Abstract
Surfaces and interfaces can significantly influence the overall response of a solid body. Their behavior is well described by continuum theories that endow the surface and interface with their own energetic structures. Such theories are becoming increasingly important when modeling the response of structures at the nanoscale. The objectives of this review are as follows. The first is to summarize the key contributions in the literature. The second is to unify a select subset of these contributions using a systematic and thermodynamically consistent procedure to derive the governing equations. Contributions from the bulk and the lower-dimensional surface, interface, and curve are accounted for. The governing equations describe the fully nonlinear response (geometric and material). Expressions for the energy and entropy flux vectors, and the admissible constraints on the temperature field, all subject to the restriction of non-negative dissipation, are explored. A particular emphasis is placed on the structure of these relations at the interface. A weak formulation of the governing equations is then presented that serves as the basis for their approximation using the finite element method. Various forms for a Helmholtz energy that describes the fully coupled thermomechanical response of the system are given. They include the contribution from surface tension. The vast majority of the literature on surface elasticity is framed in the infinitesimal deformation setting. The finite deformation stress measures are, thus, linearized and the structure of the resulting stresses discussed. The final objective is to elucidate the theory using a series of numerical example problems.
Introduction
The surface of a solid body typically exhibits properties that differ from those of the encased bulk. These differences, caused by processes such as surface oxidation, ageing, coating, atomic rearrangement, and the termination of atomic bonds, are present in comparatively thin boundary layers. Similarly, interfaces within the bulk can be viewed as two-sided internal surfaces. The mechanical and thermal properties of the interface can also differ significantly from the surrounding bulk. Surface and interface effects are especially significant for nanomaterials due to their large surface-to-volume ratio. These effects were appreciated already in the early 1900s by Pawlow [1], for example, who predicted the decrease in the melting temperature of a particle as its size reduces.
There exist two principal approaches to study the thermodynamics of surfaces and interfaces:
the zero-thickness layer or Gibbs (geometrical) method wherein a mathematical surface with zero thickness is introduced to capture excess quantities on the surface [2]
the finite-thickness layer method that dates back to van der Waals wherein a layer of finite thickness can be imagined instead of the interface
The reader is referred to Guggenheim [3] for further details and a comparison of the two approaches.
Following Gibbsean thermodynamics, various models have been proposed to endow the surface and interface with their own distinct properties, e.g., Adam [4], Shuttleworth [5], Herring [6], Bikerman [7], Orowan [8], and Cahn [9,10]. The physical chemistry of surfaces and interfaces for liquids and solids is studied in Refs. [4,11–15], among others.
The thermodynamic fundamentals of surface science were reviewed by Rusanov in Refs. [16,17]. Müller and Saul [18] presented a review on the importance of stress and strain effects on surface physics. The role of stress at solid surfaces was critically examined by Ibach [19]. Leo and Sekerka [20] investigated the equilibrium conditions for interfaces using a variational approach wherein the excess energy associated with the interface is allowed to depend on both the deformation of the interface and the crystallographic normal to the interface. Cammarata et al. [21,22,23,24] highlighted the surface and interface stress effects on thin films and nanoscaled structures. Fischer et al. [25] studied the role of surface energy and surface stress in phase-transforming nanoparticles and reported on the thermodynamics of a moving surface.
Gutman [26] recognized an inconsistency between the notion of the surface stress defined by, e.g., Shuttleworth, and that of the continuum elasticity theory. Bottomley and Ogino [27] examined an alternative to the Shuttleworth formulation of surface stress in solids (see also Ref. [28]). Kramer and Weissmüller [29] criticized Refs. [26–28] and clarified various misleading statements in the literature. Fischer et al. [25] circumvented such issues by a careful handling of the actual and reference configurations. Furthermore, Fischer et al. utilized the concept of configurational forces [30,31] as a convenient framework for distinguishing between surface energy, surface tension, and surface stress.
A widely adopted model, proposed by Gurtin and Murdoch [32,33], gives the surface its own tensorial stress measures (see, e.g., Refs. [22,34–36], for applications in nanomaterials). Murdoch [37], Gurtin and Struthers [38], and Gurtin et al. [39] extended this approach to consider interfaces within a solid. An important extension of the Gurtin and Murdoch model to account for the flexural resistance of the surface was developed by Steigmann and Ogden [40] and further studied in Refs. [41,42]. Moeckel [43] followed a different approach to Gurtin and Murdoch [32] for a moving interface within a thermomechanical solid. A master-balance equation is derived and applied to various key fields on the interface (see also Ref. [44]). The resulting equations are then restricted using the entropy principles [45] and constitutive relations for material interfaces derived. An alternative approach to develop general governing equations for the interface is to integrate the known equations for the bulk over the thickness of the interfacial layers (see, e.g., Ref. [46]). Daher and Maugin [47] used the method of virtual power [48,49] to derive the governing equations for an interface within a thermomechanical solid. This flexible approach is well suited to describing complex phenomena. For example, Daher and Maugin [50] applied the method to a model of deformable semiconductors with interfaces (see also Ref. [51]). Murdoch [52] addressed various aspects of surface modeling. Šilhavý [53] proved the existence of equilibrium of a two-phase state with an elastic solid bulk and deformation dependent interfacial energy. Park et al. [54], Park and Klein [55] developed an alternative continuum framework based on the surface Cauchy–Born model, an extension of the classical Cauchy–Born model, to include surface stresses (see also Refs. [56,57]).
The effect of surface energetics for ellipsoidal inclusions and the size-dependent elastic state of embedded inhomogeneities was investigated by Sharma et al. [58], Sharma and Ganti [59], and Sharma and Wheeler [60]. They utilized the classical formulation of Eshelby [61,62] for embedded inclusions and modified it by incorporating surface energies. Duan et al. [63] extended the Eshelby formalism for inclusion/inhomogeneity problems to the nanoscale. Effective mechanical and thermal properties of heterogeneous materials containing nanoinhomogeneities based on the generalized Eshelby formalism is investigated in Refs. [63–65]; see also related works Refs. [66–72].
Zöllner et al. [73] recently reported that to explore the biomechanical interaction during tissue expansion, one could model skin growth using a boundary energy. The thermomechanical behavior of low-dimensional systems was reviewed by Sun [74] together with a theoretical analysis elaborating on existing approximations in continuum mechanics and quantum computations. Johnson [75] studied the thermodynamics of a coherent interface separating two nonhydrostatically stressed crystals.
The objective of this presentation is to derive the equations governing the fully nonlinear coupled transient thermomechanical response of a body composed of a bulk, intersected by an interface, and encased by a surface, all of which are assumed to be energetic. Furthermore, an energetic curve that accounts for the interaction between the interface and the surface is also present. The importance of capturing such an interaction when modeling ionic nanowires was recently observed in Ref. [76].
A particular emphasis is placed here on the thermomechanical response of the interface. The governing equations for the various parts of the body are obtained from the balances of several key properties, namely linear and angular momentum, energy, and entropy. The corresponding constitutive relations then arise as thermodynamic restrictions on the dissipation. The diffusion of mass or chemical species is not accounted for in this contribution (see Ref. [77] for the case of species diffusion in thermomechanical solids with energetic surfaces).
Interfaces can be classified as follows. A material interface is one that does not move independently of the surrounding bulk material. An energetic interface is understood here to imply that the interface possesses its own thermomechanical structure in the form of an internal energy, entropy, constitutive relations, and dissipation. Such an interface is termed a thermodynamical singular surface by Daher and Maugin [47]. A thermal interface is defined as one that allows for heat conduction along the interface but possesses no energetic structure. Finally, a standard interface does not allow heat conduction along the interface and possesses no energetic structure. Daher and Maugin [47] term such an interface a free singular surface.
An interface can be classified further according to its thermal properties as follows:
Thermally perfect interface: The jump in the temperature and in the normal heat flux across the interface is zero.
Generic imperfect interface: In this general case neither the jump in the temperature nor the jump in the normal heat flux across the interface need be zero. A thermal interface is an example of a generic imperfect interface. Özdemir et al. [78], for example, developed a thermomechanical cohesive zone model for generic imperfect interfaces. The following two models are specializations of the generic imperfect interface model (see Refs. [79–81] and references therein).
- (i)
Weakly conducting (Kapitza) interface: This type of imperfect interface is modeled using Kapitza's concept of thermal resistance.1 The model allows for a temperature jump across the interface. The normal heat flux is, however, continuous across the interface. The Kapitza model for thermal interfaces has been widely investigated (see, e.g., Ref. [83] and references therein).
- (ii)
Highly conducting (HC) interface: This type of imperfect interface models the temperature as continuous across the interface, while allowing a jump in the normal component of the heat flux. Yvonnet et al. [84] have used the HC model to numerically investigate the effective conductivity of nanocomposites containing interfaces. For nonenergetic interfaces, a jump in the heat flux across the interface is only possible if there is an associated heat flux along the interface (see, e.g., Ref. [85]). This restriction does not hold for energetic interfaces [86].
- (i)
Mosler and Scheider [87] recently described the admissible assumptions for the motion of the interface for a class of damage-type cohesive models. In a similar spirit, the relation of the interface temperature to that in the surrounding bulk is discussed here in detail. In particular, it is shown that the interface temperature is not, in general, equal to the average of the temperature across the interface.
Steinmann and Häsner [88] considered an energetic interface (with zero dissipation across the interface) within a thermomechanical solid subject to the assumption of infinitesimal deformation. Steinmann [89] presented extensive details on energetic surfaces using both the deformational and configurational frameworks of continuum mechanics. That investigation did not consider thermal contributions, interfaces, or numerical aspects. Javili and Steinmann [90,91,92,93] extended the work to account for the thermomechanics of energetic surfaces and the implementation of the model using the finite element method. The current review includes an extension of these works to account for the role of the interface and curve. Related works by Yvonnet et al. [83,84] focus on stationary heat conduction across thermal interfaces.
The focus of this work is on energetic elasticity as compared to entropic elasticity. For entropic elastic materials, e.g., polymers, the temperature increases due to stretching. On the contrary, energetic elastic materials, e.g., metals, cool down when stretched. The thermomechanical coupling, whether for energetic or entropic elasticity, is termed, henceforth, the Gough–Joule effect. The Gough–Joule effect is often used for entropic elastic materials as it was originally realized for rubbers (see Gough [94] and Joule [95]). It is noteworthy, that the Gough–Joule effect is not solely due to the deformation, but it is rather the rate of a deformation that induces a change in the temperature field.
In summary, the key objectives and contributions of this work are as follow:
To systematically derive the equations governing the response of a thermomechanical body, intersected by a material interface, and enclosed by a surface, undergoing finite deformations coupled to transient heat conduction. The procedure used to derive the equations is general and could be applied to a range of complex phenomena.
To account for the interaction between the interface and the surface.
To derive the thermodynamically consistent constitutive relations, dissipation inequalities, and temperature evolution equations in the various parts of the body (bulk, surface, interface, and curve).
To briefly present the weak formulation of the governing equations for the general case of a thermomechanical solid possessing an energetic surface, interface, and curve. Various other models are shown to arise as restrictions of this general case.
To elaborate various aspects of material (constitutive) modeling and discuss the admissible range of material parameters.
To elucidate the theory with numerical examples based on a finite element approximation of the governing equations.
This manuscript is organized as follows. The notation and certain key concepts are introduced briefly in Sec. 2. The equations governing the response of the various parts of the body are derived from fundamental balance principles in Sec. 3. A thermodynamic framework is then utilized in Sec. 4 to determine the form of the thermodynamically consistent constitutive relations. The nature of the coupling between the primary fields (i.e., the displacement and the temperature) and the various parts of the body are made clear. The dissipation inequality for the interface is analyzed in detail to show the possible relations between the interface temperature and that of the surrounding bulk. The weak formulation of the governing equations is then given in Sec. 5. Various important problems, which result as restrictions of this general weak formulation, are highlighted. Aspects of material modeling are elaborated in Sec. 6. A surface Helmholtz energy is introduced and the relationship between surface material parameters is discussed. The admissibility of surface negative material parameters is investigated. A series of numerical example problems, based on a finite element approximation of the weak formulation, is presented in Sec. 7 to elucidate the theory. Section 8 concludes this work.
Preliminaries
The purpose of this section is to summarize certain key concepts in nonlinear continuum mechanics and to introduce the notation adopted here. Detailed expositions on nonlinear continuum mechanics can be found in Refs. [96–101], among others. For further details concerning the continuum mechanics of deformable surfaces and interfaces the reader is referred to Refs. [38,43,47] and the references therein.
Notation and Definitions
Direct notation is adopted throughout. Occasional use is made of index notation, the summation convention for repeated indices being implied. The three-dimensional Euclidean space is denoted 3. The scalar product of two vectors a and b is denoted = [a]i[b]i. The scalar product of two second-order tensors A and B is denoted . The composition of two second-order tensors A and B, denoted , is a second-order tensor with components []ij = [A]im[B]mj. The vector product of two vectors a and b is denoted a × b with [a × b]k = [ε]ijk[a]i[b]j, where ε denotes the third-order permutation (Levi–Civita) tensor. Analogously, the vector product of a vector a and a second-order tensor B is denoted a × B with [a × B]kl = [ε]ijk[a]i[B]jl. The nonstandard products of a fourth-order tensor and a vector b are defined by and . Other nonstandard products of a fourth-order tensor and a vector b are defined by and . The nonstandard product of a fourth-order tensor and a second-order tensor A is defined by . The nonstandard minor transposes of a fourth-order tensor are defined by . The action of a second-order tensor A on a vector a is given by []i = [A]ij[a]j. The tensor product of two vectors a and b is a second-order tensor D = ab with [D]ij = [a]i[b]j. The tensor product of two second-order tensors A and B is a fourth-order tensor = A B with []ijkl = [A]ij[B]kl. The two nonstandard tensor products of two second-order tensors A and B are the fourth-order tensors and .
Quantities or operators corresponding to the bulk, surface, interface, and curve are denoted as {•}, , , and , respectively, unless specified otherwise. Note, denotes a surface-quantity that is not necessarily tangent to the surface. Similarly, a bulk quantity {•} on the surface implies the evaluation of {•} on the surface and is not, in general, equivalent to the corresponding surface quantity .
The expression the various parts of the body implies the bulk, the surface, the interface, and the curve.
Key definitions are given in Tables 1 and 2. For the derivations thereof and extensive technical details see classical manuscripts on differential geometry [102,103,104], among others. Basic concepts and terminologies of surfaces and curves can be found in the fundamentals of differential geometry on surfaces and curves briefly reviewed in Appendix A.
Bulk | Surface | Interface | Curve | |
---|---|---|---|---|
Domain | ; | ; | ; | ; |
Normal | N ; n | , ; , | , , , ; , , , | |
Tangent | , ; , | ; | ; | |
Unit tensor | I ; i | ; | ; | ; |
Differential element | dV ; dv | dA ; da | dA ; da | dL ; dl |
Oriented element | — | dA ; da | dA ; da | dL ; dl |
Control regions | 0 ; t | 0 ; t | 0 ; t | 0 ; t |
Curvature | — | ; | ; | ; |
Covariant basis | G1, G2, G3 ; g1, g2, g3 | , ; , | , ; , | ; |
Contravariant basis | G1, G2, G3 ; g1, g2, g3 | , ; , | , ; , | ; |
Placement | X ; x | ; | ; | ; |
Tangent map | F ; f | ; | ; | ; |
Normal map | Cof F ; cof f | |||
Jacobian | J ; j | ; | ; | ; |
Velocity | V ; v | ; | ; | ; |
Divergence | Div ; div | ; | ; | ; |
Gradient | Grad ; grad | ; | ; | ; |
Determinant | Det ; det | ; | ; | ; |
Cofactor | Cof ; cof | ; | ; | ; |
Bulk | Surface | Interface | Curve | |
---|---|---|---|---|
Domain | ; | ; | ; | ; |
Normal | N ; n | , ; , | , , , ; , , , | |
Tangent | , ; , | ; | ; | |
Unit tensor | I ; i | ; | ; | ; |
Differential element | dV ; dv | dA ; da | dA ; da | dL ; dl |
Oriented element | — | dA ; da | dA ; da | dL ; dl |
Control regions | 0 ; t | 0 ; t | 0 ; t | 0 ; t |
Curvature | — | ; | ; | ; |
Covariant basis | G1, G2, G3 ; g1, g2, g3 | , ; , | , ; , | ; |
Contravariant basis | G1, G2, G3 ; g1, g2, g3 | , ; , | , ; , | ; |
Placement | X ; x | ; | ; | ; |
Tangent map | F ; f | ; | ; | ; |
Normal map | Cof F ; cof f | |||
Jacobian | J ; j | ; | ; | ; |
Velocity | V ; v | ; | ; | ; |
Divergence | Div ; div | ; | ; | ; |
Gradient | Grad ; grad | ; | ; | ; |
Determinant | Det ; det | ; | ; | ; |
Cofactor | Cof ; cof | ; | ; | ; |
Kinematics
Consider a continuum body that takes the open set at the time t = 0, as depicted in Fig. 1. The boundary of the body is denoted by := ∂. The body is partitioned into two disjoint subdomains, denoted by the open sets and , by a two-sided interface . The interface is an open set; thus, . The boundary of the interface, a two-sided curve, is denoted as := . In a similar fashion to the interface, the curve partitions the surface into two open sets and such that and = .
Remark. The boundary is closed and is assumed to be smooth. For the more general case where the boundary is nonsmooth, the curve can be understood as a geometrical entity. For the restricted case considered here where the boundary is smooth, the curve accounts for the interaction between the interface and the surface and should be understood as a physical entity. The extension of this work to nonsmooth boundaries is straightforward. Nevertheless, it involves introducing additional geometric concepts and further notation while providing little additional insight into the fundamental concepts.
The bulk is defined by := and is the reference placement of material particles labeled . The two sides of the interface are denoted and , with . Material particles on the interface are labeled and, by definition, . The outward unit normals to and are denoted and , respectively, with = . The unit normal to is denoted .
The surface is defined by with outward unit normal N. Material particles on the surface are labeled and, by definition, . The two sides of the curve are denoted and , with . Material particles on the curve are labeled , where . The outward unit normals to and , tangential to and identical at a point , are denoted . The unit normals to and , tangential to , are denoted and , respectively, with . The unit normal to , tangential to , is defined by . The unit normals to in the sense of the Frénet–Serret formulae and identical at a point are denoted . Note that does not necessarily coincide with either or .
Let = [0, T] denote the time domain. A motion of the reference placement for a time is denoted by the orientation-preserving map φ: . The current placement of the bulk associated with the motion φ is denoted with particles designated as x = φ (X, t) (see Fig. 2).2
The restriction of the motion φ to the surface is denoted . The current placement of the surface is denoted with particles . It is assumed that particles on the surface of the body constitute the surface for all times and , consequently . This assumption, referred to as kinematic slavery by Steinmann and Häsner [88], mimics coherence in interfaces.
The restriction of the motion φ to the positive and negative sides of the interface is denoted and , respectively. The placements of the positive and negative sides of the interface at time t are denoted and , respectively, with particles and . The material interface is assumed to be coherent: The motion is not necessarily smooth across the interface, however, it is continuous across the interface, smooth away from the interface, and smooth up to the interface from either side [105]. Thus, and, consequently, . The restriction of the motion φ to the interface is defined by with particles designated as .
Remark. Given that this work focuses on coherent material interfaces it may seem superfluous to distinguish between and . However, coherence concerns only the continuity of the motion across the interface and jumps in other fields, e.g., temperature, are permitted.□
Following from the aforementioned assumptions, the restriction of the motion φ to the curve at the junction of the interface and the surface is denoted . The current placement of the curve is denoted with particles . Finally, and .
For the normal vectors to the surface, interface, and curve in the spatial configuration the same convention as in the material configuration is used, however, with lowercase n instead of uppercase N.
Deformation Gradients
Control Regions
The equations governing the coupled response of the various parts of the body are obtained from the balances of linear and angular momentum, energy, and entropy performed over a control region. These balances are then localized at arbitrary points in the bulk , on the surface , on the interface , or on the curve .
The canonical control region is defined as one that possibly has as part of its boundary the surface and is denoted with boundary (see Fig. 3). The orientable external surface of the control region is defined by while the interior surface is . The outward unit normals to and are denoted M and N, respectively. The surface of the control region is defined by . The boundary of , a curve, is defined by . The unit normal to the curve is denoted and is tangent to the surface .
The control region can include a segment of the interface defined by . Thus, can be partitioned into two subdomains and . The interface has two sides and with .
The boundary of the interface is a curve denoted as . The, possibly empty, intersection of the curves and is denoted as . The interior boundary of the interface is . The outward unit normal to the curve is denoted and is tangent to the interface . The unit normal to the curve , tangent to the interface , is denoted . The unit normal to the curve , in the sense of Frénet–Serret formulae, is denoted . The unit normal to the curve , tangent to the surface , is denoted . The unit normal to the boundary of the curve , tangent to the curve is denoted . The curve has two sides and with . Analogously, and with .
Key Definitions and Identities
Various key definitions and identities that are required in the remainder of the presentation are now introduced. Several other useful relations are recorded in Appendix B.
Jump and Average Relations
Integral Relations
Piola Identity and Piola Transform
Superficial and Tangential Tensor Fields
Spatial second-order tensors and vectors can be defined in the same fashion by replacing the material quantities with the spatial ones. Note that the set of all tangential second-order tensors is contained within the set of all superficial second-order tensors.
Governing Equations
The coupled governing equations for the thermomechanical problem in the various parts of the body are now derived using an arbitrary (canonical) control region . Thereafter, the balances of momentum, energy, and entropy, i.e., the governing equations in integral form, are given. Finally, these integral balance expressions are localized at arbitrary points in the bulk , on the surface , or interface , or on the curve , thereby giving the local (strong) form of the governing equations.3
where denotes a source-like term and is the corresponding flux. The corresponding integral expressions for the balance of linear and angular momentum are derived and, subsequently, localized in Sec. 3.1. The localization process is explained in detail. In Sec. 3.2 the integral and resulting local forms of the energy balance are given. Finally, in Sec. 3.3 the process is repeated for the entropy balance.
Mechanical Power
The prescribed bulk (body) force per unit reference volume of is denoted Bp (N/m3) where the superscript p indicates a prescribed quantity. The prescribed tractions per unit reference area of the surface and the interface are denoted (N/m2) and (N/m2), respectively. The prescribed traction per unit reference length of the curve is denoted (N/m). As with the bulk force , the prescribed tractions on the surface, interface, and curve can be understood as prescribed surface, interface, and curve forces, respectively. The bulk, surface, interface, and curve Piola stresses are denoted , , , and , respectively. The terms and denote the tractions acting on and , respectively.
Remark. It assumed that the Cauchy theorem holds for the bulk, surface, interface, and curve Piola stresses. Furthermore, , , and are superficial tensor fields in their tangent spaces.□
where R, , and are vectors to the points in the bulk, surface, interface, and curve, respectively, relative to a fixed point in , e.g., the origin.
Bulk
Surface
Interface
Curve
Remark. The local balance of momentum on the interface Eq. (17) relates the interface Piola stress to the jump of the bulk Piola stress across the interface. The curve Piola stress, however, is coupled not only to the jump of the surface Piola stress across the curve but also to a contribution from the interface Piola stress according to Eq. (19).□
Material and Spatial Formulations
where the spatial velocity fields in the bulk on the surface, interface, and curve are denoted , and , respectively.
where the prescribed bulk force per unit current volume of is denoted bp := jBp. In an identical fashion, the prescribed surface, interface, and curve forces are denoted , respectively.
Thermal Power
The prescribed bulk (body) heat source per unit reference volume (and per unit time) in is denoted Qp (N/s m2). The prescribed heat sources per unit reference area (and per unit time) of the surface and the interface are denoted (N/s m) and (N/s m), respectively. The prescribed heat source per unit length (and per unit time) of the curve is denoted (N/s). The bulk, surface, interface, and curve Piola heat flux vectors are denoted Q, , , and , respectively. The terms QM, , , and denote the heat fluxes on , , and , respectively. It is assumed that the Cauchy theorem holds for the heat flux vectors in all parts of the body. Note that , , and are tangential vectors fields.
where the bulk internal energy per unit reference volume, the surface and interface internal energies per unit reference area, and the curve internal energy per unit reference length are denoted E (N/m2), (N/m), (N/m), and (N), respectively.
The Piola identities for the various parts of the body allow Eq. (26) to be expressed solely in terms of spatial quantities. The (bulk) Cauchy heat flux vector q and the surface, interface, and curve tangential Cauchy heat flux vectors, denoted , , and , respectively, are related to their corresponding Piola heat flux vectors via the Piola transforms (8) and (9).
where the prescribed heat source per unit current volume (and unit time) in is denoted qp := jQp. The bulk internal energy per unit current volume is denoted e := jE. In an identical fashion to the bulk, the prescribed surface, interface, and curve heat sources are denoted , and respectively. Similarly, denote the surface, interface, and curve internal energies, respectively. Recall that dt denotes the spatial time derivative. Thus, convective contributions of the type ev appear in the corresponding divergence theorems.
Entropy Power
The prescribed bulk (body) entropy source per unit reference volume (and unit time) in is denoted (N/sKm2). The prescribed entropy sources on the surface , and the interface are denoted (N/sKm) and (N/sKm), respectively. The prescribed entropy source on the curve is denoted (N/sK). In the global entropy input (28) the terms HM, , , and denote the entropy fluxes on , , , and , respectively. The vectors H, , , and are the bulk, surface, interface, and curve Piola entropy flux vectors, respectively. The vectors , , and are tangential tensor fields in their tangent spaces.
where II ≥ 0, ≥ 0, ≥ 0, and ≥ 0 are the entropy production rates per unit reference volume of the bulk, per unit reference area of the surface and interface, and per unit reference length of the curve, respectively, all constrained to be positive.
where the bulk entropy per unit reference volume is denoted (N/Km2). The surface and interface entropies per unit reference area are denoted (N/Km) and (N/Km), respectively. The curve entropy is denoted (N/K).
The (bulk) Cauchy entropy flux vector h and the surface, interface, and curve tangential Cauchy entropy flux vectors, denoted , , and , respectively, are related to their corresponding Piola entropy flux vectors via the Piola transforms (8) and (9).
where the bulk prescribed entropy source, entropy and entropy production per unit current volume in are denoted , respectively. In an identical fashion to the bulk, the prescribed surface, interface, and curve entropy sources are denoted , respectively. Similarly, , and denote the surface, interface, and curve entropies, respectively. Finally, the prescribed surface, interface, and curve entropy productions are denoted , respectively.
The variables > 0, > 0, > 0, and > 0 denote the absolute temperatures in the bulk, on the surface, interface, and curve, respectively. Note that the assumption (32) is not true in general but holds for simple thermodynamic processes (see, e.g., Refs. [30,106,107]). The validity and necessity of this assumption should be explored using the procedure outlined in Müller [108] and later works.
Thermohyperelastic Helmholtz Energy
Note that the surface and interface Helmholtz energies are allowed to depend on their corresponding normals in the spatial configuration to capture possible anisotropic behavior. The anisotropy of the curve, however, is modeled by dependence of the curve Helmholtz energy on its tangent in the spatial configuration. It should be emphasized that in deriving the balance equations in Sec. 3, no isotropy assumptions were made.
Constitutive Equations
where the Helmholtz energies in the spatial configuration, i.e., per unit deformed area for the surface and interface and per unit deformed length for the curve, are defined by and , respectively. The quantities , , and denote the surface shear, interface shear, and curve shear, respectively (see Refs. [105,109]). Note that the terms , , and map from the tangent spaces to the material surface, interface, and curve, to the normals to the spatial surface, interface, and curve, respectively.
Reduced Dissipation Inequalities
Note that the first terms in Eq. (39)2–4 represent the dissipation along the surface, interface and curve and are denoted , and , respectively. In contrast, the remaining terms represent the dissipation across the surface, interface, and curve, and are denoted as , , and , respectively.
Dissipation Inequalities in the Bulk and Along the Surface, Interface, and Curve
where k, , , and denote the bulk, surface, interface, and curve positive semidefinite spatial thermal conductivity tensors, respectively.
Dissipation Inequalities Across the Surface, Interface, and Curve
The requirements for these inequalities to be satisfied are now discussed.
Surface.
that is, to impose so-called thermal slavery on the surface. It should be emphasized that the thermal slavery condition on the surface is not an a priori assumption but arises, rather, as a natural consequence of the thermodynamics. Alternatively McBride et al. [77] propose a Robin-type boundary condition between the bulk and surface.
Interface.
The further reduced dissipation inequality on the interface (41)2 is satisfied if either of the two equivalent forms in Eq. (44) is satisfied.
Coldness-based discussion.
where ≥ 0 is a heat flux resistance coefficient across the interface.
which is the thermal Young–Laplace equation on the interface. Equation (49) states that the normal jump of heat flux is equal to the negative of the divergence of the heat flux along the interface. In particular, in the absence of heat flux along the thermal interface the normal jump of the heat flux, i.e., the left-hand side of the Eq. (49) vanishes, i.e., the standard interface, and, therefore, Eq. (46)1 is trivially satisfied.□
Remark. A sufficient condition for a thermal interface to satisfy Eq. (46)1 is to choose the interface coldness equal to the average of the coldness across the interface. This conclusion shall be compared with Daher and Maugin [47] who viewed it as a necessity.□
Remark. Using the relation , the condition (46)2 can be reformulated as that can be satisfied by a Fourier-type law for the average heat flux across the interface where denotes the Kapitza heat flux resistance coefficient.□
Temperature-based discussion
where is the entropy flux resistance coefficient across the interface.
Remark. For a standard interface, the jump of the entropy flux; in contrast to the jump of the heat flux, does not necessarily vanish according to Eq. (30)3. That is, for a standard interface Eq. (50)1 is not trivially satisfied. Thus, a sufficient condition to satisfy Eq. (50)1 is to choose the interface temperature equal to the average of the temperature across the interface.□
Remark. For a highly conducting interface the jump of the temperature across the interface vanishes, which corresponds to zero normal dissipation across the interface, i.e., . For this case or . Therefore, for a highly conducting interface, the interface temperature (resp. coldness) is trivially the average of the temperature (resp. coldness) across the interface since.□
Remark. Using the relation , the condition (50)2 can be reformulated as, , which can be satisfied by a Fourier-type law for the average entropy flux across the interface where denotes an entropy flux resistance coefficient.□
Summary.
Curve.
where and are heat flux and entropy flux resistance coefficients, respectively. Analogous to the surface, we can impose the thermal slavery on the curve, i.e., , as the most obvious way to guarantee that the second term in Eq. (41)3 is non-negative. A second novel option is to impose a constraint between the temperatures on the curve and interface and the heat flux that resembles a Robin-type boundary condition between the curve and the interface.
Temperature Evolution Equations
Weak Formulation
In order to establish a principle of virtual work like statement and as a prerequisite for finite element method approximations (to be presented in a subsequent contribution), the mechanical weak form, i.e., the weak form of the linear momentum balance Eq. (21) and the thermal weak form, i.e., the weak form of the temperature evolution Eq. (56), are required.
Weak Formulation—Mechanical
To derive the mechanical weak form, the local linear momentum balance equations for the various parts of the body (21)1–4 are tested from left with vector-valued test functions , , , and , respectively. The result is integrated over the corresponding domains in the material configuration to give the global weak form of the balance of linear momentum as
where the test functions satisfy the conditions given in Eq. (57).
Weak Formulation—Thermal
In order to derive the thermal weak form we proceed formally. The local temperature evolution equations for the various parts of the body (55)1–4 are tested from left with scalar test functions , , , and , respectively. The result is integrated over the corresponding domains in the material configuration to give the global weak form of the temperature evolution equation as:
where the test functions satisfy the conditions given in Eq. (60).
Comparison With Several Other Models
The objective of this section is to briefly derive several models given in the literature by simplifying the weak forms (58) and (61). In particular, the focus is on the thermal weak form (61). The following cases are examined:
Note that none of the aforementioned models account for the contributions from the curve and, therefore, all the corresponding integrals vanish identically.
Standard Thermomechanical Solids
The numerical implementation of finite strain thermoelasticity is well documented in the literature (see, e.g., Ref. [110] and references therein).
Highly Conducting Interfaces
For details of the numerical implementation see, e.g., Yvonnet et al. [84] and the references therein.
Kapitza Interfaces
The last term on the left-hand side vanishes for a standard interface; this assumption has been made by Yvonnet et al. [83].
Thermomechanical Surfaces
which is identical to the one given in Javili and Steinmann [93].
Thermomechanical Highly Conducting Interfaces
Generalized Kapitza Interfaces
Material Modeling
The objective of this section is to elaborate various aspects of material (constitutive) modeling. From the perspective of material modeling, the surface and the interface are essentially identical. Furthermore, the derivations on the surface and interface capture all the key features of the material model for the curve. Therefore, without any loss of generality, we shall focus the majority of the subsequent discussion on the material modeling of the bulk and the surface. However, in order to provide sufficient background material to explain the numerical examples in Sec. 7, we also briefly present details on the thermomechanical modeling of interfaces. The influence of the surface on the overall response of a structure at the nanoscale is one of the key applications for models of surface elasticity in the literature. We also restrict attention to thermoelastic materials. Inelastic behavior at the nanoscale is recently reported in Ref. [112].
The material modeling of bulk materials is a mature field with many standard references (see, e.g., Ref. [100]). This is not the case for the surface. Modeling the constitutive response of the surface surrounding the bulk introduces additional complexity for the following reasons:
The surface deformation gradient is rank deficient. Thus, the derivations of the surface stress measures and their linearizations are more complicated.
The surface is a two-dimensional manifold embedded in three-dimensional space. Thus, the admissible range for the surface material parameters differs from those of the bulk (see Refs. [92,113] for further details).
Surface tension, which corresponds to a constant energy per unit area of the spatial configuration, introduces a nonstandard term into the surface Helmholtz energy. The surface Helmholtz energy for the strain-free configuration no longer vanishes. Hence, the classical (bulk) growth conditions [114,115] are no longer valid.
The notions of the surface energy, surface tension and surface stress are often used inappropriately in the literature (for a detailed discussion and clarification see Ref. [25] and references therein).
The definition of the surface elasticity theory in the seminal work of Gurtin and Murdoch [32] has a serious defect. This error was reported by the authors in Ref. [116] but has been overlooked by many others.
Gurtin and Murdoch used the letter S to denote the Piola stress and called it the Piola–Kirchhoff stress. The letter S is widely used in the literature to denote the (second) Piola–Kirchhoff stress.
For these reasons, there exists considerable confusion and inconsistency in the literature on surface elasticity (e.g., Refs. [25,26,29,117] aim to report and clarify such issues). In order to clarify the theory and to present a self-contained and consistent framework, a Helmholtz energy for the surface is introduced based on measures of the finite deformation, and the resulting stresses derived. The vast majority of the literature on surface elasticity theory considers the small-strain setting. Thus, the stress measures are linearized at the material configuration to obtain their small-strain counterparts. The result of the linearization of the finite deformation stress measures needs to be interpreted carefully. It is shown here that the linearization of the surface Piola stress at the material configuration is different from that of the surface Piola–Kirchhoff stress . Recall that for the bulk the linearizations of the Piola stress P and the Piola–Kirchhoff stress S at the material configuration are identical.
First a standard hyperelastic material model for the bulk is introduced and corresponding Piola and Piola–Kirchhoff stresses given. Thereafter, in Sec. 6.2, the derivations of the surface Piola and Piola–Kirchhoff stresses corresponding to a hyperelastic material model for the surface are performed. The surface material parameters and their admissible ranges for the linear elastic case are also presented. Furthermore, in order to elucidate the role of surface tension and the dependence of the surface operators upon the curvature within a simplified setting, an example involving the Young–Laplace equation is given. Finally, the hyperelastic material models are extended to the case of thermohyperelasticity in Sec. 6.3.
Remark. The form of the Helmholtz energy on the interface, and indeed in the bulk, can be obtained from fundamental reasoning or from atomistic modeling (see, e.g., Refs. [25,118]). Dingreville and Qu [119] developed a semianalytic method to compute the surface elastic properties of crystalline materials. Moreover, an interface energy can be constructed using the surface Cauchy–Born hypothesis [56]. In Ref. [76] the surface elastic parameters are extracted from ab initio calculations. Similar strategies can be employed to extract the interface parameters. Thus, in general, thermal and mechanical constants can be obtained using inverse parameter identification.
Hyperelastic Material in the Bulk
It is of particular interest to linearize these two stress measures at the material configuration in order to obtain their small strain counterparts. In doing so, recall that F − I = Grad u where u denotes the displacement vector and the infinitesimal strain is defined by ε := [Grad u]sym. The symmetric part of a second-order tensor can be obtained via where denotes the fourth-order symmetric identity tensor. The fourth-order volumetric identity tensor is defined by . Furthermore, represents the volumetric part of a second-order tensor .
where σLin denotes the linearized stress tensor and is the fourth-order tensor of elastic moduli.
where κ := λ + 2/3μ is the bulk compression modulus. For the sake of completeness, the elastic modulus and Poisson ratio are denoted as E and v, respectively. The relation between the various material parameters in the bulk are summarized in Table 3 (see, e.g., the standard Ref. [120] for further details).
Hyperelastic Material on the Surface
where denotes the surface trace operator.
In an identical fashion to the bulk, these stress measures are linearized at the material configuration to obtain their small-strain counterparts. Recall that where denotes the surface displacement vector and the infinitesimal surface strain is defined by The symmetric part of a surface second-order tensor can be obtained as where is the surface fourth-order symmetric identity tensor. The surface fourth-order volumetric, or rather spherical5 identity tensor is defined by . Furthermore, denotes the volumetric part of a surface second-order tensor .
The mistake made by Gurtin and Murdoch [32] in their widely cited paper was to omit the surface tension in the definition of the effective quantities and . This error was rectified in an addendum to their original work [116]. Nonetheless, the error has been reproduced in numerous subsequent publications by many others.
Both of the linearized stress measures given in Eq. (75) are used in the literature. In the absence of surface tension they are obviously equivalent. In the presence of surface tension care must be taken to ensure that they are used in combination with their corresponding balance equation. In the authors' opinion the nonsymmetric version should be used in conjunction with Eq. (21)2.
which shall be compared with Eq. (13) of Ref. [117] where the author explains geometrically the displacement-gradient term. Ru [117] argues that the controversial displacement-gradient term appears due to the fact that the infinitesimal elastic deformation is indeed an incremental deformation superposed on the initial (finite) deformation due to the surface tension. In this manuscript, the reference configuration is assumed to be strain-free as well as stress-free. The surface tension introduces a residual stress in the body and deforms it into a state from which the infinitesimal deformation is externally applied. In this sense, the final deformed configuration may be understood as a hybrid formulation combining the linearized deformation of the bulk and finite (second-order) deformation of the surface (see Ref. [117] for further details of this approach). Such justifications seem to be unavoidable when explaining Eq. (75), nevertheless, the mathematical derivations of this section show rigorously the nature of each term.□
For certain solids the influence of surface tension is negligible when compared to that of surface elasticity (see, e.g., Ref. [122]). Thus, in order to study the elastic effects of the surface as compared to the bulk, it is assumed, for the majority of the remainder of this work, that . The role of surface tension will be explained further in the subsequent Young–Laplace example.
with as the surface compression modulus. Note that the definition of the surface modulus follows in the spirit of κ in the bulk. This parameter has been defined differently in the literature by incorporating a factor of two (see, e.g., Refs. [58,123]). Nevertheless, it seems that the definition here is more stringent. For the sake of completeness, the surface elastic modulus and surface Poisson ratio are denoted as and , respectively. The relation between the different surface material parameters are summarized in Table 4 (see Javili and Steinmann [92] for further details). Note that the surface incompressibility limit, i.e., the upper bound for the surface Poisson ratio , takes the value of 1 and not as in the bulk. This can be justified geometrically.
Finally, the admissible range for the surface material parameters is considered. This range can be determined by analyzing the well-posedness of the boundary value problem. The necessary and sufficient conditions for the loss of well-posedness of the boundary value problem governing linear elastic, homogeneous solids are well known (see, e.g., Refs. [125–134]). They are the loss of ellipticity of the governing equations and the boundary complementing condition, respectively.
Javili et al. [113] show that the sufficient conditions for the well-posedness of the boundary value problem governing a bulk surrounded by a surface are pointwise stability both in the bulk and on the surface. The results are summarized in Table 5, which shows that the strong ellipticity of the surface is necessary but not sufficient for the well-posedness. Shenoy [134] argues that, within the context of an atomistic model, the calculated surface elasticity tensor need not be pointwise stable as the “surface cannot exist independent of the bulk” and that it is the “total energy (bulk + surface)” that needs to “satisfy the positive definiteness condition,” where positive definiteness refers to pointwise stability. This is not entirely correct. The definition of pointwise stability is a local concept. An argument similar to that of Shenoy [134] is applicable when discussing the stability of the weak approximation to the governing equations, for example, as the basis for finite element computations (see Ref. [113] for further details).
Example: Young–Laplace Equation
and finally using the relation yields . Thus, it is clear that the curvature is embedded in the surface divergence operator.
Thermohyperelastic Material
The material models proposed in Secs. 6.1 and 6.2 describes the mechanical response of materials. They do not, however, possess enough structure to capture the thermal behavior and thermomechanically coupled phenomena. The objective of this section is to extend the discussion in the previous sections to account for thermomechanical contributions.
Bulk
The term is the classical Helmholtz energy function of neo-Hookean-type presented in Eq. (64). Thermomechanical coupling is captured by the term −3αk ln J/J, where the thermal expansion coefficient α weights the product of the bulk compression modulus k and the difference between the current and the reference temperatures . The term cF represents the purely thermal behavior in terms of the specific heat capacity cF. The final term 0 accounts for the material specific absolute entropy.
Surface
Interface
The term mimics the classical Helmholtz-energy function of neo-Hookean-type characterized through the two interface Lamé constants and . The term represents surface-tension-like behavior on the interface. Thermomechanical coupling is captured by the term −2 where is the interface thermal expansion coefficient and the interface compression modulus. The interface temperatures in the reference and current configurations are denoted as and , respectively. The interface specific heat capacity represents the purely thermal behavior. Finally, the interface specific entropy is included.
□
Curve
The term mimics the classical Helmholtz-energy function of neo-Hookean-type characterized through a constant . Due to the one-dimensional nature of the curve, one mechanical parameter suffices to model the elastic behavior of the curve. In this sense, the curve parameter can be understood as the elastic-modulus or the compression modulus of the curve and shear modulus is no longer meaningful. A surface-tension-like term is introduced by . The thermomechanical coupling is captured by the term where is the curve thermal expansion coefficient. The curve temperatures in the reference and current configurations are denoted as and , respectively. The curve specific heat capacity represents the purely thermal behavior. Finally, the curve specific entropy is included.
Numerical Examples
The objective of this section is to briefly elucidate the theory presented in Secs. 2–6 using a series of numerical examples. The roles of the surface and interface on the overall response of a body are studied.10 Specific attention is paid to the influence of the material parameters.
These examples are of particular relevance to problems involving materials at the nanoscale. The influence of decreasing specimen size on the overall response can be captured by fixing the bulk material parameters and increasing those of the surface and interface. This is equivalent to fixing the bulk Helmholtz energy and increasing the Helmholtz energy of the surface and interface.
The material behavior in the bulk is characterized by the thermohyperelastic Helmholtz energy function given in Eq. (81). Heat conduction in the bulk is assumed to be spatially isotropic, i.e., k = k i.
The material parameters for the bulk are stated in Table 6 and hold unless specified otherwise.
Lamé constant | μ | 80193.8 | N/mm2 |
Lamé constant | λ | 110743.5 | N/mm2 |
Thermal expansion coefficient | α | 1.0 × 10−5 | 1/K |
Conductivity | k | 45.0 | N/(s K) |
Specific capacity | cF | 3.588 | N/(mm2 K) |
Reference temperature | 0 | 298.0 | K |
Lamé constant | μ | 80193.8 | N/mm2 |
Lamé constant | λ | 110743.5 | N/mm2 |
Thermal expansion coefficient | α | 1.0 × 10−5 | 1/K |
Conductivity | k | 45.0 | N/(s K) |
Specific capacity | cF | 3.588 | N/(mm2 K) |
Reference temperature | 0 | 298.0 | K |
Surface
The following set of examples illustrate the role of the surface on the overall response. Thus, the surface Helmholtz energy presented in Sec. 6.3.2 is chosen. The role of the various terms in the energy is then studied by varying their corresponding constitutive coefficients. In Sec. 7.1.1 the influence of surface tension is demonstrated. In contrast to all of the other terms in the surface Helmholtz energy, it does not vanish at the reference (strain-free) configuration and acts in a fashion similar to an external force field. In this sense, it should be treated as a prescribed force and, therefore, is applied incrementally [90].
The surface theory presented here is not restricted to the isotropic case and anisotropic behavior can be captured if the form of the energy includes the anisotropy via a dependence on the spatial surface normal; see Eq. (38). For the case of surface tension a simple anisotropic model is proposed. The numerical results demonstrate the flexibility and generality of the proposed scheme.
In Sec. 7.1.2 the focus is on the thermomechanical behavior of the surface and its coupling with the bulk. The influence of surface tension is ignored. Furthermore heat conduction on the surface is assumed to be spatially isotropic, i.e.,
Surface Tension
Consider the unit-cube depicted in Fig. 4 (top-left). The cube is discretized using 1728 trilinear hexahedral elements and 2197 nodes. The cube is covered by an energetic surface with a surface Helmholtz energy containing only the surface tension terms in Eq. (82). The surface tension of = μ is prescribed in 20 equal steps.
For a nonvanishing surface tension the reference configuration does not coincide with that of an equilibrium state. Thus, the cube deforms in order to reach an equilibrium state, i.e., to minimize the total energy of the bulk and the surface. Clearly, any deformation introduces an increase in the bulk energy as it is zero at the reference configuration. Nevertheless, the total energy of the system can still decrease by decreasing the surface energy and increasing the bulk energy. Recall that surface tension is equivalent to a constant energy per unit deformed area. Therefore, the surface energy can decrease by decreasing the surface area. As a result, the cube tends to transform into a shape where the surface-to-volume ratio is minimal, i.e., a sphere. The surface also tends to shrink. The transformation of the cube into a spherical shape is illustrated in Fig. 4.
where the material parameter controls the anisotropic contribution and the vector e is a given unit vector characterizing a preferred direction. The anisotropic effects will be greatest where the normal to the surface is parallel to e and will vanish where the normal to the surface is orthogonal to e. The vector e = [0,0,1] together with the reference and deformed configurations are shown in Fig. 5. The anisotropic effect will be maximum on the top and the bottom of the model and will vanish on the lateral walls of the model (elsewhere the effect is between these two extremes). The model is expected, therefore, to shrink more on the top and the bottom and less at the lateral walls of the cube. Consequently, the cube will transform into an ellipsoidal shape instead of the spherical one shown in Fig. 4. The resulting ellipsoidal shapes for different values of are shown in Fig. 5.
Thermomechanical Surface
Consider the three-dimensional strip shown in Fig. 6. The strip is subjected to plane-strain-like conditions across the thickness. In the middle of the specimen is a hole of radius 0.2 mm. Lateral deformations are prevented, i.e., the width of the strip cannot change. A displacement of 0.5 mm is prescribed at the edges resulting in a tensile loading. Thermally homogeneous Neumann boundary conditions are applied to all boundaries of the bulk. The resulting amount of deformation is clearly in the finite strain regime. The displacement loading is applied in 20 equal steps and the total time is 10 ms. The strip is discretized using 576 trilinear hexahedral elements and 1008 nodes. Assume that the surface of the hole in the specimen is energetic. The surface Helmholtz energy is chosen to be of the form given in Eq. (82) with vanishing . The response of the material without the energetic surface is studied first. Thereafter, mechanical and thermomechanical surface effects are detailed, respectively.
Bulk effect.
Consider first a thermomechanical bulk without an energetic surface. Under the prescribed boundary conditions the tensile stress increases. This increase of the tensile stress, or more precisely the divergence of the strain rate, cools the specimen due to the Gough–Joule effect. The presence of the hole leads to a nonuniform stress distribution in the specimen as illustrated in Fig. 7. Consequently, the nonuniform temperature profile depicted in Fig. 8 is obtained.
Mechanical surface effect.
Consider now a purely mechanical surface. That is The mechanical material parameters for the surface are assumed to be The mechanical resistance of the surface influences the bulk, resulting in a different stress distribution when compared to that shown Fig. 7 obtained in the absence of surface effects. The evolution of the stress field is illustrated in Fig. 9. The increase in the tensile stress cools the specimen due to the Gough–Joule effect. Thus, the temperature decreases as the applied deformation increases. The nonhomogeneous stress distribution produces the nonhomogeneous temperature distribution shown in Fig. 10.
Thermomechanical surface effect.
The surface is now endowed with thermal properties. The mechanical properties of the surface are fixed as before by setting In order to better understand the role of the surface thermal properties, the surface conductivity heat capacity , and heat expansion coefficient are investigated separately.
The effect of surface conductivity.
The surface heat conduction coefficient is increased from zero to = 100 mm. The remaining surface thermal parameters are set to zero. Figure 11 shows the temperature distributions at the end of the applied loading for a range of . Increasing the surface conductivity clearly produces an increasingly uniform temperature distribution along the surface. Note that the result shown in Fig. 11 for = 0 is the same as shown in Fig. 10(f).
The effect of surface heat capacity.
The surface heat capacity is now varied from zero to . The remaining surface thermal parameters are set as zero. Increasing the surface heat capacity causes the surface to conserve its initial temperature and resist temperature change. A higher value of the surface heat capacity results in less temperature decrease at the surface. This can be observed in Fig. 12, which shows the final temperature distributions for different surface heat capacities.
The effect of the surface Gough–Joule contribution.
Finally, the effect of the surface heat expansion coefficient on the temperature distribution is studied. The surface heat expansion coefficient is varied from zero to while the remaining surface thermal parameters are specified as zero. According to Eq. (82), increasing the surface heat expansion coefficient increases the surface thermomechanical coupling. As a result of the Gough–Joule effect, the surface cools due to the positive strain rate applied to the surface from the surrounding bulk material. The surface cooling is proportional to . This explains the temperature distributions shown in Fig. 13.
Interface
The “toy” problem is a cube with an internal energetic interface, as illustrated in Fig. 14. The material interior and exterior to the interface is identical. For both the bulk and the interface a thermohyperelastic material model is assumed. The choice of the Kapitza or highly conducting assumption on the interface's behavior is examined.
Kapitza Interface
The interface is initially defined to be a Kapitza interface. Hence, while discontinuities in the temperature field are permitted across the interface, the normal component of the heat flux vector is continuous. The domain is loaded according to the thermal load case depicted in Fig. 14. The top and the lateral sides of the cube are thermally isolated, i.e., = 0. A temperature increase of ΔΘp = 40 K relative to the initial temperature of Θ0 = 298 K is applied to the lower face of the specimen. The resulting temperature distributions at the same instant in time for different Kapitza resistance coefficients are illustrated in Fig. 15. As expected, the jump in the temperature across the interface increases with increasing Kapitza resistance.
Highly Conducting Interface
The interface is now assumed to be highly conducting. Hence, while discontinuities in the normal component of the heat flux vector are permitted across the interface, the temperature field is continuous. The domain is subjected to the mechanical loading conditions illustrated in Fig. 14. The cube is extended by 100% by prescribing equal displacements to the top and bottom faces. Lateral contractions are prevented. The loading is applied quasi-statically and the cube is thermally isolated.
Consider first the response without an interface as shown in Fig. 16(a). In the absence of the interface, the stress state in the cube is homogeneous. The inclusion of the energetic interface results in a change in the distribution of the stress field, as shown in Fig. 16(b). The mechanical resistance of the interface causes the stress to increase above and below the region enclosed by the interface. Half of the midplane of the cube, corresponding to the cases where the interface was omitted and included, is shown alongside one another in Fig. 16(c) to facilitate the comparison of the mechanical response. The mechanical resistance of the energetic interface causes the material surrounded by the interface to deform less than the corresponding region in the absence of an interface (the spatial discretizations are identical).
The nonhomogeneous stress distribution that arises due the mechanical resistance of the interface results in a nonhomogeneous temperature field as shown in Fig. 16(d). The Gough–Joule effect causes the temperature in the body to decrease. The temperature field is clearly continuous across the interface.
Conclusion
A procedure for deriving the equations governing the response of a solid bulk, surrounded by a surface, and intersected by an interface, all of which are energetic, has been presented. The interaction between the surface and the interface has also been accounted for. The structure of the constitutive relations in the various parts of the body was made clear by assigning these regions distinct Helmholtz energies and following a Coleman–Noll-like procedure. The remaining reduced dissipation inequalities suggest the structure of the heat and entropy flux vectors and permissible restrictions on the temperature field in the various parts of the body; particular attention was paid to the interface. The weak formulation of the governing equations was then given and compared to six commonly used restrictions thereof. The theory was elucidated using several numerical examples.
The theory proposed here is often motivated by applications in nanomechanics where size-effects play a critical role. The ratio of the bulk Helmholtz energy to those in the remaining parts of the body implicitly accounts for the size-effects. However, the theory lacks a physical length scale related to the thickness of the interface or surface. Steigmann and Ogden [40] incorporate local elastic resistance to flexure into the Gurtin and Murdoch [32] theory, thereby introducing an inherent length scale and regularizing the associated variational problem. Recent work by Forest et al. [135], comparing first and second strain gradient theories for capillary effects in elastic fluids at small length scales, emphasizes the importance of nonlocality and the associated issue of length scale. An extension to account for the flexural resistance of the surface or interface is, therefore, suggested (see also Refs. [41,42,136]).
Acknowledgment
The support of this work by the European Research Council Advanced Grant MOCOPOLY is gratefully acknowledged. The second author thanks the National Research Foundation of South Africa for their support.
Geometry of Surfaces and Curves
It is enlightening to briefly review some basic terminologies and results on surfaces and curves. For further details the reader is referred to Refs. [89,102–104] among others. Here, some technicalities are borrowed from Steinmann [89].
Surfaces
Clearly the mixed-variant surface unit tensor acts as a surface (idempotent) projection tensor.
The covariant coefficients of the curvature tensor (second fundamental form of the surface) are computed by .
Curves
Clearly the mixed-variant curve unit tensor acts as a curve (idempotent) projection tensor.
The curvature tensors are computed as and
Useful Relations
Reformulation of the Global Working in Terms of the Various Stress Measures
The interface is assumed to be coherent; thus = 0 and = 0. Therefore, = and = and the last four terms vanish. What remains is the global working given in Eq. (22).
Derivation of the Interface Temperature Evolution Equation
It is enlightening to review the procedure to obtain the nonstandard interface temperature evolution Eq. (55)3.
where is the interface heat capacity. Finally, inserting this result into Eq. (D1) gives the equation for the evolution of the interface temperature as in Eq. (55)3.
Pyotr Leonidovich Kapitza (1894–1984) was a physicist and Nobel laureate who first proposed the presence of thermal interface resistance corresponding to a discontinuous temperature field across an interface for liquid helium; see Kapitza [82].
Here and henceforth, the subscripts t and 0 shall designate spatial and material quantities, respectively, unless specified otherwise.
Note that inertial forces are omitted. Furthermore, in the absence of mass flux, mass is conserved according to the standard relations given in Table 1.
The term volumetric is used in analogy to the bulk. Nevertheless, it has a different meaning on the surface. A volumetric deformation in the bulk is a deformation mode that changes the volume uniformly. A volumetric surface deformation, however, is a deformation mode that changes the area uniformly. In this sense the term spherical seems more appropriate. Nonetheless, for the sake of consistency, the term volumetric is used henceforth.
For the sake of simplicity, it is assumed that the surface is an interface between the fluid and an outside vacuum.
Often in the literature is written as sum of the principal curvatures = −[1/r1+ 1/r2] where r1 and r2 denote the principal radii of curvature. Based on this definition, the curvature is negative if the surface curves away from the normal.
For isotropic thermal conduction in the material configuration and .
For isotropic thermal conduction in the material configuration
Numerical investigations of a more general case where the surface, interface, and curve are all considered involve additional complications while providing little additional insight into the problem and, therefore, are not studied here.