An alpha spectrometer including a semi-conductor detector is commonly used for measurements of the emergent alpha particles from an air filter, on which was sampled a radioactive aerosol. The alpha spectrometry and the detection efficiency are necessary input information for real radioactivity measurements. The MCNPX code based on the Monte Carlo method has been applied to simulate the detection process in order to obtain spectrum peaks and determine the detection efficiency for modeled geometry. First simulations with MCNPX have been carried out in order to validate the alpha particle energy spectrometry of an electrodeposited solid source and an initial simulated filter model. Furthermore, to improve our first filter model, the real spatial distribution of radioactive aerosols across the filter thickness, found experimentally, is taken into account in a multi-layer filter model. Such an alpha particle distribution allows achieving an adequate simulation of the filter. Comparison between measured and simulated alpha spectra highlights the good agreement in spectral parameters and in detection efficiency even under different aerosol spatial distributions inside the volume of the filter. © 2009 Health Physics Society.