CMS-DP-2012/008
ECAL Detector Performance Plots: more on 2011 data

Abstract: Detector Performance Plots of the CMS Electromagnetic calorimeter, based on the 2011 pp data sample. Plots include: energy corrections, calibration, light monitoring, ECAL spikes.

CDS entry

iCMS entry

Figure	Caption
pdf version	Energy Correction along pseudorapidity Average MC-driven correction as a function of the pseudorapidity for W->enu electron candidates with R9>0.94 and R9<0.94. The large scale structure, particularly between 0.8< abs(eta) <2.2 is dominated by tracker material effects. Local structures correlate with inter-module boundaries in the Barrel, for example at abs(eta) =0, 0.45,0.8,1.15.
pdf version	Local containment corrections in the ECAL Barrel Dependence of the reconstructed energy as a function of the super-cluster pseudorapidity position, with respect to the local position within the crystal with maximum energy. The x-coordinate is normalized so that -0.5 and 0.5 indicate the crystal edges; the crystals within each Barrel module are superimposed in phi. MC driven corrections reduce the dependence of the scale on the local position of the cluster. A residual sub-module dependent spread in the response (due to the variation of crystal staggering along eta) of 0.3%-0.5% rms remains. Improved corrections obtained directly from data will be used in the future to compensate for this effect.
pdf version pdf version	Response dependence on pileup Dependence of the reconstructed energy on the number of reconstructed vertices in the event. The default reconstruction of the data (open red circles) and MC (filled red circles) is compared to MC-driven corrections to the energy based on a multivariate analysis (MVA) of the energy response which includes pileup sensitive global event variables, for the data (open green circles) and MC (filled green circles) .
pdf version pdf version	PiZero reconstructed peaks in Barrel (top) and Endcap (bottom) Peaks fitted with a combination of a gaussian and 4th order polynomial. The entire 2011 dataset is considered, using special online PiZero calibration streams. S/B and fitted resolution indicated on the plots. The fitted peak positions are not exactly at the nominal pizero mass values. This is mainly due to the effects of selective readout and leakage outside the 3x3 clusters used in the mass reconstruction. In the endcaps, the energy from both the EE and the preshower (ES) is used. The total signal sample size (Nsignal from the fit) is estimated to be 8.7x10^9 in the Barrel and 2.9x10^8 in the Endcaps.
pdf version pdf version	Eta reconstructed peaks in Barrel (top) and Endcap (bottom) Peaks fitted with a combination of a gaussian and 2nd order polynomial. The entire 2011 dataset is considered, using special online Eta calibration streams. S/B and fitted resolution indicated on the plots. The fitted peak positions are not exactly at the nominal eta meson mass values. This is mainly due to the effects of selective readout and leakage outside the 3x3 clusters used in the mass reconstruction. In the endcaps, the energy from both the EE and the preshower (ES) is used. The total signal sample size (Nsignal from the fit) is estimated to be 1.0x10^9 in the Barrel and 3.5x10^7 in the Endcaps.
pdf version pdf version	Pi0 rate and S/B as functions of the crystal eta index in the ECAL Barrel Top plot: Average number of pi0 decays per crystal in the barrel, as a function of the crystal eta index. Bottom plot: Signal/Background from the pi0 peak fits in the barrel, as a function of the crystal eta index. The entire 2011 dataset is considered, using special online Pi0 and Eta calibration streams. The estimated statistical precision of the pi0 calibration procedure depends on both the accumulated pi0 statistics and the S/B ratio.
pdf version	Single Electron Occupancy from W->enu decays Number of electrons per crystal per 1/fb for events used in the single-electron calibration. Each point in the Barrel and Endcap corresponds to an eta-ring one crystal wide. The red points in the EE (abs(eta)>1.48) show the data before applying a cut on R9. The green points show the EE data after applying a cut of R9>0.9 to provide a dataset of non-showering electrons in a region where the tracker material budget is significant. There is no such cut in the EB where the green and red sets of points are identical.
pdf version pdf version	Crystal Transverse Energy in different ECAL rings Distributions of the transverse energy, ET, measured online from 160 million zero-bias events, for rings of crystals at specific values of abs(eta). The online energy threshold is150 MeV in EB (ET>60-150 MeV) and 650 MeV in EE (ET>60-260MeV). Technical details: the data in each histogram are from two rings of crystals, each located at the same abs(eta) on either side of CMS. In EB the pairs of rings comprise 720 crystals. In EE the number varies with pseudorapidity. The phi symmetry method: intercalibration constants are extracted by comparing the total transverse energy deposited in one crystal with the total transverse energy mean of the crystals in the associated pair of rings. The transverse energy sum is a truncated sum, constructed by summing ET values between a lower and an upper threshold. The lower threshold is applied to remove contributions due to noise. The upper threshold is applied to avoid possible bias from very high ET deposits. Offline, the lower EB threshold is 250 MeV, which is about 6 times the rms noise. In the EE the lower threshold is parameterized to take into account the response of the Vacuum Photo-diodes as a function of pseudorapidity. The offline lower thresholds are well above the online thresholds to minimise possible effects due to trigger bias. The upper offline ET threshold is set at 1 GeV above the lower threshold, for both EB and EE.
pdf version pdf version	The precision of the laser monitoring system (Top) Method to estimate the precision of the monitoring system: - Assume that the response is stable between p(n-1) and p(n+1) and that the variations are only due to the precision of the measurement - Interpolate the response at t(n) and compare the interpolation with p(n): d(n) = interpolation(t(n)) - p(n) - Repeat every 3 points, to avoid mutual correlations - Take the distribution of d(n) and get an estimate of the precision from its rms or from a gaussian fit (comparable results) (Bottom) Example of a history plot for the APD/PN ratio of one typical channel of EB at large pseudorapidity A low irradiation period (HI run) was used in this assessment of the precision of the laser monitoring system
pdf version pdf version	The laser monitoring precision in EB Precision as estimated from a gaussian fit of the distributions of d(n) for each of the crystals in EB. The top plot shows an example of the distribution and gaussian fit for one of the EB channels. The bottom plot shows the map of the gaussian sigma for each of the crystals in EB. The violet spots correspond to dead channels. The reddish patterns are related to regions where the common reference PN receives less light than normal.
pdf version pdf version	The laser monitoring precision in EE The precision as estimated from a gaussian fit of the distributions of dn for each of the crystals in EE. The violet spots correspond to dead channels. The reddish patterns are related to regions where the common reference PN receives less light than normal.
pdf version	The laser monitoring precision in EB and EE Distribution of the precision for the whole ECAL (i.e. the projection along z of the previous 2-D plots). The peaks are around 3*10^-4, much better than the required 2*10^-3.
pdf version	Spike Event Display CMS Event Display of a pp collision event (sqrt(s)=2.36 TeV), showing an isolated ECAL spike simulating a 600 GeV transverse energy deposit in the ECAL Barrel. The CMS detector is shown in x,y projection. The outlines of the CMS tracker and muon chambers are shown. The purple histogram shows the ECAL energy deposits (length proportional to the energy of the reconstructed signals). The defining signature of a spike in the ECAL Barrel is the presence of an isolated, high energy deposit.
pdf version	Spike pulse shape Oversampled pulse shape (normalised signal amplitude versus time) for spike (solid line) and electromagnetic (dashed line) energy deposits measured from CMS data. The rise time of the electronic pulse is consistent with an instantaneous signal from the Avalanche Photodiodes (APD), and not from the typical scintillation light signal in the crystal.
pdf version	The transverse energy distribution of the highest rechit in Minimum Bias data and Monte Carlo events A lower threshold of 3 GeV is applied to the transverse energy of the rechit, and  the two distributions are normalised to the total number of events below (1-E4/E1)<0.9. There  is a clear excess of high energy hits in the data distribution, due to spikes. These anomalous signals dominate the rechit spectrum at high energies. Above 20 GeV over 98% of the rechits are  due to spikes - and are a major source of missing transverse energy in CMS triggered events.  The spike energy distribution extends up to the ECAL Barrel saturation energy of ~ 1.7 TeV.
pdf version	The distribution of the “Swiss-cross” variable (1 − E4/E1) for data and Monte Carlo minimum bias events The data distribution (points), is normalised to the same number of events as the Monte  Carlo (hatched) below a Swiss-cross value of 0.9. It shows good agreement with the shape of the Monte  Carlo distribution below this value. However, there is a significant second peak at 1.0, due  to spikes not present in the default Monte Carlo simulation. A cut on the Swiss-cross variable at 0.95 is efficient at removing spikes. For a EM shower that is well-centered on a crystal, one expects approximately 80% of the shower energy in the central crystal, and ~20% of the energy in the four adjacent crystals. The Monte Carlo distribution therefore shows a peak at approximately 0.8 in the Swiss-cross variable, with a tail extending to low values and a relatively sharp cut-off above 0.95.
pdf version	The distribution of the RecHit timing for Minimum Bias data and Monte Carlo events Distribution of the RecHit timing for Minimum Bias data and Monte Carlo, plotted for the highest energy rechit in each event, requiring a minimum rechit transverse energy of 3 GeV. The timing distribution for Minimum Bias data is inconsistent with the Monte Carlo distribution (hatched). The peak at zero is due to prompt electromagnetic showers, and is modelled by the Monte Carlo. The secondary peak at −10 ns is due to spikes. The origin of the 10 ns difference can be understood by comparing the pulse shape of ‘normal’ (Swiss-cross< 0.95) and ‘spike’ (Swiss-cross> 0.95) RecHits. The faster pulse rise time for the spike RecHits is due to the lack of a scintillation component (80% of light emitted in 25 ns), since the spikes are produced by particles directly interacting in the APD. Since the time reconstruction algorithm compares this pulse to the expected shape, this faster rise time results in an apparent ‘early’ reconstructed time for the pulse.
pdf version pdf version pdf version	Spike rate vs minbias collisions rate (top) and vs number of tracks (center) (Top) Spike rate versus Minimum Bias trigger rate (technical bit 41) for early 7 TeV data. The dashed line is a straight line fit to the data points, with a gradient of 2.7 × 10−3 spikes per Minimum Bias event. (Center) Probability to observe a spike per minimum bias event as a function of the number of reconstructed charged tracks. Spikes are defined as ECAL energy deposits (RecHits) with ET > 3 GeV and (1-E4/E1)>0.95. (Bottom) The rate of spikes in CMS is proportional to the minimum bias collision rate, and to the number of charged tracks per event.
pdf version pdf version	Swiss-cross vs timing Two-dimensional distribution of the reconstructed signal timing (in ns) vs the Swiss-cross variable (1-E4/E1). Top plot, rechit transverse energy threshold > 3 GeV. Bottom plot , rechit transverse energy threshold > 10 GeV. Only the most energetic rechit in each event is plotted. Since the swiss-cross and rechit time variables relate to independent properties of the event (the energy deposition in the affected channel relative to its neighbours, and the pulse shape of the anomalous energy deposit) they should be independent variables to first order.
pdf version pdf version pdf version	Efficiency of spike cleaning cuts Estimates from data of the spike rejection efficiency of the Swiss-cross and timing cuts using Minimum Bias events. (Top) The efficiency of the Swiss-cross cut (1 − E4/E1 < 0.95) plotted as a function of rechit transverse energy.  The Swiss-cross cut is very efficient at high RecHit ET (better than 99% for ET > 10 GeV). There is a reduction in efficiency at the lowest rechit energies due to noise. For a spike with ET=3 GeV at eta = 0, a Swiss-cross value of 0.95 corresponds to 150 MeV in the adjacent four crystals, which is approximately an upward fluctuation of 2 standard deviations in the noise (single channel noise is approximately 40 MeV). In addition, there is a reduction in efficiency for rechits with T < 0.  This is due to the phenomenon of non-isolated spikes, which are more prevalent for negative T. (Center) The efficiency of the timing cut plotted as a function of rechit transverse energy. The efficiency of this cut is approximately 90%, independent of RecHit transverse energy. The inefficiency is due to spikes that fall within the ±3 ns timing window around zero. (Bottom)The Swiss-cross efficiency as a function of RecHit time. One can again see the reduction in efficiency for rechits with T < 0, due to non-isolated spikes.
pdf version	Double Spike A class of non-isolated events with two high energy adjacent crystals has been identified from scans of high MET events. An example is shown in the event display, which shows the transverse energy of ECAL energy deposits for a LHC collisions event at sqrt(s)=7 TeV. br<> In this event, 100 GeV is shared between two crystals, with very little surrounding activity. Both crystals are significantly out-of-time, suggesting a spike-related origin for the two hits. The energy deposition is also inconsistent with the shape expected from an electromagnetic shower. Such “double spike” events are believed to be caused by particles produced in hadronic showers during LHC collisions that induce anomalous signals in neighbouring APDs.
pdf version	APD Monte Carlo: location of spike progenitors A detailed model of the APD structure has been implemented in the CMS Monte Carlo simulation. The simulation (using Geant 4) is used to further understand the origin of spikes and their rates in CMS. The simulations show that direct ionization of the APD by protons/ions produced in pp collisions can produce large apparent energy signals after amplification (direct ionization produces ~2.8x10^5 e/MeV, c.f. ~4 p.e./MeV for scintillation light in PbWO4). The plots show the location in CMS of the particles produced in CMS collisions that are the progenitors of the APD (spike) hits in the simulation: pions are produced both at pp interaction point and close to APDs - photons are produced close to APDs - neutrons are produced in PbWO4 crystals - protons are produced in epoxy close to APD - ions are produced in APD high gain layer. The simulation results (and laboratory measurements) indicate that a significant fraction of the anomalous signals are produced by np scattering in the protective epoxy coating of the APD, and the resulting proton directly ionizing the APD active volume.