Passive Reduced-Order Modeling via Nonlinear Semidefinite Programming Roland W. Freund Bell Laboratories Room 2C-525 700 Mountain Avenue Murray Hill, NJ 07974-0636, USA e-mail: freund@research.bell-labs.com web: http://cm.bell-labs.com/who/freund/ In recent years, there has been a lot of interest in reduced-order modeling of large-scale time-invariant linear dynamical systems. This surge in interest was mainly driven by the need for dimension reduction of large RCL networks arising in VLSI circuit simulation. RCL networks are passive systems, and usually, it is crucial that reduced-order models of such systems preserve passivity. However, some of the most efficient reduction techniques to not preserve passivity in general. A possible remedy is to employ post-processing techniques that turn non-passive models into nearby passive ones. In this talk, we discuss post-processing techniques for constructing passive reduced-order models via the solution of certain nonlinear semidefinite programs. We discuss a sequential semidefinite programming method for the solution of these nonlinear semidefinite programs, and we present some numerical results. This work is joint with Florian Jarre (Heinrich-Heine-Universitaet, Duesseldorf, Germany)