(A Lebesgue measure on Rd). The converse question to what extent this restricted mean value property implies harmonicity has a long history (we axe indebted to I. Netuka for valuable hints). Volterra [26] and Kellogg [20] noted first that a continuous function f on the closure U of U satisfying (*) is harmonic on U. At least if U is regular there is a very elementary proof for this fact (see Burckel [7]): Let g be the difference between f and the solution of the Dirichlet problem with boundary value f . If g#0 , say a=sup g(U)> 0, choose x6{g=a} having minimal distance to the boundary. Then (*) leads to an immediate contradiction. In fact, for continuous functions on U the question is settled for arbitrary harmonic spaces and arbitrary representing measures #x # ~ for harmonic functions. If f is bounded on U and Borel measurable the answer may be negative unless restrictions on the radius r(x) of the balls B ~ are imposed (Veech [23]): Let U = ] I , 1[, / (0 )=0 , f = i on ] -1 ,0[ , f= l on ]0, 1[, 0~B x for x?t0 (similarly in R d, d/>2)! There are various positive results, sometimes under restrictions on U, but always under restrictions on the function x~-+r(x) (Feller [9], Akcoglu and Sharpe [1], Baxter [2] and [3], Heath [17], Veech [23] and [24]). For example Heath [17] showed for arbitrary U that a bounded Lebesgue measurable function on U having the restricted mean value property (.) is harmonic provided that , for some e>0, cd(x, CU)<r(x)< (1-~)d(x, CU) holds for every x6U. Veech [23] proved that a Lebesgue measurable function f on U
Read full abstract