Application of Reduce Order Modeling to Time Parallelization Ashok Srinivasan 1 , Yanan Yu 2 and Namas Chandra 3 1 Computer Science, Florida State University, Tallahassee FL 32306, USA asriniva@cs.fsu.edu 2 Computer Science, Florida State University, Tallahassee FL 32306, USA yu@cs.fsu.edu 3 Mechanical Engineering, Florida State University, Tallahassee FL 32310, USA chandra@eng.fsu.edu Abstract. We recently proposed a new approach to parallelization, by decomposing the time domain, instead of the conventional space domain. This improves latency tolerance, and we demonstrated its effectiveness in a practical application, where it scaled to much larger numbers of pro- cessors than conventional parallelization. This approach is fundamentally based on dynamically predicting the state of a system from data of re- lated simulations. In earlier work, we used knowledge of the science of the problem to perform the prediction. In complicated simulations, this is not feasible. In this work, we show how reduced order modeling can be used for prediction, without requiring much knowledge of the science. We demonstrate its effectiveness in an important nano-materials application. The significance of this work lies in proposing a novel approach, based on established mathematical theory, that permits effective paralleliza- tion of time. This has important applications in multi-scale simulations, especially in dealing with long time-scales. 1 Introduction Many problems in science are formulated as initial value problems. The initial state of a physical system at some time is given, along with, possibly, some boundary conditions. A differential equation describes how the state changes with time, and possibly space. The problem is solved by iteratively computing the states at successive points in time, using a differential equation solver. We shall refer to each iteration as a time step. A large computational effort can be involved when the state of the system is large, or when the number of time steps is large. In order to reduce the computation time, parallelization is often used, especially with large physical systems. Even when the state space is not large, the computational effort can be large if we need to compute for a large number of time steps. This has been identified