Re: [ESPResSo-devel] [bug #46210] "analyze pressure" affecting simulatio

Hi,

Ok, I think a clean way to handle that is via the rotational degrees of freedom. Non-pair forces like this can never be handled in the same way the pair forces are handled (who are the extra forces for?).

If I remember correctly the pressure calculation via the virials works only for pair forces, so the multi-body interactions have to be excluded from the virial calculation.

For now I've moved the lj-angle out of the (inner) pair force loop and made the particles const for the pair forces for good measure. The pressure command should now have no side effects,

albeit still not returning the correct pressure (see pull request). The fact that the pressure calculation is only correct for some interactions is reflected in the documentation.

Of course this dug up other stuff. The affinity interaction from object in fluid was also manipulating particles (bond creation). I am in general not in favor of the force calculation changing the state.

Changing forces is one thing, but integrate 0 actually can do all sorts of things, like creating and removing bonds. I think it would be much cleaner to do a second run through the particle pairs to

do that if it is needed or just flag particles for modification. What do you think?

Cheers,

Florian

On Sat, Oct 17, 2015 at 9:35 AM, Marcello Sega <address@hidden> wrote:

Hi,

We're using that in Vienna! We might put some effort to implement it properly in the future, so maybe for the time being we could leave it in as is...

M

On 16 Oct 2015 23:33, "Florian Weik" <address@hidden> wrote:
Hi,
what Axel said. This is broken and can't be fixed easily in the current structure as it expects only pair interactions at this points. This probably won't fix. I am in favor of doping the interaction,
as it could actually be implemented as simple pair interaction with rotation and torques, and since this is a cruel hack I'm not motivated to put any effort into fixing it since apparently the original author was not motivated to do a clean implementation. Is anybody else using that?

See also
https://github.com/espressomd/espresso/issues/439

Cheers,
Florian

On Fri, Oct 16, 2015 at 10:30 PM, Axel Arnold <address@hidden> wrote:
Hi all,

the culprit is the LJ-angle interaction, which changes particle forces directly, rather than returning them. That works for force calculation, but when you calculate the pressure, your pressure is lacking the LJ-angle interaction, and your forces are f****ed up. You can get the simulation working by using the recalc_forces option of integrate, but your pressure values are still be wrong.

So, just don’t use that interaction or fix the interaction to behave well (that is, like any other two-particle interaction).

Best,
Axel

Am 15.10.2015 um 22:09 schrieb Nicholas Jin <address@hidden>:

Hi Peter,

I have created a barebones tcl script that replicates the simulation I am trying to run. I have attached it below (demo.tcl).
As a warning, this script writes 1000 pdb files to whatever directory it's currently in.

Commenting out the set pressure [analyze pressure] causes qualitatively massive changes in the simulation.

Espresso 3.3.0 is compiled with the following features:
/* Generic features */
#define PARTIAL_PERIODIC
#define EXTERNAL_FORCES
#define CONSTRAINTS
#define EXCLUSIONS
#define COMFORCE
#define COMFIXED
#define MOLFORCES
#define NPT
#define DPD

/* Interaction features */
#define TABULATED
#define LENNARD_JONES
#define LENNARD_JONES_GENERIC
#define LJ_ANGLE
#define SMOOTH_STEP

#define BOND_ANGLE

Thanks for your help!
Nick

On Thu, Oct 15, 2015 at 4:22 AM, Peter Košovan <address@hidden> wrote:
Hi Nick,

I believe that analyze pressure is used by many people and if none has reported an exploding simulation upon the call, I suspect this issue is related to some specific feature of your simulations.

Providing a minimal example to enable others reproduce the bug would be very helpful. Otherwise, only a person fully familiar with your simulations can diagnose the problem.

Best,

peter

On Wed, Oct 14, 2015 at 10:15 PM, Nicholas Jin <address@hidden> wrote:
URL:
<http://savannah.nongnu.org/bugs/?46210>

Summary: "analyze pressure" affecting simulation
Project: ESPResSo
Submitted by: njin
Submitted on: Wed 14 Oct 2015 08:15:00 PM GMT
Category: Simulation core
Severity: 3 - Normal
Status: None
Assigned to: None
Open/Closed: Open
Discussion Lock: Any
Release: None
Fixed Release: None

_______________________________________________________

Details:

Hi all,

The tcl command "analyze pressure" is causing misbehavior in my simulations.
Absent this analysis command everything runs fine, but once the analysis runs
the simulation explodes.

This occurs in espresso-3.3.0. Let me know what other information you might
need to diagnose this issue.

Best,
Nick (address@hidden)

_______________________________________________________

Reply to this item at:

<http://savannah.nongnu.org/bugs/?46210>

_______________________________________________
Message sent via/by Savannah
http://savannah.nongnu.org/

--
Dr. Peter Košovan

Department of Physical and Macromolecular Chemistry
Faculty of Science, Charles University in Prague, Czech Republic

Katedra fyzikální a makromolekulární chemie
Přírodovědecká fakulta Univerzity Karlovy v Praze

www.natur.cuni.cz/chemistry/fyzchem/
Tel. +420221951290
Fax +420224919752

Pokud je tento e-mail součástí obchodního jednání, Přírodovědecká fakulta Univerzity Karlovy v Praze:
a) si vyhrazuje právo jednání kdykoliv ukončit a to i bez uvedení důvodu,
b) stanovuje, že smlouva musí mít písemnou formu,
c) vylučuje přijetí nabídky s dodatkem či odchylkou,
d) stanovuje, že smlouva je uzavřena teprve výslovným dosažením shody na všech náležitostech smlouvy.

<demo.tcl>

--------------------------------------------
Axel Arnold
Richard-Wagner-Ring 16
76437 Rastatt
Email: address@hidden
Telefon: +49 173 870 6659

--
Florian Weik

address@hidden
++49 157 85939252

Florian Weik

address@hidden
++49 157 85939252

From:	Florian Weik
Subject:	Re: [ESPResSo-devel] [bug #46210] "analyze pressure" affecting simulation
Date:	Sat, 17 Oct 2015 16:12:00 +0200