Summary of Physics Analysis Perspectives and Detector Performance Interests

1. Summary of Physics Analysis Perspectives and Detector Performance Interests Peter Loch University of Arizona Tucson, Arizona USA

2. This Talk • Scope • Some input from what I heard yesterday • West Coast physics analysis interest • Jason, Anna,… • German analysis facility at DESY • Wolfgang • More emphasis on first data performance evaluation • Underlying event & pile-up • Fast access to systematics of hadronic final state energy scales • Jets, taus, missing Et • Some focus on possible analysis facility role of SLAC • Resources and critical mass • Meetings & educational efforts • Biased views • “Small” university group experience • Arizona is locally below critical mass for most studies of interest • But excellent and long standing connections to CERN and ATLAS (PL) • Personal connections to CERN > 20 years • Arizona is ATLAS member since ~15 years • Well established collaborations with INFN Pisa, INFN Milan, MPI Munich, CERN, ATLAS Canada,… • Sometimes limited to intellectual input role so far! • Longstanding (continuing) and short-lived (need-driven) collaborations • Some personal connections to theorists/phenomenologists • West Coast (SLAC – Peskin, Oregon – Soper, Washington – S.D. Ellis), FNAL/CERN (Skands, Mrenna), BNL (Paige), LPTHE/Paris VI (Salam/Cacciari),… • Not limited to client role!

3. What does critical mass mean? Sufficient local expertise around the corner Detector, software, physics analysis Peers facing similar problems and/or challenges Low communication thresholds for new graduate students etc. Common group interests Significant number of experts and soon-to-be experts work closely together on related topics E.g., Ariel’s Jet/MET/b-tag group Fast senior and peer review of new ideas and general progress Less time lost with “wrong approaches” Willingness to share new ideas at early stages and “help out” Collaborative work environment Usually not available to university groups Some exceptions! (Computing) Resources Locally available computing CPU power, storage Often limited for university groups Historically no or little dedicated funding for computing requested “Hardware group” tag with funding agents Different visions on group contribution to ATLAS Distributed computing resources often deemed sufficient Grid etc. – see next slide! Funding crunch at state level does not help Recently implemented new charges to groups at state universities Cooling & power for larger systems Not so much an issue for Arizona so far! Loss of “free” operation support Loss of system manager may become a problem for Arizona Critical Mass & Resources

4. Physics Analysis Resources ATLAS Computing Model for Physics Analysis (long term) Graph from Jim Cochran, this meeting

5. Physics Analysis Aspects • Mostly needs critical mass • Intellectual guidance, exchange and progress review essential • Assuming we want to keep significant analysis activity in the US/at the West Coast • I think we should (entry point for new graduate students, already significant investments,…)! • Resources provided by ATLAS distributed computing • Data extraction, compression by official production system… • Managed by ATLAS physics groups • …or on Tier2 “around the corner” for different filters • Reflecting local (US ATLAS etc.) group interest • Local refined and final analysis needs much less resources • DnPD as simple as root tuples • Tier3g cluster, desktop, laptop can be sufficient depending on analysis • May have somewhat relaxed time scale • Not always true, in particular for early data (competition with CMS, conferences, graduation deadlines,…) • But what about performance?

6. On the Side… • Tier3 in US ATLAS • Tier3’s are not a US ATLAS resource • They are a local group resource! • US ATLAS spends money to help with setting up Tier3’s… • Help with dataset subscriptions, some software support (my understanding)… • Hardware and configuration recommendations to simplify support role • …but they are not running it for you! • Tier2’s or larger groups may offer to host smaller group’s computing • At a price: $ or fractional use of equipment by host are discussed • Helpful if no or little local computing support • Contractual limitations in base programs may require funding agent agreements/permissions • Requests for American Recovery & Reinvestment Act (ARRA 2009) funds by universities • Not only computing • Arizona: ~25% requested for computing, rest replacement of obsolete electronic lab equipment • Not a direct competition to US ATLAS Tier1/2 funding • Invitation to apply fully supported by US ATLAS – including per group recommendations for scale of computing hardware to acquire! • Thanks to Jim! • Unique chance for university groups to get ready for analysis of first data!

7. Performance Tasks for First Data • Need flat jet response rather quickly • Basis for first physics analyses • E.g., inclusive jet cross-section • Systematics can dominate total error already at 10-200 pb-1 • Depending on final state, kinematic range considered, fiducial cuts,… • Requires understanding of hadronic response (Jason, Ariel,…) • Minimum bias • Single charged hadron response of calorimeter • Pile-up with tracks and clusters • Di-jets • Underlying event with tracks and clusters • Direction dependent response corrections • Prompt photons • Absolute scale by balancing, missing Et projection fraction, etc. • Initial focus on simple approaches • Least number of algorithms • Least numerical complexity • Realistic systematic error evaluation • Conservative ok for first data! • But also try new or alternative approaches! (plots from CERN-OPEN-2008-020)

8. JES Systematics (Example) • Systematic error evaluations for JES • Requires understanding of effect of calorimeter signal choice in real data • Towers, noise suppressed towers, topological cell clusters • Corresponding algorithm tuning • Noise selection criteria, mostly • Need to consider at least two different final states • Can be di-jets, photon+jet(s), single track response • May need repeated partly or full event re-reconstruction • Need calorimeter cells to study different signals • Dataset with most details – ESD or PerfDPD • Considerations for MC based calibrations • Study change of response from calibrated (jet) signals for different hadronic shower models • Determined with one model, applied to simulated signals from other figures from K. Perez, Columbia U.

9. JES Systematics Resource Needs (1) • Iterative reconstruction with quick turn-around • Large (safe) statistics to focus on systematic effects • Can be several 100k of ESD events each for different final state patterns (photon triggers, jet triggers, MET triggers, combinations…) • >1M for minimum bias! • CPU ~1 min/event, disk size ~1.5 Mbytes/event • Need to save at least 2 iterations • Change configurations from default reconstruction • Run on cells from default reconstruction • Need for cell signal re-reconstruction (RAW data) not obvious in jet context • Maintenance of configurations • Book-keeping • Potentially several rounds to explore parameter space(s) • Semi-private production scenarios feeding back to official reconstruction • Highly efficient access to significant resources • Speed is an issue • Maybe not quite within the 24-48 hours gap between data taking and reconstruction, but close! • Local CPU and disk space • Cannot afford dependence on grid efficiency • Need clear tracebacks in case of problems • Reports/output files lost on the grid? • Database access • Latest conditions data need to be available locally • E.g., from CAF etc.

10. JES Systematics Resource Needs (2) • Changing hadronic shower models in G4 • Not clear that appropriate resources are available in production system… • Especially after data taking starts! • … but important part of evaluation of systematic uncertainties! • All jet calibrations will likely use some MC input – admittedly with different sensitivities to hadronic shower models • Long term fully MC based? • Big project • Setting up full simulation, digitization, and reconstruction chain • Severe disk space requirements if truth level detector information is turned on (probably needed) • ~2.4 MByte/event → 5.7 Mbyte/event, 8% increase in CPU • Dedicated resources preferred • Especially if not part of the standard ATLAS production • May not be taken on by other international institutions • Maybe MPI Munich – resource limitations?

11. Potential Role of SLAC • Physics analysis • Clearly potential to provide critical mass as an analysis facility • Involvement of West Coast institutions required • See Wolfgang’s talk: existence of analysis facility needs community to fill with life (providing expertise, attendants…) • Does not seem to be a big problem! • Physics interest for initial data well aligned at the West Coast • E.g., lots of common interests in various standard model hadronic final states, minbias/pile-up, b-jets, W/Z production… • SLAC Tier2 can be resource to extract user data • Need some organization to handle request • Just the grid? Analysis groups? Individuals? • Extension of Tier2 to allow some interactive/local batch analysis capabilities? • Depends on final university Tier3 funding • Hosting computers for West Coast (and other) institutions? • Meeting and office space • Accommodate visitors working on specific aspects of an analysis • Provide meeting space for groups working on same/similar projects • Seems to be possible • Certainly desirable from my standpoint! • Some coordination with LBL? • Analysis facility extension to analysis center? • Going to SLAC instead of CERN saves $$$ for Arizona! • Reduce amount of time grad student spends at CERN by better preparation!

12. Role in Performance Evaluation • Bigger task • Needs critical mass and significant resources • Would need dedicated computing resources for a unique contribution to ATLAS • Critical mass not a problem at all • C.f. Ariel’s group • SLAC meetings have been a significant source of new ideas in hadronic final state reconstruction • Already two very productive US ATLAS Hadronic Final State Forums (2008 and 2009) • ~30 attendants in very informal environment – low threshold for contributions & discussions • Want to continue with 1-2/year meetings • New convenership for HFSF – PL likes to continue to be involved in organization • Several concepts seriously worked on today were initially discussed at these meetings • Latest: data driven jet calibration, some jet substructure focus • Workshops have “real work” portion with educational perspective • First access to ATLAS jet simulations for some attendants, thanks to huge effort of Ariel’s group – Bill L., Stephanie M. etc. may remember! • Computing resources – a wish-list? • Dedicated task or project assigned computing resources need extension of Tier2 center • Don’t believe typical Tier3 computing does the job – very similar infrastructure and maintenance levels as Tier2 may be required • Something for Richard to work on! • Resource needs not completely out of scope, even for re-reconstruction • But also needs organization, local batch management, disk space management, i.e. $$ • SLAC seems to be able to provide infrastructure – in principle! • How can potential clients help? How much do funding agents allow us to help? • Personal opinion • I see a clear potential for a very significant contribution to the timely improvement of the hadronic final state reconstruction in ATLAS with SLAC hosting the intellectual and resource center for the related activities

13. Educational Role • Education beyond software tutorials (e.g., “ATLAS 101” lectures) • I like DESY’s approach of “lower level” schools • E.g., focusing on detectors principles and reconstruction approaches, selected (early) physics & performance topics • Mostly aim at new graduate students and post-docs • Also good source to refresh knowledge for senior collaborators • Good opportunity to capture knowledge within the collaboration • Original designers or other specialists can lecture on detector physics in the context of the actual ATLAS designs • Preserve knowledge potentially lost due to people leaving ATLAS • Not everything is written down! • SLAC could help to organize and host such schools • Need to involve external experts to help with lectures • E.g., ATLAS calorimetry, some physics topics… • Probably need modest financial support • Not clear – I imagine mainly for the lecturers’ travel and some local infrastructure • Scope of support? Possibly from US ATLAS project?? • Helps grad students in smaller groups to get into ATLAS faster • Education beyond the local expertise useful for full physics analysis • Can improve and accelerate understanding of data • Knowledge of detector principles often weak point in analyses