To compute effective screened interaction U between 5f-electrons in actinides one has to vary the number of f-electrons in f-shell. To model the effect of screening the transfer between 5f-shell and 6d-shell can be considered. To compute U one has to plot the dependence of one-electron eigenvalue of f-level E{5f} as a function of the number of transferred electrons, x. The U is then equal to the derivative of E{5f} with respect to x when x=0. To evaluate derivative numerically small x's should be considered like 0.1,0.2, etc.

As a first step one has to prepare the RAT files for different x's. As an example let's take Pu. RAT.PU is the file with original configuration 7s2 6f5 6d0. Note that the original RAT.PU file does not contain 6d level information! Therefore, 6d3/2 level with number of electrons = 0 should added. To do it, one changes the input parameter J counting total number of levels from 26 to 27 and adds the following line at the end:

6D3/2 6.0 2.0 1.5 0.0 -0.100000E+00

The position of one-electron energy is just approximate guess and it is not very important.

After this RAT.PU file is ready, configurations with non-zero x can be considered. Consider x=0.1 which will be called RAT.PU1. For this file, one changes to lines in the rat.pu file:

5F5/2 5.0 3.0 2.5 5.9 -0.1802559E+00 6D3/2 6.0 2.0 1.5 0.1 -0.1802559E+00

which means that 0.1 electron is transferred from f-shell to d-shell. Also it is usefull to redirect output to DEN.PU1, INF.PU1 and TAB.PU1 files. The same can be done for other x's.

The second step is to run the atomic program and get TAB.PU and TAB.PU1 files. They will contain the information about level distributions.

The third step is to evaluate numerically the derivative of the level E{5f} with respect to x. For Pu, E{5f,x=0}=-0.223, E{5f,x=0.1}=-0.245, and the estimate for U is (0.245-0.223)/0.1=0.22 Ry.

This example can be found in /import/positron/h2/savrasov/atomdat/PU.