A Phase-Field Study of Desiccation Crack Pattern Maturation under Drying-Wetting Cycles

arXiv:2609.11048 · cond-mat.soft, nlin.PS · Submitted 2026-09-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "A Phase-Field Study of Desiccation Crack Pattern Maturation under Drying-Wetting Cycles".

Mira: Here's the gist of the paper. A phase-field model incorporating crack healing and scar effects reveals that the characteristic intersection angle of desiccation cracks relaxes from near 90° toward 120° under repeated drying–wetting cycles The standard deviation of the intersection angle from 120° relaxes exponentially with a characteristic time of about 2.85 cycles, consistent with experiments Crack energy dominates the total energy and also relaxes exponentially with nearly the same characteristic cycle as the angle relaxation, driven mainly by a shortening of the effective crack length rather than a change in effective fracture toughness.

Kai: Model Formulation The study introduces a shrinkage-fracture model that combines elasticity theory with a phase-field model, extending previous work by adding crack healing and scar effects The energy functional considers elastic energy of the paste, crack energy, and interaction energy with the substrate.

Mira: The elastic energy density is modified by a degradation function g(d) to account for weakening due to cracks.

Kai: Crack generation is modeled using a phase-field approach where crack energy is expressed via the damage field d as Ψcr

d: = Z ∫omega ψcr(∇d, d) dV.

Paper summary: Kai: So, to recap what we just heard, this paper is using a phase-field model to investigate how drying and wetting cycles reorganize the pattern of cracks in desiccating material. They found that the angle distribution develops a growing peak near one hundred twenty degrees as the number of cycles increases.

Mira: And they confirmed this behavior is consistent with experiments, but they also noted that their characteristic time for the angle relaxation is slightly smaller than what was observed experimentally.

Lev: From an engineering standpoint, if you were trying to build hardware based on these models, the fact that the decay of crack energy is driven by a shortening of the effective crack length means we need to focus on modeling how fast that length shrinks.

Kai: The model itself combines elasticity theory with a phase-field approach, adding scar effects and crack healing to previous work. They use a memory field to represent these scars, which reduces the effective fracture toughness and critical energy density based on what happened before.

Mira: That's crucial because it means the material gets weaker after each cycle, which feeds back into how new cracks form in the next drying process.

Lev: If we were running this on real hardware, we'd have to worry about that memory field construction being numerically stable and accurate when you repeat those cycles hundreds of times.

Kai: So, they also looked at the energy decomposition of the system, finding that both the total energy and crack energy relax exponentially with relaxation times around three point nine two cycles for the total energy and three point one seven cycles for just the crack energy.

Mira: It’s neat that those two relaxation times are nearly equal to each other, which suggests they are essentially describing the same underlying physical process in this system.

Lev: So, if you're building something, you can think of it like this: the energy doesn't just dissipate; it’s being channeled into making the effective crack shorter over and over.

Conclusion: Kai: So to wrap up on "A Phase-Field Study of Desiccation Crack Pattern Maturation under Drying-Wetting Cycles," the authors show that repeated drying and wetting cycles cause cracks to mature from a pattern dominated by ninety degree intersections toward one hundred twenty degrees.

Mira: They used their phase-field model to numerically reproduce this experimentally observed maturation, showing how T-shaped intersections persist because previous crack paths take priority for subsequent cracking.

Lev: The implication for real hardware is that the energy relaxation is tied directly to the effective crack length decreasing exponentially, which has a characteristic relaxation time of two point five one cycles.

Kai: So essentially, they've linked the angle relaxation time of two point eight five cycles to the energy relaxation time of three point nine two cycles, suggesting it’s one unified process.

Mira: The main caveat the authors mention is that their artificial discontinuity between the first and second cycles suggests that they need to refine how they model those scar effects if they want perfect agreement with experimental situations.

Lev: If you're looking at this for real hardware, you should focus on the modeling of those scar effects because that’s where the main source of discrepancy lies right now.

Hiroki Fukushima, Satoshi Yukawa

Department of Earth and Space Science, Graduate School of Science, The University of Osaka

cond-mat.soft, nlin.PS

Submitted: 2026-09-10

Updated: 2026-11-04

Comments: 38 pages, 10 figures

Project page: http://www.mcs.anl.gov/petsc

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 77/100

The gist: A phase-field model incorporating crack healing and scar effects reveals that the characteristic intersection angle of desiccation cracks relaxes from near 90° toward 120° under repeated…

Key concepts

Phase-Field Model
This is a mathematical tool used to simulate how materials change over time, specifically focusing on crack formation and healing. It describes the material's state using a 'damage field' that evolves based on energy minimization principles, allowing researchers to model complex fracture behaviors like crack propagation and interaction.
Crack Healing
This concept models the ability of a material to repair itself after cracking during repeated drying-wetting cycles. In this study, it is incorporated by using a memory field that records past cracks, which reduces the effective fracture toughness of the material in subsequent cycles, simulating how prior damage affects future cracking.
Angle Relaxation Dynamics
This refers to how the shape and orientation of crack intersections evolve during cycling. The model showed that the standard deviation of intersection angles relaxes exponentially, moving from a distribution centered near 90° to one centered around 120°, indicating that the crack pattern matures from T-shaped to Y-shaped within a few cycles.
Effective Crack Length
This represents the actual length of a crack as it evolves under repeated stress. The study found that this effective length decays exponentially with each cycle, and its shortening is what drives the relaxation of the total energy, linking crack energy changes directly to the observed angle maturation.

Terminology

Summary

A phase-field model incorporating crack healing and scar effects reveals that the characteristic intersection angle of desiccation cracks relaxes from near 90° toward 120° under repeated drying–wetting cycles The standard deviation of the intersection angle from 120° relaxes exponentially with a characteristic time of about 2. p. 185 cycles, consistent with experiments Crack energy dominates the total energy and also relaxes exponentially with nearly the same characteristic cycle as the angle relaxation, driven mainly by a shortening of the effective crack length rather than a change in effective fracture toughness. p. 1

Model Formulation

The study introduces a shrinkage-fracture model that combines elasticity theory with a phase-field model, extending previous work by adding crack healing and scar effects The energy functional considers elastic energy of the paste, crack energy, and interaction energy with the substrate. The elastic energy density is modified by a degradation function g(d) to account for weakening due to cracks. p. 5 Crack generation is modeled using a phase-field approach where crack energy is expressed via the damage field d as Ψcr[d] = Z ∫omega ψcr(∇d, d) dV. p. 5

Cycle Simulation and Maturation

The numerical calculation repeats the drying and wetting cycle by updating material properties using a memory field S(x) which represents scars left by past cracks. The effective fracture toughness Geffc(x) and critical energy density ψeffc(x) are reduced in proportion to S(x), modeling the weakening effect of previous cracks. p. 8 The process is formulated as a cycle where the memory field S is constructed from the final damage field d, and these effective values are used in the next drying process. p. 9

Angle Relaxation Dynamics

The angle distribution clearly differs between the first cycle and subsequent cycles, with an increase in frequency near 120° ∼ 140° indicating an increase in Y-shaped intersections. The standard deviation of the angle θ from 120°, σ(c), decays substantially at first and subsequently settles to a constant value, indicating that the crack intersection angle is substantially reorganized early on and matures from T-shaped to Y-shaped within a few cycles. p. 1 The relaxation time τθ associated with the angle is found to be 2. p. 185 ± 1.5 cycles, which is consistent with experimental observations. p. 3

Energy Relaxation Analysis

The total energy of the system, Etot, and the crack energy Ecr relax exponentially as the cycle proceeds, with nearly equal relaxation times. The relaxation time τE for both Etot and Ecr is found to be approximately 3. p. 192 ± 0.5 cycles and 3. p. 017 ± 0.12 cycles, respectively. p. 0 This energy-based maturation time scale is essentially equal to the angle relaxation time τθ. p. 1 The crack energy relaxation is attributable to the effective crack length shortening as the cycle proceeds. p. 1

Conclusion and Implications

The model reproduces the experimentally observed maturation of the crack pattern, showing that some T-shaped intersections persist because previous crack paths have priority for subsequent cracking. The energy relaxation is dominated by the crack energy term, and its shortening of the effective crack length is geometrically consistent with the increasing frequency of cracks near 120°. p. 1 The artificial discontinuity between the first and second cycles suggests that further refinement of scar modeling is necessary to fully agree with experimental situations. p. 1 The relaxation time for both angle and energy relaxation are essentially equal, indicating they are the same process. p. 1

Appendix Details

The degradation function g(d) is derived from the principle of least action, leading to the condition g'(0) = -Gc/c0l0ψc. A Lorentz-type degradation function is employed for numerical stability, satisfying g(1) = η where η is a small residual parameter. p. 5 Random fields Gc(x) and ψc(x) are generated using a log-normal distribution with specified mean and relative standard deviations. p. 8 Spatial correlation is introduced by applying a smoothing operation S[f(xi,j)] to generate the final heterogeneous fields. p. 8 The crack intersection angle θij between two edges i and j is determined by θij = cos−1(ti · tj). p. 1 The numerical solution employs an alternating iteration scheme with a Newton's method solver for the displacement field and a variational inequality Newton's method solver for the damage field.

Numerical Implementation

The finite element discretization uses four-node bilinear quadrilateral elements to approximate the domain Lx×Ly. The volume integral over each element is approximated by a 2 × 2 Gaussian quadrature formula. p. 4 The weak form integrals for the displacement field and damage field are derived using shape functions and Gaussian quadrature, resulting in residual vectors R(u) with dimension 2NxNy and R(d) with dimension NxNy. p. 9 The tangent stiffness matrix K(u) is constructed by differentiating the residual vector with respect to the displacement field components. p. 4 The constraint max(d n+1, d n) = d n+1 is imposed at each point to guarantee the irreversibility of crack generation. p. 7

Angle Measurement Algorithm

Crack intersection angles are measured by treating a snapshot of the damage field as an image and applying a Gaussian filter followed by Guo–Hall thinning. Nodes are detected as pixels whose 8-neighborhood contains three or more crack pixels, and the angle θij is calculated from the direction vectors ti = (qm−1 − q0) / (qm−1 − q0). p. 2 The intersection angle θij between edge i and j is straightforwardly determined by θij = cos−1(ti · tj). p. 1 The method constructs a representative outward vector without pixels near the node to mitigate errors from clustered pixels.

Parameter Summary

The simulation parameters include Young’s modulus E = 4 MPa, Poisson’s ratio ν = 0.2, and regularization lengthl0 = 0. p. 65 mm. The maximum shrinkage strain εmaxsh is set to 0. p. 705, and the system size Lx and Ly are both 100 mm. p. 3 The correlation coefficient of the random field ρ is set to 0. p. 85, with smoothing repetitions Ns = 64. p. 8 The weakening coefficients aG = 0. p. 515 and aP = 0.35 are used for fracture toughness and critical energy density reduction, respectively. p. 8 The number of loading steps per cycle Nstep is set to 1000, with the shrinkage strain increment ∆εsh = εmaxsh / Nstep = 5 × 10−5. p. 7 The discretization numbers Nx and Ny are both 128. The proportionality constant of frictional force κ is set to 0. p. 64 N/mm3. The mean fracture toughness Gc is 8 × 10−4 mJ/mm2, and the mean critical energy density ψc is 3 × 10−5 mJ/mm2. p. 8 The relative standard deviation of fracture toughness c(0) is set to c(0)Gc = q Var(G(0)c)/Gc, and the relative standard deviation of the critical energy density c(0)ψc is set to c(0)ψc = q Var(ψ(0)c)/ψc. p. 8 The correlation between G and ψ is Cov(G, ψ) = G ψ (e ρσGcσψ − 1). p. 8 The mean fracture toughness Gc is defined as 1/N Σ X i,j G(0)c(xi,j), and the mean critical energy density ψc is defined as 1/N Σ X i,j ψ(0)c(xi,j). p. 8 The spatial correlation length ξ grows with the number of steps n as ξ ∼ √n. p. 8 The smoothing operation S[f(xi,j)] = 1/2 f(xi,j) + 1/8 f(xi+1,j) + f(xi-1,j) + f(xi,j+1) + f(xi,j-1). p. 8 The parameters aG and aP are dimensionless parameters that adjust the degree to which the fracture toughness and critical energy density are reduced by the scar. p. 8 The mean fracture toughness Gc is defined as 1/hLcr Z Σomega 3Geffc(x) 8l0 d(x) + l20 ∇d(x)2. p. 1 The effective crack length Lcr is defined as Lcr = 3/8l0h ZΣomega d(x) + l20 ∇d(x)2. p. 7 The decomposition of Ecr is Ecr = GeffcΓhLcr. The effective crack length Lcr decays exponentially, and its characteristic relaxation time is slightly smaller than the energy-relaxation timescale. p. 1

Energy Decomposition

The total energy decomposes into elastic energy Eel, crack energy Ecr, and interaction energy Esb. The total energy Etot changes substantially from the first to the second cycle and then gradually approaches a steady-state value. p. 1 Both Etot and Ecr relax with a relaxation time of a few cycles, which is nearly the same as τθ associated with the angle. p. 1 The main contribution to Etot is the crack-energy term. p. 6

Effective Toughness and Length

The effective crack length Lcr decays exponentially, and its characteristic relaxation time is slightly smaller than the energy-relaxation timescale. The average effective fracture toughness ⟨Geffc⟩Γ is essentially constant, and changes in the effective crack length Lcr affect the crack energy. p. 1 The effective crack length also decreases monotonically with the cycle, and fitting this change with an exponential function gives a relaxation time of 2. p. 151 ± 0.1 cycles.

Summary and Discussion

The initial distribution of crack intersection angles peaks near 90° and shows a broad peak in the range 160° ∼ 180°, while the frequency near 120° increases as the drying–wetting cycle proceeds. The model reproduces this behavior, suggesting that some T-shaped intersections persist because previous crack paths have priority for subsequent cracking. p. 1 The relaxation time τθ = 2. p. 185 ± 1.5 cycles is consistent with experimental observations, and the energy relaxation time τE = 3. p. 192 ± 0.5 cycles is nearly the same. p. 1 This suggests that crack energy relaxation and angle relaxation are essentially the same process. p. 3 The model successfully captures the maturation of the crack pattern under drying–wetting cycles, although an artificial discontinuity between the first and second cycles needs improvement. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Phase field modeling for crack maturation allows AI to predict how crack intersection angles evolve under repeated drying-wetting cycles, specifically predicting a relaxation time of about 2.85 cycles for the standard deviation of the angle from 120 degrees.

  2. The model can predict that while the peak near 90 degrees persists, T-shaped intersections relax and transform into Y-shaped intersections as the cycle number increases, providing a statistical prediction of crack pattern evolution.

  3. AI systems can estimate energy relaxation times for crack maturation, finding that both total energy and crack energy relax exponentially with nearly equal characteristic cycle times as the angle relaxation time.

  4. The model enables prediction of effective fracture toughness reduction, showing that the effective crack length shortens exponentially and providing a relaxation time of 2.51 cycles for this parameter based on experimental data comparison.

  5. AI can quantify the effect of scar generation by modeling how past cracks induce a weakening effect on the next drying process arising from the memory field S, reducing fracture toughness Geff c by a factor dependent on parameters ag and aP.

Sources

Related papers