Metallic nanoparticles (NPs) have been shown to be effective in enhancing radiotherapy through photoelectric effect due to the large photoelectric interaction cross-sections of high atomic number atoms. Experiments, however, demonstrated a strong radiosensitizing effect of Gold and Gadoliniums NPs in vivo that is beyond the calculated macroscopic dose enhancement, and even beyond the microdosimetric properties of the Auger electrons following a photoelectric interaction. We propose to describe NPs in cells as a bomb to kill the cell at a probability during a photoelectric event, depending on biological properties of the cell.The biological damage caused by NPs may be categorized into local and non-local. The non-local part corresponds approximately to the macroscopic dose enhancement. Based on the reported in vivo NP concentration, the calculated contribution of this part is negligibly small, especially for MV beams. The local part of the damage is caused by the electron cascade near the NP surface following a photoelectric interaction. We model this effect as a bomb, which gives an upper limit of cell killing for a known NP concentration inside a cell or cell nucleus. We performed a Monte Carlo simulation of irradiating a Gd-based NP with a 250 kVp beam and a 6 MV FFF beam, and calculated the photoelectric interaction probability of GdNP per gray of dose to surrounding medium (water) and the distribution of the number of Auger cascade electrons generated per ionization. The increase in α of the Linear-Quadratic survival model can be calculated using the following equation for the mean number of lethal lesions due to local damage: Nlocal = Nionization·p = NNp·p1Gy·D·p = Δα·D where Δα = NNp·p1Gy·p, Nionization is the total number of NP ionizations that occur in the target (e.g., the nucleus), p is the probability of killing which is set to a value ≤1 when the number of electrons generated per ionization is > 1, and 0 otherwise, NNP is the average number of NPs in the target, D is the dose to the surrounding water due to the X-rays without NPs, and p1Gy is the probability of ionizing an NP when D is 1 Gy.Based on reported in vivo GdNP concentrations of ∼5.5 ppm, for 250 kVp X-rays, the upper limit on the increase in α will be 0.0008 Gy-1 for a 100 μm3 effective DNA volume. If the entire cell of a typical volume of 1000 μm3 is taken to be the target, the upper limit on the increase in α will be 0.008 Gy-1 and 9e-6 Gy-1, respectively for a typical 250kVp beam and a 6MV FFF beam. Radiosensitization may be predicted based on the NP concentrations in cells and the cell properties.The NP radiosensitization effect, which cannot be accounted for by macroscopic dose enhancement alone and has been failed to be described by many other microdosimetry studies, could be potentially described using a bomb model of NP. One of the consequences of this model is that the β does not change due to the presence of NP, which fits most experimental observations.