Minutes December 17th, 2008
Present: Rama Calaga, Fritz Caspers, Alexej Grudiev, Elias Métral, Jean-Luc Nougaret, Federico Roncarolo, Giovanni Rumolo, Benoit Salvant, Andreas Wagner, Carlo Zannini, Bruno Zotter.
Agenda
- Andreas Wagner (IT)
"Separating dipolar and quadrupolar wakes with Mafia" (talk ppt, PhD thesis pdf, minutes)
- Carlo Zannini
"Simulation with Particle Studio" (slides, minutes)
Minutes
Presentation by Andreas:
- Andreas did his PhD on the 3D simulations of collective effects with Bruno Zotter in the SL/AP group. He is now working with the internet services of the IT department. A few weeks ago, Bruno mentioned that his work could be very interesting for us as we need to separate dipolar and quadrupolar wakes in the Headtail simulation code.
- Andreas simulated simple structures as well as LEP equipments with MAFIA 3, and Bruno insisted his results showed very good agreement with the thresholds and tune shifts measured at LEP. Andreas did many studies on optimal mesh size to avoid numerical issues. He used the so-called "open" boundary conditions on the planes through which the beam is passing ("open" means perfect matching layer, i.e. no signal reflection, and simulates a perfect electric tube attached to the simulated device that goes to infinity).
- For axially symmetric structures (Pillbox cavity), the transverse kicks are dipolar (i.e. only depends on the position of the source beam, and not the position of the test beam).
- For non axially symmetric structures (ex: LEP bellows), the transverse kick depends on both the position of the source beam (dipolar) and the position of the test beam (quadrupolar). Andreas subtracted the quadrupolar field (obtained for a centered source beam) and got a dipolar transverse kick.
- Andreas used symmetry planes (explained in the thesis), as there was a limitation on the number of mesh cells at the time. Therefore, the results shown are the additions of 1, 2 or 4 separate simulations (see Ch 4 p. 31 of the PhD thesis) with different boundary conditions at the symmetry planes to cancel the image bunches. For instance, to get the kick factor for a beam vertically on axis with a horizontal offset, we need to add the results for the following 2 simulations:
1) magnetic boundary conditions on the x=0 symmetry plane and electric boundary conditions on the y=0 symmetry plane
2) magnetic boundary conditions on the x=0 symmetry plane and magnetic boundary conditions on the y=0 symmetry plane- The postprocessing was performed with Mathematica.
- Alexej mentioned we could check for sextupolar components with triangular cross-sections, which are not so far from some magnet shapes in the PS.
- No comparison with GdfidL was made.
- Fritz mentioned that one has to be careful not to use electric boundary conditions when simulating kickers.
Presentation by Carlo:
- Carlo presented his Particle Studio simulations of round and rectangular beam pipes in copper (simulated as a "lossy metal"). He found out that the loss factor depends on the number of mesh cells but fortunately seems to converge for high enough number of mesh cells. He also checked that increasing the thickness of copper does not change the wakes, which is expected since the skin depth in the frequency range of interest is smaller than the thickness for both cases.
- Carlo then showed the horizontal and vertical wakes for a rectangular copper beam pipe (form factor q=0.5). He obtained the general (or total) horizontal (resp. vertical) wake by displacing both the source and test beams horizontally (resp. vertically) to the same location. He obtained the detuning horizontal (resp. vertical) term of the wake by only displacing the test beam horizontally (resp. vertically), and leaving the source beam on axis. Only the general and detuning wakes were simulated, and the driving term was then obtained by subtracting the detuning wake to the general wake. As expected by theory, the general horizontal wake is very small, which means that both driving and detuning terms cancel each other. Also, the vertical general wake was found to be three times as big as the detuning term, which is also expected since the driving term should be twice the detuning term. However, the horizontal and vertical detuning terms should also be of opposite sign for this high form factor, but a 20% difference can be seen. Action 1: Carlo will clarify this issue and see whether this is related to the mesh size. Carlo also showed results for the cylindrical shape, which show that the detuning terms are negligible, and that both driving horizontal and vertical driving terms are equal, as expected. Some oscillations remain on the wake, but this is due to a coarse mesh.
- Bruno Z. asked why we would want to use Particle Studio over Mafia. The main reason is that Particle Studio is the direct successor of MAFIA 4 (quote from CST: "The CST STUDIO SUITE code family is the direct successor of the code MAFIA, combining the numerical accuracy of the Finite Integration Theory and Perfect Boundary Approximation within an intuitive, easy-to-use CAD environment."). In particular, even if it is still possible to run simulations with MAFIA 4 on LXPLUS, it was made clear to Elias that CERN is not supporting MAFIA anymore, and pushes users to switch to Particle Studio. After digging in the codes' specs, the main differences between the Wakefield Solver of Particle Studio and the T3 solver of MAFIA 4 seem to be:
- (and mainly) the GUI
- As Rama mentioned, an upgraded algorithm for modeling curved metallic boundaries in Particle Studio is used in Particle Studio (the so-called "Perfect Boundary Approximation"). Improvements in CPU time and accuracy are reported.
- Particle Studio seems to take into account finite conductivity, and not MAFIA (see for instance this paper, and the corresponding talk). More precisely, CST Particle Studio gives the choice between two options for materials conductivity ("normal" and "lossy metal"), whereas MAFIA only has one option ("normal"). This article from CST also states that there is indeed a difference between Particle Studio and Mafia for a specific application. Action 2: This question has to be clarified with CST.
AOB
- The topic "opportunity of using a finite length distribution instead of a point charge in Beam dynamics simulations will be discussed in an upcoming meeting. Here is already a link to what is thought to be Gianluca Sabbi's PhD thesis translation into English (pdf, ps) and to the TRISIM homepage, which gathers references and documentation.
Author: Benoit Salvant CERN
AB/ABP-LIS