Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Particle tracing (3D) on a 2D axial symmetric solution

Please login with a confirmed email address before reporting spam

Hi,

I'm new to COMSOL, but I didn't find any useful hints on this topic so far...

I'm trying to simulate the trajectories of electrons in a magnetic bottle drift tube (static E and B fields, basically a long selenoid with an electric potential inside). The fields are cylindrically symmetric, so I solve them for a 2D axial symmetric geometry - no problem so far.

The problems start when I try to do particle tracing with mass on the solution. The particle tracer doesn't seem to work when a particle crosses the symmetry axis or lies exactly on the axis. A purely 2D-tracing (no axial symmetry) would work for the E-field, but not for the B-field, since the Lorentz force has a perpendicular component to the 2D plane.
Now COMSOL has a nice feature to make 3D plots from 2D axial symmetric solutions, which is "Revolution 2D". Unfortunately, I had to find out that 3D particle tracing doesn't work with this revolved "3D" data set (which is not really documented), so it's only nice for visualizing the fields but is apparently not a "real" 3D solution...

So, is there an elegant way to do "real" particle tracing on the 2D axial symmetric solution without having to solve the fields in 3D (which would take forever with the desired accuracy)? I've read a bit about mapping a 2D solution to a 3D geometry (extrusion model coupling) but I haven't found so far a tutorial or example to do it (all I found was for version 3.5 and not 4.2).

Thanks for any advice on this..

10 Replies Last Post 6 nov. 2013, 11:05 UTC−5

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 oct. 2011, 14:06 UTC−4
Hi,

May I know how you came to a conclusion that the 2D axial-symmetry solution (with Revolution) is not the real 3D solution. Because, I currently have a 2D axial-symmetry model and I am under the assumption that it is a simplified representation of 3D. I am not able to verify my axial symmetry solution with the real 3D due to lack of computational power for computing 3D solutions, So I wanted to check with you.

Thanks
Hi, May I know how you came to a conclusion that the 2D axial-symmetry solution (with Revolution) is not the real 3D solution. Because, I currently have a 2D axial-symmetry model and I am under the assumption that it is a simplified representation of 3D. I am not able to verify my axial symmetry solution with the real 3D due to lack of computational power for computing 3D solutions, So I wanted to check with you. Thanks

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 oct. 2011, 15:29 UTC−4
Hi,

the 2D axial symmetric solution is indeed a representation of 3D with cylindrical symmetry. Therefore, the physics should be correct, e.g. a rectangle defined as a "coil domain" will yield a B-field of a selenoid, equivalent to a "coil-cylinder" in 3D. You will get a different result with the same geometry in "pure" 2D (without axial symmetry). So I think you don't have to worry for the physical validity of your solution.

The problem I described above is a problem with postprocessing, in particular particle tracing. Particle tracing in 3D (like electrons in E- and B-fields - the trajectories are not cylindrically symmetric) requires a 3D solution. It accepts neither a 2D axial symmetric solution nor a revolution of that around the z-axis. That means (to my understanding) that a "revolution 2D" is technically not a real 3D solution but rather a kind of "remapping" suitable for 3D-plots like slices etc. but not for particle tracing. Probabely 3D particle tracing requires a real 3D mesh or it's simply a bug, I don't know...

Anyway, I kind of solved my problem by extrusion model coupling, i.e. mapping the cylindrically symmetric (r,z)-solution to a real 3D-geometry by converting the (r,z) fields (no phi-component) to (x,y,z) via variable definitions. Still I would like to know why 3D particle tracing doesn't work with revolution 2D data sets, since that would be way more elegant and less complicated...
Hi, the 2D axial symmetric solution is indeed a representation of 3D with cylindrical symmetry. Therefore, the physics should be correct, e.g. a rectangle defined as a "coil domain" will yield a B-field of a selenoid, equivalent to a "coil-cylinder" in 3D. You will get a different result with the same geometry in "pure" 2D (without axial symmetry). So I think you don't have to worry for the physical validity of your solution. The problem I described above is a problem with postprocessing, in particular particle tracing. Particle tracing in 3D (like electrons in E- and B-fields - the trajectories are not cylindrically symmetric) requires a 3D solution. It accepts neither a 2D axial symmetric solution nor a revolution of that around the z-axis. That means (to my understanding) that a "revolution 2D" is technically not a real 3D solution but rather a kind of "remapping" suitable for 3D-plots like slices etc. but not for particle tracing. Probabely 3D particle tracing requires a real 3D mesh or it's simply a bug, I don't know... Anyway, I kind of solved my problem by extrusion model coupling, i.e. mapping the cylindrically symmetric (r,z)-solution to a real 3D-geometry by converting the (r,z) fields (no phi-component) to (x,y,z) via variable definitions. Still I would like to know why 3D particle tracing doesn't work with revolution 2D data sets, since that would be way more elegant and less complicated...

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 30 oct. 2011, 03:17 UTC−4
Hi

I would rather express it as 2D-axi is the "real" solution so long the physics and the result respect 2D-axi symmetry.

There are cases you need full 3D because you do not respect 2D-axi.

One simple example: a horizontal tube in 2D-axi, then you want to add gravity (not along the axis but perpendicular). Your prolem is no more 2D-axi, it's now true 3D

--
Good luck
Ivar
Hi I would rather express it as 2D-axi is the "real" solution so long the physics and the result respect 2D-axi symmetry. There are cases you need full 3D because you do not respect 2D-axi. One simple example: a horizontal tube in 2D-axi, then you want to add gravity (not along the axis but perpendicular). Your prolem is no more 2D-axi, it's now true 3D -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 oct. 2011, 08:40 UTC−4
I understand. My fields are axial symmetric, so real 3D is not needed. However, i want to do "general" (3D) particle tracing on the axial symmetric fields, since the force resulting from the axial symmetric fields might be not axial symmetric (e.g. the lorentz force of an axial symmetric B-field is not axial symmetric). The particle tracing with mass (in 2D plots) has 3 force components for the 2D axial symmetric solution (Fr, Fphi, Fz) - however, the particles disappear when they cross the symmetry axis. Maybe I'm doing it wrong (there is no sufficient documentation or tutorial how to do particle tracing in 2D axial symmetry), so I did it with model coupling to a 3D model, which is not so elegant but it works at least...
I understand. My fields are axial symmetric, so real 3D is not needed. However, i want to do "general" (3D) particle tracing on the axial symmetric fields, since the force resulting from the axial symmetric fields might be not axial symmetric (e.g. the lorentz force of an axial symmetric B-field is not axial symmetric). The particle tracing with mass (in 2D plots) has 3 force components for the 2D axial symmetric solution (Fr, Fphi, Fz) - however, the particles disappear when they cross the symmetry axis. Maybe I'm doing it wrong (there is no sufficient documentation or tutorial how to do particle tracing in 2D axial symmetry), so I did it with model coupling to a 3D model, which is not so elegant but it works at least...

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 oct. 2011, 10:27 UTC−4
Hi,

I think just bringing in a particle into the model makes it non-axially symmetric because the particle isn't. All radial field components in an axisymmetric model must be zero on the axis. I think this holds for the radial velocity of the particle too.

Regards
Edgar
Hi, I think just bringing in a particle into the model makes it non-axially symmetric because the particle isn't. All radial field components in an axisymmetric model must be zero on the axis. I think this holds for the radial velocity of the particle too. Regards Edgar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 oct. 2011, 10:43 UTC−4
Hmm.. I'm talking about the particle tracing post processing, which allows a "3D" force expression for the 2D axial symmetric solution. What you say about the radial field components on the axis is true, of course, and it surely would be a shortcoming (or "bug"?) if indeed the r-velocity of the particles also must be zero on the axis. I mean, the symmetry for the particles is broken already by allowing an "Fphi" force component - then why the restriction of the r-velocity on the axis?
Hmm.. I'm talking about the particle tracing post processing, which allows a "3D" force expression for the 2D axial symmetric solution. What you say about the radial field components on the axis is true, of course, and it surely would be a shortcoming (or "bug"?) if indeed the r-velocity of the particles also must be zero on the axis. I mean, the symmetry for the particles is broken already by allowing an "Fphi" force component - then why the restriction of the r-velocity on the axis?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 oct. 2011, 11:19 UTC−4
Yes it looks all a little inconsistent. I guess your workaround, calculate the fields in 2D and do the particle tracing in 3D is the way to go.
Yes it looks all a little inconsistent. I guess your workaround, calculate the fields in 2D and do the particle tracing in 3D is the way to go.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2 nov. 2011, 14:59 UTC−4
Just for completeness:
I kind of made it work in "purely" 2D axial symmetry now (without 3D mapping). You have to make sure that the particle position is never at r = 0 (I guess, because the phi-velocity would be undefined there - anyway, I consider this a "bug")... For my time-of-flight drift tube I have to give the electrons a negligible initial phi-velocity (say 1m/s, while the z-velocity is 0.3c and the r-velocity is 0), then the z-axis "crossing" is simulated correctly (projected to the r-z-plane it looks like the particles "bounce" off the symmetry axis). Also, I can't use r = 0 as a starting point, the particle has to step a nanometer or so away from the axis...
Just for completeness: I kind of made it work in "purely" 2D axial symmetry now (without 3D mapping). You have to make sure that the particle position is never at r = 0 (I guess, because the phi-velocity would be undefined there - anyway, I consider this a "bug")... For my time-of-flight drift tube I have to give the electrons a negligible initial phi-velocity (say 1m/s, while the z-velocity is 0.3c and the r-velocity is 0), then the z-axis "crossing" is simulated correctly (projected to the r-z-plane it looks like the particles "bounce" off the symmetry axis). Also, I can't use r = 0 as a starting point, the particle has to step a nanometer or so away from the axis...

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 26 juil. 2012, 13:30 UTC−4
Someone asked me about a particle tracing problem via PM - but it's better to discuss it here with everyone:


Dear Alexander,

I'm trying to use the Charged Particle Tracing module to track
particles moving through previously calculated E fields (AC/DC module).

I get this error when I run the model:

Syntax error in low-level element syntax representation.
- Index: 43
- Type: elparticle
- Field: posgrid
Error in multiphysics compilation.
- Detail: Missing field

I appreciate if anyone could help me to rid of this error.

Thank you very much,


It's hard to say what's wrong from your info (missing field?) - better attach your model or some screenshots...
It seems like your field is not calculated for the whole volume/area of your particle tracing - that might be the reason for the missing field error.. But again, hard to say anything without knowing the details...
Someone asked me about a particle tracing problem via PM - but it's better to discuss it here with everyone: [QUOTE] Dear Alexander, I'm trying to use the Charged Particle Tracing module to track particles moving through previously calculated E fields (AC/DC module). I get this error when I run the model: Syntax error in low-level element syntax representation. - Index: 43 - Type: elparticle - Field: posgrid Error in multiphysics compilation. - Detail: Missing field I appreciate if anyone could help me to rid of this error. Thank you very much, [/QUOTE] It's hard to say what's wrong from your info (missing field?) - better attach your model or some screenshots... It seems like your field is not calculated for the whole volume/area of your particle tracing - that might be the reason for the missing field error.. But again, hard to say anything without knowing the details...

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 6 nov. 2013, 11:05 UTC−5
Hi

I am wondering if this problem has found a solution since then (more than a year since the last post in this topic)...

I have a similar problem but with a plane-symmetric case (I simulate only 1/4 of the model because the other 3/4 are the symmetric copies of this field), and would like to trace particles in the field of the full geometry. Can it be done in a smart way?

Thank you
Daniel
Hi I am wondering if this problem has found a solution since then (more than a year since the last post in this topic)... I have a similar problem but with a plane-symmetric case (I simulate only 1/4 of the model because the other 3/4 are the symmetric copies of this field), and would like to trace particles in the field of the full geometry. Can it be done in a smart way? Thank you Daniel

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.