1: \begin{abstract}

2:  The Truncated Nonsmooth Newton Multigrid (TNNMG) method is a robust and efficient solution method

3:  for a wide range of block-separable convex minimization problems, typically stemming from discretizations

4:  of nonlinear and nonsmooth partial differential equations.  This paper proves global convergence

5:  of the method under weak conditions both on the objective functional, and on the local inexact

6:  subproblem solvers that are part of the method.  It also discusses a range of algorithmic choices

7:  that allows to customize the algorithm for many specific problems.

8:  Numerical examples are deliberately omitted, because many such examples

9:  have already been published elsewhere.

10: \end{abstract}

11: