Journal of the European Ceramic Society 37 (2017) 381–391

Contents lists available at www.sciencedirect.com

Journal of the European Ceramic Society

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

Spherical indentation for brittle fracture toughness evaluation by

considering kinked-cone-crack
Karuppasamy Pandian Marimuthu a , Felix Rickhey a , Jin Haeng Lee b , Hyungyil Lee a,∗
a
Department of Mechanical Engineering, Sogang University, Seoul, 04107, Republic of Korea
b
Division for Research Reactor, Korea Atomic Energy Research Institute, Daejeon, 34057, Republic of Korea

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

Article history: This work aims at evaluating the fracture toughness of brittle materials by spherical indentation. The
Received 30 March 2016 cone-cracking is simulated by the extended finite element method (XFEM) in Abaqus. The formation of a
Received in revised form 5 August 2016 kinked-cone-crack is observed when the indenter comes into (second) contact with the surface part out-
Accepted 11 August 2016
side the ring-crack. The effects of friction, Poisson’s ratio and cone-crack kinking on the Roesler’s constant
Available online 17 August 2016
c are analyzed. Based on numerical results, the Roesler’s method for evaluating the fracture toughness
is enhanced by considering kinked-cone-crack. By performing systematic XFE analyses, a database for
Keywords:
enhanced Roesler’s constant c | kink is provided for the fracture toughness evaluation of brittle materials.
Fracture toughness
Spherical indentation cracking
Finally, the proposed method is verified by conducting spherical indentation tests on soda-lime glass
Kinked-cone-cracks specimens.
XFEM © 2016 Elsevier Ltd. All rights reserved.
Roesler’s constant

1. Introduction tip. As a result, depending on the indenter shape, cone, radial, ring,
median, half-penny and lateral cracks form in the specimen. In the
Material failure under working condition often causes property evaluation of fracture toughness with formulated simple equations,
damage and accidents. Especially brittle materials exhibit catas- the dimensions of these cracks and indentation loads are related
trophic failure even under quasi-static loadings. A characterization with other material properties of specimen [3–7]. However, there
of materials based on fracture mechanics helps to prevent the is no universal formula applicable to evaluate the fracture tough-
material failure. In fracture mechanics, fracture toughness KIC is ness of all brittle materials. Also indentation test method requires
an important material property, which describes the resistance of a clear understanding of the complex deformation processes going
a material to fracture. However, evaluation of KIC of brittle mate- on in the material beneath the indenter.
rials is a challenging task, as conventional testing i.e. tensile test, This present work focuses on spherical indentation cracking,
bending test, compact tension test is tedious. Preparing those spec- which avoids micro-cracking and phase transformation in the
imens and performing controlled fracture test are difficult due to materials in contrast to the sharp indentation. The test materials
the brittleness of materials [1]. also show elastic behavior until fracture. Therefore the compli-
Compared with conventional test methods, indentation method cations associated with the residual stress caused by inelastic
for evaluation of fracture toughness [2] becomes more essential deformation in sharp indentation can be obviated with spherical
to predict the lifetime of brittle materials for engineering applica- indentation [4].
tions such as bearings, engine components, monoliths and coatings The aim of this work is to evaluate brittle fracture toughness KIC
[3]. The indentation test offers convenient evaluation of fracture by establishing a quickly and directly applicable spherical indenta-
toughness [3–7] on micro/nano scales. The indentation test is rela- tion method. The extended finite element method (XFEM) is used to
tively simple regarding specimen preparation and procedure, and simulate the cone-crack formation in Abaqus/Standard (ver. 6.14)
the indentation test results can be produced at lower cost and time. [8,9]. The formation of kinked-cone-crack (KCC), which was not
The indentation cracking test involves developing controlled frac- addressed in the spherical indentation studies [3–7], is discussed.
ture beneath the surface of brittle material by pressing an indenter The effects of friction, Poisson’s ratio and cone-crack kinking on
the Roesler’s constant c are analyzed. Based on XFEM, Roesler’s
law method [6] for evaluating brittle fracture toughness is modi-
fied by considering kinked-cone-crack. With further XFE analyses,
∗ Corresponding author.
a database for an enhanced Roesler’s constant c | kink is provided
E-mail address: [email protected] (H. Lee).

http://dx.doi.org/10.1016/j.jeurceramsoc.2016.08.014
0955-2219/© 2016 Elsevier Ltd. All rights reserved.
382 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391

only for small indenters. Experimental observations show that Pc

is proportional to the square of R (Pc ∝R 2 ) for large indenters [16].
Argon [17] also noted that the strength of glass depends on indenter
size.
In the quasi-static spherical indentations, if P > Pc , the cone-
crack substantially grows with an angle  to the surface. When
the cone-crack becomes relatively larger than ring-crack, further
propagation of cone-crack is not affected by the contact zone [18].
By conducting the spherical indentation test on soda-lime glass,
Roesler [18] reported that  depends only on Poisson’s ratio , and
thus  is a sort of invariant for a given ; soda-lime glass ( = 0.25)
resulted  = 22◦ [18].

Fig. 1. Example of evolved cone-crack in soda-lime glass under spherical indenta- 2.3. Prior fracture toughness evaluation methods
tion: indenter radius R, ring-crack radius ro , cone-crack base radius rc .
Under plane strain condition, the fracture toughness KIC is
related to the critical strain energy release rate or fracture energy
for the fracture toughness evaluation of diverse brittle materials.
 as follows
Spherical indentation tests are conducted on soda-lime glass spec-  
imens to validate the cone-crack kinking observed in XFE analyses. 1 − 2
2
 = KIC = 2 (2)
E
2. Prior spherical indentation methods where surface energy  is the energy required to create a unit area
of new surface in the materials. To evaluate fracture toughness,
In brittle materials, spherical indentation generates ring and the spherical indentation variables (P, Pc , a, ro , rc ) were related
cone-cracks beneath the indenter as shown in Fig. 1. Numerous to  based on following three basic methods: (i) Auerbach’s law
research works describe the spherical indentation system and the method, which uses the critical load Pc [5,19] (ii) minimum fracture
cone-crack formation [3–7,10–18]. Nonetheless, there are limited load method, which uses a value of minimum load Pcmin necessary to
articles which explain the complete cone-cracking in brittle mate- propagate cracks from at least 25 indentation tests [4,20] and (iii)
rials under spherical indenters [3–7]. Roesler’s law method [6,21], which uses cone-crack base radius rc
and corresponding P.
2.1. Spherical indentation (Hertzian theory) At larger indentation loads P, the cone-crack base radius rc varies
as the 2/3 power of indentation load P [6]. Based on this observation,
By assuming a frictionless contact between two elastic solids, Roesler [18] related P, rc and  as
Hertz [10] proposed an elastic theory to describe the complex stress
field in the contact materials. When an elastic spherical indenter P2 E
= (3)
with radius R indents an elastic half-space with indentation load P, rc3 
the contact radius a and the indentation depth h are given as [11] where  is a Roesler’s dimensionless -dependent coefficient for
 3PR 1/3 a2 1 1 − 2 1 − I2 well-developed cone-cracks (rc /ro > 5). Lardner et al. [6] re-wrote
a= ; h= ; = + (1) the Roesler’s law in terms of normalized energy release rate as
4Ee R Ee E EI
2 Erc3
where E, EI and , I are Young’s modulus and Poisson’s ratio of the Ĝ = = c (4)
half-space and the indenter, respectively. The above equations are P2
sufficient to specify the loading condition for all kinds of spheri- For well-developed cone-cracks, Ĝ saturates to a constant value
cal contact [12]. By extending Hertz’s theory, Huber [12] provided i.e. c (Roesler’s constant) for a given ; that is, c is also depen-
complete stress field solutions for spherical contact. Later, Hertzian dent on · Based on FE analyses, Lardner et al. [6] obtained the
theory was used to explain the initiation and propagation of ring normalized energy release rate and validated the Roesler’s law.
and cone-cracks in brittle materials [3–7]. For soda-lime glass, c values from the literature are listed in
Table 1. Due to the different experimental procedures used by var-
2.2. Cone-crack formation ious researchers, experimental scatter can be seen in c values for
the same material.
Fig. 2 schematizes the formation of the cone-crack under a The Auerbach’s law and minimum fracture load methods are gen-
spherical indenter. In the specimen surface, radial stresses  r are erally based on Pc . In the indentation test, Pc tends to vary with
tensile outside the contact circle (with radius = a) and reach a max- following factors: (i) specimen surface condition, especially sur-
imum rmax at the contact periphery (r = a) [11]. Therefore, it is face roughness, (ii) experimental setup, (iii) friction on the contact
convenient to understand the stress field in the surface as a local- interface and indenter elasticity [11] and (iv) surface residual stress.
ized tensile skin outside the contact circle [13]. A ring-crack may
be presumed to initiate at the contact periphery by satisfying the Table 1
maximum tensile stress fracture condition. But in experiments, the c values for soda-lime glass ( = 0.23) in the literature.
ring-crack radius ro is always larger than a; the spherical indenta-  (N/m) 103 × c method
tion tests resulted in the ratio ro /a = 1.1 to 1.4, which depends on
Lardner et al. [6] 8.0 14.0 FEM
flaw size [3], indenter radius, high stress gradient below the surface
Roesler [18] – 27.1a experiment
[7] and friction in the contact regions [14]. Lawn and Wilshaw [21] 8.0 12.8 experiment
At a critical indentation load Pc , the ring-crack grows into a cone- Lawn and Fuller [22] 7.8 11.7 experiment
crack. Auerbach [15] extensively studied the effect of indenter size Chaudhri and Phillips [23] 7.8 14.5 experiment
Davis et al. [24] 8.0 10.4 experiment
on cone-crack initiation in glass specimens, and gave a linear rela-
tionship between Pc and R as Pc ∝R (Auerbach’s law), which holds a
Overestimates the cone-crack radius at a given indentation load [6].
 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391 383

Fig. 2. Schematic representations of cone-crack formation.

Fig. 3. Cut view of FE model; enlarged view shows the refined inner region beneath
the indenter.

Fig. 4. Schematic representation of traction-separation law for damage initiation

Whereas Roesler’s law method is based on well-developed cone-
and propagation [29].
cracks, the above mentioned factors do not affect the propagation
of well-developed cone-cracks.
By conducting spherical indentation test,  can be calculated by ern the damage initiation and propagation on a continuum level.
Eq. (4) provided  and c are known. However, c values are hard to The traction-separation law governs damage initiation through
find in the literature for brittle materials. Only for soda-lime glass, a damage-initiating stress threshold  c . The damage is initiated
a large number of experimental resources are found to calculate when the applied tensile stress reach  c . Upon further loading
c , which is thus limited to  = 0.23. It would thus be worthwhile to the stiffness tensor is degraded according to the softening branch.
enhance the Roesler’s law method for an easy and quantitative eval- Once the separation ı reaches a critical value ıc , a crack has formed.
uation of fracture toughness. Therefore, this present work improves For linear softening,  c and ıc are related to the fracture energy 
the Roesler’s law method by analyzing the spherical indentation through = 12 c ıc . Since the considered fracture energy and other
cracking for a wide range of brittle materials. material properties are identical, we assumed that the variation of
 c (thus ıc for a given  ) has a negligible effect on the crack propa-
3. Spherical indentation cracking gation of well-developed cracks [29]. In fracture analysis, material
softening behavior and stiffness degradation may lead to severe
3.1. Finite element modelling convergence difficulties [30]. Hence, we use viscous regularization
by adding a small artificial viscosity (1 × 10−5 ) to the XFE model.
Spherical indentation cracking is simplified as an axisymmet- As a result, tangent stiffness matrix becomes to be positive definite
ric problem. An axi-symmetric XFE model, which contains about for sufficiently small time increments [9].
98000 nodes and 52000 4-node bi-linear axisymmetric elements
(CAX4) [9], is used (Fig. 3). The inner region, beneath the indenter, 3.2. Finite element analysis
is refined with square elements in order to maintain the same min-
imum element size e; the model thus can facilitate the cone-crack The formation of cone-cracks under rigid spherical indenters is
formation. The inner region elements are enriched with discontin- simulated in XFEM by considering soda-lime glass properties with
uous function for cracking analysis in XFEM. In the FE model, the f = 0.1 [25],  c = 0.3 GPa and  = 8 N/m [7]. As the indenter indents
nodes on the bottom are allowed to move only in the r-direction, the specimen surface, damage initiates ahead of the contact circle
while the axisymmetric boundary conditions are applied to the left in the surface. The ring-crack is then formed with radius ro , which
side nodes. is observed to be always larger than contact radius a. The ring-crack
The material is assumed to be isotropic and exhibit purely linear then propagates downwards with further loading. At critical load
elastic behavior. For preliminary studies, the material properties of Pc , the crack assumes a skirt-like shape and then grows very rapidly
soda-lime glass (E = 70 GPa,  = 0.23) are assigned to the FE model. as a cone-crack as shown in Fig. 5. In experiments, this rapid cone-
As the well-developed cone-cracks are independent of the contact crack formation is accompanied by acoustic emission [5]. When
condition [18] i.e. indenter deformation, a rigid spherical indenter maintaining P > Pc , the cone-crack propagates continuously with
with R = 500 m is employed. The appropriateness of the FE mesh an angle  to the surface. Note that the simulated cone-crack angle
is confirmed by comparing FE solutions of radial stress distribution  = 18◦ ± 0·5◦ for  = 0.23 is close to the experimental value for glass
and load-depth curve with Hertzian solutions. Since we consider [4].
dissimilar material properties for indenter and the specimen, the As P (>Pc ) increases, a also increases. If ro is not much larger than
friction plays significant roles in determining the ring-crack radius a, the indenter may come into (second) contact with the surface
during loading and initiating secondary ring-cracks during unload- part outside the ring-crack as shown in Fig. 5. The correspond-
ing [25]. Therefore, a non-zero friction coefficient is considered in ing indentation load is defined as second critical load Psc . As a
the contact region for cracking analysis. result, the stress field ahead of the crack tip changes as compared
XFEM is used to simulate the fracture. A more detailed descrip- in Fig. 6(a)-(b). The cone-crack extends along the new minimum
tion about XFEM can be found elsewhere [9,26–28]. We apply principle stress  3 directions. Therefore, the direction of cone-crack
the bilinear traction-separation (T−ı) law (Fig. 4) [9] to gov- propagation changes and a kink appears in the cone-crack. It is evi-
384 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391

Fig. 5. Alteration of Von Mises stress  v distribution due to cone-crack formation in soda-lime glass.  v is normalized by using maximum contact pressure po .  v and po are
function of P.

Fig. 6. Directions of maximum principle stress [arrows] before (left) and after kinking (right).

dent from Hubers’s stress field solutions that  affects the entire contact,  1 becomes tensile outside the ring-crack as shown in
stress field [12]; hence it also affects the angle of cone-crack prop- Fig. 6(b). If the fracture criterion is satisfied, the secondary (addi-
agation. The kink angle  K becomes functions of . In fine-grained tional) ring-cracks may form in the surface outside the ring-crack
silicon nitride (Si3 N4 ) monoliths and bilayers, the kink appeared in and extend along in loading direction as shown in Fig. 8. These
the cone-crack as shown in Fig. 7 [31] under WC spherical indenter secondary ring-cracks do not spread out as secondary cone-cracks,
with R = 1.98 mm at load P = 3 kN. But Lee et al. [31] didn’t state the since the primary cone-crack prevents the development of tensile
phenomena of kinking in their work. stress in the direction normal to the primary cone-crack [32]. In
Upon further loading (P > Psc ), the kinked-cone-crack continues experiments, the secondary ring-cracks may merge with the pri-
to propagate with an angle  K to the surface. Due to the second
 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391 385

Fig. 7. Hertzian contact damage in Si3 N4 monoliths (left) and bilayers (layer thickness = 600 m) (right) under WC spherical indenter with R = 1.98 mm at P = 3 kN. Contact
diameter is indicated with A-A [31].

cone-crack, which is equivalent to the real size and shape of cone-crack

observed in experiment at particular load. For that, we use calibra-
tion techniques of Cao and Zhang [33] and mean strength theory
of Bao et al. [7], which average the maximum principle stress in an
element for stress-based crack initiation in the element-failure FE
analyses.  sc is calculated by following Cao and Zhang’s [33] relation
as follows
√  1/2 √
sc =
KIC / e =
E  / e (5)

where E = E/(1− 2 ). Cao and Zhang [33] calibrated the dimension-

less parameter
as 1.2 for e = 8.17 m by comparing FE results
with those of Lardner et al. [6] for soda-lime glass. XFEM is used
here to calibrate
for different e by comparing  c | kinkless with
the literature data [6,21–24], and thereby, to select an appropriate
 sc for considered e. Note that,  sc is not an actual material con-
Fig. 8. Formation of secondary ring-cracks in spherical indentation.
stant, but equivalent to  c for considered e. From mesh refinement
study, the XFE model with e = 5 m is found to be appropriate to
mary cone-crack to produce a ‘detachable collar’ around the neck simulate the cone-cracking with better convergence at relatively
of the cone (Fig. 8) [32]. low computational cost. By simulating KLCC,
is then calibrated as
0.83 for e = 5 m (based on Table 2) by selecting soda-lime glass as
4. Results and discussion a characteristic material, which has c ≈ 12 × 10 −3 (Table 1) as a
characteristic value.
Eq. (4) was developed based on a single cone-crack regime by
assuming that  is invariant for a given material. However, the 4.2. Analyzing the characteristics of Roesler’s constant c
XFE analyses support the formation of kinked-cone-crack under
spherical indenters. Due to kinking, the cone-crack angle changes To see that c depends only on , we use two different  = 7,
 (before the kinking) to  K (after the kinking). Therefore, one can- 8 N/m and  = 0.21, 0.23 in the XFE analysis.  sc is calculated based
not accurately evaluate the  from Eq. (4) without considering the on Eq. (5). Fig. 10(a) shows the variation of Ĝ as a function of
effect of kinking on c . In the following sections, the effect of kinking rc /ro . We use ro to normalize rc , since ro is constant in spherical
on c is investigated. indentation. Above rc /ro > 7, Ĝ reaches a saturated value, which is
To analyze the effect of kinking on c , a comparison between represented as c . Fig. 10(a) also confirms that the saturated values
the values of c calculated from a kinkless cone-crack (KLCC) i.e. c vary with , not with  . When kinking occurs at Psc ,  changes to 
| kinkless and from the KCC i.e. c | kink are needed. To obtain KLCC K . As a result, Ĝ decreases [Fig. 10(b)]; however, Ĝ reaches a satu-
even at higher loads P » Pc , Lawn and Fuller [22] used truncated rated value for well-developed KCC. But the saturated Ĝ values for
spherical indenter with a base radius of 1 mm, and Lardner et al. [6] KCC are smaller than those for KLCC as shown Fig. 10(b). It confirms
used Hertzian pressure distribution (in FEM) instead of spherical that the kinking has significant effect on c .
indenter to ensure that the contact radius be less than ring-crack The effect of friction on c is analyzed by considering various
radius. Here, the indenter shape is slightly modified as shown in f = 0, 0.1, 0.12, 0.15, 0.2 in the contact interface between the rigid
Fig. 9(b) and KLCC is produced even at higher loads by avoiding the indenter and the specimen. When f increases, Pc and ro increase
second contact of indenter with cracked surface. Fig. 9(c) compares as listed in Table 3. The changes in Pc and ro change Psc and the
the load-depth curves (P-h) from kinked and kinkless cone-crack location of kinking in the cone-crack. As a result, the Ĝ and the
system. At higher loads, P-h curve of KCC deviates from those of load-depth curves are significantly affected when the cone-crack
KLCC. The value of c | kinkless , which is calculated from the simu- becomes large as shown in Fig. 11(a) and (b), respectively. We
lated KLCC, is compared with the c values in Table 1 to validate observe negligible effect of friction on load-depth curve when the
the XFE results. cone-cracks are relatively small. Unlike ,  K is not always function
of Poisson’s ratio. It gently depends on ring-crack and kinking loca-
4.1. Applying the calibration techniques to XFE analysis tion in most cases. This is the reason for the small variation in  K
for the change in friction coefficient from f = 0 to f = 0.1 in Table 3.
The element size has substantial effects on damage initiation The friction has a negligible effect on c as the change in f has trivial
(i.e. critical load Pc ). When we use the same  c for different ele- effects on  and  K . Therefore, Ĝ curves reach almost similar satu-
ment size e in the inner zone, we obtain different critical load and rated values for well-developed KCC [Fig 13(a)] even with different
ring-crack radius. In order to avoid this mesh-dependency, we use a f. For further analysis, the non-zero friction coefficient is thus taken
scaled damage initiation strength  sc for particular e to simulate the as f = 0.1.
386 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391

Fig. 9. Formed (a) KCC under spherical indentation and (b) KLCC under modified spherical indentation at Pmax = 0.60 kN. (c) Comparisons of load-depth curves of kinked and
kinkless cone-crack system.

Table 2
Calibration of
for KLCC in XFEM (E = 70 GPa,  = 0.23, f = 0.1, R = 500 m, rc /ro > 7).

e ( m)  (N/m)  sc (GPa)
Pc (kN) 103 ×  c |
kinkless

5 8 0.185 0.95 0.083 10.0

5 8 0.180 0.93 0.080 10.1
5 8 0.170 0.88 0.065 10.9
5 8 0.165 0.85 0.060 11.3
5 8 0.160 0.83 0.058 11.8

Bold values are calibrated values.

Fig. 10. Comparison of Ĝ variation (a) for KLCC with different  = 0.21, 0.23 and  = 7, 8 N/m and (b) between KLCC and KCC.

Table 3
Effect of friction on cone-cracking (E = 70 GPa,  = 0.23,  = 8 N/m,  sc = 0.16 GPa, rc /ro > 7, KCC).

f Pc (kN) ro ( m) (deg.)  K (deg.) 103 × c |

kink

0.00 0.041 87.5 18 29 9.33

0.10 0.061 107.5 18 26 9.40
0.12 0.068 112.5 18 26 9.50
0.15 0.081 127.5 18 26 9.60
0.20 0.096 142.5 18 26 9.39
 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391 387

Fig. 11. Comparison of (a) Ĝvariation and (b) load-depth curves for simulated KCC with different f = 0.0, 0.1, 0.12, 0.15, 0.2.

Fig. 12. (a)−(c) P vs. rc plots and (d)−(f) variation of Ĝ as a function of rc /ro for KCC in TiB2 , ZrB2 , PcBN (These plots are presented here for a part of considered brittle materials.
For all considered brittle materials, the same plots are separately presented in the Supplementary materials).

To check the effects of E and  sc on c , we have also simu- accommodating smaller elastic deformation and increases differ-
lated the KCC and KLCC formation for following material properties ence between  and  K . As a result the effect of kinking is significant
E = 410 GPa,  = 50 J/m2 ,  sc = 1 GPa and  = 0.14, 0.23. Compar- for smaller . Whereas a larger  enhances the failure-resisting abil-
ing c values for  = 0.23 in Tables 3 and 4, we observe negligible ity i.e. toughness of the material by accommodating larger elastic
effect of E,  sc on c . For different , the effect of kinking on c deformation [34] and decreases the difference between  and  K .
is listed in Table 4. A smaller  enhances the crack propagation by Hence, the effect of kinking on c is relatively small. Henceforth, it’s
388 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391

Fig. 13. Variation of (a) c and (b) cone-crack angles as a function of .

Table 4
Effect of  and kinking on c (E = 410 GPa, f = 0.1,  = 50 N/m,  sc = 1 GPa, rc /ro > 7).

  (deg.)  K (deg.) 103 × c | 103 × c | c

kink kinkless difference

0.14 23 – – 20.9 9.65 × 10−3

0.14 23 35 11.7 –
0.23 18 – – 11.4 1.9 × 10−3
0.23 18 26 9.50 –

suggested to use the enhanced Roesler’s constant  c | kink instead material’s , this method provides the fracture toughness based on
of  c or  c | kinkless in Eq. (4), especially for the materials with reverse engineering analysis. Therefore one can apply this method
relatively small Poisson’s ratio. to a brittle material whose fracture toughness is not known a priori.

4.4. Measurement of fracture toughness

4.3. Generating database of  c ( c | kink and  c | kinkless )
In this section, we explain the method to evaluate fracture
To generate database of c , we collect material properties (E, toughness of brittle materials by considering well-developed KCC.
, KIC ) of various brittle materials from several literatures. As By substituting maximum indentation load Pmax and c | kink instead
the microstructure of ceramics plays an important role in brit- of P and c , respectively, we re-write Eq. (4) as follows
tle responses under spherical contact, the material properties of
2
1 Pmax
homogenous and fine-grained ceramics are used in this analysis.  = c |kink (6)
Using Eq. (2) and Eq. (5), we then calculate the corresponding  E2 rc3
and  sc (
= 0.83), respectively, for cracking analyses in XFEM as
By combining Eq. (2) and Eq. (6), KIC is then given as
listed in Table 5. Spherical indentation cracking are analyzed for
various brittle materials. Since we use  sc in XFEM, we achieve E
1/2 2
1 Pmax c |kink
1/2

good convergence for considered brittle material properties. For KIC (or) KIC |kink = ( ) =[ ] (7)
1 − 2  rc3 1 − 2
2
KCC system, increasing rc is plotted against increasing P as shown in
Fig. 12. The plot P vs. rc can be regressed with 2nd order polynomial The fracture toughness of the indented material can be calcu-
or exponential functions for well-developed cone-cracks (rc /ro > 5). lated from Eq. (7) by knowing , c | kink , Pmax and rc . By considering
The regression functions for particular material can be used to pre- KCC, this research provides c | kink values for a wide range of  as
dict rc under applied P, and thereby, to validate the experimental listed in Table 5. By conducting spherical indentation test up to a
data. We then plot the variation of Ĝ against increasing rc /ro , and we specified Pmax , one can measure corresponding rc by using optical
thus get saturated Ĝ values, that is  c | kink , for each brittle material. microscope for transparent specimens or the serial cross-sectioning
By following the same procedure,  c | kinkless are calculated from methods for opaque materials (see reference [48]). In addition, the
the KLCC. Obtained  c | kink values are smaller than  c | kinkless as experimental values of rc can be compared with those calculated
listed in Table 5. based on regression in Fig. 12 (also see Supplementary data). When
The variation of  c | kink and  c | kinkless with  has two regions I we considered  c | kinkless in Eq. (7), there are some error involves
and II as shown in Fig. 13(a). In region I, c | kink is rather sensitive to in fracture toughness evaluation as shown in Fig. 14. The errors are
 K which depends on  and kinking-location, thus giving scattered estimated quantitatively as error (%) = (KIC | kink −KIC | kinkless )/KIC |
 c | kink . Also the phenomena of kinking play an important in this kink . KIC | kinkless are evaluated by substituting  c | kinkless instead
region. In the region II, c | kink is more sensitive to Poisson’s ratio, of  c | kink in Eq. (7), and Pmax and rc of kinked-cone-crack. As we
and exponentially decays with . c | kinkless variation appears not explained in the Section 4.2, the effect of kinking on KIC is relatively
to change in Region II, when compared with Region I· Regressing large when the materials Poisson’s ratio is small. The percentage of
the plot c |kink vs.  with two regression functions for each region error gradually decreases with increasing of . Hence, this research
I and II, we can obtain KIC with ± 5% accuracy. Also, both  and  K work suggests to use  c | kink in Eq. (7) while evaluating the fracture
depend on  as shown in Fig. 13(b). The c | kink values in Table 5 can toughness of brittle materials by spherical indentation.
be used to evaluate KIC from spherical indentation by considering
cone-crack kinking. For all considered brittle materials, the P vs. 4.5. Experimental validation of proposed method
rc plots are separately presented in the supplementary materials.
Known fracture energies are used in XFEM analyses only to develop Spherical indentation tests are conducted to verify the forma-
database of  c for a wide range of  as listed in Table 5. By knowing tion of kinked-cone-crack and validate the proposed method. A
 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391 389

Table 5
Database of  c ( c | kink and  c | kinkless ) values from XFE analysis.

materials literature data XFE results

E (GPa)   (N/m) 103 ×  c | kink 103 ×  c | kinkless  |  K (deg.)

Titanium diboride (TiB2 ) [35] 565 0.11 67.2 14.1 26.0 24.0 | 37.5
Zirconium diboride (ZrB2 ) [36] 346 0.11 16.5 15.5 25.0 24.0 | 36.3
Polycrystalline cubic boron Nitride (PcBN)[37] 894 0.12 51.0 14.3 24.1 23.4 | 36.1
Fused silica [38] 70 0.14 9.0 13.2 20.7 22.6 | 35.4
Silicon carbide (SiC) [39] 410 0.164 61.7 11.8 18.2 21.3 | 33.4
Boron carbide (B4 C) [40] 465 0.17 17.6 13.2 18.1 20.7 | 32.1
Magnesium oxide (MgO) [41] 317 0.17 19.2 13.0 18.8 20.7 | 32.1
Borofloat glass [42] 64 0.18 16.0 11.5 16.4 20.4 | 32.1
Titanium carbides (TiC) [43] 400 0.19 38.6 11.4 16.1 19.8 | 31.0
Vanadium carbide (VC) [19] 320 0.19 5.5 12.7 16.3 19.4 | 29.3
Starphire [42] 73 0.20 14.8 11.0 14.0 19.2 | 29.3
Pyrex (borosilicate glass) [41] 64 0.20 9.0 11.2 14.1 19.4 | 30.0
Zirconium carbides (ZrC) [41] 407 0.20 9.43 12.2 14.5 18.9 | 29.1
Tungsten carbides (WC) [19] 680 0.22 12.3 11.5 12.7 18.1 | 27.1
a
Soda-lime glass [43] 70 0.23 7.61 9.65 11.3 18.0 | 27.0
Aluminium oxynitride (AlON) [44] 323 0.24 15.4 8.50 10.0 16.9 | 25.9
Aluminium nitride (AlN) [45] 308 0.25 42.9 8.30 10.2 16.3 | 26.5
a
Silicon nitride (Si3 N4 ) [46] 335 0.27 42.1 7.90 7.40 15.5 | 22.8
Alumina (Al2 O3 ) [4] 393 0.27 47.8 7.50 8.00 15.4 | 23.1
Calcium fluoride (CaF2 ) [41] 108 0.29 5.43 6.50 6.60 14.2 | 21.6
Strontium fluoride (SrF2 ) [41] 90 0.29 10.2 6.90 5.90 14.2 | 21.3
Yttria (Y2 O3 ) [47] 176 0.30 8.74 6.32 5.70 13.4 | 20.0
a
Only for soda-lime glass and Si3 N4 [31],  and  K are compared with experimental observations.

outside the ring-crack. However KCC may not occur in a flat-ended

cylindrical indentation because the contact radius a is constant, and
always smaller than the ring-crack radius rc .
It is hard to discern the cone-crack-tip because the cone-crack
closes during unloading. Therefore, we put a drop of commercial
red ink on the surface before starting unloading (test A) and after
unloading (test B). Due to the capillary effect, the red ink is drawn
inside the crack up to the cone-crack-tip; doing so the cone-crack
tips become visible as shown Fig. 16. Radial and secondary ring-
cracks are also observed in the experiments. The dimensions of
radial crack may be used to evaluate the fracture toughness. Dis-
cussing the radial cracks is beyond the scope of present work. It will
be studied in the future work.
Propagated cone-cracks at different indentation loads Pmax = 0.1,
Fig. 14. Error involved in fracture toughness evaluation with  c | kinkless in Eq. (7). 0.2, 0.3, 0.4 and 0.5 kN are compared in Fig. 16. The experimen-
tal values of rc is also compared with those obtained in XFE
analysis. The rc = 745 ± 5 m at Pmax = 0.5 kN is used to validate
micro indentation machine (Indentation system, R&B Inc., South the proposed approach as Pmax = 0.5 kN produces well-developed
Korea) with a spherical indenter tip is used. The indentation cone-cracks (rc /ro > 7) in soda-lime glass. By using Eq. (4), the
machine has a load resolution of ± 5 N. The indenter tip is a experimental value of normalized energy release rate is calculated
tungsten-carbide sphere with R = 500 m. The experiments are asĜ = 9.14 × 10−3 , which is in close agreement with the modeled
carried out on soda-lime glass specimens with 100 mm diame- value of Ĝ = 8.96 × 10−3 at Pmax = 0.5 kN; XFE model is validated.
ter and 20 mm thickness (Seoul special glass Ind., South Korea). Hence the modeled value of the c | kink = 9.65 × 10−3 in Table 5
The glass specimens are used in an as-received condition with no is used to evaluate the fracture toughness of soda-lime glass as
further annealing or surface treatment. The indenter is pressed KIC = 0.79 MPa m1/2 , which is in good agreements with those in liter-
into the specimen surface with a loading rate of 5 m/min and atures KIC = 0.7 ∼ 0.8 MPa m1/2 for soda-lime glass [4]. This, in some
the load-depth data are recorded. The indentation machine is pro- sense, validates the generated database for  c | kink .
grammed in a way of displacement control. If P reaches the specified
maximum indentation loads P max (= 0.1, 0.2, 0.3, 0.4, 0.5 kN), the
unloading process starts with an unloading rate of 5 m/min. 2–5 5. Summary and conclusions
indentations are made on the smaller specimens, and minimum 5
indentations are made on the larger specimens. In this research work, we established a quickly and directly
An optical microscope is used for taking the images. The cross applicable spherical indentation method for evaluation of brittle
section images of KCC [Fig. 15(c)] are captured after sectioning the fracture toughness KIC by considering kinked-cone-crack (KCC). By
specimens. The sectioning process includes scratching the indented selecting soda-lime glass as a reference material, we simulated the
surface with a glass cutter and then applying 3-point bending load cone-crack formation in XFEM. The formation of KCC was observed
to the specimen. The sectioned surfaces are polished by using emery when the indenter comes into (second) contact with the surface
papers with grit size 320 and 600 to make smooth surfaces. Fig. 15 part outside the ring-crack. The element size effect on cone-crack
compares the experimental cone-crack with that from XFEM anal- propagation was reduced by scaling the damage initiation strength.
yses. The experimental observations also confirm the formation of The effects of friction and Poisson’s ratio and cone-crack kinking
KCC when the indenter comes into contact with the surface part on the Roesler’s constant c were analyzed. To apply this method
390 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391

Fig. 15. (a) ring and (b) simulated and (c) experimentally observed cone-cracks (section A−A) in soda-lime glass under spherical indenter.

Fig. 16. Top view propagated cone-cracks at different Pmax = 0.1, 0.2, 0.3, 0.4, 05 kN. Cone-crack edge is clarified by dyeing. Experimental values of rc is compared with XFE
values.

for fracture toughness evaluation to a wide range of brittle mate- References

rials with diverse , we provided a database for c (c | kink and
c | kinkless ) with further XFE analyses. Indentation tests were con- [1] B.N. Jaya, C. Kirchlechner, G. Dehm, Can microscale fracture tests provide
reliable fracture toughness values? A case study in silicon, J. Mater. Res. 30
ducted on soda-lime glass to confirm the cone-crack kinking. The (2015) 686–698.
following conclusions can be drawn from this study, [2] M. Sebastiani, K.E. Johanns, E.G. Herbert, G.M. Pharr, Measurement of fracture
→ Both XFEM and experimental studies support the formation toughness by nanoindentation methods: recent advances and future
challenges, Curr. Opin. Solid State Mater. Sci. 19 (2015) 324–333.
of kinked-cone-crack under spherical indenters. [3] B.R. Lawn, Indentation of ceramics with spheres: a century after Hertz, J. Am.
→ For a given material,  c | kink is smaller than c | kinkless . Ceram. Soc. 81 (1998) 1977–1994.
→ Friction has significant effect on cone-crack initiation, but not [4] P.D. Warren, Determining the fracture toughness of brittle materials by
Hertzian indentation, J. Eur. Ceram. Soc. 15 (1995) 201–207.
on  c . [5] T.R. Wilshaw, The Hertzian fracture test, J. Phys. D: Appl. Phys. 4 (1971)
→ P vs. rc plot can be regressed with 2nd order polynomial or 1567–1581.
exponential functions for developed cone-crack (rc /ro > 5). [6] T.J. Lardner, J.E. Ritter, G.-Q. Zhu, Spherical indentation and fracture of glass
plates, J. Am. Ceram. Soc. 80 (1997) 1851–1862.
Based on the experimental data, we obtained the fracture tough-
[7] Y.W. Bao, S.B. Su, J.J. Yang, L. Sun, J.H. Gong, Nondestructively determining
ness of soda-lime glass as KIC = 0.79 MPa m1/2 , which is in good local strength and residual stress of glass by Hertzian indentation, Acta Mater.
agreements with KIC = 0.7 ∼ 0.8 MPa m1/2 in literature. This vali- 50 (2002) 4659–4666.
dates the generated database of c for other materials. [8] D.Y. Tumbajoy-Spinel, E. Feulvarch, J.-M. Bergheau, G. Kermouche, 2D
axisymmetric X-FEM modeling of the Hertzian cone-crack system, C. R.
Mecanique. 341 (2013) 715–725.
Acknowledgment [9] Abaqus Version 6.14, User’s manual, (Dassault Systems Simulia Corp.,
Providence, RI, USA, (2014).
[10] H. Hertz, On the contact of elastic solids, J. Reine. Angew. Math. 92 (1881)
This research was supported by Basic Science Research Program 156–171.
through the National Research Foundation of Korea (No. NRF-2015 [11] K.L. Johnson, Contact Mechanics, Cambridge University Press, 1985.
R1A2A1A 15056163). [12] M.T. Huber, Zur Theorie der Berührung fester elastischer Körper (On the
theory of contacting solid elastic bodies), Ann. Phys. 319 (1904) 153–163.
[13] F.B. Langitan, B.R. Lawn, Hertzian fracture experiments on abraded glass
Appendix A. Supplementary data surfaces as definitive evidence for an energy balance explanation of
Auerbach’s law, J. Appl. Phys. 40-10 (1969) 4009–4017.
[14] K.L. Johnson, J.J. O’Connor, A.C. Woodward, The effect of the indenter elasticity
To get more insight into the proposed method, we provided Sup- on the Hertzian fracture of brittle materials, Proc. Phys. Soc. Lond. Sect. A 334
plementary data, which can be found in the online version, of P vs. (1973) 95–117.
rc and Ĝvs. rc /ro plots (Fig. 12) for brittle materials in Table 5 by [15] F. Auerbach, Measurement of hardness, Annu. Rev. Phys. Chem. 43 (1891)
61–100.
considering KCC. Supplementary data associated with this article [16] J.P.A. Tillett, Fracture of glass by spherical indenters, Proc. Phys. Soc. Lond.
can be found, in the online version, at http://dx.doi.org/10.1016/j. Sect. B 69 (1956) 47–54.
jeurceramsoc.2016.08.014.
 K.P. Marimuthu et al. / Journal of the European Ceramic Society 37 (2017) 381–391 391

[17] A.S. Argon, Investigations of the Strength and Anelasticity of Glass, Doctoral [34] J. Liu, X. Wang, Effects of Poisson’s ratio on scaling law in Hertzian fracture,
Dissertation, Massachusetts Institute of Technology, 1956. Acta Mech. Solida. Sin. 22 (2009) 474–478.
[18] F.C. Roesler, Brittle fracture near equilibrium, Proc. Phys. Soc. Lond. Sect. B 69 [35] R.G. Munro, Material properties of titanium diboride, J. Res. Natl. Inst. Stand.
(1956) 981–992. Technol. 105 (2000) 709–720.
[19] R. Warren, Measurement of the fracture properties of brittle solids by [36] S.-Q. Guo, Densification of ZrB2-based composites and their mechanical and
Hertzian, Acta Mater. 26 (1978) 1759–1769. physical properties: a review, J. Eur. Ceram. Soc. 29 (2009) 995–1011.
[20] A. Franco, S.G. Roberts, P.D. Warren, Fracture toughness, surface flaw sizes [37] M.P. D’Evelyn, T. Taniguchi, Elastic properties of translucent polycrystalline
and flaw densities in Al2 O3 , Acta Mater. 45 (1997) 1009–1015. cubic boron nitride as characterized by the dynamic resonance method,
[21] B.R. Lawn, R. Wilshaw, Review indentation fracture: principles and Diamond Relat. Mater. 8 (1999) 1522–1526.
applications, J. Mater. Sci. 10 (1975) 1049–1081. [38] T. Suratwala, L. Wong, P. Miller, M.D. Feit, J. Menapace, R. Steele, P. Davis, D.
[22] B.R. Lawn, E.R. Fuller, Equilibrium penny like cracks in indentation fracture, J. Walmer, Sub-surface mechanical damage distributions during grinding of
Mater. Sci. 10 (1975) 2016–2024. fused silica, J. Non-Cryst. Solids 352 (2006) 5601–5617.
[23] M.M. Chaudhri, M.A. Phillips, Quasi-static indentation cracking of thermally [39] K.-Y. Lim, Y.-W. Kim, K.J. Kim, Mechanical properties of electrically conductive
tempered soda-lime glass with spherical and Vickers indenters, Philos. Mag. silicon carbide ceramics, Ceram. Int 40 (2014) 10577–10582.
A. 62 (1) (1990) 1–27. [40] R.B. Leavy, R.M. Brannom, O.E. Strack, The use of sphere indentation
[24] B. Davis, H.C. Cao, G. Bao, A.G. Evans, The fracture energy of interfaces: an experiments to characterize ceramic damage models, Int. J. Appl. Ceram.
elastic indentation technique, Acta Metall. Mater. 39 (1991) 1019–1024. Technol. 7 (2010) 606–615.
[25] D. Elaguine, M.-A. Brudieu, B. Storakers, Hertzian fracture at unloading, J. [41] C.B. Carter, M.G. Norton, Ceramic Materials: Science and Engineering,
Mech. Phy. Solids. 54 (2006) 2453–2473. Springer-Verlag, New York, 2013.
[26] T. Belytschko, T. Black, Elastic crack growth in finite elements with minimal [42] A.A. Wereszczak, C.E. Anderson, Borofloat and starphire float glasses: a
remeshing, Int. J. Numer. Eng. 45 (1999) 601–620. comparison, Int. J. Appl. Glass Sci. 5 (2014) 334–344.
[27] N. Moes, T. Belytschko, Extended finite element method for cohesive crack [43] S.Y. Chen, T.N. Farris, S. Chandrasekar, Contact mechanics of Hertzian
growth, Eng. Fract. Mech. 69 (2002) 813–833. cone-cracking, Int. J. Solids. Struct. 32 (1995) 329–340.
[28] J.M. Melenk, I. Babuska, The partition of unity finite element method: basic [44] C.T. Warner, T.M. Hartnett, D. Fisher, W. Sunne, Characterization of ALONTM
theory and applications, Comput. Methods Appl. Mech. Eng. 39 (1996) optical ceramic, Proc. SPIE 5786 (2005) 95–111.
289–314. [45] J. Lankford, W.W. Predebon, J.M. Staehler, G. Subhash, B.J. Pletka, C.E.
[29] H.C. Hyun, F. Rickhey, J.H. Lee, M. Kim, H. Lee, Evaluation of indentation Anderson, The role of plasticity as a limiting factor in the compressive failure
fracture toughness for brittle materials based on the cohesive zone finite of high strength ceramics, Mech. Mater. 29 (1998) 205–218.
element method, Eng. Fract. Mech. 134 (2015) 304–316. [46] S.K. Lee, S. Wuttiphan, B.R. Lawn, Role of microstructure in hertzian contact
[30] Y.F. Gao, A.F. Bower, A simple technique for avoiding convergence in finite damage in silicon nitride: I Mechanical characterization, J. Am. Ceram. Soc. 80
element simulations of crack nucleation and growth on cohesive interfaces, (1997) 2367–2381.
Model. Simul. Mater. Sci. Eng. 12 (2004) 453–463. [47] I.C. Albayrak, S. Basu, A. Sakulich, O. Yeheskel, M.W. Barsoum, Elastic and
[31] K.S. Lee, S.K. Lee, B.R. Lawn, Contact damage and strength degradation in mechanical properties of polycrystalline transparent Yttria as determined by
brittle/quasi-plastic silicon nitride bilayers, J. Am. Ceram. Soc. 81 (1998) indentation techniques, J. Am. Ceram. Soc. 93 (1997) 2028–2034.
2394–23404. [48] T. Lube, Indentation crack profiles in silicon nitride, J. Eur. Ceram. Soc. 21
[32] F.C. Frank, B.R. Lawn, On the theory of Hertzian fracture, Proc. R. Soc. Lond. (2001) 211–218.
Ser. A 299 (1967) 291–306.
[33] Y. Cao, G. Zhang, Multiple surface crack initiation and growth in glass
ceramics loaded by a sphere, Eng. Fract. Mech. 67 (2000) 277–292.