Minutes from the MCNSI meeting, Prague 29 March 2007 KL May 13 2007 PARTICIPANTS: JS Jan Saroun, NPI Rez (CZ) UF Uwe Filges, PSI Villigen (CH) KLi Klaus Lieutenant, IFE Kjeller (N) PB Phil Bentley, HMI Berlin (D) LC Leo Cussen, HMI Berlin (D) EF Emmanuel Farhi, ILL Grenoble (F) DC Dickon Champion, ISIS (UK) PKW Peter K Willendrup, Risø/DTU (DK) EK Erik Knudsen, Risø/DTU (DK) KLe Kim Lefmann, Risø/DTU (DK) 1. GROUP STATUS - PSI: Geza Szigmond left the simulation group for a permanent position. UF is now only simulator at PSI. - NPI: The MCNSI post doc finished in september. The remaining money will pay 3 month post doc for Alexej, who will simulate a powder diffractometer. - Risø: Peter Christiansen has left, to be replaced by EK, who will work on polarised neutrons. KL is writing a common MCNSI paper on virtual experiments with input from all. - IFE: KLi is employed as an instrument responsible, not simulator. IFE is building 3 new instruments, meaning need for simulations. This work will be included into VITESS. - ILL: A new permanent scientific computing position has been filled by Miguel Gonzales. He will work on molecular dynamics and multiple scattering. This MSCAT-like work will go into McStas. - HMI: Carlos Fehr has left the group. The VITESS manual is still unfinished. PB will ensure that this work is done. LC has joined HMI for some months to simulate ISIS TS2 instruments. - ISIS: Intercomparison work going on between McStas and VITESS descriptions of ISIS moderators. TS2 instruments have moved from simulation phase into hardware design phase. There is a desire at ISIS to incorporate simulations into data analysis. 2. PRESENTATION OF WORK IN THE GROUPS PB presented his work on optimization algorithms. This fell into two categories: Genetic algorithms and particle swarms. In the genetic algorithms, the parameters are viewed as chromosomes, and the evolutionary route is taken by i) survival of the fittest ii) random mutations iii) two parents produce a number of childs. A particle swarm behaves similar to an insect swarm with i) attraction towards local optimum ii) attraction towards awerage optimum. This has e.g. been used to optimize the superior jet engine for the Boing 777. For a neutron example: Ken Andersen has designed a focusing device to get the neutron beam down to 5x5 mm^2 area. This gave a factor 7 increase by manual optimization, whereas swarm algorithm gave a factor 9. These algorithms have been put into VITESS. They have proven to be better than Simplex optimization when the dimension of the parameter space exceeds 30. It is desirable to put this also into McStas. KLi presented the simulations of the new ODIN thermal powder diffractometer at Kjeller. This instrument was optimized by global algorithms, using as a figure-of-merit the combination intensity/(line_width^3 peak_position_error). A full automatic optimization gave unphysical results, but stepwise optimization yielded a realistic solution. The resulting instrument has twice the intensity of the human-designed one, better determination of the line positions, and the same resolution width. In total, this diffractometer has a factor 10 intensity over the previous Kjeller powder diffractometer. PKW presented the developments in the McStas project. a) EK has just started as a two-year McStas post doc. b) There are plans to develop X-ray components to an X-ray version of McStas, in collaboration with ESRF. This could take over the role of the old SHADOW software, which would need to be rewritten anyway. c) The release 1.11 is coming in May. New features are the ENHANCE keyword, which "reuses" the neutron rays at a certain point. Further, there is support for some version of NeXus, and Windows support for PGPLOT. There is now genuine MacOS support. d) Using the Intel C compiler wins a factor 2 in excecution time. e) MPI on a dual core processor wins another factor 2. f) Comparison between experiment and simulation is going on at ISIS g) The American DANSE project is working on component for S(q-vector,w) (incomprehensible KL notes: Ver 1.11 update - plot ??? online plots, grid merging, DSA keys, components: source adapt, single-crystal, powderN, reflectivity tables, histogramming instrument. Monochromator: Hans Graf 1980, diploma thesis) JS presented the news on RESTRAX/SIMRES. - Release 5.1 is due in June 2007. - The code has been converted to Fortran95. - RESTRAX has been enhanced with an X-ray option for the calculation of resolution functions for inelastic X-rays. - Models for SANS scattering have been included and a double-crystal SANS instrument has been simulated. - Models for quasielastic scattering have been programmed. - NeXus is supported - The system manual has been written UF presented the simulation efforts at PSI. 2 SINQ beam lines, S52 and S82 were measured with the new Pb/Bi target (MEGAPIE) in 2006. These beam lines are for the thermal TAS EIGER and the cold imaging station ICON. Pictures taken by image plates through pinholes deviate from the simulation results. An improvement of the McStas Source_gen component with a high-energy tail provided a very good agreement. The PSI sources have been simulated with McStas and MCNPX and compared to gold foil activation measurements with good agreement. A web page with geometrical descriptions of all PSI guides has been made. http://lns00.psi.ch/montecarlo There are plans to make a web inteface to a virtual DMC instrument to be used for experiment planning. EF presented an interesting overview of sampling techniques in Monte Carlo simulation, covering importance sampling and stratified sampling. The new ENHANCE feature in McStas amplifies the sampling of "good" neutron rays and is as such an importance sampling, but not a truly stratified sampling, as the amplified number of rays are randomly distributed after ENHANCE. This feature can improve statistics by a factor 100, but it is also dangerous as it can introduce false biases. Some discussion followed. VITESS has by its construction been able to do this since its beginning, but it has almost never been utilized due to the fear for false biases. (what about RESTRAX???) It seems clear that some work is needed in order to understand the pros and cons of this feature sufficiently. 3. MILESTONE UPDATE The list of MCNSI milestones was displayed and discussed. It seems clear that all milestones can be met until the end of MCNSI. A few items behind schedule lies at HMI, where PB will take action. E.g. he will perform a common simulation of SPAN is McStas and VITESS. This will also be a benchmark for the new McStas polarization option. 4. ACTIVITIES UNDER ECNS2007 It was suggested that a MCNSI booth (or Monte Carlo arena) is set-up at the conference. PKW will take action to contact the organizers. It was suggested to have a social MCNSI gathering 29 June in Copenhagen. PKW coordinates this. 5. FUTURE OF MONTE CARLO IN FP7 There was quite some discussion about the prospect of continuing a MCNSI-type JRA in FP7. Some ideas for the topics are listed below, they fall into two cathegories: - Increased coherence of packages. . An instrument description should be given in a common format, e.g. NeXus, to be read and excecuted by all packages. . A common object-oriented library with numerical facilities for all packages. . A library with common components. (we need to standardize our coordinate systems, then) - Virtual experiments. . Facilities for experiment planning, meaning on-line simulations for all instruments at all facilities. This needs strong effort in instrument description. . Facilities for data analylsis, meaning very accurate description of the experiment. This means a strong effort in describing instrument, experiment, and sample. 6. NEXT MEETING wiil be held in HMI, autumn 2007. PB organizes the next meeting.