Elastic Scattering Corrections

In order to make a library of multigroup constants, as defined by Eqs.~(\ref{eq23}) and (\ref{eq24}), it is necessary to choose a model-weighting flux. In an actual system, the real flux may be quite different from this model. Because the flux will usually vary from region to region, the mismatch between the actual and model flux cannot always be removed by improving the library flux. For the vector cross sections (e.g., capture, fission), the size of a group can usually be made small enough so that the weighting flux difference is not too important. However, the elastic downscatter for the heavier materials always takes place from a range of energies very close to the bottom of the group. If $r$ is the ratio of the flux near the botton of a group to the group-average flux, the elastic removal cross section is very sensitive to any difference between $r$ calculated for the actual flux and $r$ calculated for the model flux.

Starting with the 1DX code (Ref. 19), various attempts have been made (Refs. 20 and 21) to correct the elastic removal based on the shape of the computed flux using an iterative scheme. However, the complicated consequences of resonances in the cross sections make this dangerous except in regions where the calculated flux varies smoothly. Therefore, TRANSX does not attempt to correct for variations that only show up in one or two groups. Instead it smoothes the flux by averaging over several groups above and below the group in question. This results in an effective slope for the flux near the bottom of the group, which can be used to correct the removal as shown below. The corrected cross-section set should give improved answers for any responses that depend on the average removal effect such as k$_{\rm eff}$, but the actual flux shape near a particular resonance may not be improved at all.

The removal from group $g$ to group $g^{\prime}$ can be written

R_{g^{\prime} \leftarrow g}^M = \int_g du\,\int_{g^{\prime}}du^{\prime} \sigma_e (u^{\prime}{\leftarrow}u)\, \phi^M (u)~~,

where $u$ is lethargy, $\sigma_e$ is the elastic scattering differential cross section, and $\phi^M$ is the model flux. If the actual flux at the bottom of the group has a different slope than the model flux,

\phi (u) = [ a-b(u_g -u) ] \phi^M (u)~~,

where $u_g$ is the lethargy at the bottom of group $g$, the removal rate becomes

R_{g^{\prime} \leftarrow g} = R_{g^{\prime} \leftarrow g} [ a-b\gamma_{g^{\prime} \leftarrow g} ]~~,

where

\gamma_{g^{\prime} \leftarrow g} = \frac{\displaystyle \int_g du \int_{g^{\prime}} du^{\prime} \,(u_g{-}u)\, \sigma_e (u^{\prime}{\leftarrow}u)\, \phi^M (u)} {\displaystyle\int_g du\int_{g^{\prime}} du^{\prime}\, \sigma_e (u^{\prime}{\leftarrow}u)\, \phi^M (u)}~~.

The quantity $\gamma_g = \sum_g \gamma_{g^{\prime} \leftarrow g}$ is available on the MATXS library, so this $\gamma$ is used for all groups and Legendre orders to obtain the corrections to the elastic scattering matrix elements.

It is then assumed that the smoothed flux ratio can be represented by

\ln (\phi/\phi^M) = x_0 + x_1 (u_g{-}u) + x_2 (u_g{-}u)^2~~,

in the group $g$. The coefficients are computed using a least squares fit to six values of the ratio assigned to the lethargy center of each of the six groups (three above $u_g$ and three below). The parameters $a$ and $b$ are determined at the bottom of the group, and the smoothed value of the ratio at the center of the group is used to find the change in cross section that will give the desired change in removal rate.

The method is very stable. It avoids the divergences in removal cross sections often seen with other methods, and it usually converges adequately in one iteration.

TRANSX HyperText Manual
TRANSX HyperText Help Package
T-2 Nuclear Information Service