An adaptive-in-temperature method for on-the-fly sampling of thermal neutron scattering data in continuous-energy Monte Carlo codes

The Monte Carlo simulation of full-core neutron transport requires high fidelity data to represent not only the various types of possible interactions that can occur, but also the temperature and energy regimes for which these data are relevant. For isothermal conditions, nuclear cross section data are processed in advance of running a simulation. In reality, the temperatures in a neutronics simulation are not fixed, but change with respect to the temperatures computed from an associated heat transfer or thermal hydraulic (TH) code. To account for the temperature change, a code user must either 1) compute new data at the problem temperature inline during the Monte Carlo simulation or 2) pre-compute data at a variety of temperatures over the range of possible values. Inline data processing is computationally inefficient while pre-computing data at many temperatures can be memory expensive. An alternative on-the-fly approach to handle the temperature component of nuclear data is desired. By on-the-fly we mean a procedure that adjusts cross section data to the correct temperature adaptively during the Monte Carlo random walk instead of before the running of a simulation. The on-the-fly procedure should also preserve simulation runtime efficiency.;While on-the-fly methods have recently been developed for higher energy regimes, the double differential scattering of thermal neutrons has not been examined in detail until now. In this dissertation, an on-the-fly sampling method is developed by investigating the temperature dependence of the thermal double differential scattering distributions. The temperature dependence is analyzed with a linear least squares regression test to develop fit coefficients that are used to sample thermal scattering data at any temperature. The amount of pre-stored thermal scattering data has been drastically reduced from around 25 megabytes per temperature per nuclide to only a few megabytes per nuclide by eliminating the need to compute data at discrete temperatures. The fits are capable of accurately reproducing the thermal scattering probabilities and accounting for thermal binding effects.;A new module was written for the continuous-energy Monte Carlo code MCNP6 to allow for the on-the-fly treatment and was tested for two moderator nuclei of interest to the nuclear community: 1) carbon bound in graphite ( grph) and 2) hydrogen bound in light water (lwtr). A series of problems sensitive to thermal scattering were simulated to test the accuracy of the new method. The thermal neutron flux and secondary energy/angle distributions were very well represented by the new on-the-fly procedure and k-eigenvalue results were comparable (within 1-2 sigma) for all problems tested. The new format only requires a few megabytes of total data storage per nuclide, which is much smaller than the current storage requirements. In addition, the runtime for the new on-the-fly method is comparable to the standard method on the pin cell level, but varies between 10-20% slower for full-core problems. Future optimizations will eliminate this slowdown.