Status of Acceptance Calculations with GSIM

 

 

• Progress since the last collaboration meeting
• List of near-term projects
• Experience to date in calculating acceptances
• Known effects included in GSIM acceptances
• Methods of calculating acceptances
• Issues and problems
• Run proposals
• Guidelines for creating run proposals
• Review process for run proposals
• Broad goals and challenges in the longer term
• Conclusions

 

 

 

 

Progress since the last collaboration meeting

 

• Prototype of electron shower parameterization now working (Dave Tedeschi, Maurik Holtrop) and being tested (Mike Vineyard)
• Prototype version of automated GSIM/CELEG control and tracking system now functional (Greg Riccardi)
• Formation of a dedicated event generator group (Dan Sober, Gabriel Niculescu, Bryan Carnahan, Stepan Stepanyan, John Price, others?)
• Created tagger banks for photon beam simulations (Franz Klein, Burin Asavapibhop)
• Comparative study of acceptances obtained from fiducial cuts on data, and on simulated data with and without fiducial cuts (Kyungseon Joo)
• Performed region-dependent smearing of drift chamber position data, and applied dead wire map to GSIM data using GSIMKO (Kyungseon Joo, Dennis Weygand, Joe Manak)
• Run simulations supporting speakers at Bonn conference and Sante Fe conference (~20 million events) (Cole Smith, Junho Yun, David Rowntree)
• Cooking of GSIM data using same methods and code as for experimental data (Costy Loukachine)
• Implemented first version of the electron library scheme (Stepan Stepanyan)
• Gained preliminary understanding of effects of decaying particles (low energy pions) on acceptance (Stepan Stepanyan, Hovanes Aznauryan)
• Implemented and tested prototype GSIM particle pre-filter (Kyungseon Joo, Cole Smith)
• Performed first study of effects of energy losses on tracking results (Stepan Stepanyan, Hovanes Aznauryan)

 

List of near-term projects

• large-scale cooking of GSIM data
• large-scale application of DOCA smearing and dead wire map
• transition to exclusive use of automated run control and accounting system, including keeping parameter banks from GSIM/CELEG.
• implement polarized target field and geometry in GSIM with new target location (z=-55 cm)
• continue to support speakers at conferences and students writing theses
• document what is and is not included in GSIM and attempt to assess the accuracy of geometry and materials for each volume
• general improvements to CELEG (radiative corrections, elastic scattering cross section, deep inelastic scattering) and to the present arsenal of event generators
• finish trigger simulator prototype
• finish installing photon beam cryotarget geometry into GSIM, add event vertex position smearing, install electron beam cryotarget geometry
• finish implementing the zero-electron-acceptance option in GSIM
• finish testing of electron shower parameterization
• perform tests evaluating the effects of `shortcuts' on acceptances (inert volumes, cut values, event pre-filtering) using CELEG events
• incorporate Cerenkov counter efficiency into GSIM based on measured properties for electrons, pions, etc.
• further study of particle decay and dE/dx on acceptance
• study accidental hits in DC - do we need to add them to replicate tracking efficiency?

 

Experience in calculating acceptances

• Who has calculated acceptances/related quantities? (that I know of)
• Kyungseon Joo and Cole Smith - extensive work with GSIM acceptance calculations as well as parameterized acceptance functions. Significant contamination of signal with radiative tail from elastic scattering. Large acceptance.
• Claude Marchand et al. - developing single-particle acceptance scheme based on GSIM single-particle events for photon beams. Recently got reconstruction to work on these events.
• Richard Thompson - extensive single-particle parameterized acceptances, extensive GSIM work underway. Fairly large acceptance, some background.
• Latifa Elouadrhiri - initial work with GSIM acceptance functions, more coming soon. Fairly small acceptance.
• Hovanes Egiyan - extensive work with GSIM acceptance calculations as well as parameterized acceptance functions. Large acceptance, little background.
• Steve Barrow & Si McAleer - extensive work with SDA acceptance calculations, GSIM calculations coming soon. Acceptance change in the interesting variables appears to change extremely rapidly.
• Costy Loukachine - initial work with GSIM acceptance functions, more work presently underway. Fairly small acceptance, large and complicated backgrounds.
• Alan Coleman/Herb Funsten - parameterized acceptances for omega electroproduction, interested in using GSIM next.

 

Known effects included in GSIM acceptances

• CLAS geometry at a relatively high level of accuracy
• Includes dead wire map, can include other dead channels
• Resolution effects reproduced in making fiducial cuts
• Reconstruction efficiencies automatically accounted for, including any systematic variations (i.e. with momentum or angle) if the simulation is accurate enough
• Multiple particle hit effect on reconstruction programs (EC, DC, SC)
• Includes events scattered into and out of the geometric acceptance by realistic materials
• Includes particle decay properly (we think) if proper binning is used
• Reproduces effects related to energy losses (e.g., momentum reconstruction errors)
• Can in principle include the effects of misalignments of detectors (not presently implemented)
• Absorption losses of particles via hadronic interactions with materials

 

Methods of calculating acceptances

• Start with the event generator: CELEG/radiated AO/uniform generator
• Run events through GSIM, then GSIMKO, then TSIM, producing a BOS format file.
• Reconstruct using your favorite reconstruction program.
• Choose cuts and bins, and apply them identically to: reconstructed data and reconstructed simulated data
• Must bin in small bins in any variable where simulated data distributions do not match real data distributions well (integrated quantities); must bin in small enough bins so that distribution change across bins is approximately linear
• Resolution from simulation must match data well to avoid divide by zero
• Multiply each data bin contents by the thrown bin contents divided by the accepted simulated data bin contents
• Lower-accuracy alternatives: single-particle acceptance table; fiducial cut plus average correction factors

 

Emerging Strategies for Acceptance Calculation

• Smear DC positions in simulated data to match resolutions attained by current status of analysis (slowly improving with time).
• Use fiducial cut (depending on angles, momentum, and particle type) on both the data and the simulation. Start with a conservative cut, evolve to broader cuts where it appears to be feasible. If resolutions are correct, these cuts will be correct.
• Use dead wire map on both the simulation and the data; use a cumulative dead wire map for a set of runs. (Or, use a fiducial cut which removes regions of dead wires entirely.)
• For electron beam experiments, explore using measured Cerenkov counter properties in GSIM. E.g., use measured photoelectron number as a function of angles, momentum, and particle type.
• Use appropriate binning:
• for decaying particles
• to correct for energy losses for heavily ionizing particles identically in simulation and in experimental data
• to ensure binning is fine enough to handle rapidly changing acceptance

 

Issues and problems

• Scheduled requests for GSIM runs is growing, while available CPU time is shrinking. Situation is ok for now, but won't be for long.
• The basic cpu power of the available distributed farm is too small.
• 1500 Hz data rate x 35/52 x 50% = 500 Hz (annualized data rate)
• GSIM: 0.75 Hz/cpu-equivalent, x fraction of availability of cpu, typically 0.4 for our 25 cpu's (40 cpu-equivalents). (with no secondary particles in inert volumes)
• Order of magnitude: simulated data production rate is a factor of 40 slower than the data rate, for continuous computing. Several mitigating factors, but can't hope to compensate for such a huge disparity.
• Not clear that a conservative fiducial cut will be suitable for all measurements. May require more attention to simulation accuracy near coils and at forward angles, or more complicated, less conservative fiducial cuts.
• Need a variety of types of simulation:
• events which are purely the final state of interest
• events which include all final states in realistic ratios with realistic distributions, particularly for small cross sections. Need to develop CELEG a lot if it is to fill this role.
• Proper handling of pre-radiated and post-radiated electrons.
• Need a review process for GSIM proposals.
• Need to re-assess what the important speed-dependent accuracy issues are, in approaching publication-quality acceptances:
• which inert volumes affect the acceptance appreciably
• appropriate values for cuts

 

Run proposals

Contact	Process	Status	Run conditions, comments
Richard Thompson/ Pitt group	eta electroproduction from the proton via S11	300,000 plus several million events processed through GSIM	4 Mevents short term; longer term, ~10 M events for 2.4 GeV and 16 M events for 4.0 GeV. Richard provides the event file.
Marco Battaglieri/Genova group	p o and h electroproduction from the proton	waiting for event file/and/or more discussion	p o and h final states with a realistic phi distribution for the meson; single/multi pion background, elastic radiated tails. 2.4 GeV: Q 2 =0.2-1.5, W=1-2; 3375 A, 2249 A. 4.0 GeV Q 2=1.0-2.5, @=1-2; 3375 A, 2249 A.
Costy Loukachine	phi electroproduction from the proton, + backgrounds	400,000 events processed (prototype + prod.)	4 GeV high field, he's providing the event file.
Latifa Elaouadrhiri	ep p + p -	some events generated, waiting for proposal for more	1.6 GeV, low field.
Franz Klein	omega electroproduction on the proton	waiting for proposal
Hovanes Egiyan	p + electroproduction on proton	4 million events recently, 5 million earlier
Steve Barrow	lambda electroproduction on the proton	performing prototype runs	~ 1 M events needed
Gabriel Niculescu	strange particle electroproduction	proposal is ready	Many events to be requested, planning to use single particle acceptances.
Kyungseon Joo/UVA	p o electroproduction + backgrounds	more than 10 million events generated
Gabriel Niculescu	strange particle electroproduction	proposal is ready	1 M events initially, planning to use single particle method
Claude Marchand	single particle acceptances, kaon photoproduction	5 million events generated

 

Guidelines for creating run proposals

 

(This is available on the web page)

 

The GSIM Focus Group's purpose is to facilitate the calculation of acceptance functions for CLAS experiments for the first publications. Since limited CPU resources are available, it is important to coordinate simulation requests to provide optimal calculations for every group which is poised to make use of them. These guidelines are intended to help with the development of the best possible run request proposals.

 

Communication with the group

 

Each proposal should be informally presented to the GSIM Focus Group at one of the weekly meetings, preferably by the requesters. This may be relaxed for overseas collaborators. The purpose is to clarify the detailed requirements of the proposal, to understand the need for the statistics requested, and to provide input and support to the acceptance calculation.

 

The group is eager to help in developing these proposals and to offer support for providing the optimal calculation.

 

Number of simulated events requested

 

The number of simulated events requested should be based on quantitative estimates for the errors required for the acceptance calculation. It should be considered whether relative or absolute errors are important. (If the number requested is less than 100,000, estimates are not needed.) In some cases, special additional studies may be needed in addition to the pure statistical count, such as for background studies.

 

Some information on error estimates is available on the web on:

acceptance error calculations acceptance error plots

 

The number of simulated events should be reasonably commensurate with the number of experimental data events, both signal and background.

 

Event generators

 

Some support is available in providing event generators and help in operating them, but this is primarily the responsibility of the proposer. Event generators used to date include the CELEG event generator, the radiated AO generator, a radiated elastic scattering generator, and a collection of special purpose phase space generators. There is a separate but related group focusing on event generator development in cooperation with the GSIM Focus Group.

 

Prototype runs

 

Every proposer should plan to begin with a prototype run (e.g. 100,000 events). The run should be analyzed to make sure all problems are worked out before investing resources in longer runs.

 

Status of analysis

 

It should be demonstrated that the data analysis for a particular reaction is sufficiently advanced so that the simulated data can be immediately analyzed.

 

Conferences and presentations

 

Calculations for people making presentations at upcoming conferences, PAC meetings, etc. will be given higher priority, as long as sufficient advance notice is given of the request.

 

Students

 

Calculations for Ph.D. students working on their theses will be given higher priority.

 

Checklist for proposals

 

• Absolute acceptance errors have been calculated in a given binning scheme.
• Relative acceptance errors have been calculated in a given binning scheme.
• Date results are needed:
• Is request in support of a formal presentation? Date and nature of presentation? (e.g., conference date)
• How many events are requested, under what running conditions? Specify magnetic field of torus and other magnets (minitorus, polarized target):
• GSIM documentation has been reviewed to make sure what is included in the simulation is appropriate (documentation coming soon)
• Photon beam or electron beam? Other special geometries to use? (Start counter in/out, minitorus in/out, polarized target geometry in/out, photon beam target or electron beam target, etc.)
• Has a prototype run been performed?
• Any other special points? (e.g, single particle runs)
• Input event files to be provided (determine optimal size from prototype run), or else what event generators should be used? (CELEG, radiated AO, radiated elastic)

 

 

 

Review process for GSIM proposals

• Still under discussion - input is solicited
• Needs to be a simple system, with rapid turnaround
• Main components:
• entry-level thresholds
• few hundred thousand events - no discussion, just run them
• < 10 million events or <10% of capacity - no feedback sought from outside GFG group, purely internal decision
• > 10 million events or > 10% of capacity - group seeks input from working group and collaboration
• GFG makes decisions on proposals based on resource loading and feedback:
• feedback from working group: get specific expertise and relative priorities of projects. email?
• feedback from collaboration, on use of collaboration resources. email?

 

 

Broad goals and challenges in the longer term

• Build up a distributed farm (at JLAB and remote institutions) of 200-300 `cpu-equivalents' (200 MHz Pentium II Linux box) utilized ~40%for GSIM work in the two-year time frame. Some movement in this direction already (I know of 4 new or upgrading farms, plus our original 3), but more is needed.
• Make the transition to having a smoothly operating system:
• can handle 5-10 proposals at once
• runs are initiated by a simple, universal mechanism
• run control and accounting is automated
• runs are scheduled and completed in a timely manner
• each file produced is fully documented on the web
• reconstruction of simulated data is integrated cleanly into the `cooking' procedure
• Move to GEANT 4
• C++ based code, compatible with new CERN developments
• reputed to be three times faster
• natural opportunity to encapsulate geometry and other constants into a controllable database
• extra manpower available - good opportunity for new contributors, especially students
• first version could be ready for testing in ~1 year, full deployment in 2 years
• Collaboration to gain a deeper understanding of CLAS acceptance

 

Conclusions and comments

• Present results look quite promising
• Need more cpu's (!) to share
• Need more people
• Need photon beam customers
• Web site