Abstract: The prompt neutron decay constant α-eigenvalue can describe quantitatively the neutron time-dependent behavior in nuclear facilities, which has significant value in dynamic research. A frequently used method of calculating α eigenvalue is k-α power iteration method. However, this method fails in deep subcritical systems with reflector or moderator, the reason of which was analyzed in this article. A preprocessing method for low energy neutron was presented. To some extent, the problem in deep subcritical systems with reflector or moderator was solved. Thus k-α power iteration method can be more widely used. The preprocessing method for low energy neutron adjusts the weight by means of variance reduction. The low weight and low energy neutrons are bet and divided, and the high weight and low energy neutrons are divided. Based on the existing neutron photon transport Monte Carlo code NPTS, the α eigenvalue calculation function was developed by using the modified k-α power iterative method. The numerical results show that the calculation results of modified method on NPTS are consistent with MCNP4C in the uncertainty range, and the variance of the same sample number is smaller. For difficulty in deep subcritical systems with reflector or moderator, the effectiveness and accuracy of modified method was validated by comparing with the result of pulse decay method.