Error in TI derivatives for staggered lambda scaling?

From: Brian Radak (
Date: Tue Jul 21 2015 - 13:41:28 CDT

I know the TI code is generally unused these days, but I'm experiencing
some behavior that, unless I'm very much misunderstanding things, seems
quite wrong.

The default settings use alchElecLambdaStart = 0.5 and alchVdwLambdaEnd =
1.0, meaning that LJ terms scale linearly with alchLambda and
electrostatics scale linearly starting at alchLambda = 0.5 with a slope of
2. With these settings, partition 1 should have no electrostatics for
lambda = [0.5,1.0] and partition 2 should have no electrostatics for lambda
= [0.0,0.5]. As such, on those intervals there is *no scaling* and the
derivative with respect to lambda should be *zero*. However, the TI code
very clearly collects a finite, non-zero value on that interval (presumably
this is just E(lambda = 1) - E(lambda = 0)).

Am I missing something or would this lead to very different results when
changing alchElecLambdaStart and alchVdwLambdaEnd?


Brian Radak
Postdoctoral Scholar
University of Chicago
Department of Biochemistry & Molecular Biology
Gordon Center for Integrative Science, W323A
929 E. 57th St.
Chicago, IL 60637-1454
Tel: 773/834-2812

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:14 CST