what
Linear Collider WorkShop '99
where
Melia Gran Sitges hotel, Sitges, suburb of Barcelona
who
Richard Dubois, SLAC
purpose of visit
present state of North American simulations and present study of calorimeter performance of the American designs
This workshop is the 4th in a series focussing efforts on fostering a next linear e+e- collider and encouraging the international high energy physics community to rally behind the project. It brought together physicists from the three regional efforts (Asia, Europe and North America). The meeting followed on closely from preparatory meetings in the three regions, at KEK, Oxford and Ann Arbor. Approximately 210 people were in attendance, with about 20 from Asia and 60 from North America. The Asian and European efforts are fast-tracked in the attempt to deliver to their governments proposals for the accelerator and detector by mid 2001, with the hope of a construction start in 2003-2004. The American effort is more modest, with timescales 1-2 years later, and a detector design study which is more exploratory than the other programs.
JC Brient gave in an interesting report on energy flow issues for calorimetry. The high B fields (3T in the TESLA case) cause ~7% of the charged energy to fail to reach the calorimeter. The dijet mass resolution is also degraded by the spreading of charged particles in the field. This all points to the necessity of a combined tracker-calorimetry energy flow measurement. ALEPH is able to get a jet energy resolution of 60%/Root(E) even with an 85%/Root(E) hadronic calorimeter. He found that the method really stresses 3D segmentation more than energy resolution. This bodes well for the American strawman designs which are highly segmented. Good clustering is also important as the 'fake photons' caused by hadronic shower splitoff being misinterpreted as neutral. Jet resolutions of 50%/Root(E) are possible. He also stressed the ability to reconstruct neutral hadrons. He has been considering W-Si as an option to get the good segmentation. Finally, he recommends tau's as a good testbed for these studies. He found large constant terms in the hadronic energy dependence (as we did in the American studies) as well EM performance in the range of 17%/Root(E).
Volker Korbel reported that the TESLA group has decided to abandon the Shaslik calorimeter design and adopt a scintillating tile device with wave-length shifted fiber readout. Fibers with 7-9 m attenuation length can be obtained now; 10m would be necessary. For the TESLA field and calorimeter size, he finds 3.75 cm as the most probable distance between particles. They believe 5 cm pads will suffice. They expect to use the Higgs to two photon mode as a test of the design; they will see if they can differentiate between single photons and pi-zeros.
Yoshiaki Fujii reported on the JLC studies. They are also using tile-fiber with a large radius and good longitudinal segmentation. They have done test beam studies at KEK in the energy range of 1-4 GeV and find resolutions of 15% and 33%/Root(E) respectively for EM and hadrons. They will do further beam tests at FNAL at higher energies soon. They are pursuing R&D on their photon detectors as they need high sensitivity light collection in their 2T field. He suggested a few references for work on W-Si calorimeters:
They have been able to simulate hadronic longitudinal shower fluctuations in their FastMC. They found an orthogonal set of parameters to deal with the correlations. They found that only the first 5 layers needed this treatment. They also make use of a transverse shower profile. This method should be investigated.
Ren Yuon Zhu of CalTech reported on the possibility of using crystals for the EM calorimeter. PbWO_4 is an excellent candidate with system resolutons of 4%/Root(E) expected. PbF_2 is a promising alternative, but requires more R&D to be feasible since it does not scintillate. The general downside to the crystal approach was the lack of longitudinal segmentation.
Steve Magill of ANL reported on requirements for forward calorimetry. His conclusion was that good segmentation (pad sizes less than 5 mm) was again more important than energy resolution. He also found that the losses down the beampipe dominated any effects of the masking scheme.
I'll give a short summary of our American studies. EM resolutions were around 15%/Root(E) for both Small and Large devices. The hadronic resolutions were in the neighbourhood of 40%/Root(E); this ignored the coil in the Small calorimeter. The coil placed between the Small EM & Hadronic calorimeters has a noticeable effect. The results were somewhat clouded by the handling of tails in the energy distributions. How these were handled caused a tradeoff between the core resolution and constant term. The constant terms reported were in the vicinity of 7%. Another effect seen was that the large field (6T) appeared to induce an increase of 10-20% in signal in the EM calorimeter for electrons. It was reported that CMS has seen such an effect in their Pb-scintillator test beam studies. The conclusions were that we need to understand the large constant terms and the effect of the B field; also to expeditiously pursue good clustering and track-cluster association algorithms in order to get the best handle on the energy flow.
It was pointed out to me by Eric Torrence that OPAL uses an energy flow strategy in which the calorimeter energy is decreased by the total amount of measured charged momentum as found by the tracker. This is done on a per-hemisphere basis. They find equivalent results to a full-blown algorithm where track-cluster association is used to exclude charged energy from the calorimeter. Presumably this method will still suffer from the jet spreading due to the much larger magnetic fields proposed for the LC detectors.
The Asian effort continues to base their full simulations on a Fortran GEANT3 package, called JIM. They have just completed a trial version of a FastMC, completely using Root, based on the ATLFAST model offered by the Root team. They have also written a Root interface to their JIM output. This Root package came as a surprise to the Root folks who had heard nothing about it until it was announced as a completed package
The Europeans have an entirely Fortran-based operation and seem to have turned the corner from tool building to tool using. They have made great progress since their November meeting, which I attended. They have adapted SLD's ZVTOP vertex finder and DELPHI's track finding and fitting packages, so they have a complete tracking reconstruction. They have been able to feed back performance specs from their BRAHMS full simulation in to the SIMDET FastMC, so that in many respects it matches results from the full simulation remarkably well.
They plan to migrate to an OO model and GEANT4 after the CDR is out. They expect to use whatever the LHC groups move to, be it LHC++ or Root. They are betting on Root; it is too early to tell whether JAS will be a significant player.
Again, for comparison, the American LCD effort has been fully OO from the beginning, using GLAST's C++ GISMO for full simulation, and dual analysis/reconstruction paths in Java Analysis Studio (JAS) and Root. Root has been eclipsed in popularity with LCD users by JAS, as JAS has improved by leaps and bounds over the past year. There are parallel FastMC's in the two packages, while in JAS there is a full tracking and calorimetry reconstruction (although the calorimeter reconstruction is still crude).
Machine backgrounds are the next issue to be taken up. This echoed the need for background overlays and the ability for at least some of the detector components to keep track of time. For example, there was a request by a Higgs speaker requesting time information from the tracker and calorimeter with 1 ns resolutions.
The American effort is also looking to a working release of GEANT4 that does not depend on commercial components (notably the Rogue Wave standard template library). This was promised in a presentation at this meeting for sometime in May.
R.Dubois Last Modified: 01/13/04 12:43