Sharp anisotropic $L^2$-Caffarelli-Kohn-Nirenberg inequalities associated with the Minkowski functional

Abstract

Let $K\subset \RN$ be a convex body containing the origin in its interior, and let $\hK{\cdot}$ be its Minkowski functional. In this paper, we develop an identity-based framework for sharp anisotropic $L^2$-Caffarelli-Kohn-Nirenberg inequalities associated with the anisotropic radial derivative $$ \mathcal R_K(u)(x)=\frac{x\cdot\nabla u(x)}{\hK{x}}, \quad x\in\RN\setminus\{o\}. $$ A key point of the present work is that $K$ is not assumed to be origin-symmetric. Consequently, the Minkowski functional $\hK{\cdot}$ need not be even, and the usual norm-based anisotropic arguments do not apply directly. The main tools are anisotropic $L^2$-Hardy and $L^2$-Caffarelli-Kohn-Nirenberg identities with explicit nonnegative remainders. These identities yield sharp anisotropic $L^2$-Caffarelli-Kohn-Nirenberg inequalities whose best constants depend on the parameter region of $(a,b)\in\mathbb R^2$. We also study the attainability of the sharp constants in a natural completion space and obtain the corresponding extremal functions. As further consequences, we derive sharp anisotropic Heisenberg-type uncertainty principles and max-type anisotropic gradient inequalities. When $K$ is the Euclidean unit ball, our results recover the classical Euclidean $L^2$ theory; when $K$ is origin-symmetric, they are consistent with the usual norm-based anisotropic framework. In particular, the present results extend the sharp $L^2$-Caffarelli-Kohn-Nirenberg theory to general convex bodies containing the origin in their interiors, for which the Minkowski functional may be non-even.