Phase-field fracture irreversibility using the slack variable approach

Ritukesh Bharali1 , Fredrik Larsson1 , and Ralf Jänicke2

1
Department of Industrial and Material Science, Chalmers University of Technology
2
Institute of Applied Mechanics, Technische Universität Braunschweig
arXiv:2202.10040v1 [math.NA] 21 Feb 2022

Highlights
• Effective equality constraint for fracture irreversibility inequality constraint using slack variable.

• The Lagrange Multiplier Method and the Penalty method are adopted to augment the energy
functional.

• Numerical experiments carried out on both brittle and quasi-brittle fracture problems.

Abstract
In this manuscript, the phase-field fracture irreversibility constraint is transformed into an equality-
based constraint using the slack variable approach. The equality-based fracture irreversibility
constraint is then introduced in the phase-field fracture energy functional using the Lagrange
Multiplier Method and the Penalty method. Both methods are variationally consistent with
the conventional variational inequality phase-field fracture problem, unlike the history-variable
approach. Thereafter, numerical experiments are carried out on benchmark problems in brittle
and quasi-brittle fracture to demonstrate the efficacy of the proposed method.

Keywords: phase-field fracture, brittle, quasi-brittle, Lagrange multiplier method, Penalty method,
COMSOL

1
1 Introduction
The phase-field model for fracture emerged from the seminal work of [1], wherein the Griffith fracture criterion
was cast into a variational setting. Later, a numerical implementation of the same was proposed in [2],
using the Ambrosio–Tortorelli regularisation of the Mumford-Shah potential [3]. In this implementation, the
fracture is represented by an auxiliary variable, that interpolates between the intact and broken material
states. Such a formulation allows the automatic tracking of fractures on a fixed mesh, thereby eliminating
the need for the tedious tracking and remeshing processes observed with discrete methods. Furthermore, the
phase-field model for fracture is also able to handle topologically complex (branching, kinking and merging)
fractures, and is able to demonstrate fracture initiation without introducing any singularity [4]. Owing to
these advantages, the phase-field model for fracture has gained popularity in the computational mechanics
community in the past decade.
A thermodynamically consistent formulation of the phase-field fracture model was proposed by [5],
adopting an energetic cracking driving force definition. Since then, over the past decade, researchers have
extended the work to thermo-mechanical problems at large strain [6], ductile failure [7, 8], fracture in thin
films [9], anisotropic fracture [10], fracture in fully/partially saturated porous media [11–13], hydrogen assisted
cracking [14], dissolution-driven stress corrosion cracking [15], fracture and fatigue in shape-memory alloys
[16], and brittle failure of Reisner-Mindlin plates [17] to cite a few. Finally, the phase-field model for fracture
has also been used for the multi-scale finite element method [18–20], asymptotic homogenisation [21], and
variationally consistent homogenisation [22]. For a detailed overview on the phase-field fracture model, the
reader is referred to the comprehensive review works [23–25].
Despite its growing popularity, the phase-field model for fracture poses several challenges when it comes
to robust and computationally efficient solution techniques [25]. These include,

(A) the poor performance of the monolithic Newton-Raphson (NR) method due to the non-convex energy
functional,
(B) variational inequality arising from fracture irreversibility, and
(C) extremely fine meshes are required in the fracture zone.

In order to alleviate the problem concerning the poor-performance of the NR method (A), [26] utilised
a linear extrapolation in time 1 for the phase-field variable in the momentum balance equation. In [27],
a novel line search technique was developed, while in [28] and [29] Augmented Lagrangian and modified
NR approaches were proposed. More recently, [30] adopted a recursive multilevel trust region method,
that resulted in improved convergence of the NR method. The development of robust, monolithic solution
technique is still an active area in phase-field fracture research, and this manuscript is contribution towards
this aspect.
The next issue pertains to the variational inequality problem arising out of the fracture irreversibility
condition (B). This aspect has been treated in different ways in the phase-field fracture literature. They
include crack set irreversibility [2, 31, 32], penalisation [4, 28, 29], and the implicit History variable based
method proposed by [33]. The lattermost method remains popular despite its non-variational nature,
over-estimation of the fracture surface energy, and the necessity for computationally expensive alternate
minimisation (staggered) solution procedure [4]. In a novel approach, this manuscript adopts the slack
variable method [34, 35], that constructs an equivalent variational equality problem, while maintaining the
variational structure of the original problem.
Another issue pertaining to the phase-field fracture model is the requirement of extremely fine meshes
in the smeared fracture zone (C). In this context, [2] and [9] advocates the use of uniformly refined meshes
together with parallel computing while [36–38] opted for error-controlled adaptive mesh refinement. Some
other approaches include pre-refined meshes when the fracture path is known [4, 23, 27], adaptive mesh
refinement in [26] based on a phase-field threshold or Kelly error estimates [39], and multi-level hp-FEM in
1
The linear extrapolation of the phase-field variable in time is a questionable assumption although it results in a
robust solution method.

2
[40]. In this manuscript, uniformly refined meshes and pre-refined meshes based on a priori knowledge of the
fracture path, are used.
The novelty of this manuscript lies in the alternative treatment of the fracture irreversibility inequality
constraint, using the slack variable approach. The inequality constraint is replaced by an equivalent equality-
based fracture irreversibility constraint. The constraint is then augmented to the phase-field fracture energy
functional using the Lagrange Multiplier method and the Penalty method. These methods are variationally
consistent, unlike the history-variable approach proposed in [33].
The manuscript is organised as follows: Section 2 introduces the phase-field model for fracture, its
underlying energy functional and the Karush-Kuhn-Tucker (KKT) optimality conditions. Thereafter, in
Section 3, the Method of Multipliers is adopted to augment the energy functional a slack variable-based
fracture irreversibility criterion. The equivalence of the new formulation with the original problem in Section
2 is established in terms of KKT conditions. Furthermore, the Euler-Lagrange equations are presented. In
Section 4, the slack variable-based fracture irreversibility constraint is introduced in the energy functional
using the Penalisation method, followed by the derivation of the pertinent Euler-Lagrange equations. The
numerical experiments on benchmark problems are presented in Section 5, and Section 6 lays down the
concluding remarks of this manuscript.

Notation
The following notations are strictly adhered to in this manuscript:
• Zero-order tensors are represented using small italicized letters, e.g., a. Bold symbols are used for first
and second-order tensors, for instance, stress σ and strain .
• A function f with its arguments x, y is written in the form f (x, y), whereas a variable g with operational
dependencies p, q is written as g[p, q].
• The Macaulay operator on a variable x is defined as hxi± = 21 (x ± |x|).

2 Phase-field fracture model

2.1 Energy functional
Let Ω ∈ Rdim (dim = 2, 3) be the domain occupied by the fracturing solid, shown in Figure 1. Its boundary
Γ is decomposed into a Dirichlet boundary ΓuD and a Neumann boundary ΓuN , such that Γ = ΓuD ∪ ΓuN and
ΓuD ∩ ΓuN = ∅. Furthermore, the fracture is represented by an auxiliary variable (phase-field) ϕ ∈ [0, 1] within
a diffusive (smeared) zone of width l > 0.

ΓuN

ΓuD

Ω
l

Figure 1: A solid Ω ∈ R2 embedded with a diffused (smeared) crack. Dirichlet and Neumann boundaries are indicated
with ΓuD and ΓuN respectively. Figure adopted from [22].

3
The energy functional of the phase-field fracture model [4, 27] is stated as,

Z Z Z Z
Gc
g(ϕ)Ψ+ ([u]) dΩ + Ψ− ([u]) dΩ − w(ϕ) + l|∇ϕ|2 dΩ,


E(u, ϕ) = tup · u dΓ + (1)
Ω Ω ΓuN Ω cw l

with u and ϕ representing the displacement and the phase-field respectively. The symbols adopted for the
additional functions and parameters in (1) are presented in Table 1.

Symbol Function/Parameter
g(ϕ) Degradation function
Ψ+ Fracture driving strain energy density
Ψ− Residual strain energy density
[u] Symmetric gradient of u
tup Prescribed traction
Gc Griffith fracture energy
cw Normalisation constant
l Fracture length-scale
w(ϕ) Local fracture energy function
Table 1: List of functions and parameters corresponding to the phase-field fracture energy functional (1).

In this manuscript, Ψ+ and Ψ− represent the tensile and the compressive strain energy density functions
respectively. They are functions of strain tensor [u], given by

[u] = (u ⊗ ∇)sym . (2)

±
The explicit expression for Ψ is obtained following [5] as,
1
Ψ± ([u]) = λh(I : [u])i2± + µ± [u] : ± [u], (3)
2
where λ and µ are the Lamé constants, I is a second-order identity tensor, u is the displacement. Moreover,
the tensile/compressive strain ± [u] is defined as
dim
X
± [u] = hi i± pi ⊗ pi (4)
i=1

where i represents the ith eigenvalue of the strain, and pi is its corresponding normalised eigenvector. The
tensile/compressive strain energy definition in (3) yield the corresponding stresses upon taking the derivative
w.r.t the strain. They are explicitly stated in a compact form as,

∂Ψ±
σ± = = λh(I : [u])i± I + 2µ± [u]. (5)
∂
The last integral in the energy functional (1) represents the fracture energy in a phase-field fracture
model. The phase-field fracture model allows flexibility in the choice of the degradation function g(ϕ) and
the locally dissipated fracture energy function w(ϕ). Table 2 presents commonly adopted functions for brittle
and quasi-brittle fracture.
For the Quasi-Brittle model in Table 2, the parameters p, a1 , a2 and a3 are chosen such that they mimic the
different traction-separation laws in cohesive zone modelling. Following [42], the constant a1 is given by
4E0 Gc
a1 = (6)
πlft2

4
 Model g(ϕ) w(ϕ) cw
Brittle-AT1[41] (1 − ϕ)2 ϕ 8/3
Brittle-AT2[31] (1 − ϕ)2 ϕ2 2
(1 − ϕ)p
Quasi-Brittle[42] 2ϕ − ϕ2 π
(1 − ϕ)p + a1 ϕ + a1 a2 ϕ2 + a1 a2 a3 ϕ3
Table 2: Degradation function, local fracture energy function and its normalisation constant for different phase-field
fracture models. The abbreviation ‘AT’ stands for Ambrosio-Tortorelli.

where, the newly introduced parameters E0 and ft represent the Youngs’ modulus and the tensile strength of
the material. Next, Table 3 presents the different traction-separation laws and the corresponding values of p,
a2 and a3 [42].

Model p a2 a3
Linear Softening 2 −0.5 0
Exponential Softening 2.5 25/3 − 3 0
Cornelissen et. al.[43] Softening 2 1.3868 0.6567
Table 3: Quasi-brittle phase-field fracture models and their parameters, based on [42].

The prediction of fracture initiation and propagation in the solid occupying the domain Ω requires solving
the constrained minimisation problem:
Problem 1. Find u(t) and ϕ(t), where n ϕ = ϕ(tn ) is known for tn < t such that,
minu(t),ϕ(t) E and h(ϕ) ≥ 0, (7)
n
where h(ϕ) = ϕ − ϕ, and the left superscript n denotes the previous time-step. Moreover, the problem is
augmented with (pseudo) time-dependent Dirichlet boundary conditions up on ΓuD and ϕp on Γϕ D , and/or
Neumann boundary conditions tup on ΓuN and tϕ ϕ u u
p on ΓN . Furthermore, the boundary Γ is split as Γ = ΓD ∪ ΓN ,
u u ϕ ϕ ϕ ϕ
ΓD ∩ ΓN = ∅ and Γ = ΓD ∪ ΓN , ΓD ∩ ΓN = ∅ respectively. 

2.2 Karush-Kuhn-Tucker conditions

The KKT triple (u, ϕ, Λ) in appropriate spaces for the Problem 1 requires
Z
 0
E(u, ϕ) − Λh(ϕ) dΩ u = 0, (8a)
ZΩ
 0
E(u, ϕ) − Λh(ϕ) dΩ ϕ = 0, , (8b)
Ω
h(ϕ) ≥ 0, (8c)
Λ ≥ 0, (8d)
Λh(ϕ) = 0 (8e)
to hold, where Λ is a multiplier. Equations (8a) and (8b) represent the Euler-Lagrange equations pertaining
to the stationary condition, while equations (8c-8e) are the primal and dual feasibility conditions, and the
complementary slackness respectively. Formally, the latter set of equations enforces fracture irreversbility.
To the same end, alternative, but equivalent formulations have been adopted in the phase-field fracture
literature. For instance, [26] adopted the primal-dual active set strategy proposed by [44], [28] utilised an
Augmented Lagrangian penalisation approach with the Moreau-Yoshida indicator function, and [4] opted for
a simple penalisation approach. The authors would like to emphasise that the rather popular history-variable
approach in [5] is not variationally consistent.

5
3 Lagrange Multiplier Method (LMM)
3.1 Fracture irreversibility and modified energy functional
In order to enforce fracture irreversibility h(ϕ) = ϕ − n ϕ ≥ 0, a slack variable is defined as

θ2 = h(ϕ) = ϕ − n ϕ. (9)
2
It is observed that θ admits value greater than or equal to zero, thereby fulfilling the fracture irreversibility
criterion. Next, the constrained minimisation Problem 1 is reformulated as:

Problem 2. Find u(t), ϕ(t) and θ(t) such that,

minu(t),ϕ(t),θ(t) E
(10)
subjected to h(ϕ) − θ2 = 0.

with suitable (pseudo) time-dependent Dirichlet and/or Neumann boundary conditions, as mentioned in
Problem 1. 

Augmenting the equality constraint (10) in (1) via the Lagrange multiplier Λ results in the modified energy
functional,

Ẽ(u, ϕ, θ, Λ) = E(u, ϕ) − Λ h(ϕ) − θ2 .

(11)

3.2 Stationary conditions

The stationary conditions for the Problem 2 are given by

0
Ẽu (u, ϕ, θ, Λ; δu) = 0, (12a)
0
Ẽϕ (u, ϕ, θ, Λ; δϕ) = 0, (12b)
0
Ẽθ (u, ϕ, θ, Λ; δθ) = 0, (12c)
ẼΛ0 (u, ϕ, θ, Λ; δΛ) = 0. (12d)

For equivalence of these conditions with those presented for Problem 1, the reader is referred to Proposition
3.1 in [45].

3.3 Euler-Lagrange equations

The Euler-Lagrange equations for Problem 2 are (12a)-(12d), obtained upon taking the first variation of the
energy functional (11) w.r.t its solution variables u, ϕ, θ and Λ. This results in,

Problem 3. Find (u, ϕ, θ, Λ) ∈ U × V × W × A with

Z  
Ru = g(ϕ)σ + [u] + σ − [u] : [δu] dΩ = 0 ∀ δu ∈ U0 (13a)
Ω
Z   Z
Gc 0 2Gc l
ϕ
R = w (ϕ) + g 0 (ϕ)Ψ+ ([u]) − Λ δϕ dΩ + ∇ϕ · ∇δϕ dΩ = 0 ∀ δϕ ∈ V0 (13b)
Ω cw l Ω cw
Z
Rθ = 2Λθδθ dΩ = 0 ∀ δθ ∈ W (13c)
ΩZ
RΛ = − h(ϕ) − θ2 δΛ dΩ = 0


∀ δΛ ∈ A. (13d)
Ω

6
 with trial function spaces

U := {u ∈ [H 1 (Ω)]dim |u = up on ΓuD }, (14a)

ϕ
V := {ϕ ∈ [H 1 (Ω)]1 |ϕ = ϕp on ΓD }, (14b)
W := {θ ∈ [L2 (Ω)]}, (14c)
A := {Λ ∈ [L2 (Ω)]}, (14d)

and test function spaces

U0 := {u ∈ [H 1 (Ω)]dim |u = 0 on ΓuD }, (15a)

0 1 1 ϕ
V := {ϕ ∈ [H (Ω)] |ϕ = 0 on ΓD }, (15b)

and pertinent (pseudo) time-dependent Neumann condition

t := (g(ϕ)σ + + σ − ) · n = tup on ΓuN . (16)



Remark 1. In the event Λ and θ are zero, their stiffness contribution is zero and the stiffness matrix is
singular. In order to avoid this situation a fictitious stiffness is introduced to guarantee solvability.

4 Penalty (Pen.) method

4.1 The energy functional
In an alternative approach, the squared slack variable equation (9) is introduced in the energy functional (1)
as a quadratic term via penalisation. This results in the modified energy functional,
Z  2
η 2
Ê(u, ϕ, θ) = E(u, ϕ) + h(ϕ) − θ dΩ, (17)
2 Ω
where, η is a penalty parameter.

4.2 Euler-Lagrange equations

The Euler-Lagrange equations are obtained upon taking the first variation of the energy functional (17) w.r.t
its solution variables u, ϕ and θ. This results in:

Problem 4. Find (u, ϕ, θ) ∈ U × V × W with

Z  
u
R = + −
g(ϕ)σ [u] + σ [u] : [δu] dΩ = 0 ∀ δu ∈ U0
Ω
(18a)
Z   Z
Gc 0 2Gc l
Rϕ = w (ϕ) + g 0 (ϕ)Ψ+ ([u]) + η(h(ϕ) − θ2 ) δϕ dΩ + ∇ϕ · ∇δϕ dΩ = 0 ∀ δϕ ∈ V0
Ω cw l Ω cw
(18b)
Z
Rθ = −2ηθ(h(ϕ) − θ2 )δθ dΩ = 0 ∀ δθ ∈ W
Ω
(18c)

7
 with trial function spaces

U := {u ∈ [H 1 (Ω)]dim |u = up on ΓuD }, (19a)

ϕ
V := {ϕ ∈ [H 1 (Ω)]1 |ϕ = ϕp on ΓD }, (19b)
W := {θ ∈ [L2 (Ω)]|}, (19c)

and test function spaces

U0 := {u ∈ [H 1 (Ω)]dim |u = 0 on ΓuD }, (20a)

V0 := {ϕ ∈ [H 1 (Ω)]1 |ϕ = 0 on Γϕ
D }, (20b)

and pertinent (pseudo) time-dependent Neumann condition

t := (g(ϕ)σ + + σ − ) · n = tup on ΓuN . (21)



Remark 2. In the event h(ϕ) − θ2 and θ are zero, their stiffness contribution is zero and the stiffness matrix
is singular. In order to avoid this situation a fictitious stiffness is introduced to guarantee solvability.

5 Numerical Study
The numerical experiments on benchmark problems are presented in this section. They include both brittle
and quasi-brittle fracture. In the brittle fracture domain, the Single Edge Notched Specimen under Tension
(SENT), and under Shear (SENS), and the Notched concrete specimen with a hole [23] is presented. For
quasi-brittle fracture, the benchmark problems include the concrete three-point bending problem [46], and the
Winkler L-shaped panel experiment [47]. For each problem, the geometry as well as the material properties
and loading conditions are presented in the respective sub-section. The load-displacement curves and the
phase-field topology at the final step of analysis is also presented.
For the penalty method, the penalty parameter η is set to 106 [N/mm2 ] throughout the study. This choice
is motivated by similarity of the load-displacement curves obtained with those from the Lagrange Multiplier
Method. Furthermore, for both the aforementioned methods, a fully coupled monolithic solution technique is
adopted using the Newton-Raphson solver in COMSOL Multiphysics. The iterations are terminated when
the error measure defined as the weighted Euclidean norm of the solution update,
r v uM N  2
1 u X 1 X |Ei,j |
err := t , (22)
M j=1 Nj i=1 Wi,j

is less than 1e − 4. In the above equation, M indicates the number of fields, j is the number of degree of
freedom for each field j = 1, 2, ..M , E represents the absolute update (say, u(m) -u(m−1) for displacement, m
being the iteration count), and Wi,j = max(|Ui,j |, Sj ). The entire solution vector is represented with U and
Sj refers to scaling of solution variables2 .

5.1 Brittle: Single Edge Notched specimen under Tension (SENT)

A unit square (in mm) embedded with a horizontal notch, midway along the height is considered, as shown in
Figure 2. The length of the notch (shown in red) is equal to half of the edge length of the plate. The notch
is modelled explicitly in the finite element mesh. A quasi-static loading is applied at the top boundary in
the form of prescribed displacement increment ∆u = 1e − 5[mm] for the first 450 steps, following which it is
changed to 1e − 6[mm]. Furthermore, the bottom boundary remains fixed. The additional model parameters
are presented in Table 4.

8
 ∆u
∆u

Parameters Value
Model Brittle-AT2
λ 121.154 [GPa]
µ 80.769 [GPa]
Gc 2700 [N/m]
l 1.5e-2 [mm]
Element size l/2
Figure 2: SENT experiment Figure 3: SENS experiment Table 4: Parameters

1 LMM Miehe [5]

Pen. Ambati [23]
0.8
Load [kN]

0.6

0.4

0.2

0
0 1 2 3 4 5 6
Displacement [mm] ·10−3 0 0.5 1
ϕ
(a) (b)

Figure 4: Figure (a) presents the load-displacement curves for the single edge notched specimen under tension. Here,
LMM and Pen. refer to the Lagrange Multiplier Method and the Penalty method respectively. Figure (b) shows the
distribution of the phase-field variable at the final step of the analysis.

Figure 4a presents the load-displacement curves obtained using the Lagrange Multiplier Method (LMM)
and Penalty (Pen.) method, along with those from the literature [5, 23]. Both methods yield a similar
behaviour compared to [5], in terms of the peak load and the post-peak behaviour. Moreover, the phase-field
fracture topology at failure in Figure 4b is also identical to those reported in the aforementioned literature.

5.2 Brittle: Single Edge Notched specimen under Shear (SENS)

In order to perform a shear test, the SEN specimen is loaded horizontally along the top edge as shown in
Figure 3. The material properties remain same as presented in Table 4. A quasi-static loading is applied to
the top boundary in the form of prescribed displacement increment ∆u = 1e − 4[mm] for the first 90 steps,
following which it is changed to 1e − 5[mm]. Furthermore, the bottom boundary remains fixed, and roller
support is implemented in left and right edges thereby restricting the vertical displacement.
Figure 5a presents the load-displacement curves obtained using the Lagrange Multiplier Method and the
2
Scaling of solution variables prevents possible ill-conditioning of the stiffness matrix.

9
Penalty method, along with those from the literature [5]. The former methods result in 9% higher peak load
estimation as compared to [5]. However, the phase-field fracture topology in the final step of the analysis is
consistent with those reported in the aforementioned literature.

1
LMM Miehe [5]
0.8 Pen.
Load [kN]

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Displacement [mm] ·10−2 0 0.5 1
ϕ
(a) (b)

Figure 5: Figure (a) presents the load-displacement curves for the single edge notched specimen under shear. Here,
LMM and Pen. refer to the Lagrange Multiplier Method and the Penalty method respectively. Figure (b) shows the
distribution of the phase-field variable at the final step of the analysis.

5.3 Brittle: Notched concrete specimen with a hole

A notched concrete specimen with a hole, shown in Figure 6 is considered in this study. The experimental
and numerical analysis of the same has been carried out earlier in [23]. The specimen has dimensions 65
x 120 [mm2 ], the hole being 20 [mm] in diameter located at (36.5 [mm], 51[mm]). Moreover, a notch, 10
[mm] in length is located at 65 [mm] height from the bottom of the plate. The model parameters used in the
simulation is presented in Table 5, while the experimentally observed fracture pattern is shown in Figure 7.
As shown in Figure 6, the plate is loaded via the upper pin (in grey) with displacement increment ∆u = 1e − 3
mm, while the lower pin (in grey) is remains fixed.
Figure 8a presents the load-displacement curves obtained using the Lagrange Multiplier Method and the
Penalty method, along with those from the literature [23, 48]. The curves corresponding to the Lagrange
Multiplier Method and the Penalty method coincide, and predict peak loads close to those obtained in [23]
and [48]. However, the load-displacement curves are significantly different from [23]. The reason for this
behaviour is not investigated in this manuscript. However, it is important to note that in this work, the
upper and lower pins are assumed to be rigid connectors. As such, the loading or fixed boundary conditions
is applied w.r.t. the centre of the pin. Furthermore, [23] opted for l = 0.1 [mm], in this manuscript as well as
in [48], l = 0.25 [mm]. Finally, from Figure 8b, it observed that the final phase-field fracture topology does
match the experimentally observed fracture patterns in Figure 7.

5.4 Quasi-brittle: Concrete three-point bending

A three-point bending experiment on a notched concrete beam reported in [46] is considered here. The beam
has dimensions 450 × 100 [mm2 ], and has a notch 5 × 50 [mm2 ]. A schematic of the beam along with the
loading conditions is presented in Figure 9. Displacement-based load increments of ∆u = 1e − 3 [mm] is
enforced throughout the simulation. The model parameters are presented in Table 6.

10
 ∆u

Symbol Value
Model Brittle-AT2
λ 1.94 [kN/mm2 ]
µ 2.45 [kN/mm2 ]
Gc 2.28e-3 [kN/mm]
l 0.25 [mm]
Element size l/2
Figure 6: Geometry and constraints Figure 7: Experimental results [23] Table 5: Material properties [48]

1
LMM Navidtehrani [48]
0.8 Pen. Ambati [23]
Load [kN]

0.6

0.4

0.2

0
0 0.5 1 1.5 2
Displacement [mm] 0 0.5 1
ϕ
(a) (b)

Figure 8: Figure (a) presents the load-displacement curves for the notched specimen with hole test, obtained from the
current implementation, [23], and [48]. Here, LMM and Pen. refer to the Lagrange Multiplier Method and the Penalty
method respectively. Figure (b) shows the distribution of the phase-field variable at the final step of the analysis in a
section of the specimen.

Figure 10a presents the load-displacement curves obtained using the Lagrange Multiplier Method and the
Penalty method, along with those obtained from the experiments [46]. It is observed that both methods yield
identical curves. These curves are close to the experimental upper bound (shaded region). Furthermore, the
phase-field fracture topology at the final step in the analysis in Figure 10b is also identical to that reported
in [46].

5.5 Quasi-brittle: Winkler L-panel

The L-shaped panel studied by [47, 49] is considered in this sub-section. Figure 11 shows the geometry as well
as the loading conditions. The longer edges of the panel are 500 [mm] and the smaller edges are 250 [mm].
The loading is applied on the edge marked in blue, 30 [mm] in length, and is in the form of displacement-based

11
 Parameters Value
Model Quasi-brittle
Softening Cornellisen et. al. [43]
∆u E0 2e4 [MPa]
ν 0.2 [-]
ft 2.4 [MPa]
Gc 0.113 [N/mm]
l 2.5 [mm]
Element size l/5
Figure 9: Three point bending test Table 6: Parameters for three point bending test

1.5 LMM
Pen.
Load [kN]

0.5

0
0 0.2 0.4 0.6 0.8 1 0 0.5 1
Displacement [mm] ϕ

(a) Load-displacement plot (b)

Figure 10: Figure (a) presents the load-displacement curves for the concrete three-point bending test. Here, LMM
and Pen. refer to the Lagrange Multiplier Method and the Penalty method respectively. The experimental range is
represented by the shaded area. Figure (b) shows the distribution of the phase-field variable at the final step of the
analysis in a section of the beam.

increments of ∆u = 1e − 3 [mm]. The additional model parameters are presented in Table 7.

Figure 12a presents the load-displacement curves obtained using the Lagrange Multiplier Method and
the Penalty method, along with those obtained from the experiments [47, 49]. It is observed that the curves
obtained using both methods are identical. These curves are also close to the experimental range (shaded
region). Furthermore, the phase-field fracture topology at the final step in the analysis in Figure 12b is also
identical to that reported in [47, 49].

6 Concluding Remarks
An alternative treatment of the phase-field fracture irreversibility constraint is presented in this manuscript.
The fracture irreversibility constraint h(ϕ) ≥ 0 is transformed into an equivalent equality constraint. This is
carried out upon introducing a slack variable θ, and equating h(ϕ) to θ2 . By definition, θ2 admits zero or
positive values, thereby ensuring the fracture irreversibility constraint. The constraint is then introduced in
the phase-field fracture energy functional using the Lagrange Multiplier Method and the Penalty method.
The slack variable approach preserves the variational nature of the phase-field fracture problem, unlike the

12
 Parameters Value
Model Quasi-brittle
Softening Cornellisen et. al. [43]
E0 2.0e4 [MPa]
ν 0.18 [-]
∆u ft 2.5 [MPa]
Gc 0.130 [N/mm]
l 5 [mm]
Element size l/5
Figure 11: Winkler L-panel Table 7: Parameters for L-shaped panel test [49]

8
LMM
6 Pen.
Load [kN]

0
0 0.2 0.4 0.6 0.8 1
Displacement [mm] 0 0.5 1
ϕ
(a) Load-displacement plot (b)

Figure 12: Figure (a) presents the load-displacement curves for the concrete Winkler L-panel test. Here, LMM and
Pen. refer to the Lagrange Multiplier Method and the Penalty method respectively. The experimental range is
represented by the shaded area. Figure (b) shows the distribution of the phase-field variable at the final step of the
analysis in a section of the beam.

history variable approach in [33]. Furthermore, numerical experiments are conducted on benchmark brittle
(AT2) and quasi-brittle fracture problems to demonstrate the efficacy of the proposed methods.
The scope for future studies include the extension towards complex multi-physics problems, large-scale
problems, that maybe require special preconditioning techniques. Other future research directions may be
towards accelerating the convergence rate using Anderson acceleration [50] or Large time increment method
[51, 52].

7 Software Implementation and Data Availability

The numerical study in Section 5 is carried out using the equation-based modelling approach in the software
package COMSOL Multiphysics 5.6. The source files would be made available in Github repository of the
corresponding author (https://github.com/rbharali/PhaseFieldFractureCOMSOL/).

13
Acknowledgements
The financial support from the Swedish Research Council for Sustainable Development (FORMAS) under
Grant 2018-01249 and the Swedish Research Council (VR) under Grant 2017-05192 is gratefully acknowledged.
The authors would also like to thank Prof. Laura de Lorenzis (ETH Zürich) for insightful comments and
suggestions.

References
[1] G.A. Francfort and J.-J. Marigo. “Revisiting brittle fracture as an energy minimization
problem”. In: Journal of the Mechanics and Physics of Solids 46.8 (Aug. 1998), pp. 1319–1342.
issn: 00225096. doi: 10.1016/S0022-5096(98)00034-9.
[2] B. Bourdin, G.A. Francfort, and J-J. Marigo. “Numerical experiments in revisited brittle
fracture”. In: Journal of the Mechanics and Physics of Solids 48.4 (2000), pp. 797–826. issn:
0022-5096. doi: https://doi.org/10.1016/S0022-5096(99)00028-9.
[3] David Mumford and Jayant Shah. “Optimal approximations by piecewise smooth functions
and associated variational problems”. In: Communications on Pure and Applied Mathematics
42.5 (1989), pp. 577–685. doi: 10.1002/cpa.3160420503.
[4] T. Gerasimov and L. De Lorenzis. “On penalization in variational phase-field models of
brittle fracture”. In: Computer Methods in Applied Mechanics and Engineering 354 (2019),
pp. 990–1026. issn: 0045-7825. doi: https://doi.org/10.1016/j.cma.2019.05.038.
[5] C. Miehe, F. Welschinger, and M. Hofacker. “Thermodynamically consistent phase-field models
of fracture: Variational principles and multi-field FE implementations”. In: International
Journal for Numerical Methods in Engineering 83.10 (2010), pp. 1273–1311. doi: 10.1002/
nme.2861.
[6] Christian Miehe, Lisa Marie Schänzel, and Heike Ulmer. “Phase field modeling of fracture
in multi-physics problems. Part I. Balance of crack surface and failure criteria for brittle
crack propagation in thermo-elastic solids”. In: Computer Methods in Applied Mechanics and
Engineering 294 (2015), pp. 449–485. issn: 00457825. doi: 10.1016/j.cma.2014.11.016.
[7] C. Miehe et al. “Phase field modeling of fracture in multi-physics problems. Part II. Coupled
brittle-to-ductile failure criteria and crack propagation in thermo-elastic-plastic solids”. In:
Computer Methods in Applied Mechanics and Engineering 294 (2015), pp. 486–522. issn:
00457825. doi: 10.1016/j.cma.2014.11.017.
[8] M. Ambati, T. Gerasimov, and L. De Lorenzis. “Phase-field modeling of ductile fracture”. In:
Computational Mechanics 55.5 (2015), pp. 1017–1040. issn: 1432-0924. doi: 10.1007/s00466-
015-1151-4.
[9] A. Mesgarnejad, B. Bourdin, and M.M. Khonsari. “A variational approach to the fracture of
brittle thin films subject to out-of-plane loading”. In: Journal of the Mechanics and Physics
of Solids 61.11 (2013), pp. 2360–2379. issn: 0022-5096. doi: https://doi.org/10.1016/j.
jmps.2013.05.001.
[10] Thanh Tung Nguyen, Julien Réthoré, and Marie-Christine Baietto. “Phase field modelling
of anisotropic crack propagation”. In: European Journal of Mechanics - A/Solids 65 (2017),
pp. 279–288. issn: 0997-7538. doi: https://doi.org/10.1016/j.euromechsol.2017.05.
002.

14
[11] Christian Miehe and Steffen Mauthe. “Phase field modeling of fracture in multi-physics
problems. Part III. Crack driving forces in hydro-poro-elasticity and hydraulic fracturing of
fluid-saturated porous media”. In: Computer Methods in Applied Mechanics and Engineering
304 (2016), pp. 619–655. issn: 00457825. doi: 10.1016/j.cma.2015.09.021.
[12] T. Cajuhi, L. Sanavia, and L. De Lorenzis. “Phase-field modeling of fracture in variably satu-
rated porous media”. In: Computational Mechanics 61.3 (2018), pp. 299–318. issn: 01787675.
doi: 10.1007/s00466-017-1459-3.
[13] A. Mikelić, M. F. Wheeler, and T. Wick. Phase-field modeling through iterative splitting of
hydraulic fractures in a poroelastic medium. Vol. 10. Springer International Publishing, 2019.
isbn: 0123456789. doi: 10.1007/s13137-019-0113-y.
[14] Emilio Martı́nez-Pañeda, Alireza Golahmar, and Christian F. Niordson. “A phase field
formulation for hydrogen assisted cracking”. In: Computer Methods in Applied Mechanics and
Engineering 342 (2018), pp. 742–761. issn: 00457825. doi: 10.1016/j.cma.2018.07.021.
[15] Chuanjie Cui, Rujin Ma, and Emilio Martı́nez-Pañeda. “A phase field formulation for
dissolution-driven stress corrosion cracking”. In: Journal of the Mechanics and Physics of
Solids 147 (2021), p. 104254. issn: 0022-5096. doi: https://doi.org/10.1016/j.jmps.2020.
104254.
[16] Marlini Simoes and Emilio Martı́nez-Pañeda. “Phase field modelling of fracture and fatigue in
Shape Memory Alloys”. In: Computer Methods in Applied Mechanics and Engineering 373
(2021), p. 113504. issn: 0045-7825. doi: https://doi.org/10.1016/j.cma.2020.113504.
url: https://www.sciencedirect.com/science/article/pii/S0045782520306897.
[17] G. Kikis et al. “Phase-field model of brittle fracture in Reissner–Mindlin plates and shells”.
In: Computer Methods in Applied Mechanics and Engineering 373 (2021), p. 113490. issn:
0045-7825. doi: https://doi.org/10.1016/j.cma.2020.113490. url: https://www.
sciencedirect.com/science/article/pii/S0045782520306757.
[18] R.U. Patil, B.K. Mishra, and I.V. Singh. “An adaptive multiscale phase field method for
brittle fracture”. In: Computer Methods in Applied Mechanics and Engineering 329 (Feb.
2018), pp. 254–288. issn: 00457825. doi: 10.1016/j.cma.2017.09.021.
[19] R.U. Patil et al. “A new multiscale phase field method to simulate failure in composites”.
In: Advances in Engineering Software 126 (2018), pp. 9–33. issn: 0965-9978. doi: https:
//doi.org/10.1016/j.advengsoft.2018.08.010.
[20] R.U. Patil, B.K. Mishra, and I.V. Singh. “A multiscale framework based on phase field
method and XFEM to simulate fracture in highly heterogeneous materials”. In: Theoretical
and Applied Fracture Mechanics 100 (2019), pp. 390–415. issn: 0167-8442. doi: https :
//doi.org/10.1016/j.tafmec.2019.02.002.
[21] F. Fantoni et al. “A phase field approach for damage propagation in periodic microstructured
materials”. In: International Journal of Fracture (2019). issn: 1573-2673. doi: 10.1007/
s10704-019-00400-x.
[22] Ritukesh Bharali, Fredrik Larsson, and Ralf Jänicke. “Computational homogenisation of
phase-field fracture”. In: European Journal of Mechanics - A/Solids 88 (2021), p. 104247.
issn: 0997-7538. doi: https://doi.org/10.1016/j.euromechsol.2021.104247.

15
[23] Marreddy Ambati, Tymofiy Gerasimov, and Laura De Lorenzis. “A review on phase-field
models of brittle fracture and a new fast hybrid formulation”. In: Computational Mechanics
55.2 (2014), pp. 383–405. issn: 01787675. doi: 10.1007/s00466-014-1109-y.
[24] Jian-Ying Wu et al. “Chapter One - Phase-field modeling of fracture”. In: ed. by Stéphane
P.A. Bordas and Daniel S. Balint. Vol. 53. Advances in Applied Mechanics. Elsevier, 2020,
pp. 1–183. doi: https://doi.org/10.1016/bs.aams.2019.08.001. url: https://www.
sciencedirect.com/science/article/pii/S0065215619300134.
[25] Laura De Lorenzis and Tymofiy Gerasimov. “Numerical implementation of phase-field models
of brittle fracture”. In: Modeling in Engineering Using Innovative Numerical Methods for
Solids and Fluids. Springer, 2020, pp. 75–101.
[26] Timo Heister, Mary F. Wheeler, and Thomas Wick. “A primal-dual active set method
and predictor-corrector mesh adaptivity for computing fracture propagation using a phase-
field approach”. In: Computer Methods in Applied Mechanics and Engineering (2015). issn:
00457825. doi: 10.1016/j.cma.2015.03.009.
[27] T. Gerasimov and L. De Lorenzis. “A line search assisted monolithic approach for phase-field
computing of brittle fracture”. In: Computer Methods in Applied Mechanics and Engineering
312 (Dec. 2016), pp. 276–303. issn: 00457825. doi: 10.1016/j.cma.2015.12.017.
[28] Thomas Wick. “An Error-Oriented Newton/Inexact Augmented Lagrangian Approach for Fully
Monolithic Phase-Field Fracture Propagation”. In: SIAM Journal on Scientific Computing
39.4 (2017), B589–B617. issn: 1064-8275. doi: 10.1137/16m1063873.
[29] Thomas Wick. “Modified Newton methods for solving fully monolithic phase-field quasi-static
brittle fracture propagation”. In: Computer Methods in Applied Mechanics and Engineering
(2017). issn: 00457825. doi: 10.1016/j.cma.2017.07.026.
[30] Alena Kopaničáková and Rolf Krause. “A recursive multilevel trust region method with
application to fully monolithic phase-field models of brittle fracture”. In: Computer Methods
in Applied Mechanics and Engineering 360 (2020), p. 112720. issn: 0045-7825. doi: https:
/ / doi . org / 10 . 1016 / j . cma . 2019 . 112720. url: https : / / www . sciencedirect . com /
science/article/pii/S0045782519306085.
[31] B. Bourdin. “Numerical implementation of the variational formulation for quasi-static brittle
fracture”. In: Interfaces and Free Boundaries 9 (2007), pp. 411–430. doi: 10.4171/IFB/171.
[32] Siobhan Burke, Christoph Ortner, and Endre Süli. “An Adaptive Finite Element Approxima-
tion of a Variational Model of Brittle Fracture”. In: SIAM Journal on Numerical Analysis
48.3 (2010), pp. 980–1012. doi: 10.1137/080741033.
[33] Christian Miehe, Martina Hofacker, and Fabian Welschinger. “A phase field model for rate-
independent crack propagation: Robust algorithmic implementation based on operator splits”.
In: Computer Methods in Applied Mechanics and Engineering 199.45-48 (2010), pp. 2765–2778.
issn: 00457825. doi: 10.1016/j.cma.2010.04.011. url: http://dx.doi.org/10.1016/j.
cma.2010.04.011.
[34] FA Valentine. “The problem of Lagrange with differentiable inequality as added side conditions.
407–448”. In: Contributions to the Calculus of Variations 37 (1933).
[35] D P Bertsekas. “Nonlinear Programming”. In: Journal of the Operational Research Society
48.3 (1997), pp. 334–334. doi: 10.1057/palgrave.jors.2600425.

16
[36] Siobhan Burke, Christoph Ortner, and Endre Süli. “An adaptive finite element approximation
of a variational model of brittle fracture”. In: SIAM Journal on Numerical Analysis 48.3
(2010), pp. 980–1012.
[37] Marco Artina et al. “Anisotropic mesh adaptation for crack detection in brittle materials”. In:
SIAM Journal on Scientific Computing 37.4 (2015), B633–B659.
[38] Thomas Wick. “Goal functional evaluations for phase-field fracture using PU-based DWR
mesh adaptivity”. In: Computational Mechanics 57.6 (2016), pp. 1017–1035.
[39] D. W. Kelly et al. “A posteriori error analysis and adaptive processes in the finite ele-
ment method: Part I—error analysis”. In: International Journal for Numerical Methods in
Engineering 19.11 (1983), pp. 1593–1619. doi: https://doi.org/10.1002/nme.1620191103.
[40] S. Nagaraja et al. “Phase-field modeling of brittle fracture with multi-level hp-FEM and
the finite cell method”. In: Computational Mechanics 63.6 (June 2019), pp. 1283–1300. issn:
1432-0924. doi: 10.1007/s00466-018-1649-7.
[41] Kim Pham et al. “Gradient damage models and their use to approximate brittle fracture”. In:
International Journal of Damage Mechanics 20.4 (2011), pp. 618–652.
[42] Jian-Ying Wu. “A unified phase-field theory for the mechanics of damage and quasi-brittle
failure”. In: Journal of the Mechanics and Physics of Solids 103 (2017), pp. 72–99.
[43] H Cornelissen, D Hordijk, and H Reinhardt. “Experimental determination of crack softening
characteristics of normalweight and lightweight”. In: Heron 31.2 (1986), pp. 45–46.
[44] M. Hintermüller, K. Ito, and K. Kunisch. “The Primal-Dual Active Set Strategy as a Semis-
mooth Newton Method”. In: SIAM Journal on Optimization 13.3 (2002), pp. 865–888. doi:
10.1137/S1052623401383558.
[45] R. A. Tapia. “A stable approach to Newton’s method for general mathematical programming
problems inRn”. In: Journal of Optimization Theory and Applications 14.5 (Nov. 1974),
pp. 453–476. issn: 1573-2878. doi: 10.1007/BF00932842. url: https://doi.org/10.1007/
BF00932842.
[46] Jan Rots. “Computational modeling of concrete fracture”. PhD thesis. Building 23, Stevinweg
1, 2628 CN, Delft, The Netherlands: Technische Universiteit Delft, 1988.
[47] Bernhard Josef Winkler. Traglastuntersuchungen von unbewehrten und bewehrten Betonstruk-
turen auf der Grundlage eines objektiven Werkstoffgesetzes für Beton. Innsbruck University
Press, 2001.
[48] Yousef Navidtehrani, Covadonga Betegón, and Emilio Martı́nez-Pañeda. “A Unified Abaqus
Implementation of the Phase Field Fracture Method Using Only a User Material Subroutine”.
In: Materials 14.8 (2021). issn: 1996-1944. doi: 10.3390/ma14081913. url: https://www.
mdpi.com/1996-1944/14/8/1913.
[49] Jörg F Unger, Stefan Eckardt, and Carsten Könke. “Modelling of cohesive crack growth
in concrete structures with the extended finite element method”. In: Computer methods in
applied mechanics and engineering 196.41-44 (2007), pp. 4087–4100.
[50] Donald G Anderson. “Iterative procedures for nonlinear integral equations”. In: Journal of
the ACM (JACM) 12.4 (1965), pp. 547–560.

17
[51] Dan Givoli, Ritukesh Bharali, and Lambertus J Sluys. “LATIN: A new view and an extension
to wave propagation in nonlinear media”. In: International Journal for Numerical Methods in
Engineering 112.2 (2017), pp. 125–156.
[52] Ritukesh Bharali. “An Adaptive Generic Incremental LATIN method”. MA thesis. Building
23, Stevinweg 1, 2628 CN, Delft, The Netherlands: Technische Universiteit Delft, 2017.

18