betaFeedback

WorksPhysics and Astronomy

A geometrically inspired constitutive framework for damage and intrinsic self-healing of elastomers

Abstract

Autonomic interfacial self-healing in elastomers enables their reprocessing and recycling, making them valuable for applications such as ballistic resistance, surface coatings, adhesives, and biomedical materials. This article prescribes a geometry-based damage-healing theory for autonomic healing in elastomers, built on a framework where damage induces an incompatibility in the Euclidean material manifold, transforming it into a Riemannian manifold. Healing restores the Euclidean state through a reversing damage variable or an evolving healing variable. The reversing damage variable models the rebonding mechanism while the healing variable accounts for healing by chain diffusion and entanglement. The model also predicts healing where rebonding is preceded by chain diffusion. The microforce balance governs the evolution of the damage and healing variables, capturing rate-dependent damage and intrinsic temperature-independent healing. The model is validated through numerical simulations, including one-dimensional and two-dimensional simulations, demonstrating accurate predictions of coupling between damage and healing and post-healing mechanical response. The model also predicts the recovery of fracture toughness with healing time in supramolecular elastomers, aligning with experimental data. With minimal parameters, the model is versatile and can easily be used for material design and structural analysis, surpassing existing models in simplicity and predictive capability.

Introduction

Self-healing behaviour is introduced into elastomers so that they can be reprocessed and recycled. The autonomic interfacial self-healing observed in a wide range of elastomers is particularly intriguing to both industry and the scientific community. Elastomers with this capability are used for ballistic resistance, tires, surface coatings, adhesives, and biomedical applications such as artificial skins, tissue engineering, etc. [43] Autonomic or stimulus-independent interfacial self-healing is a type of intrinsic self-healing that starts on its own when broken pieces touch each other without any outside help. Understanding and predicting this phenomenon is necessary for the development of efficient self-healing systems, mimicking tissue healing and having proper control on the healing parameters. While experiments are crucial, they are prohibitively expensive, fail if there are multiple healing mechanisms, and have limited capabilities for estimating healing efficiencies, determining the mobility of polymeric chains, etc. [15] To help scientists figure out how to heal things and other factors that might affect the connection between damage and healing, damage-healing constitutive models are used along with experiments. They impart information on the recovery of fracture and elastic properties with changes in healing time and other environmental conditions. This, in turn, can aid in the design and study of new self-healing concepts, as well as the development of self-healing elastomers that exhibit more attractive mechanical behaviour under various stress conditions, thereby enabling engineers to predict performance and longevity more accurately. By integrating these models into computational frameworks, researchers can simulate real-world scenarios and optimise the material design process further. [27] Furthermore, damage-healing models can give useful information with few but measurable macroscopic properties. They can be used in computer models to look at how structures made with these materials fail and how easily they can be recycled.

Chain diffusion and entanglement [40] or the rebonding of broken chains [5, 8, 20, 28] are the main processes that make autonomic interfacial healing happen. Rebonding is a chemical process, while chain diffusion is a physical process. It can happen through covalent interactions, such as the Diels–Alder reaction products, disulphide and diselenide bridging, or non-covalent interactions, such as hydrogen bonds, ionic or hydrophobic interactions, and so on. There may be different rebonding mechanisms depending on the type of reversible bonds available. Supramolecular elastomers, which are used in surface coatings and ballistic resistance [11, 15], heal by rebonding. The dynamic bonds in these materials are hydrogen bonds that can be broken reversibly. Rebonding is also observed to be preceded by the diffusion of chains in some elastomers [2].

Irrespective of the type of mechanism, autonomic healing has strong temporal and spatial dependence [32]. As the healing time increases [11, 15], chains that are diffusing across the interface move farther, and there are more re-association events. The extent of healing is also dependent on the time gap between the damage and the healing in the case of re-bonding since, with time, the number of potential connections decreases. Additionally, healing depends on the localisation and scale of the damage, whether it is a superficial, wide scratch or a deep crack. Wider damage takes a longer time to heal, according to [2], especially when the mechanism is chain diffusion, as chains take a longer time to travel. The extent of healing has a direct influence on the post-healed damage response of these materials. In general, an increase in healing time increases the stiffness and the strain of fracture of the healed material [12], but that may vary with the mechanism. For instance, in [11], the healed elastic response of the supramolecular elastomer is only dependent on the waiting time before healing commences and is completely independent of the healing time. The strain on the fracture, however, increases with healing time.

Experiments have been, so far, the primary means of understanding healing mechanisms, designing newer self-healing polymers, and predicting healing in actual engineering systems. Only a few attempts have been made to model the phenomena. Submacroscopic models, particularly molecular dynamics simulations of interfacial healing and surface welding [9, 31, 39, 44], provide only atomistic scale insights into rebonding dynamics, timescales, influence cross-linking specific to disulphide bonds or DA reactions, relaxation timescales, etc. However, they cannot predict responses to actual fracture and healing phenomena in engineering problems. Coupling these models with actual structural analysis or finite element simulations of failure analysis is infeasible, as the resulting multiscale models require a huge computational effort. These models don’t offer any understanding of post-healed stiffness and fracture properties either, which could be readily used in solving these larger-scale engineering problems. However, there are microscopic models by [37, 38] that determine some scaling laws for post-healing fracture toughness and stiffness, but they are limited to healing mechanisms involving diffusion and entanglement only, with no rebonding. The scaling laws, however, are physical, as established by the experiments of [15]. Also, they have found use in subsequent theoretical studies on the dependence of healing time on post-healed mechanical response, but not in real-life engineering damage-healing problems. Mesoscale and multiscale models that bridge submacroscopic scales may be more efficient than molecular dynamics simulations but in no way superior to continuum scale models [30] in terms of simplicity, computational efficiency, and the number of material parameters. Most continuum scale models predict damage and healing within a single framework, considering that the underlying mechanisms are strongly coupled. Such coupled damage-healing prescriptions find maximum use in the analysis of real-life engineering structures as well as in material design.

Although researchers have sufficiently explored the microscopic and molecular dynamics modelling of the healing process, there are very few continuum scale models available. The elastoplastic mesoscopically informed continuum scale models of [41, 42] predict interfacial healing in general dynamic polymeric networks, taking into account diffusion, entanglement, and rebonding mechanisms within a single framework. Being comprehensive, the model is complex, requiring a large number of material parameters that can be determined either through mesoscopic experimental investigations or rigorous and time-consuming curve fitting similar to [13]. This makes them unsuitable for both material design and implementation when analysing actual damage-healing experiments.

The continuum Damage and Healing Mechanics (CDHM) [4, 22] framework has been widely used to prescribe damage-healing models for a large range of materials. The CDHM framework employs evolving internal variables to measure the damage and healing processes. Typically, the damage variable is defined as the ratio of the damaged to the undamaged area [19, 24], while the healing variable is the fraction of the damaged area that has been recovered. The damage variable, scaled by the healing variable, defines the effective damage variable. Instead of area or volume, the variables may also be defined in terms of mechanical or electrical properties of the damaged and the undamaged material. Multiple healing and damage mechanisms can be defined by separate internal variables with independent evolution equations. The framework itself does not pose any restrictions on the evolution equations. For restrictions of thermodynamic consistency, the rational thermodynamics framework [10] can be invoked where constitutive forms can also be determined from the Coleman-Noll procedure. In addition to the definition of the internal variables, CDHM invokes the concept of scaling the stress by a factor dependent on the effective damage variable. One of the biggest advantages of the rational thermodynamics-informed CDHM framework is that it can be used to formulate local and non-local damage-healing models of any material that can account for multiple driving mechanisms, both extrinsic and intrinsic. [1, 12, 21, 26, 29, 33, 34, 35, 36] prescribe damage-healing models for polymers, polymer blends, polymers reinforced with shape memory polymers, etc. They predict damage and healing with sufficient accuracy but are parameter-intensive and complex, requiring chain-level material parameters. While the evolution equations in some of these models are phenomenological with no relevance to the actual underlying physics, there are some models with no length-scale parameters to model diffusive processes.

Local damage-healing models fail to predict diffused damage and the influence of the damage width on the rate of healing, which is extremely crucial in the context of healing of scratches and superficial damage [2]. Also, for numerical stability of algorithms, damage evolution equations must have non-local terms [14]. [17, 18] prescribe phase-field-inspired damage-healing models for microcrack-driven fracture in PDMS. The ordinary phase-field model with quadratic shape functions [14], is not appropriate for elastomers that undergo long stretches prior to brutal damage accurately. Moreover, the mathematically motivated surface energy prescription fails to reflect the timescales associated with the healing process. To alleviate these typical drawbacks of a regularised continuum gradient-based damage model, a geometrically inspired damage model is described in [14]. Damage in the form of polymeric chain scission changes the manifold from Euclidean to Riemannian, which leads to an incompatibility that can be measured by a Ricci curvature. In line with [19, 24], a scaling of the metric is proposed such that the damage variable may physically represent the ratio of damaged and undamaged areas. We use the Ricci curvature to define the resistive part of the free energy density, the variation of which determines the evolution of the damage variable. This model is numerically stable and provides a much more accurate prediction for brutal damage compared to a quadratic phase-field damage model.

In this article, we propose a geometrically inspired non-local, coupled damage-healing model that predicts damage and intrinsic self-healing in elastomers. It is built on the geometric framework [14] and borrows concepts from Continuum Damage Healing Mechanics (CDHM) [4]. The current theory, in addition to predicting rate-dependent damage, reflects the recovery of the damaged area through an independently evolving healing variable and a reversing damage variable. The healing variable represents the area recovered due to entanglement and diffusion of chains, while the reversing damage variable, which has its own timescale, models healing due to bond reversal. The Ricci curvature defines the resistive component in the free energy expression, which governs the evolution of both the healing and damage variables. The microforce balance laws [16] yield strongly coupled, non-local evolution equations for the damage and healing variables from which models for rate-dependent damage and healing models for chain diffusion and rebonding may be derived. The rational thermodynamic framework of [10] is used to determine the constitutive restrictions on the stresses and microforces.

The damage and the healing models are implemented in single material points, one-dimensional and two-dimensional simulations of single-cycle or multicycle fracture and healing experiments to investigate the influence of healing time, initial damage width, and healing timescales on the post-healing behaviour and fracture properties of the elastomer. Further, the model is used to investigate the recovery of a supramolecular elastomer’s fracture toughness with healing time from the experiments of [11].

The first section of the article II describes the kinematics, defining the different configurations. Next is the III section, which finds the damaged-healed material’s free energy. This is followed by the development of the macroforce and microforce balances in IV. The constitutive prescriptions for the macro and micro stresses are derived in V. This is followed by the section VI, which describes the numerical studies carried out using the model. Finally, the article is concluded in VII. The article also includes an appendix IX where a relationship between Griffith’s fracture toughness GcG_c and the resistive modulus MM is derived for a pure damage case.

Kinematics of the Damaged/healed Undeformed and Deformed Riemannian Manifold

[14] prescribes a geometrically inspired damage model for elastomers that shows better convergence, which is a measure of the reduction in area that can effectively contribute to the material’s stiffness. This change in area is non-conservative and may be interpreted to change the Euclidean manifold to a Riemannian manifold, as schematically shown in 1. We define three material configurations the referential or the undamaged, undeformed configuration B\mathcal{B}, the intermediate fictitious effective damaged-self-healed and undeformed configuration Cdh\mathcal{C}_{dh} and the deformed damage configuration S\mathcal{S}. As in case of [14], the configuration B\mathcal{B} is Euclidean, while the configurations Cdh\mathcal{C}_{dh} and S\mathcal{S} are Riemannian manifolds. The system’s kinematics reflects the physical change in the area primarily through a simple metric scaling. The scaling is a function of the ratio of the effective area to the undamaged area DhD_h which is a function of the damage variable DD and the healing variable hh. DD as in [14] denotes the ratio of the number of broken polymeric chains to the intact chains in a spherical REV, on the other hand, hh is the ratio of the number of recovered bonds to the number of severed bonds. It may also measure the healing process by other mechanisms, such as the increase in the number of entanglements or cross-links, which can add stiffness to the material. The effective damaged area is used to scale the metric of the undeformed, undamaged configuration, which is Euclidean, to yield the metric of the deformed, damaged manifold. The incompatibility associated with the Riemannian manifold arises due to the damage and healing. The healing contributes to the incompatibility only if the process entails some permanent change to the configuration other than the reversal of a bond-breaking process.

ϕ\phi the isomorphic map from the reference to the deformed, effective configuration is a composition of two isomorphic maps ϕdh\phi_{dh} and ϕM\phi_M defined by,

ϕdh:B→Cdh(1)\phi_{dh}:\mathcal{B}\to\mathcal{C}_{dh} \tag*{(1)}
ϕM:Cdh→S(2)\phi_M:\mathcal{C}_{dh}\to\mathcal{S} \tag*{(2)}
Schematic figure of the three configurations and the deformation maps

Figure 1. Schematic figure of the three configurations and the deformation maps

We also define the tangent maps or the deformation gradients Fdh\mathbf{F}_{dh} and FM\mathbf{F}_{M} respectively.

F=FMFdh(3)\mathbf{F}=\mathbf{F}_{M}\mathbf{F}_{dh} \tag*{(3)}

Additionally, we assume ϕdh\phi_{dh} to be an identity map; thus, the associated deformation gradient Fdh\mathbf{F}_{dh} mapping the respective tangent spaces is identity. Thus the equation (3) simplifies to

F=FM(4)\mathbf{F}=\mathbf{F}_{M} \tag*{(4)}

The corresponding inverse map ϕ−1\phi^{-1}, which maps the damaged, deformed configuration to the undamaged, undeformed configuration, is again a composition of ϕdh−1\phi_{dh}^{-1} and ϕM−1\phi_{M}^{-1} with the following definition

ϕdh−1:Cdh→B(5)\phi_{dh}^{-1}: C_{dh}\to B \tag*{(5)}
ϕM−1:S→Cdh(6)\phi_{M}^{-1}: S\to C_{dh} \tag*{(6)}

Associated with the manifolds are the respective metric tensors G\mathbf{G}, Gdh\mathbf{G}_{dh} and gdh\mathbf{g}_{dh}. The metric tensor associated with undamaged Euclidean configuration is G\mathbf{G} those with undeformed and deformed Riemannian manifold is denoted by Gdh\mathbf{G}_{dh} and gdh\mathbf{g}_{dh} respectively, The undeformed and deformed Riemannian metrics can be expressed as functions of the undeformed and deformed damaged configurations as follows,

Gdh=f(D,h)G(7)\mathbf{G}_{dh}=f(D,h)\mathbf{G} \tag*{(7)}
gdh=f(D,h)g(8)\mathbf{g}_{dh}=f(D,h)\mathbf{g} \tag*{(8)}

where G\mathbf{G} and g\mathbf{g} are Euclidean metrics of the undamaged, undeformed and undamaged, deformed configurations. f(D,h)f(D,h) is a scaling factor, a measure of the ratio of the effective damaged area to the area before the damage. In line with [12], the scaling factor is chosen as f(D,h)=(1−D(1−h))f(D,h)=(1-D(1-h)). Pulled back to the undeformed undamaged configuration, one can write the right Cauchy Green tensor associated with the deformation as

Cdh=f(D,h)FTgF(9)\mathbf{C}_{dh}=f(D,h)\mathbf{F}^{T}\mathbf{g}\mathbf{F} \tag*{(9)}

where we make use of (4).

Using Miehe’s microsphere model Miehe et al. [23], we calculate the homogenised stretches, area and volume measures in the various configurations that appear in the free energy. In the effective configuration, the homogenised stretch is given as,

Λdh02=(1−D(1−h))(10)\Lambda_{dh0}^{2}= (1-D(1-h)) \tag*{(10)}

While in the deformed configuration, the homogenised stretch is given by,

Λdh2=1−D(1−h)3tr⁡(C)(11)\Lambda_{dh}^{2}=\frac{1-D(1-h)}{3}\operatorname{tr}(\mathbf{C}) \tag*{(11)}

where C=FTgF\mathbf{C}=\mathbf{F}^{T}\mathbf{g}\mathbf{F}

The infinitesimal area measures for the deformed configuration dAdhdA_{dh} are given by,

dAdh=(1−D(1−h))dA0(12)dA_{dh}=(1-D(1-h))dA_{0} \tag*{(12)}

The volume measure in the effective configuration,

Jdh0=(1−D(1−h))3/2(13)J_{dh0}=(1-D(1-h))^{3/2} \tag*{(13)}

and for the deformed damaged manifold, it is given by,

Jdh=(1−D(1−h))3/2det⁡(FTF)(14)J_{dh}=(1-D(1-h))^{3/2}\sqrt{\det(\mathbf{F}^{T}\mathbf{F})} \tag*{(14)}

The curvature associated with the damaged undeformed manifold is of particular relevance and hence is determined in section III C.

Geometry inspired free energy for damaged solid

Free Energy of an undamaged hyperelastic solid

For compressible solids, the free energy takes on the following form,

ψ=ψ^(C,θ)=kBnNθ[ΛˉrL−1(Λˉr)+ln⁡L−1(Λˉr)sinh⁡(L−1(Λˉr))−f(Λˉref)]+K2(J−Jref)2(15)\psi=\hat{\psi}(\mathbf{C},\theta)=k_{B}nN\theta\left[\bar{\Lambda}_{r}\mathcal{L}^{-1}(\bar{\Lambda}_{r})+\ln\frac{\mathcal{L}^{-1}(\bar{\Lambda}_{r})}{\sinh(\mathcal{L}^{-1}(\bar{\Lambda}_{r}))}-f(\bar{\Lambda}_{\mathrm{ref}})\right]+\frac{K}{2}(J-J_{\mathrm{ref}})^{2} \tag*{(15)}

where, Λˉr=ΛˉN\bar{\Lambda}_{r}=\frac{\bar{\Lambda}}{N}, Λˉ=J−1/313tr⁡(C)1/2\bar{\Lambda}=J^{-1/3}\frac{1}{\sqrt{3}}\operatorname{tr}(\mathbf{C})^{1/2}, J=det⁡gdet⁡Gdet⁡FJ=\sqrt{\frac{\det\mathbf{g}}{\det\mathbf{G}}}\det\mathbf{F}, Jref=1J_{\mathrm{ref}}=1 and Λˉrref\bar{\Lambda}_{r\mathrm{ref}} is deviatoric reference strain.

For smaller strains, the cumbersome inverse-Langevin form can be replaced with the Neo-Hookean form and the final energy can be of the form,

ψ=ψ^(C,θ)=kBnNθ(Λˉr2−Λˉrref2)+K2(J−Jref)2(16)\psi=\hat{\psi}(\mathbf{C},\theta)=k_{B}nN\theta\left(\bar{\Lambda}_{r}^{2}-\bar{\Lambda}_{r\mathrm{ref}}^{2}\right)+\frac{K}{2}(J-J_{\mathrm{ref}})^{2} \tag*{(16)}

Modification in strain energy

Thus, for compressible solids, the strain energy due to the conformal stretching of polymer chains is specified by,

ψe=ψ^e(C,θ)=kBndNθ(Λˉrdh2−Λˉrdh02)+K2(Jdh0.5−Jdh00.5)2(17)\psi_{e}=\hat{\psi}_{e}(\mathbf{C},\theta)=k_{B}n_{d}N\theta\left(\bar{\Lambda}_{rdh}^{2}-\bar{\Lambda}_{rdh0}^{2}\right)+\frac{K}{2}\left(J_{dh}^{0.5}-J_{dh0}^{0.5}\right)^{2} \tag*{(17)}

where Λˉrdh=ΛˉdhN\bar{\Lambda}_{rdh}=\frac{\bar{\Lambda}_{dh}}{\sqrt{N}}, Λˉdh=J−1/313tr⁡(Cdh)1/2=J−1/313tr⁡(C)1/2\bar{\Lambda}_{dh}=J^{-1/3}\frac{1}{\sqrt{3}}\operatorname{tr}(\mathbf{C}_{dh})^{1/2}=J^{-1/3}\frac{1}{\sqrt{3}}\operatorname{tr}(\mathbf{C})^{1/2}, and J=det⁡FTgFJ=\sqrt{\det\mathbf{F}^{T}\mathbf{g}\mathbf{F}}. ndn_{d} is the number density of chains in the deformed damaged configuration expressed in material coordinates. Hence, nd=n(1−D)3n_{d}=n(1-D)^{3}.

Substituting the geometrically consistent stretches and Jacobian in (17) we obtain the following strain energy,

ψe=ψ^e(C,θ)=kBn(1−D(1−h))3/2Nθ(Λˉr2−1)+K2(1−D(1−h))3/2(J−1)2(18)\psi_{e}=\hat{\psi}_{e}(\mathbf{C},\theta)=k_{B}n(1-D(1-h))^{3/2}N\theta\left(\bar{\Lambda}_{r}^{2}-1\right)+\frac{K}{2}(1-D(1-h))^{3/2}(J-1)^{2} \tag*{(18)}

Curvature Based Resistive Energy

The total free energy in a damaged solid comprises of the elastic energy derived in the earlier section and a surface energy or a resistive energy which we motivate from the Ricci curvature that we determine in this section.

ψ^Rdh(Rdh,D)=MRdh(D,∇D,ΔD,H,∇H,ΔH)det⁡(Gdh)(19)\hat{\psi}_{Rdh}(R_{dh},D)=M R_{dh}(D,\nabla D,\Delta D,H,\nabla H,\Delta H)\sqrt{\det(G_{dh})} \tag*{(19)}

where Rdh(D,∇D,ΔD,H,∇H,ΔH)R_{dh}(D,\nabla D,\Delta D,H,\nabla H,\Delta H) is the Ricci curvature, and MM is a material parameter quantifying the resistance of the material to the creation of a non-trivial curvature or new surfaces. lRl_R is the width of the zone of non-trivial curvature. The Christoffel symbols associated with the Levi Civita connection for both manifolds may be derived from the compatibility condition associated with the undeformed, damaged Riemannian manifold given by,

∇αGd∣μν=0(20)\nabla_{\alpha}G_{d|\mu\nu}=0 \tag*{(20)}

The Christoffel symbols in the indicial notation are thus given by,

ΓdMIL=12GdKL[GdIK,M+GdKM,I−GdMI,K](21)\Gamma^{L}_{dMI}=\frac{1}{2}G_{d}^{KL}\left[G_{dIK,M}+G_{dKM,I}-G_{dMI,K}\right] \tag*{(21)}

The curvature tensor R\mathbf{R} can be calculated from the Christoffel symbols

RdIKJL=∂KΓdJIL−∂JΓdKIL+ΓdKALΓdJIA+ΓdJALΓdKIA(22)R^{L}_{dIKJ}=\partial_{K}\Gamma^{L}_{dJI}-\partial_{J}\Gamma^{L}_{dKI}+\Gamma^{L}_{dKA}\Gamma^{A}_{dJI}+\Gamma^{L}_{dJA}\Gamma^{A}_{dKI} \tag*{(22)}

The components of the Ricci tensor can be calculated from the curvature tensor as follows,

RdIKJK=∂KΓdJIK−∂JΓdKIK+ΓdKAKΓdJIA+ΓdJAKΓdKIA(23)R^{K}_{dIKJ}=\partial_{K}\Gamma^{K}_{dJI}-\partial_{J}\Gamma^{K}_{dKI}+\Gamma^{K}_{dKA}\Gamma^{A}_{dJI}+\Gamma^{K}_{dJA}\Gamma^{A}_{dKI} \tag*{(23)}

The Ricci scalar curvature, by definition, is a result of the contraction of the two lower indices in the Ricci tensor and is given by,

R=GdhIJRdIKJK(24)R=G_{dh}^{IJ}R^{K}_{dIKJ} \tag*{(24)}

With equations (21), (22), (23), and (24), the Ricci curvature results as,

Rdh(D,∇D,ΔD)=2(1−h)(1−D(1−h))ΔD+1.5(1−h)2∇D⋅∇D(1−D(1−h))3+−2D(1−D(1−h))Δh+1.5D2∇h⋅∇h−(4−D(1−h))∇D⋅∇h(1−D(1−h))3(25)R_{dh}(D,\nabla D,\Delta D)=\frac{2(1-h)(1-D(1-h))\Delta D+1.5(1-h)^2\nabla D\cdot\nabla D}{(1-D(1-h))^3} +\frac{-2D(1-D(1-h))\Delta h+1.5D^2\nabla h\cdot\nabla h-(4-D(1-h))\nabla D\cdot\nabla h}{(1-D(1-h))^3} \tag*{(25)}

The curvature is used to define the resistive energy ψRdh\psi_{Rdh} which must be positive definite and vanish when the material is fully damaged.

ψRdh=(MD2(1−h)2+12MlR2det⁡(Gdh)13Rdh)=13MD2(1−h)2+0.7513MlR2(1−h)2∇D⋅∇D+0.7513MlR2D2∇h⋅∇h(26)\begin{aligned} \psi_{Rdh}&=\left(MD^2(1-h)^2+\frac{1}{2}Ml_R^2\det(G_{dh})^{\frac{1}{3}}R_{dh}\right)\\ &=\frac{1}{\sqrt{3}}MD^2(1-h)^2+0.75\frac{1}{\sqrt{3}}Ml_R^2(1-h)^2\nabla D\cdot\nabla D+\\ &\quad0.75\frac{1}{\sqrt{3}}Ml_R^2D^2\nabla h\cdot\nabla h \tag*{(26)} \end{aligned}

The constant MM is the resistive modulus [14] that quantifies the resistance offered by the material to damage, and lRl_R is the width of the damaged area. For a case of pure damage, as described in the appendix, MM may be related to Griffith’s fracture toughness.

Modified Free Energy of the Damaged Deformed Solid

The total free energy is thus given by,

ψ=ψRdh+ψe=13MD2(1−h)2+0.7513MIR2(1−h)2∇D⋅∇D+0.7513MIR2D2∇h⋅∇h+kBn(1−D(1−h))3/2Nθ(Λ‾r2−1)+K2(1−D(1−h))3/2(J0.5−1)2(27)\psi= \psi_{Rdh} + \psi_e = \frac{1}{\sqrt{3}}MD^2(1-h)^2 + 0.75\frac{1}{\sqrt{3}}MI_R^2(1-h)^2\nabla D \cdot\nabla D + 0.75\frac{1}{\sqrt{3}}MI_R^2D^2\nabla h \cdot\nabla h + k_B n(1-D(1-h))^{3/2}N\theta\left(\overline{\Lambda}_r^2-1\right) + \frac{K}{2}(1-D(1-h))^{3/2}(J^{0.5}-1)^2 \tag*{(27)}

Macroforce and microforce balances

We use the principle of virtual power to derive the macroforce and microforce balance laws. The external power is defined as,

δWext=∫∂V0(t⋅δϕ˙+χδD˙+γδh˙)dA0+∫V0b⋅δϕ˙dV0(28)\delta W_{\mathrm{ext}}=\int_{\partial V_0}\left(\mathbf{t}\cdot\delta\dot{\boldsymbol{\phi}}+\chi\delta\dot{D}+\gamma\delta\dot{h}\right)dA_0+\int_{V_0}\mathbf{b}\cdot\delta\dot{\boldsymbol{\phi}}dV_0 \tag*{(28)}

and the internal power is given by,

δWint=∫V0P:δ∇Xϕ˙+πδD˙+βδh˙+ξ⋅δ∇D˙+ζδΔh˙+ω⋅δΔD˙+κδ∇h˙ dV0(29)\delta W_{\mathrm{int}}=\int_{V_0}\mathbf{P}:\delta\nabla_X\dot{\boldsymbol{\phi}}+\pi\delta\dot{D}+\beta\delta\dot{h}+\boldsymbol{\xi}\cdot\delta\nabla\dot{D}+\zeta\delta\Delta\dot{h}+\boldsymbol{\omega}\cdot\delta\Delta\dot{D}+\boldsymbol{\kappa}\delta\nabla\dot{h}\,dV_0 \tag*{(29)}

where π\pi and ξ\boldsymbol{\xi} are the driving forces dual to DD and ∇D\nabla D, respectively and γ\gamma and κ\boldsymbol{\kappa} are the driving forces dual to hh and ∇h\nabla h, respectively.

Equating the external and internal virtual powers and for arbitrary δϕ˙\delta\dot{\boldsymbol{\phi}}, δD˙\delta\dot{D} and δh˙\delta\dot{h}, the following balance laws and the boundary conditions result. First, the macro force balance is given by,

∇⋅P+b=0(30)\nabla\cdot\mathbf{P}+\mathbf{b}=0 \tag*{(30)}

The microforce balances for DD and hh take the form,

∇⋅ξ−π+∇⋅∇ω=0(31)\nabla\cdot\boldsymbol{\xi}-\pi+\nabla\cdot\nabla\boldsymbol{\omega}=0 \tag*{(31)}
∇⋅κ−β+∇⋅∇ζ=0(32)\nabla\cdot\boldsymbol{\kappa}-\beta+\nabla\cdot\nabla\zeta=0 \tag*{(32)}

The boundary conditions are given by,

ξ⋅n0=0at ∂V0=0κ⋅n0=0at ∂V0=0Pn0=tat ∂V0=0ζn0⋅nl=0at ∂2V0=0ωn0⋅nl=0at ∂2V0=0(33)\begin{aligned} \boldsymbol{\xi}\cdot\mathbf{n}_0&=0 &&\text{at }\partial V_0=0\\ \boldsymbol{\kappa}\cdot\mathbf{n}_0&=0 &&\text{at }\partial V_0=0\\ \mathbf{P}\mathbf{n}_0&=\mathbf{t} &&\text{at }\partial V_0=0\\ \zeta\mathbf{n}_0\cdot\mathbf{n}_l&=0 &&\text{at }\partial^2V_0=0\\ \boldsymbol{\omega}\mathbf{n}_0\cdot\mathbf{n}_l&=0 &&\text{at }\partial^2V_0=0 \tag*{(33)} \end{aligned}

Constitutive theory

The first and the second laws of thermodynamics may be stated as follows:

ψ˙+ηθ˙+η˙θ=P:F˙+πD˙+βh˙+ξ⋅∇D˙+κ⋅∇h˙+ω⋅ΔD˙+ζ⋅Δh˙−∇⋅q(34)\dot{\psi}+\eta\dot{\theta}+\dot{\eta}\theta=\mathbf{P}:\dot{\mathbf{F}}+\pi\dot{D}+\beta\dot{h}+\boldsymbol{\xi}\cdot\nabla\dot{D}+\boldsymbol{\kappa}\cdot\nabla\dot{h}+\boldsymbol{\omega}\cdot\Delta\dot{D}+\zeta\cdot\Delta\dot{h}-\nabla\cdot\mathbf{q} \tag*{(34)}
η˙+∇⋅qθ−q⋅∇θθ2≥0(35)\dot{\eta}+\frac{\nabla\cdot\mathbf{q}}{\theta}-\frac{\mathbf{q}\cdot\nabla\theta}{\theta^2}\geq0 \tag*{(35)}

Substituting the expression for ∇⋅q\nabla\cdot\mathbf{q} from Eqn. (34) in Eqn. (35), we obtain,

1θ(P:F˙+πD˙+βh˙+ξ⋅∇D˙+κ⋅∇h˙+ω⋅ΔD˙+ζ⋅Δh˙−ψ˙−ηθ˙)−q⋅∇θθ2≥0(36)\frac{1}{\theta}\left(\mathbf{P}:\dot{\mathbf{F}}+\pi\dot{D}+\beta\dot{h}+\boldsymbol{\xi}\cdot\nabla\dot{D}+\boldsymbol{\kappa}\cdot\nabla\dot{h}+\boldsymbol{\omega}\cdot\Delta\dot{D}+\zeta\cdot\Delta\dot{h}-\dot{\psi}-\eta\dot{\theta}\right)-\frac{\mathbf{q}\cdot\nabla\theta}{\theta^2}\geq0 \tag*{(36)}

While P\mathbf{P}, ζ\boldsymbol{\zeta}, ω\boldsymbol{\omega}, ξ\boldsymbol{\xi} and β\boldsymbol{\beta} are all energetic π\boldsymbol{\pi} and κ\boldsymbol{\kappa} are assumed to have both energetic and dissipative components; that is, π=πen+ηDD˙\boldsymbol{\pi}=\boldsymbol{\pi}_{\mathrm{en}}+\eta_D\dot{D}, and β=βen−ηhh˙\boldsymbol{\beta}=\boldsymbol{\beta}_{\mathrm{en}}-\eta_h\dot{h} where ηD\eta_D and ηh\eta_h are inverse of timescales associated with the evolution of damage and healing respectively. The constitutive form for the dissipative component ensures non-violation of the second law. The free energy and the thermodynamic forces are assumed to have the following constitutive dependencies:

ψ=ψ^(F,D,∇D,h,∇h,θ),P=P^(F,D,h,θ),ξ=ξ^(∇D,h),π=π^(h,D),κ=κ^(∇h,D),β=β^(h,D),ω=ω^(ΔD,h),ζ=ζ^(Δh,h)(37)\psi=\hat{\psi}(\mathbf{F},D,\nabla D,h,\nabla h,\theta),\quad\mathbf{P}=\hat{\mathbf{P}}(\mathbf{F},D,h,\theta),\quad\boldsymbol{\xi}=\hat{\boldsymbol{\xi}}(\nabla D,h), \\ \boldsymbol{\pi}=\hat{\boldsymbol{\pi}}(h,D),\quad\boldsymbol{\kappa}=\hat{\boldsymbol{\kappa}}(\nabla h,D),\quad\boldsymbol{\beta}=\hat{\boldsymbol{\beta}}(h,D),\quad\boldsymbol{\omega}=\hat{\boldsymbol{\omega}}(\Delta D,h),\quad\boldsymbol{\zeta}=\hat{\boldsymbol{\zeta}}(\Delta h,h) \tag*{(37)}

Using these relations, the second law may be rewritten as,

(P−∂ψ∂F):F˙+(πen−∂ψ∂D)D˙+(βen−∂ψ∂h)h˙+(ξ−∂ψ∂∇h)⋅∇h˙+(κ−∂ψ∂∇D)⋅∇D˙+(ζ−∂ψ∂Δh)⋅Δh˙+(ω−∂ψ∂ΔD)ΔD˙−(η+∂ψ∂θ)θ˙+ηD(D˙)2+ηh(h˙)2−q⋅∇θθ≥0(38)\left(\mathbf{P}-\frac{\partial\psi}{\partial\mathbf{F}}\right):\dot{\mathbf{F}}+\left(\boldsymbol{\pi}_{\mathrm{en}}-\frac{\partial\psi}{\partial D}\right)\dot{D}+\left(\boldsymbol{\beta}_{\mathrm{en}}-\frac{\partial\psi}{\partial h}\right)\dot{h}+\left(\boldsymbol{\xi}-\frac{\partial\psi}{\partial\nabla h}\right)\cdot\nabla\dot{h} \\ +\left(\boldsymbol{\kappa}-\frac{\partial\psi}{\partial\nabla D}\right)\cdot\dot{\nabla D}+\left(\boldsymbol{\zeta}-\frac{\partial\psi}{\partial\Delta h}\right)\cdot\dot{\Delta h}+\left(\boldsymbol{\omega}-\frac{\partial\psi}{\partial\Delta D}\right)\dot{\Delta D}-\left(\eta+\frac{\partial\psi}{\partial\theta}\right)\dot{\theta} \\ +\eta_D(\dot{D})^2+\eta_h(\dot{h})^2-\frac{\mathbf{q}\cdot\nabla\theta}{\theta}\geq0 \tag*{(38)}

The above statement of the second law gives rise to the following constitutive restrictions:

P=∂ψ∂F;ξ=∂ψ∂∇D;κ=∂ψ∂∇h;ζ=∂ψ∂Δh;ω=∂ψ∂ΔD;η=−∂ψ∂θ;πen=∂ψ∂D;βen=∂ψ∂h(39)\mathbf{P}=\frac{\partial\psi}{\partial\mathbf{F}};\quad\boldsymbol{\xi}=\frac{\partial\psi}{\partial\nabla D};\quad\boldsymbol{\kappa}=\frac{\partial\psi}{\partial\nabla h};\quad\boldsymbol{\zeta}=\frac{\partial\psi}{\partial\Delta h};\quad\boldsymbol{\omega}=\frac{\partial\psi}{\partial\Delta D};\quad\eta=-\frac{\partial\psi}{\partial\theta};\quad\boldsymbol{\pi}_{\mathrm{en}}=\frac{\partial\psi}{\partial D};\quad\boldsymbol{\beta}_{\mathrm{en}}=\frac{\partial\psi}{\partial h} \tag*{(39)}

The heat flux assumes a Fourier form q=kc∇θ\mathbf{q}=k_c\nabla\theta, where kck_c is a Fourier constant. The constitutive forms for ξ\boldsymbol{\xi}, β\boldsymbol{\beta}, κ\boldsymbol{\kappa} and π\boldsymbol{\pi} may be implemented in the microforce balances to yield the evolution equations for damage and healing.

∇⋅∂ψ∂∇D−∂ψ∂D−ηDD˙+∇⋅∇∂ψ∂ΔD=0(40)\nabla\cdot\frac{\partial\psi}{\partial\nabla D}-\frac{\partial\psi}{\partial D}-\eta_D\dot{D}+\nabla\cdot\nabla\frac{\partial\psi}{\partial\Delta D}=0 \tag*{(40)}
∇⋅∂ψ∂∇h−∂ψ∂h−ηhh˙+∇⋅∇∂ψ∂Δh=0(41)\nabla\cdot\frac{\partial\psi}{\partial\nabla h}-\frac{\partial\psi}{\partial h}-\eta_h\dot{h}+\nabla\cdot\nabla\frac{\partial\psi}{\partial\Delta h}=0 \tag*{(41)}

The explicit forms for P\mathbf{P}, ξ\boldsymbol{\xi}, κ\boldsymbol{\kappa} and π\boldsymbol{\pi} are given as follows:

P=μ(1−D(1−h))3/2J−2/3[F−tr(FTF)3F−T]+K4(1−D(1−h))3/2(J0.5−1)(1+tanh⁡p(J−1))J0.5F−T+K4(1−D(1−h))3/2p(J0.5−1)2(1−tanh⁡2p(J−1))JF−T(42)\mathbf{P}=\mu(1-D(1-h))^{3/2}J^{-2/3}\left[\mathbf{F}-\frac{tr(\mathbf{F}^{T}\mathbf{F})}{3}\mathbf{F}^{-T}\right] \\ +\frac{K}{4}(1-D(1-h))^{3/2}(J^{0.5}-1)(1+\tanh p(J-1))J^{0.5}\mathbf{F}^{-T} \\ +\frac{K}{4}(1-D(1-h))^{3/2}p(J^{0.5}-1)^2(1-\tanh^2p(J-1))J\mathbf{F}^{-T} \tag*{(42)}
ξ=1.513MlR2(1−h)2∇D(43)\boldsymbol{\xi}=1.5\frac{1}{\sqrt{3}}Ml_R^2(1-h)^2\nabla D \tag*{(43)}
κ=1.513MlR2D2∇h(44)\boldsymbol{\kappa}=1.5\frac{1}{\sqrt{3}}Ml_R^2D^2\nabla h \tag*{(44)}
πen=−32μ(1−D(1−h))1/2(1−h)(Λˉr2−1)−3K4(1−D(1−h))2/3(1−h)(J0.5−1)2+1.513MlR2D∇h⋅∇h+213MD(1−h)2(45)\boldsymbol{\pi}_{\mathrm{en}}=-\frac{3}{2}\mu(1-D(1-h))^{1/2}(1-h)\left(\bar{\Lambda}_r^2-1\right) \\ -\frac{3K}{4}(1-D(1-h))^{2/3}(1-h)(J^{0.5}-1)^2+1.5\frac{1}{\sqrt{3}}Ml_R^2D\nabla h\cdot\nabla h \\ +2\frac{1}{\sqrt{3}}MD(1-h)^2 \tag*{(45)}
βen=32μ(1−D(1−h))1/2D(Λˉr2−1)+3K4(1−D(1−h))1/2D(J0.5−1)2−1.513MlR2(1−h)∇D⋅∇D−2M13D2(1−h)(46)\boldsymbol{\beta}_{\mathrm{en}}=\frac{3}{2}\mu(1-D(1-h))^{1/2}D\left(\bar{\Lambda}_r^2-1\right) \\ +\frac{3K}{4}(1-D(1-h))^{1/2}D(J^{0.5}-1)^2-1.5\frac{1}{\sqrt{3}}Ml_R^2(1-h)\nabla D\cdot\nabla D \\ -2M\frac{1}{\sqrt{3}}D^2(1-h) \tag*{(46)}

The microforce balance results in,

323MlR2(1−h)2ΔD−33MlR2(1−h)D∇h⋅∇D−1.513MlR2D∇h⋅∇h+32μ(1−D(1−h))1/2(1−h)(λˉr2−1)+3K4(1−D(1−h))1/2(1−h)(J0.5−1)2−23MD(1+h(h−2))−ηDD˙=0(47)\frac{3}{2\sqrt{3}}Ml_R^2(1-h)^2\Delta D-\frac{3}{\sqrt{3}}Ml_R^2(1-h)D\nabla h\cdot\nabla D-1.5\frac{1}{\sqrt{3}}Ml_R^2D\nabla h\cdot\nabla h + \frac{3}{2}\mu(1-D(1-h))^{1/2}(1-h)(\bar{\lambda}_r^2-1)+\frac{3K}{4}(1-D(1-h))^{1/2}(1-h)(J^{0.5}-1)^2 -\frac{2}{\sqrt{3}}MD(1+h(h-2))-\eta_D\dot{D}=0 \tag*{(47)}
23MD2(1−h)+1.5M13lR2(1−h)∇D⋅∇D+1.5M13lR2D2Δh−ηhh˙−32μ(1−D(1−h))1/2D(λˉr2−1)−3K4(1−D(1−h))1/2D(J0.5−1)2=0(48)\frac{2}{\sqrt{3}}MD^2(1-h)+1.5M\frac{1}{\sqrt{3}}l_R^2(1-h)\nabla D\cdot\nabla D+1.5M\frac{1}{\sqrt{3}}l_R^2D^2\Delta h-\eta_h\dot{h} -\frac{3}{2}\mu(1-D(1-h))^{1/2}D(\bar{\lambda}_r^2-1)-\frac{3K}{4}(1-D(1-h))^{1/2}D(J^{0.5}-1)^2=0 \tag*{(48)}

The microforce balance laws (47) and (48) are used to define the evolution equations for DD and hh.

Damage and healing evolution laws

The evolution equations for hh and DD for pure damage and healing by the two mechanisms – diffusion and rebonding are derived from the microforce balance laws (47) and (48). The healing variable hh measures the ratio of the damaged area retrieved by chain diffusion, while DD measures the damaged area at any instant. When the material heals by rebonding, DD is assumed to spontaneously reverse, and the area retrieval is measured by the scaling factor (1−D)(1-D). We also model healing in which diffusion precludes rebonding. Both DD and hh are assumed to evolve, and the strong coupling between the two variables enables us to model realistically the simultaneous evolution of the healing variable for diffusion and the reversal of the damage variable for rebonding. Rate-dependent healing and damage can be modelled through the time derivatives of the mobility constants ηD\eta_D and ηh\eta_h. Also, both healing and damage can be stress-driven.

Rate dependent damage model

We retrieve the damage model of [14] from (47) when there is only damage but no healing with either of the mechanisms. The evolution of the damage variable results as,

323MlR2ΔD+32μ(1−D)1/2(λˉr2−1)+3K4(1−D)1/2(J0.5−1)2−23MD−ηDD˙=0(49)\frac{3}{2\sqrt{3}}Ml_R^2\Delta D+\frac{3}{2}\mu(1-D)^{1/2}(\bar{\lambda}_r^2-1)+\frac{3K}{4}(1-D)^{1/2}(J^{0.5}-1)^2-\frac{2}{\sqrt{3}}MD-\eta_D\dot{D}=0 \tag*{(49)}

This relation must satisfy the Γ\Gamma – convergence criterion [7, 14] from which the relation between MM and GcG_c (Gc=MlRG_c=Ml_R) is established. Detailed calculations have been shown in the appendix.

Intrinsic healing by diffusion

Healing by diffusion is a spontaneous intrinsic, non-stress-driven process where chains diffuse into the damaged region, healing it with a distinct mass with properties different from the virgin material. Healing by diffusion recovers the effective stress-bearing area but the existing damage is not reversed. Incorporating, these features, we modify (48)

23MD2(1−h)+1.513lR2(1−h)∇D⋅∇D+1.513lR2D2Δh−τhh˙=0(50)\frac{2}{\sqrt{3}}MD^2(1-h)+1.5\frac{1}{\sqrt{3}}l_R^2(1-h)\nabla D\cdot\nabla D+1.5\frac{1}{\sqrt{3}}l_R^2D^2\Delta h-\tau_h\dot{h}=0 \tag*{(50)}

Further we have scaled the equation by 1M\frac{1}{M}, and introduced the chain-diffusion healing timescale, τh=ηhM\tau_h=\frac{\eta_h}{M}. Several researchers have found that this timescale is similar to the stress-relaxation timescale, which controls material creep.

Rebonding of reversible bonds

The rebonding healing mechanism is a spontaneous damage reversal phenomenon that is derived from (47) by removing stress-based driving terms.

−213D+3213lR2ΔD−τDD˙=0(51)-2\frac{1}{\sqrt{3}}D+\frac{3}{2}\frac{1}{\sqrt{3}}l_R^2\Delta D-\tau_D\dot{D}=0 \tag*{(51)}

Similar to (50), the equation is scaled by a factor of 1M\frac{1}{M}, to introduce the rebonding timescale τD=ηDM\tau_D=\frac{\eta_D}{M}.

Rebonding of reversible bonds preceded by inter-chain diffusion

Where diffusion of chains and rebonding follow each other and the rebonding process is dependent on the diffusion of chains and vice-versa, both area recovery through hh and reversal of damage through DD occurs.

3213lR2(1−h)2ΔD−313lR2(1−h)∇D⋅∇h−1.513lR2D∇h⋅∇h−τDD˙=0(52)\frac{3}{2}\frac{1}{\sqrt{3}}l_R^2(1-h)^2\Delta D-3\frac{1}{\sqrt{3}}l_R^2(1-h)\nabla D\cdot\nabla h-1.5\frac{1}{\sqrt{3}}l_R^2D\nabla h\cdot\nabla h-\tau_D\dot{D}=0 \tag*{(52)}
213D2(1−h)+1.513lR2(1−h)∇D⋅∇D+1.513lR2D2Δh−τhh˙=0(53)2\frac{1}{\sqrt{3}}D^2(1-h)+1.5\frac{1}{\sqrt{3}}l_R^2(1-h)\nabla D\cdot\nabla D+1.5\frac{1}{\sqrt{3}}l_R^2D^2\Delta h-\tau_h\dot{h}=0 \tag*{(53)}

need to be solved simultaneously.

Numerical simulations

We implement the damage and the healing models in a set of numerical simulations to observe their ability to model some key features such as rate-dependent non-local damage, non-linear unloading during healing, recovery of stiffness and fracture toughness, increase of recovery with healing time and timescales of diffusion and recovery, influence on damage-width on healing rate etc. The first set of simulations reproduces the numerical experiments of [12] in an incompressible elastomer in a homogeneous stress state. This is followed by simulations of the diffusion-based healing of damage of various initial widths in a one-dimensional bar to determine the effect of width on the extent of healing. Finally, the model is used to predict fracture in virgin and healed supramolecular elastomers and understand post-healing fracture properties of these materials as observed by [11].

Coupled rate-dependent damage and healing through homogeneous simulations

A typical damage-healing model for general polymers must possess certain basic features such as stiffness-recovery, non-linear unloading under damage reversal [25] that can be established through single material point simulations of the numerical experiments of [12]. These experiments, comprising repeated cycles of loading and healing in a polymer, investigate the damage response of healed and virgin materials. We also use these simulations to carry out parametric investigations.

We subject, an incompressible elastomer to a loading-unloading-reloading cycle in which quasistatic tensile loading is applied followed by a period of combined healing followed by reloading.

Equations (49), (50), (52) are solved. The incompressibility of the elastomer is imposed through deformation gradient of the form,

F=[λ000λ000λ](54)\mathbf{F}= \begin{bmatrix} \lambda& 0 & 0 \\ 0 & \lambda& 0 \\ 0 & 0 & \lambda \end{bmatrix} \tag*{(54)}

in the equation (42), with λ\lambda being the stretch.

The simulations yield results that qualitatively compare well with [12]. The stiffness and the fracture toughness increase with both the healing time and the timescale τh\tau_h.

Figure 2. Fig (a): Stress vs strain response in an incompressible elastomer under loading-unloading and healing-reloading experiment [12] Post loading and unloading material was then allowed to heal for tht_h healing time before being reloaded again. The peak stress post-healing increases with an increase in tht_h indicating a gradual recovery of the fracture toughness. Nonlinear stress vs strain relation during unloading due to damage reversal in line with [12, 25] Fig (b): Stress-strain curves for various healing timescales τh\tau_h. As τh\tau_h increases, healing occurs at a slower rate and stiffness and fracture toughness recovery decreases.

1D simulation: Influence of width of interfacial damage on the extent of healing

In this section, the model is used to predict the influence of the damage width on the healing rate in a bar with a centrally placed damage zone. It is used to qualitatively predict the observations of [2] that in the case of diffusion-entanglement-driven healing, narrower damage heals faster than wider damage. This is because narrow damage requires a smaller distance through which chains must travel to entangle.

The equation (50) is solved in one-dimension using the finite difference method. The one-dimensional form for the equation is given as follows,

23D2(1−h)+1.513lR2(1−h)(∂D∂X)2+1.513lR2D2∂2h∂X2−τhh˙=0(55)\frac{2}{\sqrt{3}}D^{2}(1-h)+1.5\frac{1}{\sqrt{3}}l_{R}^{2}(1-h)\left(\frac{\partial D}{\partial X}\right)^{2}+1.5\frac{1}{\sqrt{3}}l_{R}^{2}D^{2}\frac{\partial^{2}h}{\partial X^{2}}-\tau_{h}\dot{h}=0 \tag*{(55)}

The centrally placed diffused damage is modelled by specifying the damage variable as, D(X,t=0)=D0exp⁡(−XlR0)D(X,t=0)=D_0\exp\left(-\frac{X}{l_{R0}}\right). At X=0X=0, the center of the rod, DD is maximum with a value of D0D_0 and lR0l_{R0} is the width of the damage. D0D_0 for a fully developed crack is 1 and less than 1 otherwise. hh is initially 0 everywhere and must satisfy the following boundary conditions - ∂h∂X=0\frac{\partial h}{\partial X}=0 at X=∞X=\infty and X=0X=0. Figure 3 shows the evolution of hh and 1−D(1−h)1-D(1-h), the stiffness degradation function during the healing of a crack of width lR0=0.05Ll_{R0}=0.05L duration τh\tau_h. With time healing increases, while the crack width remains constant. Figure 4 shows the variation of the degradation function with time for different initial lR0l_{R0}. Inline with [2], recovery is faster for a smaller lR0l_{R0}.

Figure 3. Time snapshots of healing hh vs XX and degradation (1−D(1−h))(1-D(1-h)) vx XX for healing of a crack (D0=1)(D_0=1) of width lR0l_{R0} of 0.05L0.05L with a τh\tau_h of 40 mins and a healing width lRl_R of 0.1L0.1L. The intensity of healing increases and that of the degradation decreases with time.

Figure 4. (1−D(1−h))(1-D(1-h)) at X=0X=0 with time for different LR0L_{R0} (0.002 L, 0.05 L, 0.1 L). With the decrease in LR0L_{R0} i.e. for narrower damage, the recovery of properties as measured by the degradation function is faster

Intrinsic healing of an interfacial crack in a supraelastomer

The model is now used to investigate the pre-healed and post-healed fracture properties of a supermolecular elastomer and to reproduce the experiments of [11]. [11] observed that these supramolecular elastomers heal at room temperature within observational timescales. The primary healing mechanism is observed to be the rebonding of non-equilibrated hydrogen bonds, and it is not preceded or accompanied by chain diffusion, as the relaxation timescale is in the order of weeks [11] and much larger than the observational timescale of interest [15] which is of the order of a few hours.

In the experiment, a dogbone specimen is subject to a quasistatic uniaxial tension until fracture. Following that, the fractured pieces are brought in contact and allowed to heal at room temperature for different healing times τh\tau_h - 15mins, 30mins, 60 mins, 120 mins and 180 mins. The healing process is followed by the reloading of the specimen up to fracture. The experimental stress-strain curves are shown in dotted lines in 5. The elastic response of the healed specimen is independent of τh\tau_h. The strain to fracture, however, increases with healing time. Through our simulations, we model this dependence of strain on τh\tau_h, reaching to some conclusions on the recovery of mechanical properties τh\tau_h. The simulation of the experiment comprises the following steps -

  • Stress-driven damage and fracture of the virgin specimen

  • Rebonding or reversal of damage in the fractured specimen. A chain diffusion or the stress relaxation timescale

τh\tau_h is much higher than τD\tau_D [11] compared to our observational timescale, thus healing can be simulated by damage reversal alone.

  • Stress-driven damage and fracture of the healed specimen with recovered material properties.

The complete superposition of the elastic parts of the stress-strain curves of the material with various healing times implies that healing only affects the fracture toughness, hence the resistivity modulus. Our numerical experiment first simulates the fracture of healed specimens to determine the relationship between the healing time and the resistivity modulus, and then through the healing model, which in this case is the reversal of damage, the rebonding timescale, resistivity modulus of the virgin material is determined.

Fracture in healed specimen

The linear momentum balance (30) with the constitutive equation for stress (42) and the rate-dependent damage evolution DD, (49) are solved with a 2D finite element code using FeNICS open-source packages. The geometry comprises a rectangular domain of size 50mm X 13mm X 2mm representing the gage length of the dogbone specimen. It is meshed using linear quadratic and triangular elements with maximum and minimum sizes of mm and mm, respectively. A displacement on the top boundary is applied at a constant strain rate of 0.0004/sec. Numerical stability issues arising due to the rapid damage evolution characteristic of brutal damage are solved by linearisation of terms [3] and small load step size. However, the simulation was terminated before the crack could completely propagate as the strain to fracture does not significantly change post-peak and the computational time required for each of the complete simulations is prohibitively large.

The following material constants in the equations are determined from the curve corresponding to a healing time of 180 mins - μ=2.7276×10−3MPa\mu=2.7276 \times10^{-3}MPa, lR=0.0962mml_R=0.0962mm, κ=4.50054×10−3MPa\kappa=4.50054 \times10^{-3}MPa. The recovered modulus of resistivity, MhM_h which is a function of the healing time, tht_h must be used instead of the modulus for the virgin material. MhM_h as a function of the healing time tht_h is determined from the healing model in the following section

Healing by damage reversal and recovered resistive modulus

The fracture in the dog bone specimen is a straight crack perpendicular to the longitudinal axis. The healing process appears to be uniform in width across the entire length of the crack. Assuming that the healing progresses at the same rate at all points along the crack, it is sufficient to solve the damage reversal equation (51) along the longitudinal axis. The one-dimensional form for the equation is given by,

3213lR2∂2D∂X2−213D−τDD˙=0(56)\frac{3}{2}\frac{1}{\sqrt{3}}l_R^2\frac{\partial^2D}{\partial X^2}-2\frac{1}{\sqrt{3}}D-\tau_D\dot{D}=0 \tag*{(56)}

where lRl_R represents the width of the fracture zone as well as the healed zone where the non-equilibrated hydrogen bonds are located.

Initially, a diffused damage represented by D(X,0)=exp⁡(−X3lR)D(X,0)=\exp\left(-\frac{X}{\sqrt{3}l_R}\right) is centered at X=0X=0. Given the boundary conditions D=0D=0 at X=∞X=\infty and ∂D∂X=0\frac{\partial D}{\partial X}=0 at X=0X=0, the equation is solved analytically to yield the closed-form expression for DD as,

D(X,t)=exp⁡(−X3lR)exp⁡(−32τDth)(57)D(X,t)=\exp\left(-\frac{X}{\sqrt{3}l_R}\right)\exp\left(-\frac{\sqrt{3}}{2\tau_D}t_h\right) \tag*{(57)}

DD gives the spatial distribution of the residual damage after healing has occurred for a time period of tht_h. When we substitute DD in the surface energy expression (III C), we obtain a loss in the resistive capability of the material.

∫−∞∞ψRdh dX=MlR(exp⁡(−32τDth))(58)\int_{-\infty}^{\infty}\psi_{Rdh}\,\mathrm{d}X=Ml_R\left(\exp\left(-\frac{\sqrt{3}}{2\tau_D}t_h\right)\right) \tag*{(58)}

The healed fracture toughness is MlR(1−(exp⁡(−32τDth)))Ml_R\left(1-\left(\exp\left(-\frac{\sqrt{3}}{2\tau_D}t_h\right)\right)\right) which yields the following expression for the healed modulus MhM_h,

Mh=M(1−(exp⁡(−32τDth)))(59)M_h=M\left(1-\left(\exp\left(-\frac{\sqrt{3}}{2\tau_D}t_h\right)\right)\right) \tag*{(59)}

The expression (59) indicates that MhM_h recovers completely at an infinite time. While this is in line with the fracture toughness experiments conducted by [15] where a complete recovery is never practically obtained, the variation of MhM_h with tht_h following a power law is not in line with [15] or [37]. Both studies assume chain diffusion as the primary healing mechanism, which may explain this. To obtain the variation of MhM_h with time tht_h, experiments must be carried out which is beyond the scope of the current article.

The material parameters τD\tau_D, and MM, the rebonding timescale are obtained from the stress-strain curves corresponding to 15, 60, and 180 mins. We determine them to be 198.2 mins and 23 MPa, respectively.

The MhM_h values for 120 mins and 30 mins are obtained from the expression, (59) and implemented in the finite element simulation for the fracture. Figure 5 shows the simulation and experimental stress-strain curves for various degrees of healing. Overall, the numerically obtained curves compare well with experiments and superpose with each other in the elastic regime. The strains to fracture have been accurately predicted for th=120minst_h=120mins and th=30minst_h=30mins. There is a sudden increase of stress in the vicinity of the peak due to numerical stability issues but the peak stress correlate well with those of the experiments.

Figure 5. The figure shows stress-strain curves as observed in [11] and those obtained by the finite element model. The accurate prediction of strains to fracture indicate a good estimation of the modulus MhM_h. The good correlation between experiments and simulations establish our assumption that the stiffness recovers much faster and is independent of tht_h, the healing time. An abrupt increase in stresses in the vicinity of the peak is observed due to numerical stability issues.

Conclusions

We need simple damage-healing theories that are consistent with thermodynamics and can predict autonomic interfacial self-healing in elastomers. These theories can be used to design materials and for failure and healing analysis of real-life structures. Models available in the literature are complex and may fail in their predictions as they lack the necessary timescales and lengthscales. The very few phase-field-inspired approaches use surface energy prescriptions that result in damage and healing evolution equations that lack a physical basis. This article prescribes a geometry-based damage-healing theory for autonomic healing in elastomers. The theory is built on a geometry-based framework developed in [14] where damage is postulated to induce an incompatibility in the Euclidean material manifold, transforming it into a Riemannian manifold. We can quantify the incompatibility by using a Ricci curvature that determines the resistive surface energy. We assume that healing restores the manifold to a Euclidean manifold, either through the reversal of the damage or through rescaling the damage by a healing variable that quantifies damage recovery. The microforce balances dictate the evolution of the damage and healing variables during the fracture and healing processes. The theory focuses on the following healing mechanisms: rebonding of reversible bonds, diffusion of chains across the damaged interface and rebonding preceded by chain diffusion and entanglement. The theory models rebonding through the reversal of damage, while the evolving healing variable models chain diffusion and entanglement. Both processes are intrinsic and evolve with respective timescales. Additionally, the microforce balance for the damage variable can be used to derive a rate-dependent damage model similar to [14]. The resistive modulus of the virgin material and the healed material can be related to corresponding fracture toughnesses through Γ\Gamma − convergence criteria [6].

The predictive capabilities of the model are explored through homogeneous, one-dimensional and two-dimensional finite element and finite difference simulations. The homogeneous simulations are those of the numerical experiments of [12]. In obtaining a good qualitative comparison with their results, our theory predicts a correct coupling between the healing variable and the damage variable during reversal—a feature essential to rebonding preceded by diffusion. Additionally, our theory accurately predicts non-linear stress-strain behaviours during unloading and healing, as well as the increase in peak stress with healing time. Through one-dimensional simulations of the healing of diffused damage in a bar, we show that the model predicts the influence of the width of the damage on the healing rate accurately. This is, however, limited to healing by chain diffusion and entanglement.

Finally, the model is used to predict how supramolecular elastomers heal spontaneously when cracked pieces are brought together at room temperature. The model shows that the resistive modulus and hence the fracture toughness recover exponentially with healing time during the rebonding of broken hydrogen bonds. The rate-dependent damage model with the resistive modulus-healing time relationship is validated using two-dimensional simulations of the uniaxial tension experiments of [11] using the FeNiCS finite element package. The exponential recovery of fracture toughness with healing time is in line with observations of [15], though it differs from those of [38] as the healing mechanism is different.

The model is capable of a comprehensive prediction of damage and the intrinsic healing of elastomers with just a few parameters — healing and damage timescales, resistivity moduli, and a length scale of diffusion. These parameters can easily be determined through macroscopic fracture experiments similar to [11] with cyclic loading without any mesoscale investigations or rigorous curve fitting. Such simplicity makes the formulation much superior to existing prescriptions for material design or structural and failure analysis. Further, the geometric consistency of the theory gives it a solid physical base, which eliminates the necessity of ad hoc prescriptions for degradation functions and surface energy densities.

Temperature-dependent prescriptions for the timescales can introduce thermal dependencies. Furthermore, we can extend this model to other intrinsic healing mechanisms with minimal modifications. The model can also be extended to extrinsic and autogenous healing by incorporating the effect of encapsulated healing agents and other stimuli in the free energy or as separate internal variables. Additionally, by introducing irreversibility conditions through history variables, we can decouple healing from damage.

Reference

References

Appendix: Pure Damage, no Self-Healing

To derive the relationship between MM, the resistive modulus and the fracture toughness GcG_c we adopt the process of enforcing Γ\Gamma-convergence on the surface energy density ΨRd\Psi_{Rd} in a one dimensional bar with a centrally positioned crack. The process has been discussed in some details in [14] and [6]. According to the criteria, the surface energy density must converge to the fracture toughness GcG_c in the limit of lR→0l_R \to0.

For that the equation (49) is rewritten in a one-dimensional form, ignoring the rate dependent term, as follows,

1.513MlR2∂2D∂X2−213MD=0(60)1.5\frac{1}{\sqrt{3}}Ml_R^2\frac{\partial^2D}{\partial X^2}-2\frac{1}{\sqrt{3}}MD=0 \tag*{(60)}

.

The boundary conditions are D=1D=1 at X=0X=0 and D=0D=0 at X=∞X=\infty, This equation has the closed-form solution of the form,

D=exp⁡(−2x3lR)(61)D=\exp\left(-\frac{2x}{\sqrt{3}l_R}\right) \tag*{(61)}

Substituting the solution of DD in the surface energy density and integrating the surface energy over the whole domain [−∞,∞][-\infty,\infty],

∫−∞∞ψRdh dX=∫−∞∞(13MD2+0.753MlR2(∂D∂X)2)dX=MlR(62)\int_{-\infty}^{\infty} \psi_{Rdh}\,dX = \int_{-\infty}^{\infty} \left( \frac{1}{\sqrt{3}}MD^2 + \frac{0.75}{\sqrt{3}}Ml_R^2\left(\frac{\partial D}{\partial X}\right)^2 \right)dX = Ml_R \tag*{(62)}

Since the surface energy density must converge to GcG_c as lR→0l_R \to0, we obtain Gc=MlRG_c = Ml_R.

References

  1. [1]Al-Rub, R.K.A., Darabi, M.K., Little, D.N., Masad, E.A., 2010. A micro-damage healing model that improves prediction of fatigue life in asphalt mixes. International Journal of Engineering Science 48, 966–990.
  2. [2]Amaral, A.J., Pasparakis, G., 2017. Stimuli responsive self-healing polymers: gels, elastomers and membranes. Polymer Chemistry 8, 6464–6484.DOI
  3. [3]Ang, I., Bouklas, N., Li, B., 2022. Stabilized formulation for phase-field fracture in nearly incompressible hyperelasticity. International Journal for Numerical Methods in Engineering 123, 4655–4673.DOI
  4. [4]Barbero, E.J., Greco, F., Lonetti, P., 2005. Continuum damage-healing mechanics with application to self-healing composites. International Journal of Damage Mechanics 14, 51–81.DOI
  5. [5]Bekas, D., Tsirka, K., Baltzis, D., Paipetis, A.S., 2016. Self-healing materials: A review of advances in materials, evaluation, characterization and monitoring techniques. Composites Part B: Engineering 87, 92–119.DOI
  6. [6]Borden, M.J., Hughes, T.J., Landis, C.M., Verhoosel, C.V., 2014. A higher-order phase-field model for brittle fracture: Formulation and analysis within the isogeometric analysis framework. Computer Methods in Applied Mechanics and Engineering 273, 100–118.DOI
  7. [7]Borden, M.J., Verhoosel, C.V., Scott, M.A., Hughes, T.J., Landis, C.M., 2012. A phase-field description of dynamic brittle fracture. Computer Methods in Applied Mechanics and Engineering 217, 77–95.DOI
  8. [8]Cao, L., Gong, Z., Xu, C., Chen, Y., 2021. Mechanical strong and recyclable rubber nanocomposites with sustainable cellulose nanocrystals and interfacial exchangeable bonds. ACS Sustainable Chemistry & Engineering 9, 9409–9417.DOI
  9. [9]Chen, X., Zhu, J., Luo, Y., Chen, J., Ma, X., Bukhvalov, D., Liu, H., Zhang, M., Luo, Z., 2020. Molecular dynamics simulation insight into the temperature dependence and healing mechanism of an intrinsic self-healing polyurethane elastomer. Physical Chemistry Chemical Physics 22, 17620–17631.DOI
  10. [10]Coleman, B.D., Gurtin, M.E., 1967. Thermodynamics with internal state variables. The journal of chemical physics 47, 597–613.
  11. [11]Cordier, P., Tournilhac, F., Soulié-Ziakovic, C., Leibler, L., 2008. Self-healing and thermoreversible rubber from supramolecular assembly. Nature 457, 977–980.DOI
  12. [12]Darabi, M.K., Al-Rub, R.K.A., Little, D.N., 2012. A continuum damage mechanics framework for modeling micro-damage healing. International Journal of Solids and Structures 49, 492–513.DOI
  13. [13]Das, S., Roy, D., 2019. A constitutive model for block-copolymers based on effective temperature. International Journal of Mechanical Sciences 161, 105082.DOI
  14. [14]Das, S., Sharma, S., Ramaswamy, A., Roy, D., Reddy, J., 2021. A geometrically inspired model for brittle damage in compressible elastomers. Journal of Applied Mechanics 88, 081002.DOI
  15. [15]Grande, A., Garcia, S., Van Der Zwaag, S., 2015. On the interfacial healing of a supramolecular elastomer. Polymer 56, 435–442.DOI
  16. [16]Gurtin, M.E., Fried, E., Anand, L., 2010. The mechanics and thermodynamics of continua. Cambridge university press.DOI
  17. [17]Kumar, A., 2020. Nucleation and propagation of fracture and healing in elastomers. Ph.D. thesis. University of Illinois at Urbana-Champaign.
  18. [18]Kumar, A., Lopez-Pamies, O., 2020. The phase-field approach to self-healable fracture of elastomers: A model accounting for fracture nucleation at large, with application to a class of conspicuous experiments. Theoretical and Applied Fracture Mechanics 107, 102550.
  19. [19]Lemaitre, J., 1985. A continuous damage mechanics model for ductile fracture .DOI
  20. [20]Liu, Y.L., Chuo, T.W., 2013. Self-healing polymers based on thermally reversible diels–alder chemistry. Polymer Chemistry 4, 2194–2205.DOI
  21. [21]Mergheim, J., Steinmann, P., 2013. Phenomenological modelling of self-healing polymers based on integrated healing agents. Computational Mechanics 52, 681–692.DOI
  22. [22]Miao, S., Wang, M.L., Schreyer, H.L., 1995. Constitutive models for healing of materials with application to compaction of crushed rock salt. Journal of Engineering Mechanics 121, 1122–1129. doi:doi:10.1061/(ASCE)0733-9399(1995)121:10(1122).DOI
  23. [23]Miehe, C., Göktepe, S., Lulei, F., 2004. A micro-macro approach to rubber-like materials—part i: the non-affine microsphere model of rubber elasticity. Journal of the Mechanics and Physics of Solids 52, 2617–2660.DOI
  24. [24]Murakami, S., 1988. Mechanical modeling of material damage.DOI
  25. [25]Ortiz, M., 1985. A constitutive theory for the inelastic behavior of concrete. Mechanics of materials 4, 67–93.DOI
  26. [26]Oucif, C., Voyiadjis, G.Z., Rabczuk, T., 2018. Modeling of damage-healing and nonlinear self-healing concrete behavior: Application to coupled and uncoupled self-healing mechanisms. Theoretical and Applied Fracture Mechanics 96, 216–230.DOI
  27. [27]Rahaman, M.A., Penco, M., Spagnoli, G., Grande, A.M., Di Landro, L., 2011. Self-healing behavior of blends based on ionomers with ethylene/vinyl alcohol copolymer or epoxidized natural rubber. Macromolecular Materials and Engineering 296, 1119–1127.
  28. [28]Sani, N.F.M., Yee, H.J., Othman, N., Abd Talib, A., Shuib, R.K., 2022. Intrinsic self-healing rubber: A review and perspective of material and reinforcement. Polymer Testing 111, 107598.DOI
  29. [29]Sanz-Herrera, J.A., Aliko-Benitez, A., Fadrique-Contreras, A., 2019. Numerical investigation of the coupled mechanical behavior of self-healing materials under cyclic loading. International Journal of Solids and Structures 160, 232–246.
  30. [30]Shojaei, A., Li, G., 2020. A continuum mechanics approach to the healing efficiency of extrinsic self-healing polymers, in: Self-Healing Polymer-Based Systems. Elsevier, pp. 425–454.DOI
  31. [31]Sun, Y., Yang, H., Xia, W., Guo, Y., 2020. Molecular dynamics simulations of surface welding in crosslinked networks with thermally reversible linkages. Applied Surface Science 527, 146947.DOI
  32. [32]Utrera-Barrios, S., Verdejo, R., López-Manchado, M.A., Santana, M.H., 2020. Evolution of self-healing elastomers, from extrinsic to combined intrinsic mechanisms: A review. Materials Horizons 7, 2882–2902.
  33. [33]Voyiadjis, G.Z., Shojaei, A., Li, G., 2011. A thermodynamic consistent damage and healing model for self healing materials. International Journal of Plasticity 27, 1025–1044.
  34. [34]Voyiadjis, G.Z., Shojaei, A., Li, G., 2012a. A generalized coupled viscoplastic–viscodamage–viscohealing theory for glassy polymers. International Journal of Plasticity 28, 21–45.DOI
  35. [35]Voyiadjis, G.Z., Shojaei, A., Li, G., Kattan, P., 2012b. Continuum damage-healing mechanics with introduction to new healing variables. International Journal of Damage Mechanics 21, 391–414.
  36. [36]Voyiadjis, G.Z., Shojaei, A., Li, G., Kattan, P.I., 2012c. A theory of anisotropic healing and damage mechanics of materials. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 468, 163–183.DOI
  37. [37]Wool, R., 1985. Properties and entanglements of amorphous polymer interfaces. Journal of Elastomers & Plastics 17, 106–118.DOI
  38. [38]Wool, R., O’connor, K., 1981. A theory crack healing in polymers. Journal of applied physics 52, 5953–5963.DOI
  39. [39]Xu, H., Suzuki, N., Takahashi, A., Ohishi, T., Goseki, R., Xie, X.M., Otsuka, H., 2020. Structural reorganization and crack-healing properties of hydrogels based on dynamic diselenide linkages. Science and technology of advanced materials 21, 450–460.
  40. [40]Yamaguchi, M., Ono, S., Okamoto, K., 2009. Interdiffusion of dangling chains in weak gel and its application to self-repairing material. Materials Science and Engineering: B 162, 189–194.DOI
  41. [41]Yu, K., Xin, A., Feng, Z., Lee, K.H., Wang, Q., 2020. Mechanics of self-healing thermoplastic elastomers. Journal of the Mechanics and Physics of Solids 137, 103831.DOI
  42. [42]Yu, K., Xin, A., Wang, Q., 2018. Mechanics of self-healing polymer networks crosslinked by dynamic bonds. Journal of the Mechanics and Physics of Solids 121, 409–431.DOI
  43. [43]Zhao, J., Xu, R., Luo, G., Wu, J., Xia, H., 2016. A self-healing, re-moldable and biocompatible crosslinked polysiloxane elastomer. Journal of Materials Chemistry B 4, 982–989.DOI
  44. [44]Zheng, X., Yang, H., Sun, Y., Zhang, Y., Guo, Y., 2021. A molecular dynamics simulation on self-healing behavior based on disulfide bond exchange reactions. Polymer 212, 123111.DOI

Paper details

vlnc:W_qzf4tccj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents