Re: FATAL ERROR: unknown floating-point error, errno = 4

From: Chris Chipot (
Date: Fri Dec 23 2005 - 01:22:32 CST


could you, please, provide me with additional detail about the system
you are running?

In the meantime, I can comment on the values that you used in the ABF
section of your input file:

abf dxi 0.04
abf forceConst 0.1
abf applyBias yes
abf fullsamples 20

The choice of dxi depends on how fast the free energy varies. 0.04 Å
is what I would use for a very fast changing free energy profile.

The forceconst variable controls the walls at the border of the path
you are sampling. 0.1 kcal/mol/Ų constitutes an excessively weak
restraint unlikely to prevent sampling to go beyond the bounds defined
by ximin and ximax.

More worrisome, too small a value of fullsamples is prone to drive
your system out of equilibrium. Keep in mind that the samples are
accrued in dxi-wide bins prior to application of -?F??? ??. In other
words, fullsamples controls the convergence of ?F???. It is inherently
system-dependent and varies as a function of how fast the different
degrees of freedom relax. Too small a value of fullsamples is
equivalent to applying an erroneous average force that will often
overshoot the barriers of the free energy landscape.

Chris Chipot


Chris Chipot, Ph.D.
Equipe de dynamique des assemblages membranaires
Unité mixte de recherche CNRS/UHP No 7565
Université Henri Poincaré - Nancy 1 Phone: (33) 3-83-68-40-97
B.P. 239 Fax: (33) 3-83-68-43-87
54506 Vandoeuvre-lès-Nancy Cedex


      To sin by silence when we should protest makes cowards out of men
                                                    Ella Wheeler Wilcox

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:27 CST