Rearrangement inequalities for functionals with monotone integrands (Q2491629)

The authors consider integral functionals of the form \[ I_F(u_1,\dots,u_m) = \int_{X} F(u_1(x),\dots ,u_m(x))\,dx \] and \[ J_F(u_1,\dots ,u_m) = \int_{X^m} F(u_1(x_1), u(x_2),\dots, u(x_m))\prod_{i<j} K_{ij}(d(x_i,x_j))\,dx_1\dots dx_m, \] where \(X\) is \({\mathbb R}^n,\;\) the sphere \(S^n,\) or the hyperbolic space \(H^n,\,dx\) is the canonical measure on \(X,\) and \(d(x,y)\) the canonical distance on \(X.\) The functions \(u_1,\dots ,u_m\) are nonnegative Lebesgue measurable functions on \(X\) which are small at infinity in an appropriate sense, the \(K_{ij}\) are nonnegative decreasing functions on \([0,\infty)\), and \(F\) is a real-valued Borel measurable function on \([0,\infty)\times \dots \times [0,\infty).\) Let \(u^*\) denote the symmetric decreasing rearrangement of \(u.\) When \(X = \mathbb R\) and \(F(y_1,y_2) = y_1 y_2,\) there are inequalities \(I_F(u_1,u_2) \leq I_F(u_1^*, u_2^*)\) and \(J_F(u_1,u_2) \leq J_F(u_1^*, u_2^*).\) The \(I\)-inequality is due to Hardy and Littlewood, the \(J\)-inequality to F. Riesz. The Hardy-Littlewood and Riesz inequalities have been extended and generalized in many directions. The immediate precursors of the article under review are papers by F. Brock (2000) and C. Draghici (2005) in which it is proved that \(I_F(u_1,\dots ,u_m),\) respectively \(J_F(u_1,\dots ,u_m)\), does not increase when the \(u_i\) are replaced by their symmetric decreasing rearrangements, provided \(F\) is continuous and nonnegative, \(F(0) = 0,\) and \(F\) satisfies the inequality \[ F(y + he_i + ke_j)\, + F(y) \geq F(y+he_i) + F(y+ke_j), \] for all \(y\in [0,\infty) \times \dots \times [0,\infty),\) all positive numbers \(h\) and \(k,\) and all indices \(i \neq j,\) where \(e_i\) is the \(i\)th coordinate vector in \({\mathbb R}^m.\) The authors call such functions \(F\) supermodular. The term Schur-convex has also been used in the literature. The nonnegativity assumption on \(F\) can be replaced by an integrability condition on the negative part of \(F.\) The authors' principal results assert that the theorems of Brock and Draghici still hold when \(F\) is discontinuous. This does not readily follow from the proofs in the continuous case, so a new tool is required. The authors devise one via a factorization of \(F\) into continuous and monotonic parts, inspired by a theorem in statistics of A. Sklar. The authors also prove some sufficient conditions for equality to hold in their inequalities.