[gmx-users] g_lie reproducibility (g_lie is lying)

[Mark Abraham]

On 14/06/2012 3:56 PM, Tom Dupree wrote:
> Greetings all,
> I can't manually reproduce g_lie results.
> After raging at excel for a while I think I have found a bug.
>
> Here is my first time point,
> Reported by g_lie to be 35.0073
> > From energy file
> Lj_complex =-130.762
> Coul_complex = -286.746
> My constants specified to g_lie
> Clj = Alpha = lj_const = 0.181
> Cqq = Beta = coul_const = 0.43
> Elj = Lj_solv = -166
> Eqq = Coul_solv = -263
>
> Therefore
> Lj_diff = 35.238
> Coul_diff = -23.746
>
> Hence
> Lj_diff x lj_const = 6.378
> Coul_diff x coul_const = -10.211
>
> And there is no way I can add those to get 35.0073
>
> However I can get this value using the following
> Coul_complex - lj_solv = -120.746
> Lj_complex - coul_solv = 132.2378
> -120.746 x lj_const = -120.746 x 0.181 = -21.855
> 132.2378 x coul_const = 56.86225
> Sum = 35.00726
>
> In short I think the g_lie calculation has swapped its variables, Elj instead of Eqq and vice versa.

I think it's rather more likely you've made a transcription error going 
to Excel. You can check the code of src/tools/gmx_lie.c for exactly this 
kind of purpose, but I can't see any problem with it.

>
> One other thing I have noticed is that when calculating the average g_lie uses the final value twice, is there a reason for this?
>
> e.g.
>
>   994 24.9397
>   996 43.3382
>   998 40.5714
> 1000 40.5585
> 1000 40.5585
>

Inspect your energy file with gmxdump - I think you've managed to 
duplicate the final frame at an earlier point of your workflow. eneconv 
has options to prevent such duplicates occuring during concatenation, 
etc. In any case, g_lie just eats what you give it...

Mark