Identify common interests
M4.1.1 (M30 July 2022, DESY) Identify common beam physics interests, define necessary software developments
Minutes EURIZON M4.1.1 meeting
30 June 2022 (M30)
Participants: Simone Liuzzo, Simon White, Ilya Agapov, Lukas Malina, Bianca Veglia, Joachim Keil, Thorsten Hellert, Lina Hoummi
Introduction of Eurizon, conversion of Cremlin + into collaboration of synchrotrons DESY and ESRF.
Task 4.1, reading of the task description(SL).
Specific meeting with diagnostic group TO BE planned
List of deliverables: M38 next ones†: preliminary technical report, proof of principle validation, simplified case
Milestones M4.1.1. (this meeting)
Milestones 4.1.2. next meeting on 5/7/2022
List of tasks
Ilya Agapov: MDT shared between the two facilities is of large interest.
In particular after shutdown and set up of the machine, to see the issues faced at EBS. Tests could be conducted on EBS, look at EBS to prepare for the commissioning of PETRA IV.
Ilya Agapov: each DESY colleague present will present his ongoing projects for DESY in connection to Accelerator Toolbox (AT)
Elaf Musa (not present): simulations on pyAT, implementation of orbit correction and LOCO using pyAT. The code works, but should be cleaned up before sharing. Could be integrated in pyAT.
Lukas Malina: investigation on focusing in undulators as for discussion with Simon at LEL2022. Some problems under resolution. Studies: using MADX and Elegant mainly. Easy tests with pyAT, getting to know pyAT.
Optics measurement with AC dipoles, collimation studies in Elegant. Optics studies with MADX and PTC.
Joachim Keil: user of matlab A.T., didn't look into pyAT yet. Matlab MML user. MML with pyAT?
SL: MML with pyAT could be ambitious, big project, but fits the Task4.1 description. Agreed by the members of the meeting.
Simon White: are you running epics or tango? none, DESY/PETRA specific code.
Bianca Veglia: simulations on pyAT for PETRA III, some orbit correction, undulators gaps variation correction scheme, to be merged with some machine learning.
Thorsten Ellert: numeric simulations with A.T. . pyAT for basic tracking, calculations on the cluster, etc. Developer of Simulated Commissioning toolkit, based on Matlab.
Ilya Agapov: there are several projects on pyAT at DESY: Elaf's phd project, machine learning, some optics. Biggest projects are nevertheless based on matlab AT.
Lina Hoummi: off-energy operation for EBS, would like to move to pyAT to use pytorch and all available libraries. Bayesian optimisation, unsupervised, supervised ML.
SL: work on comissioining like simulations and tolerances. work on python simulator of the control system. Would like to use pyAT as much as possible. Work on lattice modelling, make model as close as much to the real machine, for example introducing crosstalks.
SW: developments for pyAT, DA and LT modules available to use. Present effort in pyAT: Harmonic cavity project, collective effects, multibunch tracking, parallel optimisation, etc.
If other needs, discussion is open.
Slide shown by SL (available in the indico meeting page). List of topics on which we are working or would like to work on.
Going through the established list
sextupole and octupole crosstalk model ( ESRF: most impacting dipole, quads, also sext and oct could have some effect )
- multipoles along beam direction
GGE: generalised gradient expansion, complete model of the machine, more complete, longer computation.
Lattice optimisation:
MOPSO, Bayesian optimisation, online optimisation (testable in CTRM)
parallel computations, GPUs (ALBA, new tracking code to track on GPUs)
Simulations
LOCO, errors in pyAT, loss maps and collimation schemes ? in pyAT
S.W.: Laurent Nadolski (SOLEIL) want to study collimation for their upgrade and want to use pyAT for their studies, subcontracted to CEA. It may be worth dedicating our efforts to other topic as work is already going on on losses/collimation elsewhere. There will be a meeting soon, SW will report.
IA: loss maps with pyAT (already in pyAT, but missing generation of Touschek halo).
LM: taking this over from maps, not obvious, work to be done for secondary particles for instance. Interaction with matter should be improved as it works only for primary particles. Comparisons with FLUKA are envisaged.
SL: Distinction should be made among collimation scheme studies (where particles are lost) and collimator dimensioning studies.
LM: the studies are on collimator dimensioning, in particular studying the impact of the beam on the matter, etc, to dimension the collimator. Use of Elegant software under investigation.
SW: Reine Versteegen modelled the particle impact on the collimator. The collimator dimensioning studies were done with FLUKA (done by the safety group). Collimator dimension: no primary particle would go out.
EBS Collimators are made in tungsten. No idea if the collimators are presently damaged by the beam. It is highly likely, but there is no evident detriment to operation at the moment.
SL: matlab A.T. + GEANT4: deposition of energy in the collimators, and power deposit. Any mitigation of the collimator damage studied lead to damage anyway.
SW: holes in metal (dismantled septum blade (in the old machine) are visible also with the previous ESRF beam (4000pm). We observed traces in copper septum blades. Conclusion: collimators will be damaged.
LM: collimator dimension is studied within a whole machine protection system project (Touschek, betatron beating, all scenarios, collimators dimension, collimators locations, how much they can break.)
SW: discussion with SOLEIL colleagues, very concerend about this issue.
SL: python MML is added to the potential list of topics to work on. Correction toolkit in pyAT.
SW: Project on response matrix in pyAT could be a good collaboration ground. Build a generic response matrix class in pyAT, that could be used in optics, chromaticity, etc.
Integration in cluster: a class that would manage cluster submission, to a SLURM scheduler for instance.
IA: For cluster submission (for MOGA for example), DESY uses DASK. Generic for job submission. Job management already with dask, don't see how the cluster submission in pyAT would help.
SW: exemple of Touschek Lifetime, managing how to split the jobs and putting them back together.
IA: elegant has one file input, one file output.
SW: pyAT uses python multiprocessing, so it is not cluster compatible. It can use many cores of one host only.
TH agrees on the possible advantages of a cluster submission script, with the example of error seeds.
IA worries it could become a pro-level project, ambitious. Python library, etc. SW thinks it can be made in a simple way. Class that handles your MPI data ñ for instance. Convenient way to give the data back to the user.
IA: lot of things in eight months, not a lot of manpower. Prioritise the tasks in the list.
SL: many tasks listed in the presentation are already things that were planned to be done at the ESRF.
Round table of favourite topics till now
TH: error assignment, may be importing them to python. Run optimisation in pyAT and MOPSO, Bayesian optimisation intersting, but would start from zero. Error and corrections are of his personal interest and he already has background.
SL: Elaf also working on this already and interest for EBS.
SW: possibility to have BPM errors.
TH: everything is in the lattice structure. SL: put offset in a BPM, no BPM error in the orbit. TH: calculation of the BPM error is done outside. SW: BPM passmethod to include those errors, so everything is already handled.
LM: would really appreciate to work on topics that would lead to beam tests.
BV: need some time to think about it, correction toolkit perhaps.
JK: interesting long list, some points are to be done anyway, like the crosstalk. Could be a possible way to do it. Static crosstalk, dynamics would be next level.
IA: summarising. Error assignments, crosstalks, experimental side. Monthly meeting could be set up, and exchange. Maybe in-person meetings every half-year.
SL: closing the meeting, next meeting on the 5th of July, with defining the working tasks, and task assignment, schedule of next meetings.