Journal of Non-Crystalline Solids 358 (2012) 975–984

Contents lists available at SciVerse ScienceDirect

Journal of Non-Crystalline Solids

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

Review

Internal residual stresses in glass-ceramics: A review

Francisco C. Serbena a,⁎, Edgar D. Zanotto b
a
Department of Physics, State University of Ponta Grossa (UEPG), CEP 84030-900, Ponta Grossa, PR, Brazil
b
Vitreous Materials Laboratory (LaMaV), Department of Materials Engineering, Federal University of São Carlos (UFSCar), CEP 13560-970, São Carlos, SP, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Internal residual stresses arise in glass-ceramics upon cooling down from the crystallization temperature.
Received 5 December 2011 These stresses are due to the thermal expansion and the elastic mismatch between the crystalline and glassy
Received in revised form 18 January 2012 phases. Therefore, the mechanical properties of glass-ceramics are likely to depend not only on their compo-
Available online 18 February 2012
sition and microstructure but also on the type (tension or compression) and magnitude of these residual
stresses. In this work, we critically review the most commonly used theoretical models concerning residual
Keywords:
Residual stress;
stresses in glass-ceramics and glass-matrix composites, taking into consideration the effects of crystallized
Glass-ceramic; volume fraction, crystal shape and thermal expansion anisotropy. We also discuss most of the reported
Glass-matrix composite; measurements of residual stresses in these dual-phase materials using different techniques, such as X-ray
Glass diffraction, nuclear magnetic resonance, Raman and fluorescence spectroscopy, and indentation. The
available models and experimental results regarding spontaneous microcracking due to residual stresses
are also discussed. Finally, guidelines for future work are suggested.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction can arise due to non-homogeneous cooling, leading in some extreme

cases to spontaneous cracking [4,6]. Therefore, an understanding of
Important applications for glass-ceramics have been found in the the thermal residual stresses in glass-ceramics and their relationships
domestic and high-technology markets [1–3]. Glass-ceramics com- with the microstructure and overall mechanical properties of the ma-
bine the properties of crystalline ceramics with those of glasses and terials is important. The thermal residual stresses may have a signifi-
find applications in the telecommunications and optical industries, cant impact on a material's mechanical performance including its
such as opto-electronic and microwave devices, surgical implants, strength [8–11] and stresses in composites [12–14], dental glass-
dental materials, cooktops, and telescope mirrors [4–6]. ceramics [15–18] and components of fuel cells [19,20], among other
Components with complex geometries can be molded in the glass applications.
phase at relatively low cost and with relatively simple technology [3]. In this article, we critically review the most popular models for
Then, subsequent heat treatments can partially crystallize the glass thermal residual stresses in dual-phase materials and their
object in a controlled manner with a designed microstructure and applications in glass-ceramics and glass-matrix composites consider-
with very low or no porosity. The crystallized volume fraction can ing the effects of the thermal and elastic mismatch between the
be as low as a few percent or as high as 99.5%. Generally, glass- phases, crystallized volume fraction, precipitate shape, thermal
ceramics have superior optical, chemical, electrical and mechanical expansion anisotropy and microcracking. We then discuss residual
properties to those of glasses and similar ceramics that have been stress measurement using X-ray diffraction, nuclear magnetic reso-
produced by sintering. nance, Raman and fluorescence spectroscopy and indentation. Finally,
Glass-ceramics are thus produced by a controlled crystallization we comment on previous experimental studies of microcracking due
that leads to one or more phases embedded within a glassy matrix. to residual stresses. The influence of residual stresses on fracture
Their mechanical, optical and thermal properties depend not only toughness and overall mechanical strength of glass-ceramics [21]
on their composition and microstructure but also on the thermal will not be considered here.
residual stresses that arise upon cooling due to the thermal and elas-
tic mismatch between the precipitates and the glassy matrix [7]. In 2. Theoretical models for residual stresses
addition to these thermal micro stresses, residual macro stresses
2.1. The Selsing model

⁎ Corresponding author at: Universidade Estadual de Ponta Grossa, Departamento de

Física, Av. Carlos Cavalcanti, 4748, Ponta Grossa, PR, 84030-900, Brazil. Tel.: +55 42 3220
One of the simplest models to estimate internal residual stresses
3044; fax: +55 42 3220 3042. in glass-ceramics is that of Selsing [7]. It assumes that the precipitates
E-mail address: [email protected] (F.C. Serbena). (crystals) are spherical and isotropic and that the stress fields around

0022-3093/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2012.01.040
976 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984

them do not overlap, which is the case for low crystallized volume The results of calculations using these equations were compared
fractions (b10%). The stress inside the precipitate, σP, is hydrostatic with simulations carried out by finite element analysis. Higher residual
(uniform) and is: stresses are achieved by large differences in the thermal expansion co-
efficients and a high elastic modulus. The residual stresses in inclusions
Δα:ΔT decrease with increasing volume fraction in an approximately linear
σP ¼ ð1Þ
KE relationship. For the three geometries tested, the highest stresses
found are in-plane stresses in disks. The stresses perpendicular to the
where KE is (1+ vm) / 2Em + (1− 2vp) / Ep, Δα is the thermal expansion disk plane are negligible. For fibers, the high-stress component is that
mismatch between the precipitate and the matrix, ΔT is the tempera- along the length of the fiber. Stresses along and perpendicular to
ture difference between the glass transition temperature (Tg, at which their axes remain unchanged for aspect ratios larger than ~5. For
the glass ceases to flow over the laboratory time scale) and room tem- spherical inclusions, the magnitude of residual stresses is intermediate
perature, and E and ν are the elastic constants of the matrix (m) and between the highest and lowest stress components for disks and fibers.
precipitate (p), respectively. Therefore, the stress inside the precipitate
does not depend on the radius, R, of the precipitate. 2.3. Thermal expansion anisotropy
In the matrix, the radial component, σP.R 3 / r 3, has the same sign as
σP, whereas the two tangential components have the opposite sign Another factor that influences residual stresses is thermal expan-
and are equal to –σP.R 3 / 2r 3, as shown in Fig. 1. sion anisotropy. Crystals with non-cubic symmetry have different
thermal expansion coefficients along different crystallographic direc-
2.2. Effect of volume fraction and crystal shape tions. As a consequence, intergranular stresses arise to accommodate
the grain alongside its neighbors, as shown in Fig. 2. This topic has
Generally, glass-ceramics have a crystallized volume fraction been considered by Evans [26] and Davidge [27] and is described in
higher than 10%; thus, the Selsing model is no longer valid. The effect great detail by Mura [28]. The studies assume a single grain in an in-
of the volume fraction on the internal residual stresses has been con- finite isotropic matrix. Inside the grain, different directions have dif-
sidered by Mori and Tanaka [23] and Hsueh and Becher [24]. Mori and ferent coefficients of thermal expansion, while the thermal
Tanaka calculated the average stresses in a matrix with inclusions, expansion of the matrix is isotropic with average thermal properties
and Hsueh and Becher, using the Eshelby model for transformation equal to those of the grain. This is a reasonable assumption if the
strain on ellipsoidal inclusions embedded in a matrix [25], calculated grains are randomly oriented. Using the Eshelby model of transforma-
the residual stresses on inclusions in the form of spheres, fibers and tion strains, expressions are obtained for the thermal residual stres-
disks, taking their volume fraction into consideration. For spherical ses. For example, in a crystal phase with a hexagonal unit cell, the
precipitates, the stress inside the precipitate is given by: residual stresses are given by the following equations [28–30]:

Δα:ΔT Eð7−5ν Þ
σp ¼ ð2Þ σa ¼ − Δα:ΔT ð4aÞ
1
þ 1
þ f 45ð1 þ νÞð1−ν Þ
3K p 4ð1−f ÞGm 3ð1−f ÞK m

Eð17 þ 5νÞ
where G is the shear modulus, K is the bulk modulus and f is the crys- σc ¼ Δα:ΔT ð4bÞ
45ð1 þ νÞð1−ν Þ
tallized volume fraction. When f = 0, this equation becomes the Sel-
sing expression. where σa and σc are the stresses along the a and c directions and
The average stress in the matrix ( σ! m ) is calculated from the Δα = αa − αc. It is not uncommon to have tensile stresses in one di-
equilibrium condition: rection inside the grain and compressive stresses in another. These
stresses linearly increase with the thermal expansion mismatch due
f :σ p þ ð1−f Þ:σ! m ¼ 0: ð3Þ to anisotropy and with increasing elastic modulus.

Fig. 1. Stress profiles of a precipitate according to Selsing's model. The stress, σP, is con- Fig. 2. Unconstrained grain and matrix cavity shapes due to thermal expansion anisot-
stant inside the precipitate. The radial (σR) and tangential (σT) stress components out- ropy. This is a simplified diagram of surface forces acting in the grain for shape confor-
side the precipitate decay according to 1/r3. mity.
Adapted from Mastelaro and Zanotto [22]. Adapted from Evans [26].
 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984 977

Elastic anisotropy also contributes to the stresses. If the inclusion precipitate. In each case, the tensile stress and the region of lower
has elastic constants that vary with crystallographic direction, stres- fracture toughness (matrix or precipitate) control the nucleation
ses will arise to accommodate its deformation with the matrix and propagation of such microcracks. For the case in which the ther-
when under load forces. This problem can also be treated using the mal expansion of the precipitate, αP, is higher than that of the matrix,
Eshelby model. Generally, the effect of thermal expansion anisotropy αm, microcracking occurs inside the precipitate if its fracture tough-
is higher than that of elastic anisotropy [28,29]. ness is lower than that of the matrix, as shown in Fig. 4(b). If the ma-
trix is more brittle, the crack will propagate around the precipitate
2.4. Effect on microcracking and inside the matrix due to the tensile radial stress component, as
shown in Fig. 4(c). For the case in which αP is lower than αm, micro-
Residual stresses affect the toughness of glass-ceramics. Selsing's cracking occurs in the matrix and links the precipitates due to the
model reveals that the residual stresses are tensile in the precipitate tensile tangential matrix stress components, as shown in Fig. 5(a)
when its thermal expansion coefficient is higher than that of the ma- and (b).
trix, which is accompanied by tensile radial stresses and compressive Several models have been proposed for the effect of residual stres-
tangential stresses in the matrix. A propagating crack will deviate ses on microfracture. An energy balance exists between the elastic
from the precipitates, as shown in Fig. 3(a) [31]. strain energy stored in the precipitate and in the matrix and the ener-
If the residual stresses are compressive on the precipitate, com- gy needed to create new crack surfaces. If the amount of stored me-
pressive radial stresses and tensile tangential stresses will develop chanical energy is greater than the surface energy, spontaneous
in the matrix. The propagating crack will then be directed toward cracking occurs; this is related to a critical precipitate radius, RC.
the precipitates in the matrix and eventually cracking of the precipi- This critical radius has been calculated for some cases. When the ther-
tates is observed, as shown in Fig. 3(b) [31]. In both cases, residual mal expansion and the fracture toughness of the precipitate are higher
stresses can change the path of the crack and produce spontaneous than those of the matrix [31], circumferential cracking will occur around
microcracking. These observations have important implications on the precipitate, as shown in Fig. 4(c). The critical condition is:
the fracture toughness models for glass-ceramics [32,33].
Residual stresses may thus affect the microcracking of glass- 4γ S
RC ≥ : ð5Þ
ceramics. Figs. 4 and 5 show the stress distribution inside and outside K E :σ 2P
of a precipitate and the different fracture patterns observed for differ-
ent combinations of the thermal expansion of a glass and a If the precipitate has a lower KIC value than the matrix, the precip-
itate will crack as shown in Fig. 4(b), and the critical radius is [30]:

2γ S
RC ≥ : ð6Þ
K E :σ 2P

For the case in which the thermal expansion of the precipitate is

lower than that of the matrix, cracking will occur radially in the ma-
trix, as shown in Fig. 6. The critical radius, as a function of the crystal-
lized volume fraction, was estimated by Todd and Derby [29] based
on Green's model [35] and is given by [30]:
!2
2
π ð1−f Þ K IC
RC ¼ ! "! "2 : ð7Þ
f 1−f 2=3
1 þ 2f 1=3 σ 2P

If an external stress is applied to a glass-ceramic body, the critical

radius for cracking decreases drastically, typically by one order of
magnitude, as demonstrated by Green [35].
Another source of microcracking in glass-ceramics (due to residu-
al stresses) is the anisotropic thermal expansion of crystals, which
can produce grain-boundary cracking. Defects, such as voids at grain
boundaries, act as nucleation sites. Evans [26] estimated the critical
grain size (dg) above which intergranular cracking is observed:

5:2 ð1 þ ν Þ2 γGB
dg ¼ ð8Þ
EP ðΔα:ΔT Þ2

where γGB is the grain-boundary surface energy and Δα is half of the

maximum difference in the thermal expansion due to anisotropy.
Below the critical grain size, no cracking is observed.
Evidence of grain-boundary microcracking during cooling has
been provided by the reduction in the apparent thermal expansion,
events of acoustic emission and thermal expansion hysteresis in mag-
nesium and aluminum titanate ceramics [26,36,37].

3. Residual stress measurements in glass-ceramics

Fig. 3. Schematic crack patterns when (a) the thermal expansion of the precipitate is
higher than that of the matrix or (b) the thermal expansion of the precipitate is
lower than that of the matrix. Residual stresses in glass-ceramics have been measured using var-
After Davidge and Green [31]. ious techniques; among them are X-ray diffraction (XRD), nuclear
978 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984

Fig. 4. (a) Thermal residual stress distribution and microcracking for cases in which the thermal expansion of the precipitate is higher than that of the matrix and the matrix surface
energy is (b) higher or (c) lower than that of the precipitate. Microcracking of the larger precipitates is represented in (a), and semi-spherical microcracking of the matrix at the
precipitate–matrix interface of the larger precipitates is shown in (b).
Adapted from Lange [34].

magnetic resonance (NMR), fluorescence and Raman spectroscopy displacement, roughness, sample transparency, and peak asymmetry,
and indentation. can be considered in the refinement. In addition, all reflections are
considered in this method. The unit cell dimensions of the embedded
3.1. Residual stress measurements using XRD crystals in a monolithic piece of glass-ceramic (the stressed sample)
are compared with the unit cell dimensions of the powdered sample
Residual stresses can be measured indirectly using XRD. Reviews (the stress-free sample). In this manner, strains and stresses along
on residual stress measurements by XRD can be found in the litera- any crystallographic direction are obtained.
ture [38–41]. Diffraction is based on Bragg's law, which relates the Zevin et al. [42] measured the residual stresses in β-eucrypite and
interplanar distance of a particular hkl set of planes with the diffrac- β-spodumene solid-solution based glass-ceramics with high crystal-
tion angle, θ. Residual strains shift the reflection peaks, which are lized volume fractions. Residual stresses were determined by the
measured by comparison with those of a stress-free reference sample, peak displacements of several reflections. A powder sample was
usually a finely ground and annealed powder. Sometimes an internal used as a stress-free reference sample. The measured stresses were
standard, such as alumina or silicon powder, is also used. In this man- anisotropic, and the measured values were much lower than those es-
ner, residual strains normal to the hkl planes can be measured. Resid- timated by Selsing's model. The authors attributed these differences to
ual stresses can then be calculated from the residual strains using a pre-existing network of microcracks in the residual glass, but the ef-
Hooke's law if the elastic constants of the crystal are known. fect of the high volume fraction of crystal phases was not considered.
The full stress tensor can be calculated by measuring the variation Mastelaro and Zanotto [22] measured residual stresses in partially
of the interplanar distance dhkl of a specific hkl plane along at least six crystallized 1Na2O·2CaO·3SiO2 + 3%P2O5 glass-ceramics using con-
independent directions as defined by an azimuthal angle, ϕ, and a tilt- ventional XRD equipment. Their glass-ceramics had a sufficiently
ing angle, ψ. If the principal stress directions are known, only three di- low crystallized volume fraction, in the range of 5–12%, to avoid
rections are necessary for the calculation of the full stress tensor. inter-crystal interaction. Residual strains were measured using the
The other technique involves the use of the Rietveld refinement of (404) reflection of the low combeite Na2Ca2Si3O9 phase. The experi-
high-quality diffractograms of the glass-ceramic in bulk and powder mental residual stress was 150 ± 50 MPa and agreed with the residual
forms. The lattice constants of the crystallized phase are calculated stress predicted by Selsing's model of 160 MPa.
by Rietveld refinement. Several sources of errors, such as sample Subsequently, the same authors [43] measured residual stresses in
a partially crystallized Li2O·2SiO2 glass-ceramic using synchrotron ra-
diation for several crystallographic planes. Their glass-ceramic had a
low crystallized volume fraction to avoid percolation of the stress
fields. Residual stresses were highly anisotropic and were in agree-
ment with Selsing's model if the thermal expansion anisotropy was
considered.
Residual stresses in partially crystallized Li2O·2SiO2 glass-
ceramics were also studied by Pinto et al. [44]. High and low energy
synchrotron XRD was used to measure residual stresses in the sam-
ples with simultaneous internal and surface crystallization. The
glass-ceramics investigated had a low crystallized volume fraction.
The residual stresses were confirmed to be highly anisotropic in the
volume studied, and a good agreement with Selsing's model was
obtained if the thermal expansion anisotropy of the Li2Si2O5 crystal
phase was considered. At the sample surface, the stresses were iso-
tropic and were well predicted by a film model. The average stress
was − 50 ± 15 MPa in the bulk, and − 120 MPa at the surface. The
Fig. 5. (a) Thermal residual stress distribution and (b) microcracking in cases in which
authors suggested that the compressive residual stress at the surface
the thermal expansion of the precipitate is lower than that of the matrix. could be used as a strengthening mechanism for improving the me-
Adapted from Lange [34]. chanical performance of surface-crystallized glass-ceramics [44].
 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984 979

Fig. 6. An annular crack in the matrix growing radially under the tangential residual stress components.
Adapted from Todd and Derby [29].

Peitl et al. [45] measured the residual stresses in a partially crystallized and Becher's model, [47] that considers the effects of the crystallized
1Na2O·2CaO·3SiO2 +3%P2O5 glass-ceramic using XRD. The measured re- volume fraction, and the thin film model [30,44]. Conventional XRD is
sidual stresses confirmed both the previous results of Mastelaro and a suitable technique to measure stresses near the sample surface. To
Zanotto [22] and the validity of Selsing's model for glass-ceramics with measure residual stresses in the sample interior, high energy syn-
low crystallized volume fractions. chrotron radiation must be used. The use of XRD for stress determina-
Residual stresses in low expansion Li2O–Al2O3–SiO2 (LAS) glass- tion is not suitable for amorphous samples.
ceramics produced by sintering and in a commercial LAS glass-ceram-
ic, CERAN®, were measured by Serbena et al. [30] using synchrotron 3.2. Residual stress measurements using NMR
radiation. The crystallized phase is virgilite (LixAlxSi3 − xO6). Due to
its hexagonal unit cell structure, its thermal expansion is highly an- To the best of our knowledge, the only work published thus far on
isotropic with contraction along the c-direction and almost no ther- the measurement of residual stresses by NMR is that of Zwanziger et
mal expansion along the a-direction with increasing temperature. al. [48], in which the authors have exploited the variations of the NMR
Strains calculated by the Rietveld refinement were in agreement resonance frequencies due to variations in the electron distribution
with this thermal expansion anisotropy, showing almost negligible for a particular nucleus, the chemical shift. If an applied stress modifies
compressive strains along the a-directions and fifteen-times higher the local bond geometry and bond lengths, the electron distribution is
compressive strains in the c-direction. The crystallized volume frac- affected and alters the total magnetic field experienced by the nucleus.
tions in the sintered and commercial LAS glass-ceramics were high, This, in turn, alters the resonance frequencies in a magnetic field.
approximately 84% and 67%, respectively. The overall residual stres- The glass-ceramic studied was the PTR glass-ceramic described
ses were low, in the range of − 30 MPa to −90 MPa (compression) previously with a low volume fraction of NaF crystals embedded in
and could be predicted by taking into account the effect of the volume an oxyfluoride glass matrix. The shifts in the Na and F peaks in the
fraction according to Eqs. (2) and (3). Residual stresses were also NMR spectra of a partially crystallized glass-ceramic were compared
measured in a glass with a few isolated crystals nucleated at the sur- with those of an NaF powder used as a stress-free reference sample.
face and in another with a fully covered crystallized surface. The First principle calculations were performed and considered the NaF
stresses were much higher than in the sintered glass-ceramics but structure under hydrostatic pressure and the Na and F NMR chemical
agreed with those calculated by the Selsing and thin film models, shifts were calculated as a function of pressure. By comparing the ex-
respectively. perimental chemical shifts with those predicted by the calculations,
Residual stresses in a photo-thermo-refractive (PTR) glass- stresses in the range of 600–800 MPa were obtained. These values
ceramic with a very low volume fraction crystallized (b1%) were agreed with those predicted by Selsing's model.
measured by Serbena et al. [46] using synchrotron XRD. The crystal- Thus, this is a powerful technique that relies on comparing the ex-
lized phase was an NaF nanocrystal, and due to its cubic structure, it perimental chemical shifts with those predicted by quantum me-
is isotropic and suitable for investigating Selsing's model with no in- chanical calculations. Further experiments with other glass-ceramics
fluence of crystal anisotropy. In addition, the precipitates were nearly must be performed to fully explore the capabilities of this technique.
spherical and the crystallized volume fraction was very small. The
measured thermal residual stresses were quite high, approximately 3.3. Residual stress measurements using Raman and
1 GPa. The experimental results agreed with the calculations of the fluorescence spectroscopy
Selsing model if the significant changes in the glass composition dur-
ing crystallization – due to Na and F depletion around the crystals – Raman and fluorescence spectroscopy techniques have also been
were considered. These changes in composition changed the residual used to measure residual stresses. When a lattice is strained under
glass thermal expansion coefficient and its Tg value. For glasses heat- an applied stress, the energy levels of its electronic and vibrational
treated at higher temperatures, larger dendritic crystals were ob- states are modified, which affects the transition energies and causes
served, and the measured residual stresses decreased to 640 MPa. slight shifts in frequency as a function of stress [49–52]. Raman spec-
This decrease was attributed to stress relief due to the microcracking troscopy is capable of measuring residual stresses and has been used
of the glass around the precipitates. extensively for this purpose in semiconductors [53–55].
In summary, all of the above papers demonstrate that XRD is a Mastelaro and Zanotto [43] attempted to use Raman spectroscopy
successful and accurate technique for measuring the residual stresses to measure residual stresses in partially crystallized lithium disilicate
in glass-ceramics. The results confirm the validity of Selsing's model glass-ceramics. A calibration curve was constructed by applying pres-
[7] for low crystallized volume fraction, and the validity of Hsueh sure to a fully crystallized powder sample using a diamond anvil cell.
980 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984

However, the stress detection limit of their equipment was 150 MPa, results were closer to the sphere model than to the platelet model;
which was at the limit of the residual stress variation in their samples this observation was attributed to the breaking of the platelets during
(between −120 MPa and + 150 MPa). the milling process.
Raman spectroscopy was also used by Yang and Young [56] to One advantage of this technique is its capability to sample very
study thermal residual strains in SiC and alumina fibers (with the ad- small areas (a few micrometers squared) using a confocal micro-
dition of 20% (wt.%) partially stabilized tetragonal zirconia) embed- scope. Stress gradients and stresses in very small regions can thus
ded in soda-lime silica or Pyrex glass matrices (these materials are be measured, including inclusions or fibers embedded in a glass ma-
not real glass-ceramics but have similar microstructures). Fibers trix (if the glass does not have a Raman or fluorescence peak in the
were sandwiched between two plates of glass and hot-pressed at dif- frequency range of interest).
ferent temperatures in an argon atmosphere. The measured residual
strains were proportional to the difference in the thermal expansion 3.4. Residual stress measurements using indentation
of the fiber and the glass matrix, being under either compression or
tension depending on whether the thermal expansion of the fiber Another technique used to determine residual stresses is indenta-
was smaller or larger than that of the matrix. The residual strains tion. This method, first proposed by Zeng and Rowcliffe [64], consists
did not depend on the hot-pressing temperature, confirming that of comparing the length, c1, of radial cracks produced by an indenta-
the important temperature difference is that between Tg and room tion in a stressed region of a specimen with the radial crack length, c0,
temperature. One important observation was made for a Pyrex glass made in a stress-free region. The residual stress, σm, is given as:
with SiC fiber hot-pressed at 780 °C and subsequently heat-treated
at 930 °C for 1 h. The average thermal expansion coefficients of
1−ðco =c1 Þ3=2
the SiC and Pyrex are nearly the same, 3.1 × 10 − 6 °C − 1 and σ m ¼ K IC 1=2
ð9Þ
Φ:c1
3.2 × 10 − 6 °C − 1, respectively, suggesting very low residual stresses,
but the measured strain in the fiber was 1.8%, the highest measured
in the study. Thus, there was some discrepancy between the experi- where KIC is the stress intensity factor at the indentation crack tip and
mental results and the calculations. However, observations via optical Φ is a crack geometry factor, which is related to the crack geometry
microscopy indicated that the second thermal treatment at 930 °C and loading condition and is assumed to be equal to π 1/2. If c1 > c0,
resulted in the crystallization of the Pyrex glass, matrix microcracking the residual stress, σm, is tensile. If c1 b c0, σm is compressive.
near the fiber-glass interface and a buildup of stress in the fiber. The Fig. 7 shows a Vickers indentation near a precipitate in a low vol-
residual strains along this fiber had a large scatter pattern, indicating ume fraction crystallized 17.2Na2O–32.1CaO–48.1SiO2–2.5P2O5
that the stresses were not homogeneous along the fiber. It is expected (mol%) glass-ceramic [65]. The different crack lengths are related to
that matrix cracking partially relieves the residual stress. In this case, the tensile radial stresses and the compressive tangential stresses
it seems that glass crystallization near the fiber had induced even near the precipitate as predicted by the Selsing model.
higher stresses. However, further studies are necessary to identify Nanoindentation has been used by Soares and Lepienski [66] to
the crystallized phase, possible changes in the glass matrix and the measure residual stresses around the precipitates in the glass matrix
fiber and reactions between the fiber and the glass. in a low crystallized volume fraction lithium disilicate glass-ceramic.
In another study, Dassios et al. [57,58] used in situ Raman spec- Cracks produced by a Berkovich indenter near the crystals were
troscopy to study the mechanical behavior of SiC fibers in a LiO2– used, but the calculated residual stresses were lower than those pre-
MgO–Al2O3–SiO2 glass–ceramic matrix under tensile loading. Strain dicted by the Selsing model. It was argued that the method was not
profiles along individual fibers at different loading stages were mea- appropriate to measure the stresses in a region of approximately
sured. Insight into the failure of micro-mechanisms was obtained, 100 μm near the crystal–glass interface because one indentation
and using this technique, the fiber interfacial shear strength was di- crack interacted with that of a previous indentation, and smaller
rectly measured. cracks (thus smaller indentation loads) were necessary to measure
Another technique that has been used is fluorescence spectrosco- the stresses in that region.
py. Stress gradients and stresses in very small regions, such as inside Peitl et al. [45] used the indentation technique to estimate the re-
grain bridges at a crack surface, were measured using fluorescence sidual stress distributions in the glass matrix around the precipitates.
spectroscopy [59–61]. One example is the luminescence lines of
Cr + 3 that substitute for Al + 3 as a dopant in the alumina lattice.
Strains in the lattice alter the electronic transitions of the Cr + 3 ions,
and the frequency shift is proportional to the stress. Its piezospectro-
scopic coefficients have been measured [50].
Young and Yang [62] used fluorescence spectroscopy to measure
the residual stresses in single alumina fibers with 20% by weight of
tetragonal-stabilized zirconia embedded in a soda-lime glass. The ef-
fect of the carbon coating in the fiber was also investigated. The orig-
inal stress profile, resulting from the thermal expansion mismatch
between the fiber and glass matrix, was measured. By applying an ex-
ternal stress, the evolution of the stress along the fiber was observed
at different strain levels up to the interface failure and matrix frac-
ture. The applied strain promotes a reduction in the stress level of
the fiber. The carbon coating was found to decrease the break-down
stress due to the weaker interface when compared with the uncoated
fiber.
This technique has also been used by Todd et al. [63] to measure
the residual stresses in alumina platelets embedded in a borosilicate
Fig. 7. Optical micrograph of a 0.5 N Vickers indentation around a crystalline precipitate in a
glass matrix at different volume fractions up to 30%. The results 17.2Na2O–32.1CaO–48.1SiO2–2.5P2O5 (mol%) glass-ceramic used for the determination of
were compared with the platelet and spherical models of Hsueh residual stresses.
and Becher [47] with varying volume fractions. The experimental Reproduced from Peitl et al. [45].
 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984 981

A low crystallized volume fraction 17.2Na2O–32.1CaO–48.1SiO2–

2.5P2O5 (mol%) glass-ceramic was used. Both radial and tangential
stress components were measured. The residual stress in the precip-
itates was also measured using XRD for comparison with the results
obtained via the indentation technique. The calculated stresses in
the precipitate using Eq. (9), with a value for Φ equal to π 1/2, were ap-
proximately 30% of those measured by XRD and predicted by the Sel-
sing model. A careful analysis by sequential polishing revealed that
the radial cracks were not semicircular but had a semi-elliptical
shape with a length-to-depth ratio of 9.5. This crack geometry re-
quired a new crack geometry factor Φ*. Using fracture mechanics
analysis [67], the new crack geometry factor proposed was as follows:

# $% # $&vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
%di di u
u π
Φ ¼ : 1:243−0:099 t ( )1:65 ð10Þ
ci ci 1 þ 1:464 di ci

where di and ci were the crack depth and radial crack length at the
surface, respectively. A reanalysis of the data using the new crack ge-
ometry factor restored agreement with the residual stress measured
by XRD and predicted by the Selsing model, as shown in Fig. 8(a)
and (b).
Therefore, the indentation technique is adequate for measuring
residual stresses in glass matrices. It is a fast, inexpensive and non-
destructive technique. However, several precautions must be taken
for reliable results: the crack shape must be known for the determina-
tion of the correct crack geometry factor, indentations must be suffi-
ciently separated so that the stress field of one indentation does not
interact with that of another indentation, and the load used must be
above a certain threshold to produce cracks because of the finite
crack length. This technique has been used in glass-ceramics with
large precipitates, but it may not be suitable for glass-ceramics con-
taining very small crystals (b1 μm) or in regions of high stress
gradients.
Fig. 8. (a) Radial and (b) tangential stresses as a function of the normalized distance r/a
4. Effects of residual stresses on microcracking calculated using the indentation technique for indentations in air and water and curve
fitted to air data according to Selsing's model.
Reproduced from Peitl et al. [45].
The magnitude and sign of the internal residual stresses affect the
type of microcracking in glass-ceramics. In Section 2.4 of this paper,
we showed that if the precipitate is above a certain critical size, spon- are richer in Na and depleted of Ca, and the glass “backyard” surround-
taneous microcracking of the precipitate or matrix is observed ing the precipitate is richer in Ca and depleted of Na when compared
(depending on the stress sign and the weaker phase). with the stoichiometric Na2Ca2Si3O9 composition. This fact alters the
In their classic work, Davidge and Green [31] studied the strength thermal expansions of the glass and precipitate, increasing the residual
and microcracking of samples consisting of thorium oxide spheres stresses in the precipitate [45] and decreasing the critical radius for
dispersed in a glass matrix of varying compositions. They observed spontaneous microcracking. However, even this correction is not
that if the thermal expansion of the glass was lower than that of the
spheres, the cracks contour the spheres and cracks were nucleated
around the spheres for precipitates above a certain critical radius. If
the thermal expansion of the glass was higher than that of the
spheres, radial cracks at the precipitate edges in the matrix were ob-
served, as shown in Fig. 9. Larger precipitates decreased the bending
strength, and an externally applied load decreased the sphere's criti-
cal radius for microcracking.
Mastelaro and Zanotto [22] also studied the critical radius for spon-
taneous cracking in a partially crystallized 1Na2O·2CaO·3SiO2 +
3%P2O5 glass-ceramic. Their glass-ceramic had a low crystallized
volume fraction. Different thermal treatments produced particles
from 650 μm to 985 μm in diameter. The crystallized phase was a low
combeite Na2Ca2Si3O9. Using Eq. (5), a critical size for spontaneous
cracking of 88 μm was calculated. However, no spontaneous micro-
cracking was observed in the specimens despite the fact that the exper-
imental precipitate sizes were approximately one order of magnitude
larger than the critical size. Though at the time of the experimentation
there was no firm explanation for this surprising result, it is now Fig. 9. Optical micrograph of radial cracks generated by a lithium niobate–lithium
known that there is a significant glass composition gradient around disilicate double crystal in a glass matrix.
the precipitates in this particular glass-ceramic [69,70]. The crystals Courtesy of V. O. Soares and M. Crovace [68].
982 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984

enough to quantitatively explain the observed discrepancy. Therefore, of 0.4–1.4 μm. When the specimens were tested, an increase in frac-
no satisfactory explanation exists, and further experimental work ture toughness with grain size was initially observed. However, for
must be performed on this glass-ceramic. specimens with a grain size larger than ~ 1 μm, spontaneous cracking
The spontaneous radius for microcracking in PTR glass-ceramics was observed. Fracture strength tests also revealed a complete loss of
has been studied by Serbena et al. [46]. Two different heat treatments strength for samples with grain size larger than 1 μm. The initial in-
produced two types of precipitates: NaF cuboidal nanocrystals crease in toughness with grain size for the sample without ZrO2 addi-
(shown in Fig. 10(b)) and larger NaF dendritic micrometer-sized tion was explained by the formation of microcracks ahead of the crack
crystals (shown in Fig. 10(d)). The critical diameter for spontaneous during testing, as proposed by Green [35] and described in Section 2.4.
microcracking in this glass-ceramic is 2.6 μm, as calculated by This phenomenon is microcrack toughening, where the microcracks
Eq. (5), which is about the size of some of the larger dendritic crystals cause dilatation and reduction of the elastic modulus, and both in-
observed by scanning electron microscopy (Fig. 10(d)). Microcrack- crease fracture toughness [73].
ing relieves the residual stresses from ~ 1 GPa for the cuboidal crystals Microcracking was also observed on cooling a sample from the
to 640 MPa for the dendritic crystal (Fig. 10(a) and (c)). Evidence of processing temperature for a composition with 15% ZrO2. The ob-
microcracking was also observed by the presence of two diffraction served microcracking originated from the stresses caused by differ-
peaks in the XRD pattern of the glass-ceramics with dendritic crystals. ences between the thermal expansions of lithium metasilicate and
One of the peaks corresponds to a stress of 640 MPa and another to a β-eucryptite, and possibly due to their thermal expansion anisotropy.
stress of nearly zero, which is attributed to the microcracked precip- The tetragonal to martensitic transformation of ZrO2 did not play a
itates. The Rietveld refinement analysis reveals that the amount role on the mechanical properties of this glass-ceramic.
of the “unstressed” NaF precipitates was 25%–30% of the total Sakamoto et al. [74] evaluated the thermal expansion, microstruc-
precipitates. ture and bending strength of β-spodumene glass-ceramics of the LAS
Spontaneous microcracking has been observed in some glass- system. Two glass-ceramics with slight differences in Li, Ti and Zn ox-
ceramics of the Li2O–Al2O3–SiO2 (LAS) system. The main crystalline ides were investigated. The crystalline volume fractions were rela-
phases formed in this system are metastable solid solutions of high- tively low, around 43%. One of the glass-ceramics had a high value
quartz or keatite structure, such as β-spodumene, β-eucryptite and of thermal expansion coefficient that agreed with calculations based
virgilite. All these phases have hexagonal or tetragonal crystal struc- on the thermal expansion coefficients of the residual glass and of β-
tures and their thermal expansion coefficients are highly anisotropic, spodumene reported in the literature [71,75]. The other glass-
with very low or even negative values [71]. These differences cause a ceramic had a very low thermal expansion, which did not agree
crystallographic dependence of the residual stresses, and intergranu- with theoretical predictions. Furthermore, it showed mass increase
lar stresses in highly crystalline glass-ceramics. after water immersion, and very low bending strength of 13 ±
For instance, Sarno and Tomozawa [72] have investigated the me- 2 MPa. SEM observations revealed β-spodumene crystals of 3 to
chanical properties of glass-ceramics with 22.2Li2O–18.9Al2O3– 5 μm in diameter immersed in the residual glass matrix. All these re-
55.9SiO2–3P2O5 (wt.%) with addition of 5 and 15% ZrO2. By using a sults indicate microstructure microcracking due to the thermal ex-
two stage heat treatment, 95% crystallinity or higher was achieved. pansion mismatch between the residual glass and the β-spodumene
Orthorhombic lithium metasilicate, hexagonal β-eucryptite and crystals.
tetragonal and monoclinic ZrO2 were the crystalline phases nucleat- Microcracking in virgilite-based low expansion LAS glass-ceramics
ed. No information was given about the volume fraction of each has been observed by Serbena et al. [30]. A glass-ceramic produced
phase. Different heat treatments produced grain sizes in the range by sintering and a commercial LAS glass-ceramic, CERAN®, were

Fig. 10. (a) The XRD pattern of PTR glass, a UV-exposed sample, heat treated for 450 °C—1 h/520 °C—2 h, and the corresponding NaF nano-precipitates (indicated by the
arrows); (b) a bright-field TEM image of an exposed PTR glass, heat treated for 483 °C—1 h/515 °C—1 h; (c) the XRD pattern of PTR glass, a UV-exposed sample, heat treated for
450 °C—1 h/650 °C—20 min; and (d) a scanning electron microscopy image of the fractured surface of a PTR sample, unexposed, heat treated at 450 °C—1 h/650 °C—20 min,
showing a fractured NaF precipitate and cracks in the surrounding matrix indicated by the arrows.
Reproduced with modifications from Serbena et al. [46].
 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984 983

studied. The microstructure of the sintered glass-ceramic consists of

large glass particles and smaller rounded virgilite crystals, with diam-
eters of approximately 800 nm, embedded in the residual glass, as
shown in Fig. 11(a) and (b). Larger residual glass islands exhibit ex-
tensive crack formation and a network of thin cracks. The microstruc-
ture of the CERAN® samples consists of very small crystals of virgilite
with diameters of 60 nm. No cracks were observed in the CERAN®
samples.
Both glass-ceramics exhibit thermal expansion hysteresis. The
hysteresis increased after submitting the samples to subsequent ther-
mal shock cycles, as shown in Fig. 12(a) and (b). In each cycle, the
sample was heated at 600 °C and quenched to 0 °C for five times.
The average thermal expansion of the crystalline phase, as estimated
by dilatometry, was different from that measured using high-
temperature X-ray diffraction. These facts are all evidence of micro-
structure microcracking.
The critical size for microcracking of the glass precipitates in the
sintered glass-ceramics was estimated using Eq. (6). A critical radius
of 11 μm was obtained, which is in good agreement with the size of
the observed cracked glass precipitates. Two other possible mecha-
nisms for microcracking of the microstructure were investigated:
thermal expansion anisotropy of virgilite and microcracking of the re-
sidual glass phase between the virgilite crystals. As the anisotropic
thermal expansion of virgilite crystals induced stresses that could
produce grain-boundary microcracking, the critical grain size was
also estimated for the sintered glass-ceramic. With Eq. (8), a critical
grain size of 170 μm was obtained. The critical radius of the virgilite
precipitate for spontaneous microcracking was calculated using
Eq. (7); a critical radius of 27 μm for the sintered glass-ceramic and
19 μm for CERAN® were calculated. These values were all much
higher than the observed virgilite grain size. Therefore, grain-boundary
Fig. 12. (a) Thermal expansion hysteresis of the sintered glass-ceramic and CERAN
after the first thermal-shock cycle of heating-and-quenching and (b) the accumulated
hysteresis as a function of the number of thermal cycles. The lines are added as a guide
to the eyes.
Reproduced from Serbena et al. [30].

microcracking and microcracking of the residual glass phase are unlikely

in these glass-ceramics.

5. Conclusions

Experiments conducted on some glass-ceramics have shown that

Selsing's model can successfully predict the values of internal residual
stresses in glass-ceramics containing low crystallized volume frac-
tions. Increasing the crystallized volume fraction reduces the stresses
in the crystals, but increases the stresses in the residual glass matrix.
The residual stress varies almost linearly with the crystallized volume
fraction and depends on the crystal shape. Finally, for non-cubic crys-
talline phases the thermal expansion anisotropy must be considered,
especially for glass-ceramics with high elastic moduli.
Internal residual stresses affecting crack propagation and micro-
cracking is commonly observed in crystals larger than a certain (cal-
culable) critical size. Models based on the strain energy release rate
have been largely (but not always) successful in predicting the critical
size for microcracking, leaving room for further study.
Several techniques have been used to measure residual stresses in
glass-ceramics. XRD and Raman and fluorescence spectroscopy are
powerful tools for determining the residual stresses in the crystalline
phase, whereas the indentation technique is key for estimating resid-
ual stresses in the glass matrix. However, care must be exercised in
the determination of the proper conditions under which indentation
can be correctly applied. NMR is another powerful technique that re-
Fig. 11. (a) Microcracking of the larger glass particles as revealed by optical microscopy and
(b) the virgilite grains embedded in a glass matrix in the LAS sintered glass-ceramic revealed
quires further exploration.
by SEM. Both samples were etched in a dilute HF solution. Valuable lessons can be learned from residual stress determina-
Reproduced from Serbena et al. [30]. tions and modeling in ceramic composites for use in the design and
984 F.C. Serbena, E.D. Zanotto / Journal of Non-Crystalline Solids 358 (2012) 975–984

control of nano- and microstructures in glass-ceramics. New tech- [29] R.I. Todd, B. Derby, Acta Mater. 52 (2004) 1621–1629.
[30] F.C. Serbena, V.O. Soares, O. Peitl, H. Pinto, R. Muccillo, E.D. Zanotto, J. Am. Ceram.
niques, such as high-energy X-ray microbeam from high-energy syn- Soc. 94 (2011) 1206–1214.
chrotron beam lines and neutron diffraction should be tested for use [31] R.W. Davidge, T.J. Green, J. Mater. Sci. 3 (1968) 629–634.
in the investigation of residual stresses in glass-ceramics. [32] K.T. Faber, A.G. Evans, Acta Metall. Mater. 31 (1983) 565–576.
[33] M. Taya, S. Hayashi, A.S. Kobayashi, H.S. Yoon, J. Am. Ceram. Soc. 73 (1990)
Despite its great importance, there have been few quantitative 1382–1391.
studies of residual stresses in glass-ceramics. This is a very interesting [34] F.F. Lange, Fracture and Fatigue, Academic Press, L.J. Broutman, New York, 1974,
and important subject with potential applications including optically p. 1.
[35] D.J. Green, J. Am. Ceram. Soc. 64 (1981) 138–141.
transparent, dental and bioactive glass-ceramics. Finally, the effects of [36] J.A. Kuszyk, R.C. Bradt, J. Am. Ceram. Soc. 56 (1973) 420–423.
internal residual stresses on the overall (macroscopic) mechanical [37] Y. Ohya, Z. Nakagawa, K. Hamano, J. Am. Ceram. Soc. 70 (1987) C-184.
properties of glass-ceramics must still be established. [38] I. Noyan, T. Huang, B. York, CRC Cr. Rev. Sol. State 20 (1995) 125–177.
[39] V.M. Hauk, E. Macherauch, Adv. X-Ray Anal. 27 (1983) 81–99.
[40] J. Lu, D. Retraint, J. Strain Anal. Eng. 33 (1998) 127–136.
Acknowledgments [41] P.J. Withers, H. Bhadeshia, Mater. Sci. Tech. Ser. 17 (2001) 355–365.
[42] L. Zevin, E. Levi, Z. Bessmertnaya, Inorg. Mater. 13 (1977) 1511–1514.
The authors acknowledge CNPq, FAPESP/Brazil-contract nos. 07/ [43] V.R. Mastelaro, E.D. Zanotto, J. Non-Cryst. Solids 247 (1999) 79–86.
[44] H. Pinto, L. Ito, M. Crovace, E.B. Ferreira, F. Fauth, T. Wroblewski, E.D. Zanotto, A.R.
08179-9 and 05/53241-9 and the Brazilian Synchrotron Light Laborato- Pyzalla, J. Non-Cryst. Solids 353 (2007) 2307–2317.
ry/MCT research proposals XRD1-5824, XRD2-5322, and XRD1-6712 for [45] O. Peitl, F.C. Serbena, V.R. Mastelaro, E.D. Zanotto, J. Am. Ceram. Soc. 93 (2010)
financial support. We are also indebted to Valmor Mastelaro, Oscar 2359–2368.
[46] F.C. Serbena, G.P. Souza, E.D. Zanotto, J. Lumeau, L. Glebova, L.B. Glebov, J. Am.
Peitl, Joe Zwanziger, Viviane Soares, Murilo Crovace, Haroldo Pinto, Julien Ceram. Soc. 94 (2011) 671–674.
Lumeau and Leon Glebov for most useful discussions on the different [47] C.-H. Hsueh, P.F. Becher, Mater. Sci. Eng., A 212 (1996) 22–28.
measurement techniques that we have tested jointly and used in our [48] J.W. Zwanziger, U. Werner-Zwanziger, E.D. Zanotto, E. Rotari, L.N. Glebova, L.B.
Glebov, J.F. Schneider, J. Appl. Phys. 99 (2006) 083511.
previous papers on residual stresses in glass-ceramics. [49] Q. Ma, D.R. Clarke, J. Am. Ceram. Soc. 76 (1993) 1433–1440.
[50] J. He, D.R. Clarke, J. Am. Ceram. Soc. 78 (1995) 1347–1353.
References [51] L. Schadler, C. Galiotis, Int. Mater. Rev. 40 (1995) 116–134.
[52] I.D. Wolf, Semicond. Sci. Technol. 11 (1996) 139.
[1] W. Pannhorst, J. Non-Cryst. Solids 219 (1997) 198–204. [53] Y.B. Gerbig, S.J. Stranick, R.F. Cook, Scr. Mater. 63 (2010) 512–515.
[2] G.H. Beall, L.R. Pinckney, J. Am. Ceram. Soc. 82 (1999) 5–16. [54] R. Ghisleni, J. Liu, R. Raghavan, P. Brodard, A. Lugstein, K. Wasmer, J. Michler,
[3] E.D. Zanotto, Am. Ceram. Soc. Bull. 89 (2010) 19–27. Philos. Mag. 91 (2011) 1286–1292.
[4] P.W. Mc Millan, Glass-Ceramics, 2nd ed. Academic Press, New York, 1979. [55] P. Zorabedian, F. Adar, Appl. Phys. Lett. 43 (1983) 177–179.
[5] J.M.F. Navarro, El Vidrio, 2nd ed. Consejo Superior de Investigaciones Cientificas, [56] X. Yang, R. Young, Composites 25 (1994) 488–493.
Madrid, Spain, 1991. [57] K.G. Dassios, C. Galiotis, V. Kostopoulos, M. Steen, Acta Mater. 51 (2003)
[6] W. Holand, G.H. Beall, Glass-Ceramic Technology, 1st ed. American Ceramic 5359–5373.
Society/Wiley, New York, 2002. [58] K.G. Dassios, C. Galiotis, J. Mater. Res. 21 (2006) 1150–1160.
[7] J. Selsing, J. Am. Ceram. Soc. 44 (1961) 419-419. [59] G. Pezzotti, O. Sbaizero, V. Sergo, N. Muraki, K. Maruyama, T. Nishida, J. Am.
[8] D.G. Grossman, J. Am. Ceram. Soc. 55 (1972) 446–449. Ceram. Soc. 81 (1998) 187–192.
[9] R.F. Cook, B.R. Lawn, T.P. Dabbs, P. Chantikul, J. Am. Ceram. Soc. 64 (1981) C-121–C- [60] H.Z. Wu, S.G. Roberts, B. Derby, Acta Mater. 56 (2008) 140–149.
122. [61] S. Guo, R.I. Todd, J. Eur. Ceram. Soc. 30 (2010) 2533–2545.
[10] R. Chaim, V. Talanker, J. Am. Ceram. Soc. 78 (1995) 166–172. [62] R. Young, X. Yang, Compos. A 27 (1996) 737–741.
[11] M. Kotoul, J. Pokluda, P. Sandera, I. Dlouhı, Z. Chlup, A.R. Boccaccini, Acta Mater. [63] R.I. Todd, A.R. Boccaccini, R. Sinclair, R.B. Yallee, R.J. Young, Acta Mater. 47 (1999)
56 (2008) 2908–2918. 3233–3240.
[12] A.W. Pryce, P.A. Smith, J. Mater. Sci. 27 (1992) 2695–2704. [64] K. Zeng, D.J. Rowcliffe, Acta Metall. Mater. 43 (1995) 1935–1943.
[13] K.T. Faber, Annu. Rev. Mater. Sci. 27 (1997) 499–524. [65] O. Peitl, F.C. Serbena, V.R. Mastelaro, E.D. Zanotto, J. Am. Ceram. Soc. 93 (2010)
[14] K.-L. Choy, P. Duplock, P.S. Rogers, J. Churchman-Davies, M.T. Pirzada, Mater. Sci. 2359–2368.
Eng., A 278 (1999) 187–194. [66] J. Soares, C.M. Lepienski, J. Non-Cryst. Solids 348 (2004) 139–143.
[15] M. Fischer, R. Marx, J. Dent. Res. 80 (2001) 336–339. [67] M. Murakami, Stress Intensity Factors Handbook, 1st ed. Pergamon, Oxford, 1987.
[16] B. Taskonak, J.J. Mecholsky Jr., K.J. Anusavice, Biomaterials 26 (2005) 3235–3241. [68] E.D. Zanotto, Cristais em Vidros: Ciência e Arte, EdUFSCar, Brazil, 2011 (photo in
[17] M.V. Swain, Acta Biomater. 5 (2009) 1668–1677. pg. 37 by V.O. Soares and M. Crovace).
[18] I. Denry, J.A. Holloway, Materials 3 (2010) 351–368. [69] V.M. Fokin, O.V. Potapov, E.D. Zanotto, F.M. Spiandorello, V.L. Ugolkov, B.Z.
[19] C.-K. Lin, T.-T. Chen, Y.-P. Chyou, L.-K. Chiang, J. Power Sources 164 (2007) Pevzner, J. Non-Cryst. Solids 331 (2003) 240–253.
238–251. [70] V. Fokin, O. Potapov, V. Ugolkov, E.D. Zanotto, F. Spiandorello, Russ. J. Phys. Chem.
[20] J. Malzbender, R.W. Steinbrech, L. Singheiser, Fuel Cells 9 (2009) 785–793. 77 (2003) 1639–1641.
[21] O. Peitl, E.D. Zanotto, F.C. Serbena, L.L. Hench, Acta Biomater. 8 (2012) 321–332. [71] G. Müller, Low Thermal Expansion Glass Ceramics, Springer-Verlag, H. Bach,
[22] V.R. Mastelaro, E.D. Zanotto, J. Non-Cryst. Solids 194 (1996) 297–304. Berlin, 1995, pp. 13–25.
[23] T. Mori, K. Tanaka, Acta Metall. Mater. 21 (1973) 571–574. [72] R. Sarno, M. Tomozawa, J. Mater. Sci. 30 (1995) 4380–4388.
[24] C.H. Hsueh, P.F. Becher, Mater. Sci. Eng., A 212 (1996) 22–28. [73] A.G. Evans, J. Am. Ceram. Soc. 73 (1990) 187–206.
[25] J.D. Eshelby, Proc. R. Soc. London, Ser. A 241 (1957) 376–396. [74] A. Sakamoto, Y. Himei, Y. Hashibe, Adv. Mater. Res. 39 (2008) 381–386.
[26] A.G. Evans, Acta Metall. Mater. 26 (1978) 1845–1853. [75] W. Ostertag, G.R. Fischer, J.P. Williams, J. Am. Ceram. Soc. 51 (1968) 651–654.
[27] R. Davidge, Acta Metall. Mater. 29 (1981) 1695–1702.
[28] T. Mura, Micromechanics of Defects in Solids, 2nd ed. Martinus Nijhuff Publisher,
Dortrecht, 1987.