Description
In negative ion neutral beam injection (NBI), H⁻/D⁻ ions are produced in the source region and subsequently neutralised to form the neutral beam. The dominant production mechanism is the conversion of H atoms to H⁻/D⁻ ions on a caesiated plasma grid (PG). To date, fully kinetic particle-in-cell (PIC) simulations have struggled to self-consistently reproduce the H⁻/D⁻ ion density observed in the bulk plasma, typically underpredicting experimental values. This discrepancy arises from the common assumption of a uniform background H atom density, imposed for computational efficiency, which leads to premature H⁻/D⁻ emission and the formation of an artificially deep virtual cathode near the PG. The resulting negative space-charge potential suppresses ion transport into the bulk plasma. In this work, two modifications were explored to achieve a more self-consistent treatment of virtual cathode formation and evolution: (i) modelling emitted H⁻/D⁻ ions with a truncated Maxwellian distribution, and (ii) introducing an artificial proton drift toward the PG to offset early ion emission. For a plasma density of 2 x 10^17m-3, these approaches yield an H⁻ density of 4.3 x 10^16m-3 in the bulk plasma, representing an 89% increase relative to the baseline model. While experiments report a corresponding density of 6 x 10^16m-3, simulations employing only the proton drift while retaining a Maxwellian emission model produce an H⁻ density of 4.8 x1016 m-3, substantially closer to experimental observations than previously achieved. These results indicate that the artificial early formation of the virtual cathode in simulations is the main reason for the bulk plasma H⁻/D⁻ ion density discrepancy between simulations and experiments.