1: \begin{abstract}

2:   We establish sparsity and summability results for coefficient

3:   sequences of Wiener-Hermite polynomial chaos expansions of

4:   countably-parametric solutions of linear elliptic and parabolic

5:   divergence-form partial differential equations with Gaussian random

6:   field inputs.

7: 

8:   The novel proof technique developed here is based on analytic

9:   continuation of parametric solutions into the complex domain.  It

10:   differs from previous works that used bootstrap arguments and

11:   induction on the differentiation order of solution derivatives with

12:   respect to the parameters.  The present holomorphy-based argument

13:   allows a unified, ``differentiation-free'' proof of sparsity

14:   (expressed in terms of $\ell^p$-summability or weighted

15:     $\ell^2$-summability) of sequences of Wiener-Hermite coefficients

16:   in polynomial chaos expansions in various scales of function spaces.

17:   The analysis also implies corresponding analyticity and sparsity

18:   results for posterior densities in Bayesian inverse problems subject

19:   to Gaussian priors on uncertain inputs from function spaces.

20: 

21:   Our results furthermore yield dimension-independent convergence

22:   rates of various \emph{constructive} high-dimensional deterministic

23:   numerical approximation schemes such as single-level and multi-level

24:   versions of Hermite-Smolyak anisotropic sparse-grid interpolation

25:   and quadrature in both forward and inverse computational uncertainty

26:   quantification.

27: \end{abstract}

28: