinese standard ofrefractory materials—determination of true density). The dilatationcurves up to 1400 °C were measured using connecting rod dilatometer(DIL 402 Supreme, NETZSCH, Germany) for cylindrical specimens withdiameter of 8mm. Room temperature and high temperature Young’sModulus were measured via impulse excitation technique, using a re-sonant frequency damping analyzer (RFDA-HTVP1600, IMCE,Belgium). The impulse excitation measurements were performed up to1200 °C with a heating and cooling rate of 3 °C/min, the dwell time at1200 °C for each specimen was 30min.

Wedge splitting tests (WST) were performed at room temperature toinvestigate the fracture behaviour of fused silica refractory specimens.The experimental set-up is shown in Fig. 2: the applied vertical force FVis transformed to a higher horizontal force FH with the help of a loadtransmission equipment (wedge, rollers, load transmission pieces) [21].The specimen size used is 100mm×100mm×65mm, the initialnotch length a0 equals to 12mm, and the wedge angle is 10°. Theloading rate was 0.5 mm/min. Due to the application of the loadtransmission components and a high ratio of fracture area to specimenvolume, the WST enables stable crack propagation for brittle or quasi-brittle materials, such as refractory, concrete, rock etc. [22]. A con-tacting extensometer was used to measure the horizontal displacementat the load points. A load-displacement curve was recorded by the loadcell for each specimen and the fracture parameters of the fused silicarefractory specimens. The specific fracture energy GF’ is determined byintegrating the area under the load-displacement curve divided by theligament area A, as shown in Eq. (1). The ligament area A is calculatedas b× h, which are the width and the height of the ligament, respec-tively. δH is the horizontal displacement, δ is the horizontal displace-ment at 15% of the maximum load (FH, max). The wedge splitting testsare terminated before the force reaches zero to avoid the contact be-tween the wedge and the specimens.

Fig. 1. XRD patterns of the studied silica refractory specimens (C: cristobalite;T: tridymite).

Table 1Physical properties of silica refractory specimens.

Specimen 0H 0.5H 2H 5H 10H

Bulk density (g/cm3) 1.81 1.79 1.71 1.78 1.78Apparent porosity (%) 17.8 21.1 24.4 21.7 22.1True density (g/cm3) 2.27 2.29 2.31 2.31 2.31E0 (GPa) 26.6 1.8 1.7 1.8 1.4

Fig. 2. Experimental set-up of the wedge splitting test with a digital camera.

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5602

∫=GA

F dδ´ 1F

δ

H H0 (1)

The nominal notch tensile strength σNT were calculated from themaximum horizontal load FH,max with the following equation:

= +σF

bhy

h(1 6 )NT

H ,max

(2)

Here, y is the vertical distance between the initial load point to thecenter of ligament. The nominal notch tensile strength includes both thetensile stress and the flexural stress.

A monochrome CCD camera with 5 megapixels was placed in frontof the specimen for acquiring the digital images successively during theentire mechanical test. The image acquisition frequency was two sec-onds per image. The images analysis was performed by digital imagecorrelation (DIC) technique and the post-processing software adoptedwas the MatchID DIC-2D software developed by the University ofLeuven. The Digital image correlation technique is recognized for itsability of in-situ monitoring the fracture process of the materials underdifferent loading stages [23]. It is a non-contact full-field measurementmethod which will not interfere the fracture process. The main prin-ciple of DIC is determining the full field displacement by tracking thehomologous points on the specimen surface. Thus, the contrast andrandom texture of gray level on the specimen surface is rather im-portant. In this study, a thin layer of white spots distributed on a blacksubstrate was applied instead of the original texture to increase thecontrast. DIC has already been used to study the fracture behaviour ofmetals, concrete, polymers etc. [24,25]. In the domain of refractories,this technique was successfully used to locate the crack tip as well as thefracture process zone, such as for magnesia refractory materials etc.[26,27]. The DIC correlation parameters are critical to the correlationprocess. In this study, a subset size of 32× 32 pixels2, a step size of 8pixels, and a strain window size of 11×11 pixel2 were chosen. Theinfluence of these parameters on the correlation performance has beeninvestigated widely [23,28].

3. Results and discussion

3.1. Influence of the crystallization and phase transformation on themicrostructure of fused silica refractory

The virgin fused silica refractory was amorphous, and the silicaglass grains can be clearly seen. Once it was exposed to the heattreatment process as mentioned in Section 2, the devitrification oc-curred. As shown in Fig. 3, the fused silica specimen after heat treat-ment turns into pure white as observed by naked eye. The devitrifica-tion of the fused silica refractory is the first hint of the crystallizationand the formation of cristobalite.

Fig. 4 exhibits the microstructure of each studied fused silica re-fractory specimens. In specimen 0H, the mm-sized fused silica grainsare embedded in the matrix with fine grains and pores. As for specimen0.5H, 2H, 5H and 10H, the regions transformed from silica glass intocristobalite within the grains show conchoidal microcracking and the“roof tile” structure on margin of fused silica grains are observed. Dueto the phase transformation and shrinkage of grains, the porosities ofspecimens 0.5H, 2H, 5H and 10H are higher than that of the specimen0H and the specimens are getting looser. The porosity reached aminimum for specimen 2H and reduced for specimen 5H and 10H(Table 1) as these two specimens were better sintered by prolonging thedwell time at 1400 °C. Thus, the microstructures of heat treated speci-mens are the result of the combined action of phase transformation anddensification caused by the sintering.

As shown in Fig. 5, the micrographs of specimen 0H and 2H in Fig. 4are enlarged for better illustrating the change of microstructure. Theunique “roof tile” structure within cristobalite grain is consequence ofthe microcracking, which is caused by the phase transformation fromcubic high-temperature to tetragonal low-temperature cristobalite.According to a review paper, the densities of cubic and tetragonalcristobalite are 2.20 g/cm3 and 2.32 g/cm3, respectively [2]. As a re-sult, a remarkable shrinkage by 5.2% is caused by the transformation.Besides the microcracking inside the cristobalite grains, the micro-cracking may also occur along the grains attributed to the shrinkage ofgrain (as seen in Fig. 4-specimen 2 H) during the cooling process. Asknown from the XRD patterns (Fig. 1), the tridymite phase contents instudied specimens are rather low. Furthermore, the influence of thephase transformation from the high-temperature to the low-tempera-ture tridymite is less crucial for the microstructure as their theoreticaldensities are close (2.24 g/cm3 and 2.26 g/cm3 for hexagonal high-temperature and monoclinic low-temperature tridymite, respectively)and shows only 0.9% displacive low-/high- transformation [29]. Thus,the effect of the tridymite on the fracture behaviour will be ignored andthis paper will focus on the influence of phase transformation of cris-tobalite on the fracture behaviour of fused silica refractories.

3.2. Temperature dependence of Young’s Modulus and dilatation behaviour

The Young’s Modulus of specimen 0H and 2H at room temperatureare 26.6 GPa and 1.7 GPa, respectively. Even if the porosity is taken intoaccount (17.8% and 21.1%), the Young’s Moduli are much smaller thanthe theoretical values of the dense materials (Young’s modulus of glassand tetragonal cristobalite are 72.2 GPa and 65.2 GPa [2]), especiallyfor specimen 2H. With only 3.3% increase of porosity, less than 10% ofthe Young’s Modulus are maintained for specimen 2H compared to theoriginal specimen 0H. As it is well known, the Young’s Modulus is af-fected not only by the porosity but also the pore shape [30]. Actually,the significant reduction of Young’s Modulus of the heat treated

Fig. 3. Surface of fused silica specimen before (a) and after (b) heat treatment as observed by naked eye.

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5603

specimens is caused by the formation of the extremely oblate pores, i.e.the microcracks and crack-like flaws. The microcrack and crack-likeflaws increase slightly the porosity but shown significant influence onthe Young’s Modulus at room temperature. It is evident that the mi-crostructure dominates the Young’s Modulus instead of the phasecomposition itself.

Besides the Young’s Modulus value at room temperature, the tem-perature dependent Young’s Modulus are shown in Fig. 6 for specimen

0H and 2H. For each specimen, two complete heating-cooling cycleswere performed. As the used heating/cooling rate were 3 °C /min anddwell time at 1200 °C was 0.5 h, the first heating-cooling cycle of spe-cimen 0H can imitate and monitor the influence of crystallizationprocess and phase transformation on the evolution of Young’s Modulusduring the sintering process. During the first heating, the Young’sModulus increases slowly with temperature until 900 °C for specimen0H as it is amorphous.

Fig. 4. Microstructures of each studied silica refractory specimen (0H: virgin fused silica refractory; 0.5H, 2H, 5H and 10H: fused silica refractory sintered at 1400 °Cfor 0.5, 2, 5 and 10 h, respectively).

Fig. 5. Microstructures of silica refractory specimen 0H (a) and 2H (b).

Fig. 6. Temperature dependence of Young’s Modulus for specimen 0H (a) and 2H (b).

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5604

With the gradual crystallization, the Young’s Modulus reaches35 GPa at the maximum temperature. The Young’s Modulus curveduring the subsequent cooling process does not follow the same path asfor heating. A steep degradation of Young’s Modulus is observed ataround 200 °C and it drops into 12.4 GPa at room temperature (47% ofthe initial room temperature Young’s Modulus). The difference betweenthe room temperature Young’s Modulus before and after first heating-cooling cycle confirms the damage accumulation in the microstructureof fused silica refractory, which is caused by the crystallization andlow-/high subpolymorphs phase transformation. The further behaviourshould be interpreted commonly with Fig. 7, which characterizes thedilatation behaviour. According to this, up to approximately 1050 °Cspecimen 0H shows a small length increase which is associated with thelow coefficient of thermal expansion of silica glass. Simultaneously theYoung´s Modulus increases. It is assumed, that in this temperature re-gion of decreasing specimen length softening of the residual glass phaseoccurs, and simultaneously crystallization of high temperature cristo-balite. Any length or volume change due to the density difference be-tween cristobalite and glass is still equilibrated by the viscous beha-viour of the residual glass phase, thus avoiding significantmicrocracking in this high temperature range. The considerable drop inYoung’s Modulus shown for the cooling during the first circle clearlyproofs that high temperature cristobalite has been formed. This hap-pens during the high temperature period in the heating and coolingcircle of specimen 0H, and the extent of cristobalite formation evenrises in the circle two. The high Young’s modulus at temperatures abovethe cristobalite transformation shows that microcracks are not yetformed during the first crystallisation, due to the accommodation ofstrains by the glass phase. The transformation from high to low tem-perature cristobalite is associated with a large decrease of Young’sModulus. This shows clearly that microcracks occur at that occasion.Now, at relatively low temperatures, glass phase – if it is still present –is by far too viscous to accommodate length changes. Microcrack for-mation, as observed by the drop of Young’s modulus, necessitates anuneven volume or length change in the whole specimen volume. Twofurther options remain. Transformation of glass phase to cristobaliteoccurs first in the matrix and at the grain boundaries, due to the con-tribution of surface and grain boundary diffusion. It might be the casethat because of the same reason transformation of high to low tem-perature cristobalite also occurs first in the matrix and close to thesurface of grains. Another reason might be that anisotropy of the lengthchange is responsible for microcrack formation. This may be caused bythe fact that high temperature cristobalite is of cubic symmetry, whilelow temperature cristobalite is tetragonal. While this investigationcannot give a clear evidence for one or the other option, the first oneseems more plausible according to the curve shape of Fig. 6(a): the

transition temperature range shows a considerable width, which mightgive evidence for an uneven transformation, possibly starting in thematrix prior to the bulk volume of the grains. By comparing the secondheating-cooling cycle to the first one, the room temperature Young’sModulus reduction is more severe in the first cycle. It is assumed thatthe Young’s Modulus at room temperature will gradually decrease withthe increasing degree of crystallization.

Specimen 2H was already partially transformed into cristobalite.The curve shape of the temperature dependent Young’s Modulus offused silica refractory specimen 2H is similar to that of a conventionalsilica refractory which shows the critical transformation temperature ofcristobalite [7]. When heating to 200 °C, a steep increase of Young’sModulus is observed, and the Young’s Modulus steadily increases to635% of the initial value at room temperature until 1200 °C. During theholding period at 1200 °C, the Young’s Modulus keeps growing due tothe further closure of the microcracks. For silica refractories, the in-elastic behaviour is considerable at high temperature region(> 1000 °C). After the first cooling, specimen 2H lost 62% of the ori-ginal stiffness. However, after completing the second heating-coolingcycle, no significant further Young’s Modulus decrease occurs. Asproven by W. Pabst et al., further irreversible changes might be pre-vented by crack healing at 1200 °C [7].

The cristobalite has a higher coefficient of thermal expansion (CTE)than that of amorphous silica glass [14]. Thus, fused silica refractoryshows lower dilation compared to that of the traditional silica re-fractory. As shown in Fig. 7, the relative length change of original fusedsilica refractory is rather close to zero. At temperature above 1050 °C,the relative length change begins to decrease and becomes negativeafter 1200 °C due to the softening of glass. As the thermal expansioncoefficients of materials are less influenced by the porosity and mi-crocrack, the phase compositions play a key role in the dilatation be-haviour. The length change of pure polycrystalline cristobalite due tothe low-/High- temperature cristobalite transformation occurs around260 °C [6]. As the specimen 0.5H, 2H, 5H and 10H are partially crys-tallized to cristobalite, the relative length changes increase significantlyaround the transformation temperature of cristobalite. With the in-crease of crystallization degree, the 5H and 10H show higher plateauvalue of relative length change compared with specimens 0.5H and 2H.

3.3. Fracture behaviour of fused silica refractory specimens

3.3.1. Fracture parameters measured by WSTFig. 8 shows the load-displacement curves obtained by the WST for

fused silica refractory specimens. The maximum horizontal load forvirgin fused silica refractory specimen 0H is 2431 N and is significantlyreduced for the remaining specimens (0.5H, 2H, 5H and 10 H) due tothe microcracking caused by the low-/high- temperature cristobalitephase transformation. However, the specimen 0.5H, 2H, 5H and 10Hexhibit much longer post-peak region compared to specimen 0H.

Table 2 shows the maximum horizontal force Fh,max, nominal notchtensile strength σNT, ultimate horizontal displacement at 15% of themaximum load δ, specific fracture energy GF’, ratio of GF’ to σNT andcharacteristic length lch for each specimen. The nominal notch tensilestrength of heat treated specimen 2H is only 21% of the original fusedsilica refractory specimen 0H, yet 1.9 times the specific fracture energyis consumed for failure of specimen 2H. The ultimate displacement ofheat treated specimens is much higher, which indicate a better strain-to-rupture of these specimens compared with 0H. The parameters GF’/σNT and lch are adopted to evaluate the brittleness of studied materials.The higher these two parameters are, the less brittle the material is.Furthermore, the quantity of lch is double of the thermal shock re-sistance parameter R”” according to Hasselman [31,32]. This confirmsthat a reduction of brittleness is favourable for a better thermal shockresistance. The microstructure of heat treated specimens is simulta-neously influenced by two counteracting processes, which are on theone hand microcracking caused by phase transformation and on the

Fig. 7. Comparison of the dilatation behaviour of silica refractory specimens.

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5605

other hand densification due to the sintering. As the degree of crys-tallization is higher than specimen 0.5H and degree of sintering is lowerthan specimen 5H and 10H, the fused silica refractory 2H shows thehighest porosity and the most reduction of brittleness. Therefore, moreattention will be paid to the comparison between specimen 0H and 2Hto illustrate the influence of phase transformation on the fracture be-haviour of fused silica refractories.

3.3.2. Evaluation of the fracture process by digital image correlationIn this work, the strain in the horizontal direction of each studied

material was evaluated by DIC technique and used as the indicator ofthe fracture process. As demonstrated in Fig. 9, the strain maps of eachspecimen at their peak load are shown. Normally, the fracture initiationof brittle material is onset at its peak load and followed by a brittlefailure. By reducing the brittleness, the fracture initiates in the pre-peakregion and the crack path tends to become more tortuous and showsincreasing crack branching. For specimen 0.5H, 2H, 5H and 10H, thefracture process zones are already developed to a large part at max-imum load. The crack branching is observed.

To further investigate the fracture behaviour of fused silica re-fractories with different microstructure and brittleness, the evolution ofstrain field at different loading stages of specimen 0H and 2H are re-presented in Fig. 10. A lower limit of strain in the X direction (εxx) of0.003 was determined as threshold by evaluating the scatter caused bythe numerical correlation uncertainty and image distortion. More de-tails of this procedure can be found elsewhere [26]. The fracture ofspecimen 0H initiates at the maximum load and only a single crack ispropagated. For specimen 2H, the fracture initiation is detected at 41%of the maximum load in pre-peak region and a remarkable fractureprocess zone is developed due to the microcracks induced by the low-/high- temperature cristobalite phase transformation especially occur-ring at the surface of the grains. The development of sizable fractureprocess zone in heat treated fused silica refractories is rather energyconsuming and enables that the specific fracture energy even rises whenthe strength decreases by microcracking for the heat treated specimens,thus decreasing brittleness. Crack propagation in heterogeneous mate-rials is very complex and the local microstructure governs pre-dominantly the crack trajectory and its branching. Besides the primarycrack and crack branching, the pre-existing imperfections in specimen2H, such as the pores and microcracks, and may form a discontinuousset of micro-cracks under certain stress. These discontinuous micro-cracks may merge into the primary crack and guide the crack propa-gation. Furthermore, because of the friction, grain bridging and inter-locking in the wake of process zone, there is still force transferred be-tween the crack faces. Once the macro-crack is onset, the stress betweenthe crack faces reduces to zero [26]. The surrounding material is re-leased and neighbouring microcracks may re-close as shown inFig. 10(b), i.e. at 40% and 15% of maximum load in the post-peak re-gion.

Fig. 11 shows the evolution of the crack mouth opening displace-ment (CMOD) for specimen 0H and 2H, which is measured at the initialnotch tip of the WST specimen. The X axis is the loading stage, wherethe origin represents the peak load. The pre-peak region locates on theleft of the Y axis and the post-peak is on the right. For specimen 0H, theCMOD in the pre-peak region is in the scale of micrometre and theultimate CMOD at 15% of maximum load in the post-peak region equalsto 0.13mm. The CMOD of specimen 2H increases slowly from 0.01mmat the 41% Fh,max (pre-peak) to 0.1 mm at maximum load. Oncereaching the maximum load, the increase of CMOD speeds up due to thedamage accumulation in the microstructure. The ultimate CMOD forspecimen 2H is 0.87mm, which is almost seven times higher than thatof specimen 0H. It is evident that the specimen after heat treatmentshows a larger strain-to-rupture. For a given displacement or strain (forinstance 0.1 mm), specimen 0H nearly failed due to extensive crackpropagation and specimen 2H still maintains as the further crack ex-tension is limited by the development of a sizable fracture process zonein the structure and the associated lower material brittleness.

3.3.3. Crack trajectories investigated by a fractographic methodThe specimens after wedge splitting test have been used for frac-

tographic investigation to illustrate the fracture behaviour on micro-scale [33]. Together with the DIC evaluation, the whole fracture

Fig. 8. Load-displacement curves of fused silica refractory samples.

Table 2Fracture parameters of studied samples.

Specimen 0H 0.5H 2H 5H 10H

Fh.max (N) 2431 740 509 664 556σNT (MPa) 5.8 1.8 1.2 1.6 1.3δ (mm) 0.2 0.9 1.4 1.1 1.2GF´ (N/m) 54.1 88.5 100.6 96.5 78.8GF´/σNT (μm) 9.3 50.0 84.2 60.9 59.7lch (mm) 35.7 50.9 114.7 69.7 65.4

Fig. 9. Strain maps of each specimen at their maximum load.

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5606

process of studied material will be documented. After wedge splittingtests, the specimens were impregnated with a synthetic resin and then aslice with 10mm thickness was cut parallel to the painted frontal

surface. In order to correlate the crack propagation with the DIC ob-servation (Fig. 10), the paint was removed by polishing and the mi-crostructure of the frontal surface containing the whole fracture tra-jectory was documented with a scanning electron microscope, as shownin Figs. 12 and 13. Three different portions of the crack path have beendistinguished: crack propagation within matrix, cracking along thegrain/matrix interface and transgranular crack propagation. Generally,it is expected that the ratio of crack length along the grain/matrix in-terface to transgranular crack length is a critical figure representing thematerial brittleness: the higher the ratio is, the less brittle the refractorywill be. However, the percentage of transgranular crack length as wellas the crack length along the grain/matrix interface for both specimen0H and 2H are similar. Besides, the tortuosity of specimen 0H, which isdefined as the ratio of crack length measured along the tortuous pri-mary crack path to the primary crack height, is like that of specimen2H. Thus, one-dimensional investigation along the lengthwise directionof the primary crack path may not be sufficient for specimen showingsignificant fracture process zone. As shown in Fig. 12, a single crackpropagates in the structure and the crack faces are rather sharp-edgedfor specimen 0H.

Compared with specimen 0H, the separation of crack faces forspecimen 2H is wider, which increases the displacement and

Fig. 10. Evolution of strain field at different loading stages for specimen 0H (a) and 2H (b).

Fig. 11. Comparison of CMOD at different loading stages for specimen 0H (a)and 2H (b).

Fig. 12. Crack path for specimen 0H (a), enlarged location (b), and crack path along a grain (c).

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5607

contributes to the consumed fracture energy according to Eq. (1). Inaddition, crack branching, partly resulting in cracks propagating par-allel to one another, are observed for heat treated specimen 2H asshown in Fig. 13. The local crack trajectories of specimen 2H are zigzagand their crack faces appear coarser compared with the virgin fusedsilica refractory specimen 0H. It can be observed that the crack pathwithin the micro-structure of heat treated specimens propagates par-tially along the “roof tile” structures where they are close to the grainboundary. This means, that contrary to other grain/matrix structures,cristobalite transformation causes that the peripheral grain regions maybehave as a weaker link than the grain/matrix boundary itself. Thismay additionally contribute to brittleness reduction. As shown inFig. 13(c), parts of former grains are located on two sides of the crack,and the pairs are marked with red double arrows.

4. Conclusions

Fused silica refractory materials showing different thermo-me-chanical behaviour were assessed for demonstrating the effect of crys-tallization and phase transformation on the fracture behaviour. Thevirgin fused silica refractory 0H, which is amorphous, was heated to1400 °C and dwelled at target temperature for different time (0.5H, 2H,5H and 10 H) to enable the crystallization expected to occur during itsservice. The heat-treated specimens were partially devitrified andcrystallized to cristobalite. It has been proven that the crystallizationinitiates at temperature above 1100 °C from the outer edges of thegrains. Due to the low- and high- temperature subpolymorphs of cris-tobalite, the “roof tile” structure microcracking are formed on thesurface of grains which weaken the fused silica refractory. The Young’sModulus and dilatation behaviour, which are influenced by the cristo-balite phase transformation, depend on temperature and heat treat-ment. Steep slopes of Young’s Modulus and dilatation curves are ob-served around 200 °C, which is close to the transformation temperatureof cristobalite phase. The Young’s Modulus at room temperature de-crease after different heating/cooling cycles due to the damage accu-mulation caused by the increase of crystallization degree.

For virgin fused silica refractory, no pronounced fracture processzone is detected, only a single crack propagates. Crack initiation occursat the maximum load. Due to the phase transformation induced mi-crocracking, the strength of the heat treated specimens is lower thanthat of the virgin fused silica refractory. However, the microcrackingsupports the development of sizable fracture process zones, whichcontribute to the deviation from pure linear elastic mechanical beha-viour. The formation of the fracture process zone in heat treated spe-cimens is rather energy consuming even with the reduced stiffness.Consequently, the specimens 0.5H, 2H, 5H and 10H behave less brittle

than the original fused silica refractory. The phase transformation in-duced microcracks together with the sintering and the thereby causeddensification govern the microstructure of fused silica refractory afterheat treatment: both the degree of crystallization and of sintering in-crease with the prolongation of dwell duration. By combined action ofmicrocracking and densification, specimen 2H shows the highest por-osity and lowest brittleness compared with the other specimens. Forspecimen 2H, the fracture process zone development is onset in the pre-peak region and its crack faces appear rougher. Besides, the strainbearing capacity of specimen 2H is much larger than that of specimen0H, which indicates that the heat treated specimen exhibits a betterthermal shock resistance.

Special attention should be paid to the processes of microcrackingcausing brittleness reduction. Formation of high temperature cristoba-lite out of the glass phase during first heating up does not yet causemicrocracking and brittleness reduction. Obviously, softening of stillpresent residual glass phase accommodates for all possible volumechanges. Microcracking originates from first cooling below the tem-perature of transformation from high to low temperature cristobalite.This necessitates that associated volume contraction is not homo-geneously distributed in the whole volume. It seems likely that duringcooling this transformation occurs at first in the matrix, and later and atlower temperatures in the bulk volume of the grains. It remains un-answered whether a possible anisotropy of the length change also playsa role. This consideration seems important, because it gives evidencethat the brittleness reducing microcrack network is not yet presentbefore the first cooling of the material. This means that it behaves stillbrittle under service conditions if the furnace has not been cooled downbefore.

Acknowledgements

This work was supported by the State Key Laboratory of Refractoriesand Metallurgy (No. 2018QN20) and the Special Funding forTechnological Innovation of Hubei Province (2016AHB026), China.

References

[1] EN ISO10081-1: Classification of Dense Shaped Refractory Products - Part 1:Alumina-Silica Products, European Committee for Standardization, Brussels, 2005.

[2] W. Pabst, E. Gregorová, Elastic properties of silica polymorphs – a review, Ceram.-Silikáty 57 (3) (2013) 167–184.

[3] J.S. Tse, D.D. Klug, The structure and dynamics of silica polymorphs using atwo‐body effective potential model, J. Chem. Phys. 95 (12) (1991) 9176–9185.

[4] E. Gregorová, M. Černý, W. Pabst, L. Esposito, C. Zanelli, J. Hamáček,Jaroslav Kutzendörfer, Temperature dependence of Young’s modulus of silica re-fractories, Ceram. Int. 41 (1) (2015) 1129–1138.

[5] E. Gregorová, W. Pabst, P. Diblíková, V. Nečina, Temperature dependence of

Fig. 13. Crack path for specimen 2H (a), observed parallel crack propagation and crack branching (b), illustration of transgranular fracture in peripheral region of thegrains (c). (For interpretation of the references to colour in text, the reader is referred to the web version of this article.)

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5608

damping in silica refractories measured via the impulse excitation technique,Ceram. Int. 44 (2018) 8363–8373.

[6] W. Pabst, E. Gregorová, J. Kloužek, A. Kloužková, P. Zemenová, M. Kohoutková,et al., High-temperature Young’s moduli and dilatation behavior of silica re-fractories, J. Eur. Ceram. Soc. 36 (1) (2016) 209–220.

[7] W. Pabst, E. Gregorová, J. Kutzendörfer, Elastic anomalies in tridymite- and cris-tobalite-based silica materials, Ceram. Int. 40 (3) (2014) 4207–4211.

[8] R.B. Heimann, Classic and Advanced Ceramics: From Fundamentals to Applications,Wiley-VCH, 2010, pp. 76–81.

[9] M.D. Allendorf, K.E. Spearb, Thermodynamic analysis of silica refractory corrosionin glass-melting furnaces, J. Electrochem. Soc. 148 (2) (2001) B59–B67.

[10] K. Andreev, N. Shetty, E. Verstrynge, Acoustic emission based damage limits andtheir correlation with fatigue resistance of refractory masonry, Constr. Build. Mater.165 (2018) 639–646.

[11] P. Pilate, V. Lardot, F. Cambier, E. Brochen, Contribution to the understanding ofthe high temperature behavior and of the compressive creep behavior of silica re-fractory materials, J. Eur. Ceram. Soc. 35 (2) (2015) 813–822.

[12] K. Andreev, V. Tadaion, J. Koster, E. Verstrynge, Cyclic fatigue of silica refractories– effect of test method on failure process, J. Eur. Ceram. Soc. 37 (4) (2016)1811–1819.

[13] W. Wan, J. Luo, J. Yang, Y. Feng, Y. Ouyang, D. Chen, et al., High-temperatureablation properties of nano zirconia reinforced fused silica ceramics, Ceram. Int. 44(6) (2018).

[14] W. Wan, J. Yang, Y. Feng, W. Bu, T. Qiu, Effect of trace alumina on mechanical,dielectric, and ablation properties of fused silica ceramics, J. Alloy. Compd. 675(2016) 64–72.

[15] J.J. Liang, Q.H. Lin, Z. Yo, Y.Z. Zhou, et al., Effects of alumina on cristobalitecrystallization and properties of silica-based ceramic cores, J. Mater. Sci. Technol.33 (2) (2017) 204–209.

[16] H. Kimizuka, H. Kaburaki, Y. Kogure, Mechanism for negative poisson ratios overthe α- β transformation of cristobalite, SiO2: a molecular-dynamics study, Phys.Rev. Lett. 84 (24) (2000) 5548.

[17] C.Y. Guo, L. Wheeler, Extreme poisson’s ratios and related elastic crystal properties,J. Mech. Phys. Solids 54 (4) (2006) 690–707.

[18] W. Wan, C. Huang, J. Yang, T. Qiu, Study on gelcasting of fused silica glass usingglutinous rice flour as binder, Int. J. Appl. Glass Sci. 5 (4) (2015) 401–409.

[19] H. Salmang, H. Scholze, 6th edition, Keramik vol. 2, Springer, 1982, pp. 186–194.[20] R.C. Breneman, J.W. Halloran, Effect of cristobalite on the strength of sintered fused

silica above and below the cristobalite transformation, J. Am. Ceram. Soc. 98 (5)(2015) 1611–1617.

[21] E. K. Tschegg, Equipment and appropriate specimen shape for tests to measurefracture values. Patent AT-390328, 1986.

[22] H. Harmuth, Stability of crack propagation associated with fracture energy de-termined by wedge splitting specimen, Theor. Appl. Fract. Mech. 23 (1) (1995)103–108.

[23] P. Lava, S. Cooreman, D. Debruyne, Study of systematic errors in strain fields ob-tained via DIC using heterogeneous deformation generated by plastic FEA, Opt.Laser Eng. 48 (4) (2010) 457–468.

[24] A. Tatschl, O. Kolednik, A new tool for the experimental characterization of micro-plasticity, Mater. Sci. Eng. A-Struct. 339 (1) (2003) 265–280.

[25] Z. Wu, H. Rong, J. Zheng, F. Xu, W. Dong, An experimental investigation on the FPZproperties in concrete using digital image correlation technique, Eng. Fract. Mech.78 (17) (2011) 2978–2990.

[26] Y. Dai, D. Gruber, H. Harmuth, Observation and quantification of the fractureprocess zone for two magnesia refractories with different brittleness, J. Eur. Ceram.Soc. 37 (6) (2017) 2521–2519.

[27] Y. Belrhiti, O. Pop, A. Germaneau, P. Doumalin, J.C. Dupré, H. Harmuth, M. Huger,T. Chotard, Investigation of the impact of micro-cracks on fracture behavior ofmagnesia products using wedge splitting test and digital image correlation, J. Eur.Ceram. Soc. 35 (2) (2015) 823–829.

[28] M. Bornert, F. Brémand, P. Doumalin, J.C. Dupré, M. Fazzini, M. Grédiac, F. Hild,S. Mistou, J. Molimard, J.J. Orteu, L. Robert, Y. Surrel, P. Vacher, B. Wattrisse,Assessment of digital image correlation measurement errors: methodology and re-sults, Exp. Mech. 49 (3) (2009) 353–370.

[29] S.J. Kitchin, S.C. Kohn, R. Dupree, C.M.B. Henderson, K. Kihara, In situ29Si MASNMR studies of structural phase transformations of tridymite, Am. Mineral. 81(5–6) (1996) 550–560.

[30] W. Pabst, E. Gregorová, Young’s modulus of isotropic porous materials withspheroidal pores, J. Eur. Ceram. Soc. 34 (13) (2014) 3195–3207.

[31] D.P.H. Hasselman, Unified theory of thermal shock fracture initiation and crackpropagation in brittle ceramics, J. Am. Ceram. Soc. 52 (11) (1969) 600–604.

[32] D. Gruber, M. Sistaninia, C. Fasching, O. Kolednik, Thermal shock resistance ofmagnesia spinel refractories – Investigation with the concept of configurationalforces, J. Eur. Ceram. Soc. 36 (16) (2016) 4301–4308.

[33] H. Harmuth, R.C. Bradt, Investigation of refractory brittleness by fracture me-chanical and fractographic methods, Keram. Z. 62 (4) (2010) 264–269.

Y. Dai et al. Journal of the European Ceramic Society 38 (2018) 5601–5609

5609