From b4919756d47a28c35e060a5e957d9c507db99889 Mon Sep 17 00:00:00 2001 From: Sievers Date: Tue, 10 Mar 2020 16:58:47 -0600 Subject: [PATCH] Fixed up delta note --- doc/src/fix_numdiff.rst | 25 +++++++++++++------------ 1 file changed, 13 insertions(+), 12 deletions(-) diff --git a/doc/src/fix_numdiff.rst b/doc/src/fix_numdiff.rst index 8d27346ec8..c538bb618c 100644 --- a/doc/src/fix_numdiff.rst +++ b/doc/src/fix_numdiff.rst @@ -52,18 +52,19 @@ by two times *Delta*. It is important to choose a suitable value for delta, the magnitude of atom displacements that are used to generate finite difference approximations to the exact forces. For typical systems, a value in - the range 1e-xxx to 1e-yyy will probably work well. However, the - best value will depend on a multitude of factors including - the stiffness of the interatomic potential,the - thermodynamic state of the material being probed, and so on. The only - way to be sure that you have made a good choice is to do a - sensitivity study on a representative atomic configuration, sweeping - over a wide range of values of delta. If delta is too small, the - output forces will vary erratically due to truncation effects. If - delta is increased beyond a certain point, the output forces will - start to vary smoothly with delta, due to growing contributions from - higher order derivatives. In between these two limits, the numerical - force values should be largely independent of delta. + the range of 1 part in 1e4 to 1e5 of the typical separation distance + between atoms in the liquid or solid state will be sufficient. + However, the best value will depend on a multitude of factors + including the stiffness of the interatomic potential, the thermodynamic + state of the material being probed, and so on. The only way to be sure + that you have made a good choice is to do a sensitivity study on a + representative atomic configuration, sweeping over a wide range of + values of delta. If delta is too small, the output forces will vary + erratically due to truncation effects. If delta is increased beyond a + certain point, the output forces will start to vary smoothly with + delta, due to growing contributions from higher order derivatives. In + between these two limits, the numerical force values should be largely + independent of delta. .. note::