The collision of a low-energy positron, which impinges on a crystalline surface, with a valence electron may result in the emission of a spatially separated time-correlated electron-positron pair. We present a method for calculating the cross section for this positron surface reaction channel, which we briefly refer to as (p, ep) in analogy to electron-induced pair emission (e, 2e). The two-particle final state is represented by a product of an electron and a positron diffraction state coupled by a 'correlation factor', which accounts for the screened Coulomb interaction. The electron-solid and positron-solid quasi-particle potentials are based on first-principles calculations within density functional theory. Numerical (p, ep) results are presented for Cu(111) and compared to their (e, 2e) counterparts. Energy distributions for constant emission angles reflect, to a large extent, the valence electron density of states. In equal-energy (p, ep) angular distributions, the Coulomb interaction produces a central accumulation zone-in contrast to a depletion zone for (e, 2e)-the relative weight and the extension of which are subject to 'matrix element effects'. At larger angles sharp features arise from single-particle surface resonances.