An algorithm is presented for discrete element method simulations of energy-conserving systems of frictionless, spherical particles in a reversed-time frame. This algorithm is verified, within the limits of round-off error, through implementation in the LAMMPS code. Mechanisms for energy dissipation such as interparticle friction, damping, rotational resistance, particle crushing, or bond breakage cannot be incorporated into this algorithm without causing time irreversibility. This theoretical development is applied to critical-state soil mechanics as an exemplar. It is shown that the convergence of soil samples, which differ only in terms of their initial void ratio, to the same critical state requires the presence of shear forces and frictional dissipation within the soil system. K E Y W O R D S discrete element method, geomechanics, granular media, solids 1 INTRODUCTION Most simulations aim to model a real system with maximal fidelity, limited by practical considerations such as the finite nature of computational resources. However, there is a second category of simulations in which some nonphysical element is deliberately introduced into a simulated system in order to further our understanding of the real system. One relatively commonplace example is discrete element method (DEM) simulations of frictionless particles. These serve as a valuable limiting case for real particle systems which can vary significantly in interparticle friction but are never completely frictionless. The subject of this paper, reversed-time DEM, is within this second category of simulations. While Einstein's theory of special relativity allows for time to speed up or slow down, reversing time is not thought possible according to our current understanding of physics. An algorithm which is deterministic, such as DEM, is not necessarily time-reversible. 1 To the best of the author's knowledge, since the development of DEM by Cundall and Strack 2 more than 40 years ago, no one has explored the possibility of setting DEM within a reversed-time framework. This is somewhat surprising as the time reversibility of molecular dynamics, which is algorithmically related to DEM, has been explored in depth, beginning with Orban and Bellemans 3 in the 1960s. More recent investigations of time reversibility in molecular dynamics have involved the use of integer arithmetic to eliminate round-off errors in floating-point computations. 4 The development of symplectic, reversible integrators for molecular dynamics has been an area of significant research activity. 5-8 These integrators have been proposed for application to both Hamiltonian and non-Hamiltonian dynamical systems. 8 However, one of the simple variants of Verlet, for example, velocity-Verlet, is almost invariably chosen as the This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.