“…(1) for a log-normal coefficient a(x, x) is square integrable in the probability space, the strong ellipticity still fails since a log-normal random field is not strictly positive from below, which implies that the Lax-Milgram lemma cannot be applied directly. Then we have to study the solution in a space larger than L 2 ðFÞ, where we usually associate a weight either to the probability measure or to each term of the orthonormal basis of the L 2 ðFÞ space [7,12,9,13], see Section 3.2, where F indicates a complete probability space.…”