1. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Florian Goebel Max-Planck-Institut für Physik (Werner-Heisenberg-Institut) München for the MAGIC collaboration EGEE Generic Applications Advisory Panel CERN, 14. June 2004 The MAGIC Telescope

2. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Outline What is MAGIC? What is MAGIC? The collaboration The collaboration The Telescope The Telescope Commissioning Status Commissioning Status The physics case The physics case Computing Requirements Computing Requirements Data Acquisition Data Acquisition Data analysis Data analysis MC production MC production First steps to the GRID First steps to the GRID Conclusions Conclusions

3. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN The MAGIC Collaboration Barcelona IFAE, Barcelona UAB, Crimean Observatory, U.C. Davis, U. Lodz, UCM Madrid, INR Moscow, MPI München, INFN/ U. Padua, INFN/ U. Siena, U. Siegen / U. Berlin, Tuorla Observatory, Yerevan Phys. Institute, INFN/ U. Udine, U. Würzburg, ETH Zürich Major Atmospheric Gamma-Ray Imaging Cherenkov Telescope n International collaboration of n > 100 physicists n 16 institutes n 11 countries

4. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN The MAGIC telescope Largest Imaging Air Cherenkov Telescope (17 m mirror dish) Largest Imaging Air Cherenkov Telescope (17 m mirror dish) Located on Canary Island La Palma (@ 2200 m asl) Located on Canary Island La Palma (@ 2200 m asl) Lowest energy threshold ever obtained with a Cherenkov telescope Lowest energy threshold ever obtained with a Cherenkov telescope n Aim: detect  –ray sources in the unexplored energy range: 30 (10)-> 300 GeV

5. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Imaging Air Cherenkov Telescopes ~ 10 km Particle shower ~ 1 o Cherenkov light ~ 120 m Gamma ray Cherenkov light Image of particle shower in telescope camera

6. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Standard Analysis hadron shower (background) gamma shower raw imagecleaned image Shower reconstruction and background rejection based on image shape analysis Hillas parameters: Length, width, distance, alpha

7. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN First source observations Mkn 421 (AGN) February 2004 (in flaring state) Alpha distribution 1200 excess events 800 background events Source position 100 minutes observation => Significance: 23 sigma

8. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Future of MAGIC observatory Second telescope MAGIC type telescope under construction (more observation time, background rejection & better event reconstruction in coincidence mode) Second telescope MAGIC type telescope under construction (more observation time, background rejection & better event reconstruction in coincidence mode) Plans for 30 m telescope for gamma astronomy down to E = 5 GeV Plans for 30 m telescope for gamma astronomy down to E = 5 GeV MAGIC I

9. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN The unexplored spectrum gap Satellites give nice crowded picture of  – ray energies up to 10 GeV. Satellites give nice crowded picture of  – ray energies up to 10 GeV. Ground-based experiments show very few sources with energies > ~300 GeV. Ground-based experiments show very few sources with energies > ~300 GeV. Effective area < 1 m 2 Effective area > 10 4 m 2 Close gap with MAGIC expect discovery of many new sources

10. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN n AGNs n SNRs n Cold Dark Matter n Pulsars n GRBs n Tests of Quantum Gravity effects Cosmological  -Ray Horizon Cosmological  -Ray Horizon The MAGIC Physics Program The MAGIC Physics Program n Origin of Cosmic Rays

11. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN  -rays travelling cosmological distances interact Extragalactic Background Light (EBL)  -rays travelling cosmological distances interact with the Extragalactic Background Light (EBL) Absorption of extragalactic  - rays Attenuated flux is function of  -energy and source distance (redshift z). For IACTs energies (10 GeV-10 TeV), the interaction takes place with infrared  (0.01 eV-3 eV, 100  m-0.5  m). For IACTs energies (10 GeV-10 TeV), the interaction takes place with infrared  ’s (0.01 eV-3 eV, 100  m-0.5  m). MAGIC EBL

12. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN MAGIC phase I MAGIC phase II Gamma Ray Horizon  The EBL absorption limits the maximum observable distance of  -ray sources. A lower energy thresholds allows a deeper look into the universe Gamma Ray Horizon

13. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Mkn 501 (z=0.034) Measurement of the IR background Absorption leads to cutoff in AGN spectrum Absorption leads to cutoff in AGN spectrum Need several sources (@ similar z) to disentangle source intrinsic effects. By measuring AGN spectra up to z=1, MAGIC can help determining the IR background IR background light is result of history of the universe IR background light is result of history of the universe (Star formation, Radiation of stars, Absorption and reemission by ISM)

14. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Pulsars Where do  -rays come from? Outer gap or polar cap? Where do  -rays come from? Outer gap or polar cap? n 30 – 100 GeV decisive energy range 7  -ray pulsars seen by EGRET (E < 10 GeV) 7  -ray pulsars seen by EGRET (E  < 10 GeV) Only upper limits from present IACTs for pulsed emission (spectral cut-off) Only upper limits from present IACTs for pulsed emission (spectral cut-off)

15. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Gamma Ray Bursts Origin and acceleration mechanism not yet fully understood. Origin and acceleration mechanism not yet fully understood. Do GRBs have E > 10 GeV counterparts? Do GRBs have E > 10 GeV counterparts? MAGIC can observe 1-2 GRB/year GRBs are short (10 – 100 sec) GRBs are short (10 – 100 sec) => Need fast repositioning after GRB alert If GRB origin very far If GRB origin very far => High energy  -rays will be absorbed by EBL => Need low energy threshold

16. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Search for Dark Matter Particles n Neutralino (lightest SUSY particle) is attractive Cold Dark Matter candidate  -flux from  annihilations:  -line E  = m   - line E  = m   m  2  m   continuum Particle physics:  -continuum dominates  -  - lines suppressed CDM density:  -ray flux ~  2 => search for CDM clumps galactic center (high diffuse  bkg), dwarf spheroidal and nearby galaxies, globular clusters observe: galactic center (high diffuse  bkg), dwarf spheroidal and nearby galaxies, globular clusters

17. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Key Elements of the MAGIC Telescope 17 m diameter reflecting surface (240 m 2 ) 17 m diameter reflecting surface (240 m 2 ) Analog signal transport via optical fibers Analog signal transport via optical fibers 2-level trigger system & 300 MHz FADC system 2-level trigger system & 300 MHz FADC system IPE CENET Active mirror control Active mirror control Diamond milled aluminum mirrors Diamond milled aluminum mirrors Light weight Carbon fiber Structure for fast repositioning 4 o FOV camera 577 high QE PMTs 4 o FOV camera 577 high QE PMTs

18. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN The Signal Processing Camera: 577 PMTs (up to 30% QE) ~ 2 nsec short pulses in Counting House: in Counting House: Stretch pulse to 6 nsec Stretch pulse to 6 nsec Split to high & low gain Split to high & low gain Digitize with 300 MSamples/s 8 bit FlashADCs Digitize with 300 MSamples/s 8 bit FlashADCs (in future 2GS/s) DAQ: DAQ: Linux PC with multithreaded C++ DAQ program Linux PC with multithreaded C++ DAQ program FPGA based PCI readout card FPGA based PCI readout card 15 high + 15 low gain slices/channel 15 high + 15 low gain slices/channel Typical dead time < 1 % Typical dead time < 1 %

19. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Two Level Trigger Discriminators L0 Discriminators L0 Software adjustable threshold for minimum number of photoelectrons per pixel Level 2 L2 Level 2 L2 Slower (50-150 nsec) but advanced topological pattern recognition Level 1 L1 Level 1 L1 Fast (2-5 nsec) coincidence device performing simple n-next-neighbor logic To FADC Total trigger rate: so far ~ 200 Hz (@ 60 - 80 GeV threshold) reduce threshold to 30 GeV => 500 Hz expected rate rate dominated by hadronic background

20. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Data Acquisition Rate & Storage Event Size: Event Size: 577 PM x 1 Byte x 30 samples 577 PM x 1 Byte x 30 samples  ~ 20 kByte/event Data Acquisition Rate: Data Acquisition Rate: 500 Hz typical trigger rate 500 Hz typical trigger rate  ~ 10 MByte/sec Data Storage Requirements: Data Storage Requirements: ~ 1000 h / year useful moonless observation time ~ 1000 h / year useful moonless observation time  ~ 36 TByte/year

21. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Data Flux Scheme La Palma: FileserverLTO2 Tapes (x 2) DAQ MERPP Data Center : Wuerzburg (+ Barcelona) Fileserver Tape archive Tape transfer Preprocessing & Data reduction (c++, root) raw data Regional Data Centers Physics analysis preprocessed data

22. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Standard Data Processing data processing rate data size (events/sec)(% of raw data) Rootification400 ~ 53 % (event building & compression) Rootification400 ~ 53 % (event building & compression) Pedestal subtraction, signal extraction,200~ 32 % calibration Pedestal subtraction, signal extraction,200~ 32 % calibration Image cleaning,400 ~ 0.5 % Hillas parameter calculation Image cleaning,400 ~ 0.5 % Hillas parameter calculation Total:100 Total:100 @ 500 Hz data acquisition rate => need ~ 5 Xeon 3GHz type processors @ 500 Hz data acquisition rate => need ~ 5 Xeon 3GHz type processors On 3 GHz Xeon Processor

23. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Standard Physics Analysis Methods Main challenge: Reject background from cosmic ray hadrons Main challenge: Reject background from cosmic ray hadrons Background rate: ~ 500 Hz compared to ≤< 1 Hz signal rate Background rate: ~ 500 Hz compared to ≤< 1 Hz signal rate dynamical cut methods dynamical cut methods Neural network, random forest Neural network, random forest => Significant additional computing needed

24. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Low energy analysis methods Hillas analysis fails since Hillas analysis fails since shower shape not well reconstructed shower shape not well reconstructed Strong dependence on suppression of night sky background (“image cleaning”) Strong dependence on suppression of night sky background (“image cleaning”) Model analysis Model analysis fit mean shower shape to complete camera data fit mean shower shape to complete camera data precise statistical tests precise statistical tests better shower reconstruction using shower tail information better shower reconstruction using shower tail information very promising results obtained from CAT & HESS telescopes very promising results obtained from CAT & HESS telescopes

25. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Model Analysis computing requirements CPU requirements CPU requirements Fit very CPU intensive Event reconstruction rate only several 10’s of Hz Fit very CPU intensive Event reconstruction rate only several 10’s of Hz Computation of mean shower shape Computation of mean shower shape Use MC or semi-analytical approach Use MC or semi-analytical approach Shape depends on: energy, impact parameter, zenith angle Shape depends on: energy, impact parameter, zenith angle Integrate over: shower depth, energy, angular, lateral distribution Integrate over: shower depth, energy, angular, lateral distribution 4 x 10 11 steps ≈ 500 days x CPUs 4 x 10 11 steps ≈ 500 days x CPUs Data storage requirements Data storage requirements Need to keep calibrated data Need to keep calibrated data no zero suppression (image cleaning) possible no zero suppression (image cleaning) possible

26. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN MC generation Events needed: Events needed: Signal (gamma) events: > 1 x data => ~ 3 M events Signal (gamma) events: > 1 x data => ~ 3 M events Background events:> 1/100 x data => ~ 16 M events Background events:> 1/100 x data => ~ 16 M events Shower development in atmosphere CORSIKA (f77) Atmospheric absorption & Reflection on mirror Reflector (c) Detector response Camera (c++)

27. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN MC CPU requirements Gammas (Signal) Gammas (Signal) Trigger efficiency:7% Trigger efficiency:7% 3 M events => generate: 43 M events 3 M events => generate: 43 M events Hadrons (background): Hadrons (background): Trigger efficiency:0.15 % Trigger efficiency:0.15 % 16 M events=> generate: 10 G events 16 M events=> generate: 10 G events Production rate (Xeon 3GHz): Production rate (Xeon 3GHz): Shower simulation:900 events/h/CPU Shower simulation:900 events/h/CPU (x ~100) reuse event for various impact parameters (x ~100) reuse event for various impact parameters Mirror & detector simulation:60 kevents/h/CPU Mirror & detector simulation:60 kevents/h/CPU CPU power needed: CPU power needed: 10 Gevents/year / 60 kevents/h/CPU => need 50 CPU 10 Gevents/year / 60 kevents/h/CPU => need 50 CPU

28. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN MC storage requirements Corsika output: 28kB/event10.8kB/event Corsika output: 28kB/event10.8kB/event Reflector output: 7.6kB/event 1.3kB/event Reflector output: 7.6kB/event 1.3kB/event Keep only Reflector output Gammas Hadrons Gammas: 45 M events => 320 GB Hadrons: 11 G events => 13 TB

29. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN MAGIC: steps towards the GRID Start with MC production in Italy (CNAF in Bologna) Start with MC production in Italy (CNAF in Bologna) 3 years of experience with use of CONDOR (mainly INFN, Italy) 3 years of experience with use of CONDOR (mainly INFN, Italy) 90 M events produced using up to 100 CPUs 90 M events produced using up to 100 CPUs Connect main computing centers inside MAGIC for MC production & data analysis and storage Connect main computing centers inside MAGIC for MC production & data analysis and storage Wuerzburg, Barcelona, INFN (Padova, Bologna), ETH Zuerich, MPI Munich Wuerzburg, Barcelona, INFN (Padova, Bologna), ETH Zuerich, MPI Munich

30. F. Goebel, MPI München, 14. June 2004, EGAAP, CERN Conclusions MAGIC: is a new generation gamma ray Cherenkov telescope is a new generation gamma ray Cherenkov telescope has large discovery potential both in astrophysics and fundamental physics has large discovery potential both in astrophysics and fundamental physics just started data taking just started data taking has large computing requirements has large computing requirements > 100 CPU > 100 CPU > 50 TB / year > 50 TB / year is well suited to join and test GRID technology with 16 participating institutions over all Europe (and beyond) some with strong links to mayor GRID sites (Bologna, Barcelona) is well suited to join and test GRID technology with 16 participating institutions over all Europe (and beyond) some with strong links to mayor GRID sites (Bologna, Barcelona)