A strong dependence on “fglen” is expected: it is a combination of boundary artifacts and errors due to the low grid resolution. The best test for this system is to fix fglen at a reasonable value (e.g., 10) and then systematically increase dime.
Pacific Northwest National Laboratory
From: Ling Liu [mailto:leo.mail.cu@...]
Sent: Monday, March 26, 2012 10:23 PM
Subject: [apbs-users] Issues in the APBS Verification with a Two Atom Problem
I tried to use a very simple two-atom problem to verify the force output by APBS. A +e charge and a -e charge are separated by 2A in sapce. Ion concentration is assumed low. Dielectric constant is assumed to be 1.
Simple Colomb's calculation shows the force should be 347 kJ/mol/A.
Only a short range of fglen gives 347 kJ/mol/A (e.g. fglen 9.6 9.6 9.6 as in the attached input). When I change it to fglen 9.66 9.66 9.66, for example, the electrostatic force is varied very significantly. Some results are listed below.
fglen 10 10 10: 1.179e+004
fglen 9.66 9.66 9.66: 5.557e+003
fglen 9.6 9.6 9.6: 3.477e+002
fglen 9.4 9.4 9.4: -1.533e+004
I also tried APBS for calculating biomolecules such as the MscL channel. The force results also show very large fluctuation with respect to the parameters such as fglen.
Did I make any mistake?