Re: Gradient tolerance nan for A three-atom simulation

From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Mon Jan 12 2009 - 12:17:49 CST

Hi Bin,
I would expect that you do indeed have some bonds stretching abnormally.
What CG model are you using? Have you inspected the trajectory around
the crash to see what is going on?
What cg model are you using?
1 fs timesteps are quite short for cgmd, so I doubt that that is the
problem.
Best,
Peter

BIN ZHANG wrote:
> Hi, Peter and all:
>
> I have a related question about exclusion:
>
> If I got the error " Bad global exclusion error " even exclusion 1-2 is
> used, does it mean some of the bonds between 2 atoms are longer than the
> cutoff? That would be really weird. ( I'm using CGMD with timestep 1 fs.
> What's more, I got this error after 5ns' run)
>
> Also, do you think decreasing the timestep would help in that situation?
>
> Thanks a lot.
>
> Bin
>
>
> On Jan 8, 2009, at 11:19 AM, Peter Freddolino wrote:
>
>> Hi Jiaqi,
>> it is important to understand the source of the bad global exclusion
>> count error: Two atoms which should have an exclusion are not on
>> neighboring patches. This happens a lot in coarse grained systems
>> because they have such long bonds. What do you have exclude set to? And
>> how far apart are your particles?
>> Peter
>>
>> accomp lin wrote:
>>> Hi Gumbart
>>>
>>> Thanks very much for your help.The problem had been solved when I
>>> altered the angle a little from 180.(My theta0 is also 180 degree).
>>>
>>> However , I got another error , it said:
>>>
>>> Reason: FATAL ERROR: Bad global exclusion count!
>>>
>>> The problem occurs at different time step with different set of
>>> initial coordinates of the atoms. I then I changed the cutoff ,
>>> switchdist, and pairlistdist to be very little and set the parameter
>>> of LJ potential zore, the simulation ran successfully. But when I set
>>> the cutoff, swithdist and pairlistdist to be normal, the error came
>>> back again. Wht does these parameter matter, since the LJ potential is
>>> already zore? How come it to be a problem in my three-atoms system?
>>>
>>> Thanks, best.
>>>
>>> Jiaqi.
>>>
>>>
>>> --- *09年1月6日,周二, JC Gumbart /<gumbart_at_ks.uiuc.edu
>>> <mailto:gumbart_at_ks.uiuc.edu>>/* 写道:
>>>
>>> 发件人: JC Gumbart <gumbart_at_ks.uiuc.edu <mailto:gumbart_at_ks.uiuc.edu>>
>>> 主题: RE: namd-l: Gradient tolerance nan for A three-atom simulation
>>> 收件人: accomplin_at_yahoo.com.cn <mailto:accomplin_at_yahoo.com.cn>,
>>> namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>
>>> 日期: 2009,16,周二,1:24下午
>>>
>>> I wonder if this happens because theta=180 is an indeterminate
>>> problem? What if you try offsetting the third atom from the line
>>> of the other two? What is the theta0 you’ve set in the parameter
>>> file?
>>>
>>>
>>>
>>>
>>>
>>> *From:* owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu]
>>> *On Behalf Of *accomp lin
>>> *Sent:* Monday, January 05, 2009 11:10 PM
>>> *To:* namd-l_at_ks.uiuc.edu <mailto:namd-l_at_ks.uiuc.edu>
>>> *Subject:* namd-l: Gradient tolerance nan for A three-atom simulation
>>>
>>>
>>>
>>> Dear NAMD developer:
>>>
>>>
>>>
>>> I've modifed a namd source code with New bond energy
>>> potentials for my Coarse Grained models. After I recompiled it
>>> successfully on linux, I constructed a three-atom model to test my
>>> new NAMD. Only bond and angle are included in the psf file, the
>>> three atoms are aligned straight, each of them have a mass of 400
>>> amu. But when I run the program, the simulation held still and had
>>> the following error.
>>>
>>>
>>>
>>> /INITIAL STEP: 1e-06 //
>>> /GRADIENT TOLERANCE: nan //
>>> /BRACKET: 0 0 nan nan nan// /
>>>
>>>
>>>
>>> After several try and visiting the Mailing list I found I was the
>>> Angle term tha caused the problem. When I turned it off or set the
>>> /Ktheta /zore the problem dissappeared. I already set a large
>>> enough boundry box the contain all the atoms within it. I know it
>>> might be a little inappropriate to use NAMD do to such a simple
>>> simulation, for it can cause various problems. But I really want
>>> to know how does it work and how to solve these problems. So
>>> please help, thank you.
>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> 好玩贺卡等你发,邮箱贺卡全新上线!
>>> <http://cn.rd.yahoo.com/mail_cn/tagline/card/*http:/card.mail.cn.yahoo.com/>
>>>
>>>
>>> ------------------------------------------------------------------------
>>> 好玩贺卡等你发,邮箱贺卡全新上线!
>>> <http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>
>>
>
>
>
>
>
>
>
>
>
> -------------------------------------------------------------
> The tree of liberty must be refreshed from time to time with the blood
> of patriots and tyrants.
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:50:22 CST