A new whole-core transport method is described for 3-D hexagonal geometry. This is an extension of a stochastic-deterministic hybrid method which has previously been shown highly accurate and efficient for eigenvalue problems. Via Monte Carlo, it determines the solution to the transport equation in sub-regions of reactor cores, such as individual fuel elements or sections thereof, and uses those solutions to compose a library of response expansion coefficients. The information acquired allows the deterministic solution procedure to arrive at the whole core solution for the eigenvalue and the explicit fuel pin fission density distribution more quickly than other transport methods. Because it solves the transport equation stochastically, complicated geometry may be modeled exactly and therefore heterogeneity even at the most detailed level does not challenge the method. In this dissertation, the method is evaluated using comparisons with full core Monte Carlo reference solutions of benchmark problems based on gas-cooled, graphite-moderated reactor core designs. Solutions are given for core eigenvalue problems, the calculation of fuel pin fission densities throughout the core, and the determination of incremental control rod worth. Using a single processor, results are found in minutes for small cores, and in no more than a few hours for a realistically large core. Typical eigenvalues calculated by the method differ from reference solutions by less than 0.1%, and pin fission density calculations have average accuracy of well within 1%, even for unrealistically challenging core configuration problems. This new method enables the accurate determination of core eigenvalues and flux shapes in hexagonal cores with efficiency far exceeding that of other transport methods.

The internationally circulated Pool Critical Assembly (PCA) Pressure Vessel Benchmark was analyzed using the PENTRAN Parallel SN code system for the geometry, material, and source specifications as described in the PCA Benchmark documentation. Improvements to the benchmark are proposed through the application of more representative flux and volume weighted homogenized cross sections for the PCA reactor core, which were obtained from a rigorous heterogeneous modeling of all fuel assembly types in the core. A new source term definition is also proposed based on calculated relative power in each core fuel assembly with a spectrum based on the Uranium-235 fission spectra. This research focused on utilizing the BUGLE-96 cross section library and accompanying reaction rates, while examining both adaptive differencing on a coarse mesh basis, as well as the sole use of Directional Theta-Weighted (DTW) SN differencing scheme in order to compare the calculated PENTRAN results to measured data. The results show good comparison with the measured data, which suggests PENTRAN is a viable and reliable code system for calculation of light water reactor neutron shielding and dosimetry calculations. Furthermore, the improvements to the benchmark methodology resulting from this work provide a 6 percent increase in accuracy of the calculation (based on the average of all calculation points), when compared with experimentally measured results at the same spatial location in the PCA pressure vessel simulator.