Abstract: Tens of thousands of chemical warfare agents were abandoned in China by the Japanese army after their defeat of invasion of China. Most of the chemical warfare agents have been buried in the soil for more than half a century and the risk of chemical toxicants leakage has increased sharply due to corrosion, so the disposal of these weapons is urgent. The effective disposal and destruction of chemical warfare agents are based on the accurate identification. Prompt gamma neutron activation analysis (PGNAA) is a multi-element analysis technology with low invasion, good safety and high accuracy. In this paper, to study the discrimination potentiality of PGNAA technology on the Japanese chemical projectiles abandoned in China, which can’t be identified by the visual inspection and X-ray imaging technology, the discriminant logic tree of the projectiles was established by investigating and analyzing the data of these abandoned chemical warfare agents, and the simulation model of neutron activation analysis of these projectiles was established by the use of Geant4 toolkit. The prompt gamma spectra of incendiary, yellow projectile and blue projectile excited by ²⁵²Cf source, D-D neutron generator and D-T neutron generator with neutron yield of 1×107 s^-1 were calculated respectively. For the generation of neutrons from ²⁵²Cf source, the Watt fission spectrum was used in our simulations. Monoenergetic 2.5 MeV neutrons and 14.1 MeV neutrons were set as D-D neutron generator and D-T neutron generator respectively in our simulations. The reflection effect of neutron reflectors made of lead or polyethylene on neutrons has also been studied in the simulation. To verify the simulation results, an experimental platform was set up with a D-T neutron generator and a N-type HPGe detector with relative efficiency of 55%. The samples under test were made up of non-toxic and harmless chemical agent, the elements composition and mass fraction were closed to the real chemical warfare agents. The results show that for characteristic elements P and S, the main detected prompt gamma ray energies are 1 266 keV and 2 230 keV, respectively. The analysis of closely-spaced overlapping peaks requires nonlinear least-squares peak fitting, such as the multiple gamma-ray peaks from the hydrogen region. The peak position counts ratio between 882/1763 keV and 1 951/1 959 keV represent an alternative to the existence of S, which is a key criterion for distinguish yellow projectiles from blue projectiles. This study provides the necessary references from the aspects of neutron source selection, neutron reflector material selection, nuclide library establishment and projectile discrimination method for the chemical warfare agents identification system based on PGNAA technology.