Re: [ESPResSo-users] Particle in Fluid, periodic boundary conditions

[Ulf Schiller]

Hi Markus,

glad to hear that it's working now. And yes, you're right, the walls
were partly permeable at the periodic boundary. As far as I can see,
only LB populations are ever communicated to halo regions, so
everything else has to be taken care of explicitly - this is also why
ghost particles need to be coupled to the fluid, for example.

Cheers,
Ulf

On 02/12/14 14:58, Wink, Markus wrote:
> Hi Ulf,
> 
> extending the rhomboids as you proposed seems to solve the problem. Now, I 
> don’t' see the jumps in the velocity or z-position anymore. Thanks a lot for 
> your help. 
> 
> So the problem was, that the halo nodes were not correctly marked as boundary 
> nodes and that resulted in a wrong hydrodynamic behavior of the fluid? Did I 
> get it right? 
> Good to know.
> 
> KR, Markus
> 
> 
> -----Ursprüngliche Nachricht-----
> Von: address@hidden [mailto:address@hidden Im Auftrag von Ulf Schiller
> Gesendet: Dienstag, 2. Dezember 2014 15:04
> An: address@hidden
> Betreff: Re: [ESPResSo-users] Particle in Fluid, periodic boundary conditions
> 
> Hi Markus,
> 
> try the following: make the rhomboids extend from -1 to boxY+1 in the 
> y-direction to ensure that the halo nodes are correctly marked as boundary. 
> It can be debated whether this is a bug or a feature.
> 
> Cheers,
> Ulf
> 
> On 02/12/14 12:52, Ulf Schiller wrote:
>> Hi Markus,
>>
>> I tried another thing and replaced the boundaries by plane walls 
>> normal to the x and z direction. With that, I do not see any peculiarities.
>> This suggests that the issue is related to the more complex boundaries 
>> in your system. I'll have a look at your boundaries.tcl tonight.
>>
>> Cheers,
>> Ulf
>>
>> On 01/12/14 17:08, Wink, Markus wrote:
>>> Hello everybody,
>>>
>>>  
>>>
>>> I performed simulations as Ulf proposed to help to track the problem 
>>> down. In detail three different simulations are performed, graphs on 
>>> that can be found attached:
>>>
>>> 1)      I removed the boundaries, putting a particle with an initial
>>> velocity v_0y. I can see the exponential decrease of the velocity in 
>>> y-direction (as expected). The z-position is pretty stable, 
>>> nevertheless I see an (uncorrelated (?)) change of the velocity in z 
>>> direction. It seems to be directed towards the –z direction, although 
>>> it is very small (of the orders of 1E-24 ), so I guess it is an numerical 
>>> artifact (?).
>>> Nevertheless I don’t see the jump in the z-position nor in the z-velocity.
>>>
>>> 2)      A simulation as in 1), in addition the particle has an initial
>>> velocity in the +z direction (of the order I got for the lift force). 
>>> I still see the exponential decrease in both the z- and y-velocity. 
>>> No jumps of the z-position at the boundaries can be found.
>>>
>>> 3)      The same as simulation 1), but this time with boundaries. I see
>>> the migration towards the middle of the channel as expected (notice, 
>>> that this time, the initial z-velocity of the particle is set to zero).
>>> Neither a jump in the z-velocity nor in the z-position can be seen.
>>>
>>>  
>>>
>>> In combination of the first script I posted (which was the same as in 
>>> simulation 3 here, but furthermore an external force is exerted to 
>>> the
>>> fluid) I see the jump in the z-component of the position and the 
>>> z-component of the velocity only if there is a combination of 
>>> boundaries and an external force acting on the fluid. What is missing 
>>> is a simulation with no boundaries and an external force acting on the 
>>> fluid.
>>>
>>>  
>>>
>>> I furthermore played around with the friction coefficient. If I lower 
>>> it by three orders of magnitude, I still notice a jump in the 
>>> z-component of the velocity, although it is not as steep as before. I 
>>> am not sure what this means.
>>>
>>>  
>>>
>>> I hope this helps to isolate the problem.
>>>
>>>  
>>>
>>> Kind regards
>>>
>>>  
>>>
>>> Markus
>>>
>>>  
>>>
>>>  
>>>
>>>  
>>>
>>> -----Ursprüngliche Nachricht-----
>>> Von: Ulf Schiller [mailto:address@hidden
>>> Gesendet: Montag, 1. Dezember 2014 16:41
>>> An: address@hidden
>>> Cc: Joost de Graaf; Wink, Markus
>>> Betreff: Re: [ESPResSo-users] Particle in Fluid, periodic boundary 
>>> conditions
>>>
>>>  
>>>
>>> Hi Joost, Markus, all,
>>>
>>>  
>>>
>>> after a quick inspection I can't see any relevant changes to the 
>>> coupling on that branch (I've not gone through the CUDA 
>>> implementation though). After merging and running the script Markus 
>>> provided, I'm afraid the issue persists. I don't have time to track 
>>> this down, but according to my experience it is likely to be due to one of 
>>> the following:
>>>
>>> - problem with pos-lattice mapping (less likely since only halo 
>>> affected)
>>>
>>> - halo not up-to-date
>>>
>>> - forces missing in halo upon redef of fluid momentum
>>>
>>>  
>>>
>>> Note that this bug potentially affects any system with particle-fluid 
>>> coupling and periodic boundary conditions.
>>>
>>>  
>>>
>>> Cheers,
>>>
>>> Ulf
>>>
>>>  
>>>
>>> On 28/11/14 15:27, Joost de Graaf wrote:
>>>
>>>> Dear Markus,
>>>
>>>>
>>>
>>>> Sorry about that. The ENGINE branch on hmenke's git account
>>>
>>>>
>>>
>>>> https://github.com/hmenke/espresso/tree/engine
>>>
>>>>
>>>
>>>> has a fix for the bug. As I said, I recalled seeing something like 
>>>> it,
>>>
>>>> and at that time we apparently fixed the problem (which I cannot
>>>
>>>> remember doing). We are still working on one last testcase for the
>>>
>>>> ENGINE branch, before we have it pulled into the Master, which will
>>>
>>>> take about a week. Then a lot of the LB bugs should be fixed.
>>>
>>>>
>>>
>>>> Kind Regards, Joost
>>>
>>>>
>>>
>>>> On 28 November 2014 at 16:24, Joost de Graaf 
>>>> <address@hidden
>>>
>>>> <mailto:address@hidden>> wrote:
>>>
>>>>
>>>
>>>>     Dear Markus,
>>>
>>>>
>>>
>>>>
>>>
>>>>
>>>
>>>>     On 24 November 2014 at 21:24, Joost de Graaf
>>>
>>>>     <address@hidden <mailto:address@hidden
>>> <mailto:address@hidden:address@hidden
>>>>>>
>>> wrote:
>>>
>>>>
>>>
>>>>         Dear Markus,
>>>
>>>>
>>>
>>>>         I remember spotting something like this in one of the older
>>>
>>>>         versions of the master, but I tried to find it and could not
>>>
>>>>         reproduce the bug with the latest version, which one are you
>>> using?
>>>
>>>>
>>>
>>>>         KR, Joost
>>>
>>>>
>>>
>>>>         On 24 November 2014 at 19:35, Ulf Schiller 
>>>> <address@hidden
>>>
>>>>         <mailto:address@hidden>> wrote:
>>>
>>>>
>>>
>>>>             Hi Markus,
>>>
>>>>
>>>
>>>>             On 24/11/14 18:12, Wink, Markus wrote:
>>>
>>>>             > can explain it to me? It seems, that the periodic
>>> boundary condition in
>>>
>>>>             > the LB Fluid doesn’t work. If I remove the fluid I 
>>>> don’t
>>> see that.
>>>
>>>>
>>>
>>>>             Yes, looks like something is at odds with the 
>>>> periodicity
>>>
>>>>             here. Can you
>>>
>>>>             plot the z-velocity over x-position? That may point 
>>>> towards
>>>
>>>>             what is
>>>
>>>>             going wrong.
>>>
>>>>
>>>
>>>>             Cheers,
>>>
>>>>             Ulf
>>>
>>>  
>>>
>>> --
>>>
>>> Dr Ulf D Schiller
>>>
>>> Centre for Computational Science
>>>
>>> University College London
>>>
>>> 20 Gordon Street
>>>
>>> London WC1H 0AJ
>>>
>>> United Kingdom
>>>
>>>  
>>>
>>> Phone: +44 (0)20 7679 5300
>>>
>>>  
>>>
>>
>>
> 
> 
> --
> Dr Ulf D Schiller
> Centre for Computational Science
> University College London
> 20 Gordon Street
> London WC1H 0AJ
> United Kingdom
> 
> Phone: +44 (0)20 7679 5300
> 


-- 
Dr Ulf D Schiller
Centre for Computational Science
University College London
20 Gordon Street
London WC1H 0AJ
United Kingdom

Phone: +44 (0)20 7679 5300