Abstract: ECAL upgrade performance plots for 2013 LHCC and ECFA.
The CMS electromagnetic calorimeter (ECAL) is made of about 75000 scintillating lead tungstate crystals arranged in a barrel and two endcaps. The scintillation light is read out by avalanche photodiodes in the barrel and vacuum phototriodes in the endcaps, at which point the scintillation pulse is amplified and sampled at 40 MHz by the on-detector electronics. The fast signal from the crystal scintillation enables energy as well as timing measurements from the data collected in proton-proton collisions with high energy electrons and photons. The stability of the timing measurement required to maintain the energy resolution is on the order of 1ns. The single-channel time resolution of ECAL measured at beam tests for high energy showers is better than 100 ps. The timing resolution achieved with the data collected in proton-proton collisions at the LHC is presented. The timing precision achieved is used in important physics measurements and also allows the study of subtle calorimetric effects, such as the timing response of different crystals belonging to the same electromagnetic shower. In addition, we present prospects for the high luminosity phase of the LHC, where we expect an average of 140 concurrent interactions per bunch crossing (pile-up). It is speculated that time information could be exploited for pileup mitigation and for the assignment of the collision vertex for photons. In this respect, a detailed understanding of the time performance and of the limiting factors in time resolution will be important.
NB: There is no associated detector note.
At test beam, in 2008, the ECAL time intrinsic resolution was measured to be better than 100 ps. For collisions there are several effects that can worsen the resolution: run by run variations, intercalibration, effects vs energy, radiation, etc... The calorimeter has been properly calibrated to take care of part of such effects. The goal is now to estimate the resolution and compare it with the design one coming from TB.
For the following plots we use high pt photon-like ECAL deposits by using the following selection criteria: 1) they pass cluster shape strict requirements to look like a real em deposit based on Sminor and Smajor variables, 2) no isolation requirements (π0 are fine) are applied
We use a method which is almost identical to the one of the TB analysis. We compare the timing of neighbouring crystals of an ECAL cluster which have a very similar energy. This is to minimize shower propagation effects. We require:
The resolution is estimated from a gaussian fit, taking the core of the distribution. The fit is in the range mean±2RMS.
Results are for 2011 + 2012 data. Barrel only results are shown.
A study based on Z reconstruction has been also performed. For electrons we apply
The time of the electron corresponds to the time of the cluster seed crystal. When comparing the time of the two electrons we correct for time of flight differences due to primary vtx position.
Figure | Caption |
---|---|
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the effective amplitude, normalized to the noise in the ECAL Barrel for 2011+2012 data. The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. Bottomline: noise term consistent with TB. Constant term about 70ps. |
pdf version |
Resolution of time difference between the two electrons from Z->ee decays, as a function of the effective amplitude, normalized to the noise in the ECAL Barrel for 2011+2012 data. The selection applied, the method and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. Bottomline: noise term consistent with TB. Constant term about 150ps, much larger than the one obtained with the neighbouring crystals method. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the effective amplitude, normalized to the noise in the ECAL Barrel for 2011+2012 data, for crystals belonging to the same readout unit (trigger tower). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. Bottomline: noise term consistent with TB. Constant term smaller than 70ps. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the effective amplitude, normalized to the noise in the ECAL Barrel for 2011+2012 data, for crystals belonging to different neighbouring readout units (trigger towers). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. Bottomline: noise term consistent with TB. Constant term about 130 ps, quite larger than the one obtained when the two crystals belong to the same readout unit. This explains why Z method and neighbouring crystals method give such a different constant term. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the run number for crystals belonging to the same readout unit (trigger tower). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. The resolution here is estimated by the spread of t1-t2 distribution after placing a cut on Aeff (Aeff > 30GeV). This is because there is not enough statistics to perform the sigma vs Aeff/sigma_n fits per run. As a consequence, it can be slightly larger than the constant term. Bottomline: resolution quite stable vs run. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the run number for crystals belonging to different neighbouring readout units (trigger towers). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. The resolution here is estimated by the spread of t1-t2 distribution after placing a cut on Aeff (Aeff > 30GeV). This is because there is not enough statistics to perform the sigma vs Aeff/sigma_n fits per run. As a consequence, it can be slightly larger than the constant term. Bottomline: resolution seems to increase with time in 2011. At the beginning of 2011 it was not very different from the one obtained for crystals in the same readout unit. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the pseudorapidity for crystals belonging to the same readout unit (trigger tower). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. The resolution here is estimated by the spread of t1-t2 distribution after placing a cut on Aeff (Aeff > 30GeV). This is because there is not enough statistics to perform the sigma vs Aeff/sigma_n fits per run. As a consequence, it can be slightly larger than the constant term. Bottomline: resolution is quite stable vs eta. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster as a function of the pseudorapidity for crystals belonging to different neighbouring readout units (trigger towers). The selection applied and the resolution are the ones specified in the introduction. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. The resolution here is estimated by the spread of t1-t2 distribution after placing a cut on Aeff (Aeff > 30GeV). This is because there is not enough statistics to perform the sigma vs Aeff/sigma_n fits per run. As a consequence, it can be slightly larger than the constant term. Bottomline: resolution is quite stable vs eta. |
pdf version |
Resolution of time difference between the two most energetic crystals of an ECAL cluster for crystals belonging to same readout unit (trigger tower), for each readout unit. The selection applied here differs from the previous plots and it is much looser (no requirement on E1/E2 and loose isolation) to increase statistics. The effective amplitude, Aeff, corresponds to A1A2/sqrt(A1^2+A^2), where A1 and A2 are the amplitude of the two crystals. The noise corresponds to 42MeV. The resolution here is estimated by the spread of t1-t2 distribution after placing a cut on Aeff (Aeff > 15GeV). This is because there is not enough statistics to perform the sigma vs Aeff/sigma_n fits per run. As a consequence, it can be slightly larger than the constant term. White spots correspond to dead towers. Bottomline: resolution is quite stable vs the full barrel, there are local variations which show that regions of the detector (in particular some SMs) seem to behave better. |
The results proposed here are aimed to determine the physics performance of a precision timing detector integrated in the electromagnetic calorimeter. They start from a very basic study at Geant level to understand
These are aimed to find the best information to be stored in Digis and Reco in order to use a possible new timing determination for PU mitigation purposes. This approach has the advantage of integrating this new timing information with full reconstruction and the particle flow algorithm. Clearly, the use of full simulation makes the conclusions on the physics performance more realistic.
The current ECAL geometry is used. Crystals are divided in different longitudinal sub-cells. In each sub-cell the new timing information is extracted as the average of all Geant hits time within that volume, weighted by the energy of each deposit (only hits with E>5 KeV used).
The thickness and the position of the subcell has been studied in terms of intrinsic spread of this new timing information. The smallest spread is obtained for 1 cm thick layer when there are about 8-10 radiation lengths in front of it, i.e. at the maximum of the shower. The dependence vs depth is shown in the plot below (already approved).
In the following studies we use such new timing information extracted for a 1 cm layer after 10 radiation lengths. To emulate a real detector timing is further smeared with different gaussian resolutions.
In the different MCs used for the following studies the simulated interaction region corresponds to the configuration which reproduces Run1 conditions, i.e. the spread in z is about 5-6 cm. In addition, no spread in the time of the interaction is simulated. This means that the time of the interaction is constant.
Figure | Caption |
---|---|
pdf version |
Vertex determination with timing in H -> gamgam events, when both photons are in the barrel. The timing corresponds to the new timing (see introduction) of the crystal seed of each of the two photon clusters. We require pT(photon)>5GeV. The vertex position is determined by imposing that the photons originate from the same primary vertex. In this sample the time of the primary interaction is constant. The resulting resolution (RMS of the z vertex) is plotted vs different smearing resolutions. Bottomline: with a 30ps detector the resulting resolution on the vertex is about 1cm. |
pdf version |
Vertex determination with timing in H -> gamgam events, when both photons are in the endcap. The timing corresponds to the new timing (see introduction) of the crystal seed of each of the two photon clusters. We require pT(photon)>5GeV. The vertex position is determined by imposing that the photons originate from the same primary vertex. In this sample he time of the primary interaction is constant. The resulting resolution (RMS of the z vertex) is plotted vs different smearing resolutions. Bottomline: with a 30ps detector the resulting resolution on the vertex is about 0.6 cm. |
pdf version |
Effect of a timing cut on ECAL based SumET. SumET is here calculated by summing all ECAL hits with E>1GeV. In this sample the time of the primary interaction is constant. The three histograms correspond to: 1) no PU: here SumET is about ET(photon1)+ET(photon2) 2) ave nPU ~ 20 with no additional cut: clear bias due to PU additional energy 3) ave nPU ~ 20 with a requirement that hit timing has to be within a given 90ps window: most of the PU contribution is gone. A small fraction of the photon ET is also removed due to show propagation effects. This will be improved in the future when the shower propagation will be taken into account. |
pdf version |
TOF-corrected timing for jets coming from the primary interaction and PU jets. The timing corresponds to new timing (defined in the introduction) of the ECAL crystal with the largest energy within a jet. This timing is subtracted of the TOF under the hypothesis that the jets come from the primary interaction. In this sample the time of the primary interaction is constant. We then expect this variable to be about zero for jets from primary vertex and wider for jets from PU. The asymmetric distribution for jets from the primary interaction is due, for instance, to charged tracks which are bent and arrive, in general, later than a photon. Bottomline: visually this variable can be used to reject PU jets. |
pdf version |
Impact of a selection based on a TOF-corrected timing for jets compared to a tracker based selection (beta*). The timing corresponds to new timing (defined in the introduction) of the ECAL crystal with the largest energy within a jet. This timing is subtracted of the TOF under the hypothesis that the jets come from the primary interaction. The sample used corresponds to QCD events. In this sample the time of the primary interaction is constant. Different selections are applied based on this variable and the efficiencies for jets from primary vertex and PU are evaluated and shown in a ROC curve. beta* is the variable used in several CMS analysis and corresponds to the ratio of the sum of the pT of tracks of the jet not pointing to the primary vertex w.r.t. the sum of the pT of all tracks. This variable is evaluated in the endcap region where the tracker starts to be inefficient. |
-- FedericoFerri - 20 Mar 2014