Nonlocal Coulomb Friction with Unbounded Coefficients: Variational Selection in Multi-Body Elastic Contact, an Existence–Uniqueness Dichotomy, and a Sharp L^∞ Obstruction for the Local Law
The paper, authored by Sami Mehennaoui of the Université 20 août 1955-Skikda and entitled "Nonlocal Coulomb Friction with Unbounded Coefficients: Variational Selection in Multi-Body Elastic Contact, an Existence–Uniqueness Dichotomy, and a Sharp L∞ Obstruction for the Local Law", addresses two questions that arise naturally when the classical theory of static contact with Coulomb friction in linear elasticity is pushed toward physically realistic rough interfaces. The first question asks whether the classical nonlocal dichotomy — existence for arbitrary friction coefficients, uniqueness and convergent iteration for small coefficients — can be established in a coefficient class that contains pointwise unbounded functions, with smallness measured in a norm weaker than L∞, such as an L^p norm or a bounded mean oscillation norm. The second question asks whether, when several elastic bodies interact, the contact problem can be formulated and solved as a non-cooperative game among the bodies in the sense of generalized Nash equilibrium problems, and what exactly such a formulation proves. The paper answers both questions with results that are more subtle than a naive reading suggests. The governing classification is 74M15, 74M10, 74B05, 49J40, 91A10, 35J86, 42B20, 46E35. The introduction distinguishes carefully between two theories whose hypotheses differ in an essential way. For the local law, in which the friction bound is the product of the friction coefficient and the true contact pressure, the classical existence theory of Nečas, Jarušek and Haslinger and of Eck and Jarušek requires the friction coefficient to be small in L∞. The proofs rest on local rectification and shift techniques for the unilateral problem, and the smallness restriction is genuinely needed in the available schemes; non-uniqueness and non-existence for large friction are documented. For the nonlocal law, in which the bound is generated by a mollified pressure, the situation is different and more favourable: the smoothing map yields existence of a solution for an arbitrary friction coefficient, while uniqueness requires smallness. In all existing work the coefficient is a bounded function, and typically continuous or L∞. Pointwise unbounded coefficients are excluded, although they are natural for rough interfaces where tall asperities occupy a very small fraction of the nominal contact area. The setting is fixed with care. The geometry consists of a finite collection of bounded Lipschitz domains, each with a boundary decomposition into a Dirichlet part, a Neumann part, and contact patches, all disjoint up to surface measure zero. Each contact patch is assumed to be of class C^{1,1} so that the outward unit normal is Lipschitz. Matched interfaces are imposed through bi-Lipschitz bijections between contacting patches, so that all fields on one patch are pulled back to a common parameter domain. Initial gaps are prescribed. Elasticity tensors are uniformly elliptic, and the bilinear forms satisfy the second Korn inequality and Poincaré's inequality, giving coercivity on the appropriate subspaces. The trace inequality gives control of surface norms by bulk norms. The global admissible set is defined by unilateral non-penetration constraints at every interface, and it is closed and convex. For each player, the admissible set is exactly the slice of the global set obtained by fixing the displacements of all other bodies, so the constraints are jointly convex and shared, which is the defining structure of a jointly convex generalized Nash equilibrium problem. A Tresca friction bound is a family of nonnegative functions, one per interface. The total energy is the sum of elastic energies, load terms, and friction terms, where the friction term at each interface is the integral of the bound times the magnitude of the relative tangential displacement. The payoff of each player is its own elastic and load energy plus the friction terms at its interfaces. Both are finite on the admissible set because the loads are in the dual space and traces lie in the appropriate Sobolev space. A Tresca generalized Nash equilibrium is defined in the standard game-theoretic sense: no player can decrease its payoff by changing its own displacement while keeping the others fixed. A variational Tresca equilibrium is a displacement that minimizes the total energy over the global admissible set, equivalently one that satisfies the associated variational inequality. The first structural result is Lemma 3.1, the exact potential identity. It states that for every player, every admissible displacement of that player, and every fixed displacement of the others, the difference between the payoff of the player at the two configurations equals the difference of the total energy, plus a bracket that depends only on the fixed displacements of the other players. In particular the bracket does not depend on the displacement of the player, so the game is an exact potential game with potential equal to the total energy, in the sense of Monderer and Shapley. Because the constraints are shared and jointly convex, it is also a generalized potential game in the sense of Facchinei, Piccialli and Sciandrone. The proof observes that the interface friction term for a pair appears in the payoff of one player as a function of the relative displacement, and identically in the payoff of the other player, because the relative displacement changes sign but its magnitude does not. The elastic and load terms of one player appear identically in both the payoff and the total energy. All remaining terms of the total energy involve only the other players and sum to the bracket. Theorem 3.2 establishes existence and uniqueness of the variational Tresca equilibrium. The total energy is strictly convex because the elastic part is coercive and the friction term is convex, being an L¹-type norm of a linear map of traces. It is continuous, hence weakly lower semicontinuous, and coercive because the elastic part dominates. The admissible set is nonempty, closed and convex. The direct method of the calculus of variations gives a unique minimizer, characterized by the variational inequality. For the Nash property, one fixes a player and an alternative displacement; the slice property ensures the alternative is admissible, and the potential identity converts the minimality of the total energy into the minimality of the player's payoff. No invariance condition, Mosco stability hypothesis, or fixed-point theorem is required. Proposition 3.3 gives a global a priori bound and a global Lipschitz dependence on the friction bound. The a priori bound states that the energy norm of the equilibrium is controlled by the load norm divided by the coercivity constant, because the elastic part dominates and the friction term is nonnegative. The Lipschitz estimate states that the difference of two equilibria corresponding to two different friction bounds is controlled in the energy norm by the L∞ norm of the difference of the bounds, times a constant involving the trace constant and the coercivity constant. The proof writes the variational inequalities for the two equilibria, uses each as a test function in the other, adds, and applies the reverse triangle inequality, Cauchy–Schwarz, and the trace bound. Corollary 3.4 derives traction bounds and stability from the preceding estimates. The normal traction functional at each interface is bounded in the dual of the appropriate boundary space by the energy norm times the lifting constant, and the difference of traction functionals corresponding to two different friction bounds is controlled by the Lipschitz estimate. The normal traction is defined rigorously as a boundary functional. At any Nash equilibrium the interior equilibrium equations hold: testing the minimality of a player with variations supported away from the interface gives the weak form of the elasticity equation in the bulk. A Lions–Magenes-type boundary space is introduced, with a bounded lifting from the boundary to the bulk, constructed using the Lipschitz regularity of the normal vector, Kirszbraun's theorem for componentwise extension, and the bounded right inverse of the trace operator on Lipschitz domains. The normal traction functional is then defined by duality: it pairs the bulk stress with the lifting of a boundary test function. This is well defined because two liftings differ by an element of the kernel of the trace, where the interior equilibrium equation applies. The pressure functional is the negative of the normal traction. A measure representation is noted: since the functional is nonnegative on nonnegative tests, it is a positive Radon measure by the Riesz–Schwartz theorem. Proposition 3.5 proves action–reaction balance and weak Signorini conditions for the variational equilibrium. The action–reaction balance states that the normal traction functionals computed from the two bodies at an interface are negatives of each other as functionals on the boundary space. The proof fixes a test function, constructs a variation that moves one body normally while keeping the other fixed, observes that all constraints and friction terms are unchanged, and uses minimality. The compression condition states that the pressure functional is nonnegative on nonnegative tests, proved by testing with variations that open the gap. The support condition states that the pressure is concentrated on the contact set, proved by a localization argument using uniform slack and a density approximation. The tangential balance states that the friction tractions admit balanced selections with opposite signs on the two bodies, which follows because the friction term is the same function of the relative slip for both players, and the Euclidean norm has an odd subdifferential. Definition 3.6 formalizes what it means for a Nash equilibrium to be mechanically admissible: for every interface there must exist multipliers and measurable
Authors
- Mehennaoui Sami (ORCID: https://orcid.org/0009-0000-2483-7604)
Institutions
- University of Skikda (DZ)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-16
- DOI
- https://doi.org/10.5281/zenodo.24287711
- Primary Topic
- Contact Mechanics and Variational Inequalities
- Type
- preprint