qwolf
| Format: | qwolf R1 R2 |
| Description: | Turns on direct summation method for the electrostatic interactions as described here. In this approach by using charge neutralization scheme one achieve reasonable accuracy for the coloumbic interactions while only considering short-range contributions. It is very reasonable for energy and forces and is much faster than Ewald summation. However, as we showed here, phonons computed with this approach lack long-range effects, which are responsible, for example, for TO-LO splitting. Thermal conductivity error is not very large but can reach 20-30%, depending on parameters. The meaning of parameters is: R1 – exponential decay factor that dumpens charge oscillations (in inverse angstroms), R2 – cut-off radius (in angstroms). The rule of thumb is R1*R2~3, this is because the effective short-range potential within this approach turns out to be an error function, and error function decays to zero when its argument is about ~3. The longer the cutoff, the smaller exponential decay is, and summation is closer to the exact 1/r. |
| Examples: | qwolf 0.3 10.0 |
| Default: | Not set |
rspeed
| Format: | rspeed R |
| Description: | this parameter controls partition between real and reciprocal sums in Ewald summation. Most of the time, it should be left unset, and PhonTS will compute its value for the most efficient calculation of the Ewald sum. Note that this calculation is not a bottleneck, unless one uses low_mem=2, in which case diagonalizations performed on the fly within the main loop of the thermal conductivity calculations. The greater this parameter, the more of the summation is shifted into the reciprocal sum, and shorter the real space cut-off. |
| Examples: | rspeed 0.5 |
| Default: | Computed automatically to optimize performance. |
Ewald_prec
| Format: | Ewald_prec R |
| Description: | Determines the accuracy of the Ewald summation. |
| Examples: | Ewald_prec 10-6 |
| Default: | 10-8 |