On the equivalence of parabolic Harnack inequalities and heat kernel estimates (Q1928336): Difference between revisions

This paper presents a thorough investigation of heat kernel estimates and their equivalence to certain parabolic Harnack inequalities on metric measure spaces which satisfy a volume doubling property. The authors do \textit{not assume}: that the metric is a geodesic metric, that the heat kernel is continuous, nor that the Dirichlet form is conservative. The introduction summarizes parabolic Harnack inequalities and equivalences to heat kernel estimates in more classical and restrictive geometric settings. The geometric setting considered here and the notion of heat kernel in that setting is presented in \S 2.1. In \S 2.2 the generalized notion of a solution to the heat equation, known as a caloric function, is defined, and preliminary results for these functions are demonstrated. The statement of the main results, Theorems 3.1 and 3.2, is contained in \S 3. Theorem 3.1 gives the equivalence of three conditions for metric measure Dirichlet spaces such that all metric balls are relatively compact and a volume doubling condition is satisfied. Any of these three conditions imply the continuity of the heat kernel. Theorem 3.2 specializes to the case in which the metric is geodesic and gives the equivalence of six conditions. The proof of these results is contained in \S 4 and \S 5. The deduction of the Harnack inequality from heat kernel estimates is based on a modification of the arguments of \textit{E. B. Fabes} and \textit{D. W. Stroock} [Arch. Ration. Mech. Anal. 96, 327--338 (1986; Zbl 0652.35052)]. On the other hand the deduction of the heat kernel estimates from the Harnack inequality requires an adaptation of the techniques of \textit{D. G. Aronson} [Bull. Am. Math. Soc. 73, 890--896 (1967; Zbl 0153.42002)] for the on-diagonal bound, a new argument based on \textit{T. Coulhon} and \textit{A. Grigor'yan} [Duke Math. J. 89, No. 1, 133--199 (1997; Zbl 0920.58064)] for the lower bounds, and a modification of the arguments of \textit{W. Hebisch} and \textit{L. Saloff-Coste} [Ann. Inst. Fourier 51, No. 5, 1437--1481 (2001; Zbl 0988.58007)] for the off-diagonal upper bounds. Finally in \S 6 an example is presented which shows that the stronger statement of Theorem 3.2 is not true in general for the weaker geometric setting of Theorem 3.1.