Re: ABF simulation with extendedLagrangian on keeps crashing

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Wed Apr 20 2011 - 11:22:37 CDT

Indeed the agonizing moments look quite scary.

You have dummy atoms in the dihedral definitions, which stay fixed in
space. The rest of the system, however, can drift under the effect of the
PME truncation error, Langevin forces, etc as you know already. Do the
other atoms fly very far away from the dummy ones? Maybe that could be why
you get the dihedrals out of control?

Giacomo

---- ----
  Giacomo Fiorin - Postdoctoral Researcher
  ICMS - Institute for Computational Molecular Science - Temple University
  1900 N 12 th Street, Philadelphia, PA 19122
  giacomo.fiorin_at_temple.edu
---- ----

2011/4/20 Ajasja Ljubetič <ajasja.ljubetic_at_gmail.com>

> Hi,
>
> Just for fun, I run from the last restart file, writing each step to the
> DCD. The last agonizing moments (in graphical form) can be found here<http://www.ijs.si/ijs/dept/epr/ajasja/mails/ABF/crash.gif>.
> The two atoms included in the ABF are the CB atom and the last oxygen of the
> side-chain. The potential and kinetic energies of the fictive particles
> are getting larger, but stay finite. If to much energy is transited into the
> spring, then I should try with a heavier particle and a weaker
> spring constant, right?
>
>
>
> theta
> step potential kinetic
> 11000 2.15652342996330e-002 3.18990492159784e-004
> 792187 1.10846675128305e+002 2.83213123555579e+001
>
> phi
> step potential kinetic
> 11000 1.31686077436764e-001 1.69791759804012e-002
> 792187 1.78364015718248e+002 2.96619241022563e+002
>
>
> The simulation normally crashes somewhere between 2 ns and 15 ns (I have
> done many runs allready:)
> Perhaps the fixed dummy atoms in the definition of the colvars could be
> problematic as well?
>
> Best regards,
> Ajasja
>
> colvar {
> name theta
>
> width 5
> lowerboundary 0
> upperboundary 180
> lowerWallConstant 10
> upperWallConstant 10
>
> extendedLagrangian on
>
> angle {
>
> group1 { dummyAtom (3.283, -0.213, 99.366) } ; #100 A above CB
> group2 { atomnumbers 204 } ;#CB
> group3 { atomnumbers 173 } ;#O1
> }
> }
>
> colvar {
> name phi
>
> width 5
> lowerboundary -100
> upperboundary +100
> lowerWallConstant 10
> upperWallConstant 10
>
> extendedLagrangian on
>
> dihedral {
> group1 { atomnumbers 173 } ;#O1
> group2 { atomnumbers 204 } ;#CB
> group3 { dummyAtom (3.283, -0.213, -100.63) } ; #100 A below CB
> group4 { dummyAtom (13.283, -0.213, -100.63) } ; #100 A below and 10 in
> the x direction from CB
> }
> }
>
>
>
>
>
>> I still found no reason why the ABF simulation is crashing though.
>>
>> Best regards,
>> Ajasja
>>
>> 2011/4/19 Ajasja Ljubetič <ajasja.ljubetic_at_gmail.com>
>>
>>> Hi,
>>>
>>> I'm not getting any output of the PMF in the metadynamics simulations,
>>> for which I have also kept extendedLagrangian on.
>>> These simulations run OK, but don't give me a PMF output. (I should
>>> have probably mentioned this in a separate e-mail as it is unrelated to the
>>> ABF crashing).
>>>
>>> For the ABF simulations: I looked through the colvar traj file, but I
>>> don't notice anything unusual. The energies are low and the r_XXX are
>>> similar to the colvar values. I'm attaching the traj file.
>>>
>>> Thank you & best regards,
>>> Ajasja
>>>
>>>
>>>
>>> On Tue, Apr 19, 2011 at 18:41, Giacomo Fiorin <giacomo.fiorin_at_gmail.com>wrote:
>>>
>>>> Hi Asjasja, did you check the trajectory of the moving restraint centers
>>>> aka fictitious masses? Because you have few degrees of freedom in an in
>>>> vacuo system, the chance of transmitting energy to the harmonic spring may
>>>> be high.
>>>>
>>>> So you may have too much energy transmitted to the harmonic spring, and
>>>> the centers may have gone far from the actual values of the colvars. This
>>>> could explain why you got no PMF output (the colvars are way off the grid)
>>>> and in some cases instability (harmonic forces too strong).
>>>>
>>>> Giacomo
>>>>
>>>> ---- ----
>>>> Giacomo Fiorin - Postdoctoral Researcher
>>>> ICMS - Institute for Computational Molecular Science - Temple
>>>> University
>>>> 1900 N 12 th Street, Philadelphia, PA 19122
>>>> giacomo.fiorin_at_temple.edu
>>>> ---- ----
>>>>
>>>>
>>>>
>>>>
>>>> On Tue, Apr 19, 2011 at 12:05 PM, Ajasja Ljubetič <
>>>> ajasja.ljubetic_at_gmail.com> wrote:
>>>>
>>>>> This is strange: I sent the whole mail, but it seems only half got sent
>>>>> though the list. Trying again...
>>>>> Dear all,
>>>>>
>>>>> I have a 2D ABF simulation that keeps crashing after several ns.
>>>>> The error is:
>>>>> ERROR: Constraint failure in RATTLE algorithm for atom 202!
>>>>> ERROR: Constraint failure; simulation has become unstable.
>>>>> ERROR: Exiting prematurely; see error messages above.
>>>>> (this atom is the hydrogen connected to the beta carbon)
>>>>>
>>>>> if I turn rigid bonds of I get:
>>>>> FATAL ERROR: High global exclusion count! System unstable or
>>>>> pairlistdist
>>>>> or cutoff too close to periodic cell size.
>>>>>
>>>>> FATAL ERROR: See http://www.ks.uiuc.edu/Research/namd/bugreport.html
>>>>>
>>>>> The metadynamics simulation with the same colvars appears to be ruining
>>>>> OK
>>>>> (but it's not writing a PMF)
>>>>>
>>>>> The system is an alanine alpha helix in vacuum, spin-labelled at the
>>>>> middle
>>>>>
>>>>
>>>>
>>>
>>
>

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:09 CST