The Cutoff File for Jaguar Calculations

The cutoff file specifies parameters to be used for the various iterations of an SCF calculation. The file to be used is determined by the CUTOFFFILE entry in the input file, as described in General Description of the Jaguar Input File. If the input file has no such entry, Jaguar uses the file default.cutoff from the data directory. If the CUTOFFFILE entry is accurate.cutoff, solvent.cutoff, or quick.cutoff, the program interprets the setting as default.cutoff.

The first line of a cutoff file contains a character string that includes the version number of Jaguar. This should be cutv followed by four digits giving the version number times 100. Leading zeroes are added if necessary. A comment on the same line can follow the version string.

The next five lines each have five numbers. Each line describes a particular level of accuracy to be used for the calculation. The first line provides the information necessary to run a calculation with all ultrafine pseudospectral grids and with “tight” cutoffs, and corresponds to an accuracy level setting of Ultrafine from the GUI, as described in SCF Accuracy Level, or to the keyword setting iacc = 1 in the gen section of the input file, as described in Keywords for SCF Methods in the Jaguar Input File. The second line gives the parameters for the accurate level (iacc = 2), while the third line provides information for the quick level (iacc = 3). The last two lines are filled with zeroes, since they are required, but are not yet used.

In each of these rows, the columns describe which cutoff sets are used for various SCF iterations. The cutoff sets themselves are provided later in the file, and dictate the level of analytic corrections, the grid, and the non-default values of the gen section cutoff keywords (cut1, for example). The cutoff sets are described in more detail below. The columns reflect a scheme in which calculations are broken down into preliminary and final sets of iterations. The iterations from the beginning of the first SCF calculation in a run are considered to be part of the preliminary set, while the iterations from the end of the first SCF calculation, or from any subsequent set of SCF iterations, are considered to be part of the final set. For instance, for a solvation calculation, the SCF iterations for the analysis of the converged gas phase wave function are preliminary iterations followed by final iterations, while the SCF iterations for all subsequent SCF calculations (those including the solvent effects) are final iterations. Jaguar determines how many iterations are preliminary and how many are final for the initial SCF calculation.

The number in the first column in each of the five accuracy level lines dictates the cutoff set used for the first iteration in the preliminary sequence: if the number is a 1, the first cutoff set listed in the file is used; if it is a 5, the fifth is used, and so on. The number in the second column provides the cutoff set used for updates during the preliminary sequence of iterations. The third and fourth columns describe the cutoff sets used for the first and updating iterations in the final sequence, respectively. Finally, the last column dictates the cutoff sets used for non-SCF calculations, as for gradient calculations.

The first six lines of the default.cutoff file, which illustrate these points, are:

cutv0300
1 1 1 1 7   max. accuracy (prelim,prelim update,final,final update,gradient)
3 5 1 4 7   accurate
5 6 2 6 8   quick/solvent
0 0 0 0 0
0 0 0 0 0

The rest of the .cutoff file consists of the cutoff sets. Each set is specified by one line with four integers, sometimes followed by lines containing explicit cutoff keyword values, and ending with a blank line. The four integers represent the variables jcor and kcor (described below), the grid number, and the number of cutoff values to follow immediately below. The grid number should be 1 for the coarse grid, 2 for the medium grid, 3 for the fine grid, and 4 for the ultrafine grid, 5 for the charge grid, 6 for the gradient grid, 7 for the electron density cubic grid, 8 for the DFT medium grid, or 10 for the DFT gradient grid, where these grids are specified by the keywords gcoarse, gmedium, gfine, gufine, gcharge, ggrad, geldens, gdftmed, and gdftgrad. Grid and Dealiasing Function Keywords in the Jaguar Input File contains more information on these keywords.

The next lines specify each cutoff by number (e.g., 22 for the variable cut22) and value. Thus, the cutoff set:

5 2 4 3       set 3
  21 1.0e-3  
  22 3.0
  24 1.0e-2

means that jcor is 5, kcor is 2, the ultrafine grid is used, and that three cutoff values which differ from the defaults follow. The next three lines set the cutoff values cut21, cut22, and cut24. If you need more information on cutoffs, contact Schrödinger.

The variables jcor and kcor determine what analytic corrections are calculated for a particular SCF iteration. The meanings of their possible values are shown in Table 1. The variables a, b, and c in the table refer to distinct atoms.

To perform an all-analytic calculation, you can set the keyword nops in the gen section of the input file to 1. All-analytic calculations use the cutoff keyword values in the .cutoff file, but ignore the jcor, kcor, and pseudospectral grid information.