Engineering Fracture Mechanics 162 (2016) 232255

Contents lists available at ScienceDirect

Engineering Fracture Mechanics

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

Analytical fractal model for rugged fracture surface of brittle

materials
Lucas M. Alves a,, A.L. Chinelatto b, Edson Cezar Grzebielucka b, Eduardo Prestes b,
L.A. de Lacerda c
a
GTEME Grupo de Termodinmica, Mecnica e Eletrnica dos Materiais, Departamento de Engenharia de Materiais, Setor de Cincias Agrrias e de
Tecnologia, Universidade Estadual de Ponta Grossa, Av. Gal. Carlos Calvalcanti, 4748, Campus UEPG/Bloco CIPP Uvaranas, Ponta Grossa, PR CEP. 84030.900, Brazil
b
Departamento de Engenharia de Materiais, Setor de Cincias Agrrias e de Tecnologia, Universidade Estadual de Ponta Grossa, Av. Gal. Carlos Calvalcanti, 4748,
Campus UEPG/Bloco CIPP Uvaranas, Ponta Grossa, PR CEP. 84030.900, Brazil
c
LACTEC Instituto de Tecnologia para o Desenvolvimento, Universidade Federal do Paran - PPGMNE, Centro Politcnico da Universidade Federal do Paran,
Cx. P. 19067, Curitiba, PR, Brazil

a r t i c l e

i n f o

Article history:
Received 28 August 2015
Received in revised form 15 May 2016
Accepted 17 May 2016
Available online 26 May 2016
Keywords:
Fractal dimension
Fracture profile
Mortar
Heavy clay
Roughness
Self-affine surface

a b s t r a c t
The fractal modeling of a rugged fracture surface has received different purposes. However
none definitive model for the most of materials has been reached. Therefore, a general selfaffine fractal model is proposed for fracture surfaces and applied to heavy clay and mortar.
An analytical expression for the rugged crack length is obtained for application on fractal
fracture mechanics. Stereoscopic images are obtained for each tested specimens. Image
processing filters are used to extract the rugged profile of the cracks. The box-counting
and sand-box methods are used on the crack profile to obtain the local and the global
roughness exponents. Specimens prepared under different conditions validated the model.
Mortars and heavy clay specimens were characterized by measuring their modulus of rupture and the rugged crack profile under 3-point bending tests. A good agreement between
the model and the experimental results was observed. A strong correlation between the
fractal dimension and the sintering temperature for heavy clay specimens was verified.
The results also showed that the increasing rugged crack length of the profile of the fractured mortar specimens is well correlated with the increase in water/cement ratio. These
results validate the application of the proposed model for estimating the fracture strength
of brittle materials.
2016 Elsevier Ltd. All rights reserved.

1. Introduction
The fracture surface is a record of the information left by the fracture process. Generally, the rugged fracture surface profile has fractal geometry, so it is possible to establish a relationship between its topology and physical quantities of the fracture mechanics using fractal characterization techniques. However, the classical fracture mechanics (CFM) was developed
idealizing a flat, smooth, and regular fracture surface, as the geometry of crack surfaces is usually rugged and cannot be
easily described by the Euclidean geometry [1]. In this sense, the mathematical basis of CFM considers an energetic
equivalent between the rugged and the projected fracture surfaces [2]. Besides the mathematical complexity, part of this

Corresponding author.
E-mail addresses: lucasmaximoalves@gmail.com (L.M. Alves), adilsonchinelatto@gmail.com (A.L. Chinelatto), alkimin@lactec.org.br (L.A. de Lacerda).
http://dx.doi.org/10.1016/j.engfracmech.2016.05.015
0013-7944/ 2016 Elsevier Ltd. All rights reserved.

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

233

Nomenclature

Latin characters
0
index to denote measurements taken on the projected plane
a
minimal area of fracture surface
DA
area of fracture surface
A1; A2; A3; Ar1; . . . ; Ar30 mortar and ceramic samples
B1; B2; B3; Br1; . . . ; B30 mortar and ceramic samples
C1; C2; C3; Cr1; . . . ; Cr30 mortar and ceramic samples
d
infinitesimal increment
D1; D2; D3; Dr1; . . . ; Dr30 mortar and ceramic samples
DB
fractal box dimension
Df
fractal dimension
divider dimension
DD
DBC
fractal dimension measured by box-counting method
e0
fractal cell size or minimal crack length in transversal direction
E
elastic or Young s modulus
crack length measured in transversal direction
E0
DE0
variation of crack length measured in transversal direction
f
general functions
g
general functions or subscript index for global quantity
G
elastic energy release rate
h0
fractal cell size or minimal crack length in vertical direction
H
Hursts exponent
H0
plane projected crack height
DH0
variation in the plane projected crack height
K
multiplicative factor
Kx
multiplicative factor in x-direction
multiplicative factor in y-direction
Ky
l
subscript index for local quantity
L
rugged crack length
LBC
rugged crack length measured by box-counting method
lx ; ly ; lz rulers length or scale in the x-direction, y-direction, z-direction, respectively
l0
fractal cell size or minimal crack length in propagation direction
L0
plane projected crack length measured in propagation direction
L0S
saturation plane projected crack length measured in propagation direction
Griffiths critical crack length
L0C
DL
rugged crack length
DL0
distance between two points of the crack (the projected length of the crack)
DL0C
critical crack length
NL
number of units of the crack length in longitudinal or propagation direction
NT
number of units of the crack length in transversal direction
NV
number of units of the crack length in vertical or height direction
number of units of the crack length in the growth direction
Nx
Ny
number of units of the crack length in the perpendicular direction
sat
subscript index indicate saturation
Y0
shape function
x; y; z
spatial coordinates
Dx; Dy; Dz, incremental measure length in x-direction, y-direction, z-direction
Dzx ; Dzy fluctuation or roughness mean squares or incremental measure of crack height in x-direction, y-direction,
respectively
w
width of specimen test
Greek letters
exponent
b

d
incremental measure
@
partial derivative
eL ; eT
horizontal and vertical scale of the fractal scaling
ce ; cp
elastic and plastic specific energy surface, respectively
kx
amplification factor scale in x-direction

234

ky
n
fg ; fl
fx ; fy

rf
t
v

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

amplification factor scale in y-direction,

local roughness
global and local roughness exponent, respectively
roughness exponent in x-direction, y-direction, respectively
fracture stress
dynamic roughness exponent
Poissons ratio

background is associated with the difficulties of making an accurate measurement of the real fracture area. Although there
are several methods capable of quantifying the fracture area, the results are dependent on the size of the ruler of measurement [3], which did not contribute for its insertion in the CFM in the past.
With the increasing interest in fractal theory, it became possible to describe and quantify almost any apparently irregular
structure in nature [4]. In fact, many theories based on Euclidean geometry are being reviewed. It has been experimentally
proved that fracture surfaces have fractal scaling, so the fracture mechanics is one of the scientific areas included in this
context.
The work herein deals with the mathematical description of the roughness of cracks in the fracture mechanics, using the
fractal geometry to model its irregular profiles. Considering that the fracture surface is a record of the information left by the
fracture process, it may be possible to establish a relationship between the topology of this surface and the physical quantities of the fracture mechanics using fractal characterization techniques. Mandelbrot et al. [5] developed the Island Slit
method by searching for correlations between fractal dimensions and well-known physical quantities of fracture mechanics.
Following this pioneering work, other authors also made theoretical and geometrical considerations with a similar purpose
[614], as described briefly below.
1.1. Brief review of fractal scaling models of fracture surfaces
Mosolov [15] and Borodich [14] were the pioneers to associate the strain with the surface energies involved in the fracture process with the roughness exponents of the surfaces produced during the breaking process and the splitting of the
crack surfaces. They related these quantities using the singularity exponent of the stress field at the crack tip and the fractionary dependence of the fractal scaling exponent in fracture surfaces, postulating the equivalence between the changes in
the strain and the surface energies. Bouchaud and Bouchaud [16] proposed an alternative model from Mosolov [15] who
showed fracture parameters in terms of the height fluctuations of the rugged fracture surfaces, whose fluctuations are perpendicular to the crack propagation direction. Bouchad proved the relation between the fracture critical parameters as fracture toughness K IC and the relative fluctuations in the height of the rugged surface. The universality of the roughness
exponent for fracture surfaces had been assumed, as it did not depend on the type of material tested [17]. However, this
assumption generated a lot of controversy [18], which led to the discovery of anomalies [19] in the scaling exponents for
the local and the global scales in the fracture surfaces of brittle materials.
Family and Vicsek [20] and Barabsi and Stanley [21] showed a fractal scaling model for rugged surfaces formed in ballistic deposition films. This fractal scaling model was an inspiration for subsequent models of fractal surfaces present in other
phenomenologies. Since then, the modeling of fracture surfaces was done based on the models of Family and Vicsek [22],
receiving different supplements [19,23,24]. Based on this work, Lpez and Schmittbuhl [25] proposed an analogous model
for the fracture surfaces, which accounted for anomalies in the fractal scaling, with critical crossover dimensions for the transition in the roughness behavior of these surfaces in brittle materials. Morel et al. [26] observed experimental anomalies in
the roughness of wood fracture surfaces, and using the fractal model of Lpez et al. [19], they determined the roughness and
the dynamic exponent values. They also showed that the anomalies could only be explained by a dynamic scaling fractal
model involving local and global roughness exponents [25].
From the initial ideas of Lpez and Schmittbuhl [25] and following their fundamental hypothesis, Morel et al. [27] proposed a GR curve model for fracture surfaces in wood. Morel applied this model to describe the relief presented by the
rugged fracture surfaces whose topology is characterized by local and global roughness exponents, relating them with the
behavior of the elastic energy release rate and fracture resistance, called GR curve [28] for a brittle material. He generalized the Lpez and Schmittbuhl model including situations that have size effect [27].
This present work presents a generalization of the previous models, elaborating a fractal scaling in the longitudinal direction similar to that in the transversal direction for the crack propagation direction in a coherent way. The fractal surfaces
always exceeds the Euclidean dimensions of its projection and does not fulfill the immediately superior Euclidean dimensions in which it is immersed. For instance, a fracture profile has a fractal dimension in the interval 1 6 D 6 2, so it exceeds
the dimensions of a smooth line, but it does not fulfill those of a smooth surface. Morel et al. [24] shows that the fractal nature of a fluctuation Dz in a direction must be scaled using lengths and parameters of the same direction, although there can
be other factors related to the perpendicular direction that can be coupled to the model. In this sense, it is adopted the fol


lowing relationship for the roughness mean square Dzx lx ; x and Dzy ly ; y in the transversal and longitudinal directions,
respectively:

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Dxfg fl lxl

Dxfg
8 f f
g l f


 < Dy t lyl
Dz y l y ; y
: ftg
Dy
Dz x l x ; x

235

if lx  Dx
if lx  Dx
if ly  Dy

if ly  Dy

where x; y are the coordinates, lx ; ly are the used ruler lengths, Dx; Dy is the interval of measure of roughness, Dzx ; Dzy is the
height variation in the transversal and longitudinal directions on the fracture surface respectively, and fg ; fl ; t are the global,
local and dynamic roughness exponents, respectively.
It is also shown that from a general and analytical definition of a rugged surface, whose height z f x; y is a fluctuation
function of the coordinates of the mean plane of the projected surface, it is possible to establish a general expression that
involves the local and the global self-affine fractal scaling for the rugged fracture surface.
Many researchers have developed fractal models for the fracture surface, by employing the roughness dimension and
introducing it to the CFM [16,2932]. In particular, the work of Alves [32] introduced a self-affine fractal model of a fracture
surface in the mathematical formalism of the CFM through a local roughness derivative term. However, it lacked a more
detailed explanation about the origin of that model. It is shown in the present paper the basic mathematical premises of this
self-affine fractal model derived from Voss [33], who presented a fractal description for the noise in Brownian motions. To
calculate the rugged dimensions and define the model, the box-counting method is used [34]. A general expression for the
rugged crack length as a function of the projected length and fractal dimensions is shown. Also, a roughness expression is
derived, which can be directly introduced into classical fracture mechanics.
Despite the generality of the model presented in the work herein, applications were focused on brittle material and experimental tests were carried out for its validation. Mortar and heavy clay specimens were prepared under different conditions and
characterized by measuring their modulus of rupture and the rugged crack profile after a 3-point bending test. Obtained results
showed that the fractal dimension was capable of representing the roughness of the fracture surface of the different materials. A
strong correlation between the LipfshitHlder exponent for the longitudinal and the vertical directions, and the sintering temperature were verified on the heavy clay material. The results also showed that the increasing rugged crack length of the profile
of the fractured mortar specimens is well correlated with the increasing water/cement ratio. Overall, the application of the proposed self-affine fractal model was validated for estimating the fracture strength of brittle materials.
2. Analytical model for rough surfaces
Consider a rough surface z f x; y, as shown in Fig. 1.
The area DA of the surface can be parameterized and calculated from Greenberg [35],

DA

DE0 dE0

DE0

DL0 dL0

D L0

s
 2  2
@f
@f

1
dxdy:
@x
@y

If this rough surface is discretized with a mesh lx by ly elements with l0 6 lx 6 dL0 and e0 6 ly 6 dE0 , a box-counting method
can be used to define a relationship between the surface height fluctuations and the area within the limits dL0 and dE0 . Thus,
for a non-differentiable rough surface the rough area can be approximately calculated from the following:

DA
since

@f
@x

DE0 dE0

DE0

DDzxx and

DL0 dL0

D L0
@f
@y

s




Dzx x; y 2
Dzy x; y 2

1
dxdy
lx
ly

D zy
.
Dy

The area da of each projected element lx by ly is given by the following:

s
 2 

Dz x
Dzy 2
da  lx ly 1

lx ly n
lx
ly

It is observed that the term n da=lx ly represents the local roughness of the rugged surface and the whole area DA is given by,
Ny X
Nx
X
DA 
daij dL0 dE0
j1 i1

"

y X
Nx
1 X
n
Nx Ny j1 i1 ij

with N x dLlx0 and N y dEly0 .

Analogously, the length of a rough profile can be calculated with

DL 

"
#
Nx
Nx
X
1 X
dli dL0
ni
Nx i1
i1

which is a function of the projected length dL0 in the Euclidean space.

236

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 1. (a) Generic rough surface z = f(x, y); (b) close view of the rough surface limited by dL0 and dE0 .

3. The self-affine fractal model of a rugged fracture surface

Fracture surfaces are considered fractal objects with statistical invariance by scale transformation (self-similarity or selfaffinity). A fractal is an object whose measure of geometrical extension depends on the size of the measurement ruler and
has a so-called HausdorffBesicovitch dimension [21,3639].
Consider now a generic rugged fracture surface (Fig. 1) with self-affine properties [4,17]. As its morphology can be characterized by its fractal nature [5,8,14,40], a scale transformation along the orthogonal directions x and y with scale factors kx
and ky , respectively, results in the following mathematical relationship:
f

zkx x; ky y kfxx kyy zx; y

where fx is the roughness exponent of the surface in direction x and fy is the rough exponent of the surface in direction y. The
self-affinity property allows formulation of Eq. (7) in an uncoupled form and in each perpendicular direction, one has the
following:

zx kx x; y kfxx zx; y

zy x; ky y kyy zx; y
3.1. Computing the area of the rugged fracture surface

The height fluctuations Dzx; y of the fracture surface can be correlated with their fractal behavior in a window of observation in directions x and y in the following manner,

Dzx kx x; y kfxx Dzx; y

f

Dzy x; ky y kyy Dzx; y

To calculate the area effectively, it is useful to uncouple the mutual dependence of the transversal and the longitudinal
fractal properties of the height fluctuations. The following general scaling relationships for the height fluctuations are proposed, dividing Eq. (9) by the unit scale sizes lx and ly , respectively, one obtains
Dzx lx ;x
lx

 Dz

Dzy ly ;y
ly

kxx
lx

 Dz

kyy
ly

Dlxz

fx

Dlyz

Dx
lx

fy
Dy
ly

10

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

237

Fig. 2. The changing fractal dimension of a fracture profile of a brittle material.

3.2. Local and global roughness dimensions

Measurements of the fractal dimension Df of a fracture surface profile in a brittle material with the sand-box method [41]
show an asymptotic behavior with the growing crack length L0 . Fig. 2 shows this behavior, where the fractal dimension
decreases from an initial local value Df > 1 with L0  L0S and L0S being the saturation length, to a global value Df  1 for
L0  L0S . In this sense, the rough length L of the profile is not dependent on the measuring scale lx when L0  L0S , being closer
in size to the projected length L  L0 [25].
To cope with this behavior, Eqs. (10) can be adapted according to these intervals, and the roughness fluctuations in the
transversal and the longitudinal directions [24,26,27] are, respectively, given by

8
fx
>
< Dlxx
Dzx lx ; x K y
f
>
: Dx gx

if x  xsat

8
fy
>
< Dlyy



Dz y l y ; y K x
f
>
: Dy gy

if y  ysat

lx

ly

11

if x  xsat

if y  ysat

12

where K x ; K y are functions that relates the interaction of the roughness of one direction to another direction. Note the cross
terms in this equation where the roughness in one direction affects the roughness in the other direction. Although the model
separates contributions from every direction, the fracture occurs at the same time in both directions. Thus the fibers or particles in the microstructure to break up, suffer local strains that have normal and tangential components affecting perpendicular directions. fgx and fgy are the global rough fractal exponents in directions x and y, respectively.
3.3. Relationship between transversal and longitudinal roughness exponents
As the self-affine crack presents rough exponents with local and global aspects, the following expressions can be used to
account for the roughness transition from local to global scales in x and y directions [24,26,27].

fx fgx  flx

13

fy fgy  fly

14

Considering, for example, that the crack propagates in y direction, the rough exponent fy differs from fx by a dynamic factor
1=t [22],

238

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

fy fx =t b

15

where b
characterizes crack propagation in y direction along time.

Thus, Eqs. (11) and (12) become,

8

fgx flx
>
>
< Dl x
x
Dzx lx ; x K y
f
> Dx gx
>
:

if x  xsat

8

fgy fly
>
>
< Dl y



y
Dzy ly ; y K x
f
> Dy gy
>
:

if y  ysat

16

if x  xsat

lx

17

if y  ysat

ly

3.4. Defining the quantities K x and K y

The dependency of quantities K x and K y must be analyzed. In isotropic materials, the functions that define the self-affine
fractal scaling have the same nature in the transversal and the longitudinal directions. Therefore, the height fluctuations are
defined by isomorphic1 functions in both directions, and

@f
@f
$
@x
@y

18

which results in,

Ky
lx
Ky
lx

fgx flx
Dx
lx

fgx
Dx
lx

Kx
ly

$ Klyx

fgx tflx
Dy
ly

fgxt
Dy
ly

if DA0  DA0sat

19

if DA0  DA0sat

Assuming the scale relationship used by Morel et al. [24,2628],

   1t
Dx
Dy

lx
ly

20

Based on Eq. (19), it is possible to express K x and K y as functions of the same scaling factors and roughness exponents such
as,

Kx K
Ky K

fgx flx
Dx
lx

21

fgx f
lx
t
Dy
ly

where K is a scaling ratio defined in the z direction.

Substituting Eq. (21) in Eqs. (16) and (17), one has,

Dz x l x ; x K

Dy
ly

8
> Dx
fgx flx
lx >
fgx f
<
t
lx

if x  xsat

lx

if x  xsat

fgx
>
>
: Dx

22

and

8
fgx flx
t
>
 fgx flx >
< Dl y



Dx
y
Dz y l y ; y K

fgy
>
lx
>
: Dy t
ly

if y  ysat
if y  ysat

3.5. Computing the K factor

Consider a section of a rough fracture surface as illustrated in Fig. 3.
1

$ this symbol is used to denote isomorphic functions.

23

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

239

Fig. 3. Scheme of a window lx ; ly on a rough fracture surface showing the minimal crack lengths e0 ; l0 for the transversal and the longitudinal directions.

Height fluctuations also depend on the scaling of vertical direction, suggesting that K depends on scaling between directions z and x and directions z and y of the crack. In this way, to calculate K f h0 ; DH0 as a power law function of parameters
in directions x; y, and z, one must analyze the crack profile on the planes xz or yz, as illustrated in Fig. 4.
Considering square units with sizes l0 h0 , covering the crack length DL0 within the area DL0 DH0 , and e0 h0 , covering
the crack thickness DE0 within the area DE0 DH0 , the number of necessary units to cover the self-affine profile in each
direction is obtained from,
fly

longitudinal

24

flx

transv ersal

25

NL N V
eL

NT NV
eT

where N V is the number of minimal boxes on the z-perpendicular direction to the plane of the crack. flx ; fly is the local roughness exponent, related to the called divider dimension as fl D1D [42], DE0 is the transversal projection of the crack, DH0 is its
vertical projection,

eT and eL are the transversal and the longitudinal scale factors, and e0 6 lx 6 DE0 and l0 6 ly 6 DL0 .

Fig. 4. Self-affine fractal of WeierstrassMandelbrot, where eek 1=4, Dx = 1.5, and H 0:5, is used to represent a fracture profile. Source: Modified from
FAMILY and VICSEK [20, p. 7].

240

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Taking a cut plane of the fracture surface in the longitudinal direction the profile of Fig. 4 is obtained. In the profile shown
in Fig. 4, the basic measuring units l0 , and h0 are different in the longitudinal y, and the vertical z directions, respectively.
If h0 is the minimum scaling factor in vertical direction, one has,

NV



DH 0
v ertical
h0

Choosing

26

eL eT and NL NT for counting the number of cells in the profile at the xz and the yz planes, one has





DH 0
DH 0
f
f

eL ly

eT lx
h0
h0

27


 
fly 
 
flx
DH 0
l0
DH 0
e0

h0
DL 0
h0
DE 0

28

or

Thus,

l0
DL0

e0

DE 0

fflx

ly

29

For each xz and yz plane across the crack surface, there is a projected crack profile. The number of cells covered by the
fractal scaling in each of these planes in the horizontal and the vertical directions is given by,

Nxz
L
Nxz
V
and

Nyz
T
Nyz
V

fy 9
>
=DL fy DH 
0
0

>
h0
DH 0
; l0

D L0
l0

30

h0

fx 9
>
=DE fx DH 
0
0
;

> e0
h0
DH 0
;

D E0
e0

31

h0

respectively. From Eqs. (30) and (31) and

t fx =fy , one has,

9
DL0
=DL 1=t DE 
DhH00 >
l0
0
0

fx
> l
e0
0
DE0
DH 0 ;
h0
e0

fy

32

which is equivalent to Eq. (20). Therefore, from Eq. (31) for K  h0 , one has

K DH 0

e0
DE0

flx

33

Substituting Eq. (33) in Eqs. (22) and (23), one can now drop the x subscript of all equations. Therefore the height fluctuations are as follows:

8

fl
fg fl
fg tfl
>
>
@f
Dy
Dx
>
< @x
Dlzx x DlHx 0 DeE00
lx
ly

fl
fg
>
>
>
Dy
Dx
: @f Dzx DH0 e0
lx
lx
lx
D E0
@x
ly

fg fl
t

if x  xsat

34

if x  xsat

and

8

fl
fg fl
fg tfl
>
Dz
@f
Dy
>
Dx
< @y
lyy DlHy 0 DeE00
ly
lx

if y  ysat

fg
>
>
: @f Dzy DH0 e0 fl Dx fg fl Dy t
lx
ly
ly
DE0
@y
ly

if y  ysat

35

Substituting Eqs. (34) and (35) in Eq. (3) for lx e0 and ly l0 , the area DA is given by:

8
s


>

2 
2fl
2fg fl
2fgtfl
2
>
> R DE0 dE0 R DL0 dL0
DH 0
DH0
e0
Dy
Dx
>

dxdy if DA0  DA0sat

>
DL0
e0
l0
e0
DE0
l0
< D E0
DA
s


>

2 
2fl
2fg fl
2ftg
>
2
R
R
>
e0
Dy
>
Dx
> DDEE0 dE0 DDLL0 dL0 1 DlH0 DeH0
dxdy
if DA0  DA0sat
:
e0
DE0
l0
0
0
0
0

36

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

241

This equation is in agreement with the model presented by Morel et al. [24] and can be seen as a
generalization.
Alternatively, Eq. (36) for the rugged area can be rewritten like,

8
s


>

2 
2fl
2fg fl
2fgtfl
2
>
>
DH 0
DH 0
e0
D E0
D L0
>
D
A

if DA0  DA0sat
>
l0
e0
DE0
e0
l0
< 0
DA
s


>



2fg
>
>
> DA 1 DH0 2 DH0 2 e0 2fl DE0 2fg fl DL0 t
>
if DA0  DA0sat
: 0
l0
e0
DE0
e0
l0

37

3.6. Crack profile model

For the analysis of a crack profile DE0 e0 , Eq. (37) becomes:

8
s


>

2 
2fgtfl
2
>
>
DH 0
DH 0
DL0
>
D
L

if DL0  DL0s
>
< 0
l0
e0
l0
DL
s


>


2fg
>
>
> DL 1 DH0 2 DH0 2 DL0 t
>
if DL0  DL0s
: 0
l0
e0
l0

38

Observe that the lengths DL0 and DH0 are not necessarily equal and correspond to the projected lengths of the crack, in the
horizontal and the vertical directions, respectively.
To describe a growing crack, one can define a fixed origin coincident with the crack onset at the left (or right) side of the
counting box. The size of the counting box can be adjusted to cover the growing crack. This is the basis of the sand-box
method [41] for crack length measurement. Mathematically, this is defined with DL L; DL0 L0 and the rugged length
is given by,

8 s


>

2 
2fgtfl
2
>
>
DH0
L0
>
DeH00
if L0  L0s
>
< L0 1
l0
l0
L
s


>

2 
2ftg
>
2
>
L0
>
> L0 1 DlH0 DeH0
if L0  L0s
:
l0
0
0

39

Applying the logarithm on the both sides of Eq. (39) one obtains an expression that relates the fractal dimension to the projected crack length:

Df 

8
>
>
>
>
>
>
>
<1 1

ln L=l0

ln L0 =l0 >
>
>
>
>
>
>
:

(

ln

(

ln

1 12

DH0
l0

DH 0
e0

ln L0

DH0
l0

DH 0
e0

ln L0

2 
2fgtfl

L0
l0

2 
2ftg

if L0  L0s

40

L0
l0

if L0  L0s

It is important to point out that DL0 denotes the projected distance between two points of the crack (the projected crack
length) and l0 is the minimum possible length of a microcrack, which defines the scale l0 =L0 under which the crack profile is
scrutinized.
Fig. 5 shows the influence of the height DH0 of the box on the rugged crack length measure, as a function of the projected
length L0 , for L0  L0S and characteristic values for the parameters fl 0:8; fg 1:2; t 4:0, and l0 0:1. Considering a
fixed roughness exponent fl 0:8, a linear relationship between L and L0 is seen in lower boxes DH0 ! 0, while in higher
boxes DH0 ! 1 a nonlinear one is observed.
Fig. 6 shows the influence of the roughness dimension of the rugged crack length, as a function of the projected length L0 .
Observe that for fl ! 1, which corresponds to a flatter profile, the relationship between the rugged and the projected lengths
is more linear. Whereas for fl ! 0, corresponding to a more rugged profile, the relationship between the rugged and the projected lengths is increasingly nonlinear.
In near-flat cracks, the box height can be chosen as DH0 l0 e0 , and Eq. (39) simplifies to,

8 s
>

2fgtfl
>
>
L
>
>
if L0  L0s
< L0 1 2 l00
L
s
>
>

2ftg
>
>
>
if L0  L0s
: L0 1 2 Ll00

41

242

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 5. Rugged length, L as a function of the projected length, L0 and the varying the box height, DH0 , width l0 0:1, representing the box height (rectangular
box DL0 DH0 with local roughness dimension, fl 0:8, global roughness dimension fg 1:2, and dynamic exponent, t 4:0.

Fig. 6. Rugged length L as a function of the projected length L0 and the varying the local roughness dimension, fl representing the box height for flat cracks
with rectangular boxes DH0 1:0; l0 0:1, global roughness dimension, fg 1:2, and dynamic exponent, t 4:0.

In analogous way to the case of Eq. (39), i.e., applying the logaritm on the both sides of Eq. (41) one obtains an expression
that relates the fractal dimension to the projected crack length,

Df 

8
>
>
>
>
>
>
>
>
<1 1

2
ln L=l0

ln L0 =l0 >
>
>
>
>
>
>
>
:1 1
2

"
ln

2fgtfl

12

ln L0

"
ln

2fg

12

L0
l0

if L0  L0s

42

L0
l0

ln L0

if L0  L0s

In the study of a self-affine fractal, two limits can be verified. The first is called local limit, when the height of the boxes
is increased compared to its length DH0 ! L0 . The second is the global limit, when the height of the boxes is small compared to its length DH0 ! h0 . These limits are presented as follows.

243

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

3.7. Case 1: The initial apparent self-similar limit of the local fractality
The local self-similar limit of the crack fractal measure can be obtained from Eq. (39), considering DH0 L0  l0 ,

L L0

 1fg tfl
l0
L0

43

It is a self-similar relationship between the projected crack length Lo and the height ho of the unit box. This expression shows
that all self-affine crack fractals can represent a local self-similarity when the square areas L0 L0 are considered instead of
the rectangular ones L0 DH0 . This is commonly observed in the onset of cracks [12,13,43], where one can extract the following result,
fg fl

L
fg fl

L0

l0

1

cte

44

which is a constant relationship between the micro- and macroscales.

This equation is analogous to the self-similar mathematical relationship described by different authors [79,14,44].
3.8. Case 2: The global self-affine limit of the fractality
On the other side, the global self-affine limit of the crack fractal measure can be obtained from Eq. (39), considering
DH0 l0  L0 ,

L L0

45

Ductile materials presenting a higher fractality possess a crack profile that can be better adjusted by Eq. (43), while brittle
materials presenting a lower fractality are better adjusted by Eq. (45), which is closer to the classical fracture mechanics
model with a flat fracture profile.
3.9. The roughness of a fracture profile
The classical definition of roughness in the fracture mechanics is given by n L=L0 [2,3]. Alternatively, a local roughness is
defined in this work considering a limiting approach for the counting box dimensions,

n  lim

DL0 !l0

LL0 DL0  LL0

DL0

46

In practice, Eq. (46) can be seen as the derivative of the rugged length L f L0 with respect to its projected length Lo , written
as

dL
dL0

47

This definition provides a full characterization of fractal profiles and can be useful to describe the crack growth phenomena.
Eq. (47) can be used to calculate the local roughness of the profile, if an expression that relates the rugged length L to the
projected length L0 is known. For instance, for the monofractal case with a square counting box, the measure of the rugged
crack length is given by Eq. (39). Therefore, the expression for the local roughness is given by,

v

2
2fgtfl
u



DH 0
L0

2  2fgtfl
u
fg  fl
l0
l0
DH 0
L0
t
s
n 1
P1
l0
l0
t

2
2fgtfl
DH 0
L0
1 l0
l0

48

or simply

1
n


fg fl



2fg fl
2
t
DH 0
L0

1
t
l0
l0
s
P1

2
2fgtfl
DH0
L0
1 l0
l0

49

4. Materials and methodology

The mortar and heavy clay brittle specimens were prepared under different conditions and molded and tested in laboratory to obtain the rugged fracture surfaces. The specimens were characterized by measuring their modulus of rupture under
3-point bending tests.

244

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

4.1. Mortar specimens

To prepare the mortar specimens, Portland cement ASTM type V (high initial resistance) and natural sand with controlled
grain sizes [45] were used. The Portland cement ASTM was used to compare the results obtained in the work herein with the
values obtained by other authors in the technical literature [31]. Four weight proportions of cement, sand, and water c=s=w
were applied, as shown in Table 1. It can be seen that from types A to D that water/cement and sand/cement ratios increase.
Twelve specimens in total with standard sizes 40 40 160 mm were prepared according to ASTM C 348 [46] and mixed
according to ASTM C 305 [47].
For the specimens of Table 1, the relative content of cement was decreasing from A to D, while the proportion of water
was increased. That was made to compensate the increase in surface area of sand excess in order to get an effect in porosity
to obtain a more rugged surface in the fracture testing.
After seven days of curing, 3-point bending tests were carried out using an EMIC DL10000 to evaluate their modulus of
rupture. Those results are shown in Table 2.
4.2. Heavy clay specimens
A homogenized dry clay powder with 5% added water was used to prepare ninety heavy clay specimens in total of dimensions 60:2 20:3 mm and thickness of 4 mm, approximately. This thickness was a result of a conforming process of nearly
8.5 g of clay under compressive forces applied in two stages an initial load of 20 kN, followed by a period of rest and a final
load of 30 kN. After that, the specimens were dried along one day and sinterized for two hours in an electrical furnace under
temperatures of 800, 900 and 1000 C (see Table 3).
After sintering, the specimens were weighed, and their dimensions were measured, and taken for the 3-point bending
tests. Their average modulus of rupture is shown in Table 4.

Table 1
Weight proportions of cement, sand and water used for molding the mortar
specimens.
Mortar type

Weight proportion c/s/w

Number of specimens

A1, A2, A3
B1, B2, B3
C1, C2, C3
D1, D2, D3

1:1:0.30
1:2:0.45
1:3:0.60
1:4:0.75

3
3
3
3

Table 2
Average modulus of rupture of mortar specimens obtained after seven days
curing.
Modulus of rupture (MPa)
Mortar type

Number of specimens

Average stress

Error

A1, A2, A3
B1, B2, B3
C1, C2, C3
D1, D2, D3

3
3
3
3

9.6
7.8
6.3
4.6

0.7
0.9
0.2
0.5

Table 3
Heavy clay specimens prepared with different sintering temperatures.
Specimen type

Sintering temperature (C)

Number of specimens

Ar1; Ar2; . . . Ar30

Br1; Br2; . . . Br30
Cr1; Cr2; . . . Cr30

800
900
1000

30
30
30

Table 4
Average modulus of rupture of heavy clay specimen types.
Modulus of rupture (MPa)
Specimen type

Number of specimens

Average stress

Error

Ar1; Ar2; . . . Ar30

Br1; Br2; . . . Br30
Cr1; Cr2; . . . Cr30

30
30
30

2.0
4.2
9.0

0.2
1.3
1.0

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

245

4.3. Crack profiles

After the rupture tests, the fractured surfaces of the specimens were selected for digital imaging with an optical stereomicroscope (Leica MZ6) connected to a digital CCD camera (KODO KC-512NT). For each fractured side of each specimen, a complete fracture surface image was formed by juxtaposing the images. Through software analysis (MOCHA Image Analysis
v1.2.10) the crack boundary profiles were obtained. Figs. 710 show the crack profiles from specimens mortar D2 side 1,
mortar D2 side 2, heavy clay Ar8, and heavy clay Br3, respectively.
The fracture surfaces were digitalized and characterized with the box-counting and sand-box methods to obtain the local
and the global roughness exponents. It can be seen that all profiles can be well covered by a rectangular box for the application of the box-counting method. The bottom side of the box (L0 represents the thickness of each specimen. The rugged
profile length and other model parameters were calculated after fitting the analytical expression.

5. Experimental results
The proposed fractal model was used to describe the roughness of ruptured mortar and heavy clay specimens.
The following specimens had their two sides modeled: mortars A2; B2; C1; D2 and heavy clay Ar8; Br3; Cr25. For each profile, the roughness dimensions were obtained applying the box-counting and sand-box methods and fitting the proposed
monofractal model given by Eq. (39). These results are listed in Table 5, represented by DBC ; LBC ; L0 ; L0s and Table 6 repre-

Fig. 7. Fracture profile of mortar specimen D2 side 1: digital images side by side and crack profile from processed image with a comparative total ruler
length equal to 1.0 mm.

Fig. 8. Fracture profile of mortar specimen D2 side 2: digital images side by side and crack profile from processed image with a comparative total ruler
length equal to 1.0 mm.

246

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 9. Fracture profile of heavy clay specimen Ar8: digital images side by side and crack profile from the processed image with a comparative total ruler
length equal to 1.0 mm.

sented by DH0 ; l0 ; fl ; fg ; b
and t. The fitting parameters L0 40 mm for mortar and L0 4 mm for heavy clay ceramic, l0 and fl
in Eq. (39) are also shown in Table 6.
Despite the different roles in the morphological description of the fracture surface in x and y directions, experimental
measurements showed that the roughness in these two directions has similar scaling properties for very different materials
[48]. A unique roughness exponent is then generally considered and it has been claimed that it has a universal value of
fl 0:8 [17]. This behavior has been confirmed for a large variety of experimental situations [23,49]. However, other studies
have also shown that a different value for fl 0:5 may be applicable [5052]. These two values of the roughness exponent
are connected to the length scale at which the crack is examined. In particular, at small length scales, one observes a roughness exponent fl 0:5, whereas at large length scales, the larger value fl 0:8 is found. Results presented in Table 6 show
rough exponent values between 0.59 and 0.99.
The value of l0 for mortar D2 specimen is very small because its roughness global dimension is very next of unity. This
means that the crack profile is almost a straight line. In this case a straight line can be scrutinized by any l0 value that is
infinitely small.
Fig. 11 shows the asymptotic behavior of the fractal dimension with the growing crack length L0 , measured with sand-box
method. The local L0  L0S and global L0  L0S ranges are depicted in the figure.
In Figs. 12 and 13, a good agreement is observed in the curve fitting of Eq. (40) to the fractal analyses of the mortar specimen A2 side1 and the heavy clay specimen A8, respectively.
In Figs. 14 and 15, a good agreement is observed in the curve fitting of Eq. (39) to the fractal analyses of the mortar specimen A2 side1 and the heavy clay specimen Ar8, respectively.
In Figs. 16 and 17, a comparison of the rugged length L projected length L0 of each crack profile is made for mortars and
heavy clay, respectively.
The general behavior is consistent within each group of results. This trend is observed in the LBC =L0 ratios in Table 5.
In Fig. 18, the local roughness dL=dL0 was plotted versus the square value of the modulus of rupture r2f , for the mortar
specimens. A linear trend can be observed where the values plotted are correlated for a R-square correlation coefficient of
0.7012.

247

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 10. Fracture profile of heavy clay specimen Br3: digital images side by side and crack profile from the processed image with a comparative total ruler
length equal to 1.0 mm.

Table 5
Results from fractal analysis using box-counting method of brittle specimens.
Specimen

DBC

LBC (mm)

L0 (mm)

LBC =L0

L0s (mm)

Mortar A2 side 1
Mortar A2 side 2
Mortar B2 side 1
Mortar B2 side 2
Mortar C1 side 1
Mortar C1 side 2
Mortar D2 side 1
Mortar D2 side 2
Heavy clay Ar8
Heavy clay Br3
Heavy clay Cr25

1.0500 0.0001
1.0534 0.0001
1.0431 0.0005
1.0380 0.0001
1.0607 0.0003
1.0504 0.0001
1.0240 0.0002
1.0514 0.0001
1.1511 0.0004
1.2292 0.0001
1.1627 0.0001

59.2064 0.0001
61.0022 0.0002
65.8381 0.0003
58.5383 0.0002
67.1098 0.0008
72.9876 0.0004
69.3074 0.0003
73.6647 0.0005
11.4017 0.0003
12.7726 0.0002
13.7940 0.0008

37.2288 0.0001
39.1765 0.0006
38.2453 0.0001
37.8984 0.0001
35.3309 0.0008
40.0000 0.0003
40.0000 0.0003
38.3006 0.0005
3.5490 0.0004
3.950 0.001
3.7148 0.0004

1.59034
1.55711
1.72147
1.54461
1.89947
1.82469
1.73269
1.92333
3.21282
3.23357
3.71326

3.616 0.001
3.0123 0.0003
8.5710 0.0003
8.7261 0.0003
9.9510 0.0006
13.700 0.001
12.1460 0.0004
8.2860 0.0002
0.5400 0.0004
0.0728 0.0001
0.7232 0.0003

In Fig. 19 it is shown the increasing water/cement ratio in the mortar specimens in function of the decreasing modulus of
rupture. Comparing this result with Fig. 18 it is observed that the water/cement ratio affects directly the roughness of the
fracture profile. The nonlinear fitting of this curve shown a result given by the following equation:

r2f r2f 0 eax

where

50

r2f 237:59 GPa2 , a 3:1205 and x = water/cement ratio for a R-square correlation coefficient value of 0.9515.

248

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Table 6
Results from fitting the model for brittle specimens.
Specimen

DH0 (mm)

l0 (mm)

fl 1=DD

fg

b

Mortar A2 side 1
Mortar A2 side 2
Mortar B2 side 1
Mortar B2 side 2
Mortar C1 side 1
Mortar C1 side 2
Mortar D2 side 1
Mortar D2 side 2
Heavy clay Ar8
Heavy clay Br3
Heavy clay Cr25

0.3740 0.0003
0.3635 0.0001
0.7071 0.0001
0.3040 0.0003
0.84250 0.0003
0.83470 0.0004
1.4160 0.0001
1.4618 0.0003
4.1542 0.0003
1.6998 0.0001
3.2733 0.0001

1.3249 0.0004
2.2526 0.0003
0.0800 0.0003
0.7610 0.0001
0.8261 0.0001
0.0623 0.0004
(1.6040 0.0001)E6
(3.75 0.0001)E10
0.1376 0.0003
0.3641 0.0001
0.1555 0.0001

0.800 0.001
0.800 0.001
0.5954 0.0001
0.6148 0.0001
0.8613 0.0001
0.8705 0.0001
0.9927 0.0001
0.9854 0.0001
0.7062 0.0001
0.7322 0.0001
0.6579 0.0001

1.0652
1.0652
1.109095
1.051606
1.0000
1.0000
1.0016
1.0031
1.179023
1.171628
1.197972

0.86097
0.92510
0.402303
0.96519
2.233964
2.294475
1.321737
1.257799
0.287048
0.091712
0.227454

0.0991 0.0001
0.0997 0.0001
0.1945 0.0001
0.3555 0.0001
0.0621 0.0001
0.0564 0.0001
0.0067 0.0001
0.0141 0.0001
0.1531 0.0001
0.1554 0.0001
0.1829 0.0001

Fig. 11. Determination of crack length saturation by fractal analysis of mortar specimen A2 side 1 fractal dimension projected length, L0 .

Fig. 12. Fractal analysis of mortar specimen A2 side 1 fractal dimension, D versus projected length, L0 . Source: 2012 Alves LM. Published in [short
citation] under CC BY 3.0 license. Available from: http://dx.doi.org/10.5772/51813 [52, p. 61].

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

249

Fig. 13. Fractal analysis of heavy clay specimen Ar8 side 1 fractal dimension, D versus projected length, L0 . Source: 2012 Alves LM. Published in [short
citation] under CC BY 3.0 license. Available from: http://dx.doi.org/10.5772/51813 [52, p. 61].

Fig. 14. Fractal analysis of mortar specimen A2 side 1 rugged length, L versus projected length, L0 . Source: 2012 Alves LM. Published in [short citation]
under CC BY 3.0 license. Available from: http://dx.doi.org/10.5772/51813 [52, p. 61].

6. Discussion
6.1. The relation with the LopezMorel fractal model of fracture surface
Morel et al. [24,2628] describe the variations of roughness with the distance x measured from the notch in the transversal direction to the crack propagation. In fact, observing the microstructural aspects of rugged fracture surfaces of woods
used in their work [53], it appears as a wave beach or as a brazilian roof, where in the direction of crack propagation
there are almost no fluctuations in the height of the rugged surface. In spite of this fact, to adapt his scaling to the experimental reality, prefactors A and B were presented whose nature depends on the material and are obtained by fitting of experimental curves. This means that if the material to be modeled presents fluctuations of height in the growth direction, these
prefactors must be explained mathematically in terms of the relations of fractal scaling for the distances of rugged length in

250

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 15. Fractal analysis of heavy clay specimen Ar8 rugged length, L versus projected length, L0 . Source: 2012 Alves LM. Published in [short citation]
under CC BY 3.0 license. Available from: http://dx.doi.org/10.5772/51813 [52, p. 61].

Fig. 16. Comparative plot of the rugged crack length, L projected crack length, L0 for the sides A and B among main mortars specimen profiles.

the direction parallel to the crack growth. The model represents the dependencies in the transversal x and the longitudinal y
directions in a unique mathematical term.
However, for materials such as glass, alumina and mortar, that present fluctuations in the transversal and the longitudinal
directions, the model cannot portray authentically the fluctuations in the height observed in the fracture surface of these
materials in both directions.
The work herein, it has been shown that the dependence of the fluctuations in the height Dzxt ; yt can be decoupled
and written in a simple way as,

Dzxt; yt f xt
g yt

51

without changes in the final result and obtaining the Morels model [24,2628] as a particular case.
From Eq. (36), it is possible to reduce dimensionally the model of a fracture surface proposed by Morel et al. [24,2628] to
a fracture profile model, neglecting the fluctuations of the roughness in the transversal crack propagation direction, as in the
case of wood fracture surfaces considering a fracture propagation in the transversal direction to the fibers. In that case the
fracture surface is given by:

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

251

Fig. 17. Comparative plot of the rugged crack length, L projected crack length, L0 among main heavy clay specimen profiles.

Fig. 18. Plot of square of modulus of rupture versus the derivative of the rugged crack length, dL=dL0 .

8
s
>

2
2fl
2fg fl
2fgtfl
>
R
R
>
D
E
dE
D
L
dL
e0
Dy
0
0
0
0
>
Dx
>
1 DeH00
dxdy if DA0  DA0sat
< DE0
DL0
lx
DE0
ly
DA
s
>
>

2
2fl
2fg fl
2ftg
>
> R DE0 dE0 R DL0 dL0
e0
Dy
Dx
>
1 DeH00
dxdy
if DA0  DA0sat
: DE0
DL0
lx
DE0
ly

52

where DA0 dL0 dE0 and DA0sat L0s E0 . For minimal subdivisions of the profile in the transversal and the longitudinal directions, i.e., lx e0 and ly l0 , Eqs. (34) and (35) are rewritten as,

Dz

f f
fl
fg fl Dy
g t l
>
>
Dx
< DeH0 DeE0
e
l
0

fg
>
>
: DH0 e0 fl Dx fg fl Dy t
e0

DE0

e0

l0

53

252

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

Fig. 19. Plot of modulus of rupture versus the water/cement ratio for a mortar cured by seven days.

Substituting Eq. (53) in (52) one has:

8
s
>

2
2fl
2fg fl
2fgtfl
>
R
>
D
L
dL
e0
DE0
Dy
0
0
>
>
1 DeH00
dy if DA0  DA0sat
< dE0 DL0
D E0
e0
l0
DA
s
>
>

2
2fl
2fg fl
2ftg
>
R DL dL
>
e0
DE0
Dy
>
dy
if DA0  DA0sat
: dE0 DL00 0 1 DeH00
D E0
e0
l0

54

Eq. (52) is analogous to the approximations presented by Morel et al. [24] considering the crack propagation direction as
being the y direction and the constants A; B, and ny in their equation given by,

A DeH00
B

e0
D E0

fl
fg fl
1
e0

E0

55

1=t

ly

ny E0

1=t
Dy
ly

The saturation length of the fracture surface in the transversal crack propagation is given by,


nysat E0

Dy
L0S

1=t

E0

56

6.2. Analysis of the values obtained by the fitting

The proposed fractal model presented a good fitting agreement with the experimental results, allowing to obtain the
Hurst roughness dimension of the profiles. The plots of the rugged crack length of the fracture profiles versus the projected
length shown in Fig. 16 for mortar and in Fig. 17 for heavy clay were fitted, using the multiparameter nonlinear minimal
squares fitting method for curves, and LevenbergMarquardt Iteration Algorithm, with the model represented by Eq. (39).
Some plots of the rugged crack length as a function of the projected crack length show a clear transition between the local
and the global regimes for the crack growth. This is particularly observed in Fig. 15, after the crack length L0 0:75 mm,
approximately. Beyond this length, the relationship between the rugged and the projected crack lengths becomes
quasilinear.
The saturation length L0S was determined in two different ways. The first way was made directly by fitting from the plots
of the rugged length as a function of the projected length. The second way was made by a tangent line from the plot of fractal
dimension as a function of the projected crack length by sand-box method. In this second method, a straight tangent line is
drawn from the value 2.0 in the vertical axis (which represents the maximal fractal limit dimension of a rugged line), until
the horizontal axis, passing by a unique tangent point of the curve fractal dimension versus projected crack length, Df L0 .
From Table 5 it is possible to observe that the saturation length, L0S for mortar is larger than heavy clay. This result means
that the rugged crack length of the mortar is smother than the heavy clay which can be easily verified comparing Figs. 7 and
8 with Figs. 9 and 10.

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

253

The local and the global roughness dimensions determined in this work, for the mortar and the heavy clay specimens, are
compatible with results in the literature [24,2628,30,31,52,53]. However, for some specimens, the values seem to exceed
the expected number, as in the case of mortar C1 (13.699 for the sides 1 and 2) and D2 (12.1456 for the sides 1 and 2).
The fractal box-dimension, DB shown in Table 5 compared to the rugged global dimension, fg in Table 6 presents compatibles results one with other. The mortar fractal dimensions are in the interval between 1.0 and 1.1 and the heavy clay has its
fractal dimension in the interval between 1.15 and 1.23. This results is in agreement with the more rugged aspect of the
crack in the heavy clay as shown in Figs. 9 and 10. Therefore, it is possible to distinguish mathematically a crack in this
two materials using its geometrical characteristics which can be portraited by different values of the exponents in the relations (39) and (41).
A linear relation is observed between the square of modulus of rupture r2f and the values of the local roughness of the
crack given by n dL=dL0 , for the mortar specimens, as shown in Fig. 18. A linear fitting is shown in that figure. Observe that
the data more below of the linear fitting presented a constant displacement compared to data closer to the straight line. Thus
it was due to systematic deviation of the points in the roughness values.
Some authors [20,22,42,50,53] suggest a relation between the local, global, and Hurst exponents, for the tridimensional
case, such as:

1H 

3b
;
2

57

where b f=t and in this case b  1 2H.


Considering that this equation is also valid for b
fg  fl =t, then


3  fg  fl =t
1H 
2

58

and one finds an approximate relation for the dynamic exponent t given by:

t


fg  fl
1 2H

59

If the local roughness dimension fl is equal to Hurst exponent H 2  DB 1=DD where DB is the box-dimension and DD is
the divider dimension. Then one has the following relation between the obtained exponents,


fg  fl
1 2fl

60


fg  fl  1  H

61

t
where

and

t

1H
1 2H

62

Values presented in Table 6 were derived from this expression. Values of t are lower than those in the literature. This was
expected as the tests were conducted to a fast rupture instead of crack growth.
6.3. Mechanical results
The increasing water/cement ratio in the mortar specimens resulted in a decreasing modulus of rupture as shown in
Fig. 19. Comparing this result with Fig. 18 it is observed that these results are also well correlated with the roughness of
the crack profile.
In Tables 3 and 4 for heavy clay, the increasing sintering temperature resulted in higher modulus of rupture. However, the
roughness cannot be associated due to the small number of temperatures values of tested specimens.
From the digitalized images of mortar and heavy clay, it can be observed that the fracture surface of the heavy clay presented bigger tortuosity compared to the mortar material. This is also evident in the fitting results where the rugged length is
near 3.5 times the projected length for heavy clay, and only 1.7 for mortar.
It can also be seen in Figs. 14 and 15 that for each type of material the slopes of the asymptotic curves are in very close
agreement with the model proposed.
This work presents several future perspectives in terms of correlating the rugged profiles with the material granulometry,
chemical composition, and cement curing time. The mathematical model can still be explored for understanding the minimal
crack length and the fractal dimension according to the mechanical testing parameters and material properties.

254

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

7. Conclusions
A general self-affine fractal model was presented and applied to brittle fracture surfaces. From the proposed equation, an
analytical expression for the rugged profile length was derived. Also, an analytical expression for the local roughness was
derived, which can be directly introduced into the classical fracture mechanics.
Comparing the fractal box-dimension, DB shown in Table 5 and the rugged global dimension, fg shown in Table 6 for the
mortar and heavy clay it was possible to distinguish mathematically a more rugged crack in this two materials using its geometrical characteristics which can be portraited by different values of those exponents in the relations (39) and (41).
The small ratio between roughness dimension for transversal and longitudinal directions, t, presented in Table 6 and
given by Eqs. (15), (60) and (62) shown that these two directions are weakly coupled for the analyzed materials, mortar
and heavy clay.
The rugged crack length is a response to the interaction of the crack tip with the microstructure of the material. By the
model presented in work herein it is possible mathematically to portray the peculiar rough behavior of a crack in Portland
cement mortar and red ceramics using fractal geometry.
The experimental technique of obtain crack profiles, proved very able to present satisfactory results that are very close to
reality.
A good agreement between the fractal model and the experimental results was observed. A strong correlation between
the fractal dimension and the sintering temperature was verified. The results also showed that the increasing rugged crack
length of the profile of the fractured mortar specimens is well correlated with the rising water/cement ratio.
Acknowledgments
The authors acknowledge contributions of Interdisciplinary Laboratory of Ceramic Materials LIMAC-CIPE-UEPG, Prof.
Dr. Vicente Campitelli (Civil Engineering Laboratory) and PIBIC/CNPq/UEPG.
References
[1] Underwood EE, Banerji K. Fractal analysis of fracture surfaces. ASM Handbook 1987;12:2115.
[2] Anderson TL. Fracture mechanics: fundamentals and applications. CRC Press; 1995.
[3] Dos Santos SF. Aplicao do conceito de fractais para anlise do processo de fratura de materiais cermicos. Master dissertation. Universidade Federal
de So Carlos. Centro de Cincias Exatas e de Tecnologia, Programa de Ps-Graduao em Cincia e Engenharia de Materiais, So Carlos; 1999.
[4] Mandelbrot BB. The fractal geometry of nature. New York: W.H. Freeman and Company; 1983. http://dx.doi.org/10.1119/1.13295.
[5] Mandelbrot BB, Passoja DE, Paullay AJ. Fractal character of fracture surfaces of metals. Nature 1984;308:7212.
[6] Mu ZQ, Lung CW. Studies on the fractal dimension and fracture toughness of steel. J Phys D: Appl Phys 1988;21(5):848.
[7] Mecholsky JJ, Passoja DE, Feinberg-Ringel KS. Quantitative of brittle analysis fracture surfaces using fractal geometry. J Am Ceram Soc 1989;72(1):605.
[8] Heping X. The fractal effect of irregularity of crack branching on the fracture toughness of brittle materials. Int J Fract 1989;41(4):26774.
[9] Chelidze T, Gueguen Y. Evidence of fractal fracture. Int J Rock Mech Min Sci Geomech Abstr 1990;27(3):2235.
[10] Lin GM, Lai JKL. Fractal characterization of fracture surfaces in a resin-based composite. J Mater Sci Lett 1993;12(7):4702.
[11] Nagahama H. A fractal criterion for ductile and brittle fracture. J Appl Phys 1994;75(6):32202.
[12] Lei W, Chen B. Fractal characterization of some fracture phenomena. Eng Fract Mech 1995;50(2):14955.
[13] Tanaka M. Fracture toughness and crack morphology in indentation fracture of brittle materials. J Mater Sci 1996;31(3):74955.
[14] Borodich FM. Some fractal models of fracture. J Mech Phys Solids 1997;45:23959. http://dx.doi.org/10.1016/S0022-5096(96)00080-4.
[15] Mosolov AB. Mechanics of fractal cracks in brittle solids. Europhys Lett 1993;24:6738. http://dx.doi.org/10.1209/0295-5075/24/8/009.
[16] Bouchaud E, Bouchaud JP. Fracture surfaces: apparent roughness, relevant length scales, and fracture toughness. Phys Rev B 1994;50(23):17752.
[17] Bouchaud E, Lapasset G, Planes J. Fractal dimension of fractured surfaces: a universal value? EPL (Europhys Lett) 1990;13(1):73.
[18] Bouchaud E. Scaling properties of cracks. J Phys: Condens Matter 1997;9(21):431944.
[19] Lpez JM, Rodrguez MA, Cuerno R. Superroughening versus intrinsic anomalous scaling of surfaces. Phys Rev E 1997;56(4):3993.
[20] Family F, Vicsek T. Dynamics of fractal surfaces. World Scientific; 1991.
[21] Barabsi AL, Stanley HE. Fractal concepts in surface growth, vol. 83. Cambridge University Press; 1995.
[22] Vicsek T, Meakin P, Family F. Scaling in steady-state clustercluster aggregation. Phys Rev A 1985;32(2):1122.
[23] Schmittbuhl J, Roux S, Berthaud Y. Development of roughness in crack propagation. EPL (Europhys Lett) 1994;28(8):585.
[24] Morel S, Schmittbuhl J, Bouchaud E, Valentin G. Scaling of crack surfaces and implications for fracture mechanics. Phys Rev Lett 2000;85(8):1678.
[25] Lpez JM, Schmittbuhl J. Anomalous scaling of fracture surfaces. Phys Rev E 1998;57(6):6405.
[26] Morel S, Schmittbuhl J, Lpez JM, Valentin G. Anomalous roughening of wood fractured surfaces. Phys Rev E 1998;58(6):6999.
[27] Morel S, Bouchaud E, Schmittbuhl J, Valentin G. R-curve behavior and roughness development of fracture surfaces. Int J Fract 2002;114(4):30725.
[28] Morel S, Bouchaud E, Valentin G. Size effect in fracture: roughening of crack surfaces and asymptotic analysis. Phys Rev B 2002;65(10):104101.
[29] Alves LM, Da Silva R, Mokross B. Influence of crack fractal geometry on elasticplastic fracture mechanics. Phys A Stat Mech Appl 2001;295:1448.
http://dx.doi.org/10.1016/S0378-4371(01)00067-X.
[30] Mourot G, Morel S, Valentin G. Comportement courbe-r dun matriau quasi-fragile (le bois): Influence de la forme des specimens dessai.
Colloque_MATERIAUX 2002:14.
[31] Mourot G, Morel S, Bouchaud E, Valentin G. Scaling properties of mortar fracture surfaces. Int J Fract 2006:3954. http://dx.doi.org/10.1007/s10704005-3471-4.
[32] Alves LM. Fractal geometry concerned with stable and dynamic fracture mechanics. Theor Appl Fract Mech 2005;44:4457. http://dx.doi.org/10.1016/
j.tafmec.2005.05.004.
[33] Voss RF. Dynamics of fractal surfaces. Singapore: World Scientific; 1991.
[34] Vicsek T. Fractal growth phenomena. Singapore: World Scientific; 1992.
[35] Greenberg MD. Curves, surfaces, and volumes. Advanced engineering mathematics. Prentice-Hall; 1998.
[36] Barnsley MF. Fractals everywhere. New York: Academic Press; 1988.
[37] Besicovitch AS. On the sum of digits of real numbers represented in the dyadic system (on sets of fractional dimensions II). In: Edgar GA, editor. Classics
on fractals. Boston: Addison-Wesley Reading; 1993. p. 1559.

L.M. Alves et al. / Engineering Fracture Mechanics 162 (2016) 232255

255

[38] Besicovitch AS. Sets of fractional dimensions (IV): on rational approximation to real numbers. In: Edgar GA, editor. Classics on
fractals. Boston: Addison-Wesley Reading; 1993. p. 1618.
[39] Besicovitch AS, Ursell HD. Sets of fractional dimensions (V): on dimensional numbers of some continuous curves. In: Edgar GA, editor. Classics on
fractals. Boston: Addison-Wesley Reading; 1993. p. 1719.
[40] Dauskardt R, Haubensak F, Ritchie R. On the interpretation of the fractal character of fracture surfaces. Acta Metall Mater 1990;38(2):14359.
[41] Bunde A, Havlin S. Fractals in science. Springer-Verlag New York, Inc; 1995.
[42] Katsuragi H. Multiscaling analysis on rough surfaces and critical fragmentation [Ph.D. dissertation]. Fukuoka, Japan: Kyushu University; 2004.
[43] Lung CW. Fractals and the fracture of cracked metals. Fractals in physics. Elsevier Science Publishers; 1985. p. 18992.
[44] Mishnaevsky Jr LL. A new approach to the determination of the crack velocity versus crack length relation. Fatigue Fract Eng Mater Struct 1994;17
(10):120512.
[45] NBR. Areia normal para ensaio de cimento. ABNT, Normas Tcnicas. 7214; 1982.
[46] ASTM C348-97. Standard Test Method for Flexural Strength of Hydraulic-Cement Mortars. West Conshohocken, PA: ASTM International; 1997. www.
astm.org.
[47] ASTM C 305-06. Standard Practice for Mechanical Mixing of Hydraulic Cement Pastes and Mortars of Plastic Consistency. PA, West Conshohocken:
1994. www.astm.org.
[48] Plourabou F, Kurowski P, Hulin JP, Roux S, Schmittbuhl J. Aperture of rough cracks. Phys Rev E 1995;51(3):1675.
[49] Mly KJ, Hansen A, Hinrichsen EL, Roux S. Experimental measurements of the roughness of brittle cracks. Phys Rev Lett 1992;68(2):2135.
[50] Milman VY, Blumenfeld R, Stelmashenko NA, Ball RC. Comment on Experimental measurements of the roughness of brittle cracks. Phys Rev Lett
1993;71(1):204.
[51] Alves LM, da Silva RV, De Lacerda LA. Fractal modeling of the JR curve and the influence of the rugged crack growth on the stable elasticplastic
fracture mechanics. Eng Fract Mech 2010;77:245166. http://dx.doi.org/10.1016/j.engfracmech.2010.06.006.
[52] Alves LM. Foundations of measurement fractal theory for the fracture mechanics. In: Belov A, editor. Applied Fracture Mechanics. Croatia: Intech; 2012,
ISBN 978-953-51-0897-9. http://dx.doi.org/10.5772/51813.
[53] Ponson L, Bonamy D, Auradou H, Mourot G, Morel S, Bouchaud E, et al. Anisotropic self-affine properties of experimental fracture surfaces. Int J Fract
2006;140(14):2737.