Performance of the ALICE Electromagnetic Calorimeter

The performance of the electromagnetic calorimeter of the ALICE experiment during operation in 2010-2018 at the Large Hadron Collider is presented. After a short introduction into the design, readout, and trigger capabilities of the detector, the procedures for data taking, reconstruction, and validation are explained. The methods used for the calibration and various derived corrections are presented in detail. Subsequently, the capabilities of the calorimeter to reconstruct and measure photons, light mesons, electrons and jets are discussed. The performance of the calorimeter is illustrated mainly with data obtained with test beams at the Proton Synchrotron and Super Proton Synchrotron or in proton-proton collisions at $\sqrt{s}=13$ TeV, and compared to simulations.


JINST 18 (2023) P08007
e-Print: arXiv:2209.04216 | PDF | inSPIRE
Figure group

Figure 1

Schematic view of the \gls{EMCal} (left) illustrating the module position on two approximately opposite locations in azimuth. The \gls{PHOS} calorimeter inside the \gls{DCal} is indicated in brown. The right figure shows a cross section of the \gls{ALICE} barrel detectors.

Figure 2

Photo and drawing of \gls{EMCal} module showing all components.

Figure 3

Schematic view of the \gls{EMCal} full-size super modules (\glspl{SM}) illustrating the strip structure made of 24 strips.

Figure 4

\gls{EMCal} full-size (\gls{SM}) in the $(\eta,\varphi)$ plane including visualizations of sub-components and their tower coverage.

Figure 5

Geometric overview of the \gls{EMCal} and \gls{DCal} detectors in the $\eta$-$\varphi$ plane. The drawing outlines the full \gls{LHC} Run 2 setup with all 20 \glspl{SM} as well as the \gls{PHOS} detector in the \gls{DCal} gap.

Figure 6

Example of a fit to a digitized pulse from electron test beam data using the function defined in \Eq{HW_Eq_Gauss2} with a fixed shaping time parameter $\tau = 235 $ns.

Figure 7

Distribution of the opening angle $\theta_{12}$ of two decay photons from \piz\ (left) and $\eta$ (right) mesons decays as a function of the meson energy obtained at generator level from a \gls{MC} simulation. The horizontal lines indicate the opening angle corresponding to a width of approximately 1, 3 and 5 cells separating the two photons Two photons completely merge into one cluster if they fall below the one cell distance, while they start merging at a width of approximately 3 cells, depending on the clusterizer The colored vertical lines correspond to the energy limits where two photons could still be fully resolved using the V1 (blue) and V2 (red) clusterizers.

Figure 8

Schematic comparison of different clusterization algorithms. Only one dimension is shown for simplicity Yellow boxes represent the energy in each cell. $E_\mathrm{agg}$ is the clusterization threshold as defined in the text. The different clusters are indicated by blue and red hatched areas Each panel represents a clusterization algorithm: a) V1, b) $3 \times 3$, c) V2, d) V1+unfolding.

Figure 9

Invariant mass distribution of cluster pairs in pp collisions at \sthirteen\ for different intervals of pair energy. The differently colored lines correspond to different clusterizer types, using the same aggregation $E_\mathrm{agg}=100$ MeV and seed $E_\mathrm{seed} = 500$ MeV thresholds. The lowest bin in energy uses the data sample with minimum bias trigger, while the others are obtained from the \gls{EMCal} \gls{L1} triggered data with thresholds at about $E \approx 4$ and $9$ GeV, respectively.

Figure 10

Fraction of neutral pions (left) and eta mesons (right) for which the showers from their decay photons are merged into a single cluster and can not be reconstructed using an invariant-mass analysis.

Figure 11

Distance between a cluster and the closest projected track in $\eta$ (left) or $\varphi$ angle (middle) versus the track momentum and for matched track-cluster pairs (right) in pp collisions at \sthirteen\ collected with the minimum bias trigger. Clusters are reconstructed using the V2 clusterizer. The black lines in the left and middle panels indicate the suggested selection criteria expressed in \Eq{Eq:TMVetoCriteria}.

Figure 12

Left: cluster-veto efficiency for primary particles (dark blue open circles) and electrons (green squares) as well as conversion electrons with a production vertex below 180 cm (green closed diamonds) and above 180 cm (green open diamonds) and other secondary particles (cyan open circles) as obtained from simulations of pp collisions at \sthirteen . Right: fraction of fake track-to-cluster matches for clusters originating from photons (yellow open squares) and other neutral particles (red open circles). Additionally, the same categories are shown for clusters that have additional charged particle contributions for photon clusters (orange squares) and other neutral particles (light red circles).

Figure 13

Schematic representation of the shower shape and the ellipse axes. The different colors indicate the amount of energy deposited in each cell, the darker the more energy.

Figure 14

\shshlo\ (left) and \shshsh\ (right) distributions in three energy intervals for photons, electrons, hadrons, \piz\ and $\eta$ mesons. The distributions are obtained using the V1 clusterizer from a simulation of pp collisions at \sthirteen\ performed with the \gls{PYTHIA} event generator, in which events are required to contain either two jets or a jet and a high-energy direct photon. Each distribution is normalized to its integral. A model simulating the effect of the cross talk was applied as discussed in \Sec{sec:crosstalk}. Distributions of \shshlo\ versus cluster energy in pp collisions at \sthirteen\ triggered by the \gls{EMCal} \gls{L1} at approximately 9 GeV for the V2 (left) and V1 clusterizer (right). Each energy bin is normalized to its integral and exotic clusters were rejected (\Sec{sec:exotics}).

Figure 15

Left: number of cells as a function of the cluster energy found with the V2 clusterizer in \pp\ collisions at \sthirteen\ using the \gls{EMCal} high threshold \gls{L1} $\gamma$ trigger The region below the lines is populated by exotic clusters The distribution for each energy bin is normalized to its integral. Right: comparison of $n_{\text{cells}}$ probability distributions for measured data (black), projection of the left plot, to simulated (blue) collisions for two different cluster energy bins. Each distribution is normalized by the integral of the distribution for $n_{\rm cells}$ > 10.

Figure 16

Left: ``exoticity'' ($F_{\rm +}$) as function of the cluster energy found with the V2 clusterizer in \pp\ collisions at \sthirteen\ using the \gls{EMCal} high threshold \gls{L1} $\gamma$ trigger The region above the line is populated by exotic clusters The distributions are normalized to have an integral of unity for each energy bin. Right: comparison of $F_{+}$ probability distributions for measured data (black), projection of the left panel, and simulated (blue) collisions for two different cluster-energy intervals. Each distribution is normalized by the integral of the distribution for $F_{+}<0.85$.

Figure 17

Left: cluster time for exotic ($F_{+}>$ 0.97 and non-exotic ($F_{+}<$ 0.97) clusters. The additional peaks in the time distribution beyond $100$ ns arise from additional bunch crossings, which could not be rejected by the online Past-Future protection using the \gls{V0} detector Right: difference in time of the most and second most energetic cell in the cluster for exotic and non exotic clusters. Both distributions are obtained for V2 clusters with energy in the interval $8 < E < 10$ GeV from data taken from pp collisions at \sthirteen.

Figure 18

\shshlo\ as a function of the exoticity parameter $F_{+}$ obtained from V2 clusters with $12< e the="" distributions="" are="" shown="" for="" pp="" collisions="" at="" using="" trigger="" l1-trig="" and="" central="" collisions.="">< /e>

Figure 20

Correlation between energy from the \gls{FEE} and FastOR readout based on the \gls{L1} \glspl{ADC} for towers corresponding to the same FastOR module. The black line indicates the mean energy in the FastOR for a given energy interval at \gls{FEE} level.

Figure 21

Trigger rejection factor (\gls{RF}) for the single-shower trigger in \pp{} collisions at \sthirteen{} for the single samples for 2018 (left) and the combined \gls{EMCal} and \gls{DCal} triggers (right).

Figure 22

Trigger efficiency for \gls{EMCal} (left) and \gls{DCal} (right) in \pp{} collisions at \sthirteen{}.

Figure 23

Left: trigger rejection of the jet triggers obtained from calorimeter-based jets with $R$ = 0.2 in \pp{} collisions at \sthirteen{} collected in 2017 and 2018. Ratios are with respect to minimum-bias events (red) or to events triggered by the low-threshold jet trigger (yellow). Right: corresponding trigger efficiency of the jet triggers in \pp{} collisions at \sthirteen{} obtained with fast simulations on cell level and with full simulations including the trigger response.

Figure 24

Trigger efficiency for different jet resolution parameters for the low-threshold (left panel) and high-threshold (right panel) trigger for calorimeter based jets as a function of their transverse momentum.

Figure 25

Trigger rejection for single-shower triggers in \pPb\ collisions at \seightplead, obtained from cluster energy spectra.

Figure 26

Correlation of the average energy density $\rho$ in \gls{EMCal} and \gls{DCal} scaled with the area of the L1-jet patch, determined with the L1 trigger electronics.

Figure 27

Left: Centrality percentile distribution of \gls{EMCal} \gls{L1} $\gamma$ triggered events (red) in comparison to the pure minimum bias distribution and minimum bias triggered events with a 10 GeV cluster in the event. Right: Trigger \glspl{RF} for the \gls{EMCal} or \gls{DCal} \gls{L1} $\gamma$ triggers in different centrality classes for \PbPb\ collisions at $\snn= 5.02$ TeV. Only statistical uncertainties of the trigger \gls{RF} are given in the legend.

Figure 28

Comparison of the trigger \glspl{RF} based on clusters and full jets for the \gls{EMCal} \gls{L1} $\gamma$ and jet trigger in 0--10\% (left) and 50--90\% (right) central \PbPb\ collisions at \sfivelead. Only statistical uncertainties of the trigger \gls{RF} are given in the legend.

Figure 29

Raw distribution of the invariant mass of cluster pairs in \gls{EMCal} (left) and \gls{DCal} (right) for one run of 2018 data taking obtained during the \gls{QA} process. The red line corresponds to a fit to the invariant mass distribution with a Gaussian function for the \piz\ signal and a second-order polynomial for the background. The fit parameters are used to monitor the performance of the reconstruction and of the detectors. In the displayed run, 819535 events were collected.

Figure 30

Mean number of clusters per event as a function of the run index for \pp{} collisions at \sthirteen{}. Example runs with similar data taking conditions are displayed Only clusters with energy above 0.5 GeV were used for the mean estimation. The gray vertical lines correspond to the start of different data-taking years. Mean number of cells per cluster (left) and mean cluster energy (right) for a selection of \glspl{SM} as a function of the run index in \pp{} collisions at \sthirteen{}. Example runs with similar data taking conditions are displayed. Only clusters with energy above 0.5 GeV were used for the mean estimation. The vertical red line indicates the run at which the 2 last \gls{DCal} \glspl{SM} were inserted into the readout. The vertical gray lines correspond to the start of different data-taking years.

Figure 31

Data flow and architecture from the raw-data collection by \gls{ALICE} to the eventual graphical representation in the \gls{Overwatch} web application .

Figure 32

Comparison of \gls{EMCal} and \gls{DCal} median energies of the trigger patch from the online \gls{QA} on the \gls{HLT}. The linear correlation indicates that both subdetectors are measuring similar event activity.

Figure 33

Schematic view of the \gls{ALICE} \gls{EMCal} mini-module at the \gls{PS} T10 beam line. The beam enters from the right. The Cherenkov detector was used for identification of the beam particle. The mini-module could be moved in the directions indicated by the red arrows in order to scan different towers.

Figure 34

Schematic view of the \gls{ALICE} \gls{EMCal} mini-module at the \gls{SPS} H4 beam line. The beam enters from the right. The mini-module could be moved in the directions indicated by the red arrows in order to scan different towers.

Figure 35

Left: Measured pulse amplitude ($A_{\rm out}$) as a function of input pulse amplitude obtained from laboratory measurements The dashed gray line indicates the case of a linear shaper. Right: Comparison of laboratory measurements with the \gls{TB} data on missing energy ($E_{\rm miss}$) as a function of the measured energy.

Figure 36

Left: energy distribution of single cell clusters obtained from scans with a 6 GeV muon-beam. Right: energy distribution of clusters obtained from scans with a 6 GeV electron beam. For both cases the data are shown with black markers and compared with the predictions from \gls{MC} simulations with \gls{GEANT}3 and \gls{GEANT}4 transport codes.

Figure 37

Cluster reconstruction/finding efficiency (left) and energy nonlinearity (right) as a function of hit position obtained from \gls{MC} simulations for 1 GeV electrons. Red markers stand for single cell clusters ($n=1$), blue makers stand for the clusters made of at least two cells ($n>1$), and the black markers stand for the clusters with any number of cells ($n>0$).

Figure 38

The reconstruction efficiency for the clusters made of at least two cells and for 1 GeV electrons as a function of hit position measured using the \glspl{MWPC} (left) and as a function of the incoming particle energy (right).

Figure 39

Energy nonlinearity correction ($E_{\rm rec}/E_{\rm beam}$) as a function of reconstructed energy for electrons obtained from \gls{TB} data (black points), and from \gls{MC} simulations with \gls{GEANT}3 (red points) and \gls{GEANT}4 (cyan points) transport codes.

Figure 41

Schematic overview of the positions of the survey points and how the vectors used for the alignment were constructed.

Figure 43

The probability distribution of the position of electron tracks propagated to the \gls{EMCal} \glspl{SM} surface and their associated cluster in $\eta$-direction (left) and $\varphi$-direction (right). The distributions are normalized by their integral. Mean differences between the position in $\eta$ (left) and $\varphi$ (right) of clusters and electron tracks propagated to the \gls{EMCal} / \gls{DCal} as function of the \gls{SM} number. The colors represent electrons (red) and positrons (blue).

Figure 44

Relative energy resolution as a function of the energy of the incident particle. Displayed are the target resolution (black, dashed-dotted) and the measured energy resolution from the test beam (green, dashed), \Sec{sec:TBNonLin} Additionally, three different bands are added showing the intrisic resolution with added 1, 2 and 3\% miscalibration, considering the residual miscalibration during the test beam to be between 0\% (upper band limit) and 1\% (lower band limit).

Figure 45

Left: Width of the distribution of $\Delta \mu_i$ (see text) as a function of the number of \piz\ mesons in sample {\tt S2}, for \glspl{SM} 0-9 (blue, green, cyan) and \glspl{SM} 10-17 (red, orange, yellow), for various $\sigma_{\piz}$ cell selection: $9.6 < \sigma_{\piz} < 11.0$ MeV/$c^2$ (circles), $9.0 < \sigma_{\piz} < 12.0$ MeV/$c^2$ (diamonds) and none (stars). Right: Statistical uncertainty on the \piz\ meson mass as a function of the number of \piz\ mesons collected in the cell, for \glspl{SM} 0--9 with the tightest (blue) and without (green) selection on $\sigma_{\piz}$ as well as for \glspl{SM} 10--17 with the tightest (red) and without (yellow) selection on $\sigma_{\piz}$.

Figure 46

Distribution of the ratios of the \piz\ meson masses found by a fit in the narrow interval ($\mu^{\text{fit}}_n$) of $70 <\mu_\pi^{0} < 220$ MeV/$c^2$ with respect to the standard range ($\mu^{\text{fit}}_s$) of $50 <\mu_\pi^{0} < 300$ MeV/$c^2$, for cells located behind the zones with more material (filled distribution) and for the other cells (blue distributions) for \glspl{SM} 0--17. The latter histogram is fit with a Gaussian (red).

Figure 47

Left: Distribution of the number of \piz\ mesons collected in an ideal calorimeter (black). The respective contributions of the 10 \gls{EMCal}, 6 2/3-sized \gls{DCal} and 4 1/3-sized \glspl{SM} are displayed in different colors (see \Sec{sec:hardware-supermoduledesign}) Right: Distributions of the smeared \piz\ meson masses normalized by $M_{\piz}^{\rm PDG}$, when only the statistical uncertainty is applied (closed symbols), and when also the uncertainty due to the fit range is applied (opened symbols).

Figure 48

Left: Energy distribution of all cells (blue), and all good cells (green) of the \gls{EMCal}. Right: Map of cells classified as good (yellow), bad (red) and dead (gray) for the part of the data taking period LHC18m as a function of row and column.

Figure 49

Hit distribution for 0.5 GeV $ \leq E_{\text{cell}} < $ 1.0 GeV before (left) and after (right) the scaling procedure. All cells are shown in blue, cells behind the TRD support structure in green, and cells not behind the TRD support structure in orange.

Figure 50

Left: Time distribution of \gls{EMCal} cells for different bunch crossings before the time calibration. Right: Aligned time distribution after the time calibration for cells with $E>$2 GeV.

Figure 52

Illustration of the fitting procedure of the normalized \gls{LED} signal for a good cell in \gls{EMCal} (left) and a problematic one in \gls{DCal} (right). The raw distribution obtained for the full 2018 data sample is shown in gray scales in the background, while the maxima in each temperature slice are indicated by green open circles. As there might be multiple clusters of points (as seen on the right) the distribution that is considered as the dominant cluster is marked by black closed circles, while the blue squares represent the shifted distributions after the correction for their offset is applied The final fit to the combined distribution of black and blue points is given as a dashed red line. Points marked in red were iteratively excluded from the fit as they were considered outliers.

Figure 53

Comparison of the obtained temperature calibration parameters in the \gls{EMCal} (left) and \gls{DCal} (right). The same cells were chosen as for \Fig{fig:emcalTempExtr} The calibration parameters were obtained separately for all years during which the corresponding \gls{SM} was installed and the cell considered good. The accessible temperature ranges for each year are indicated by the shaded areas in the same colors as the corresponding fits for the respective years.

Figure 54

Left: Radial distance from the \gls{IP} of photon conversions in the detector material for different detector configurations in 2010, 2012 and 2015-2018. The distributions are obtained for \gls{PYTHIA}8 simulations and only for photons whose conversion products were reconstructed as clusters in \gls{EMCal} and formed, when paired with another cluster, a signal in the $\pi^0$ invariant mass window. For 2011 the same number of super modules was installed as in 2012 - 2013, but two fewer modules had the \gls{TRD} installed in front of them Right: Neutral pion invariant mass peak position as a function of \pT\ for different magnetic field configurations for \pp\ collisions at \sthirteen. Invariant mass distribution of reconstructed $\pi^0$ mesons in \gls{MC} simulations. Contributions from pure photon pairs as well as from clusters which contain converted photon contributions are shown separately for $B = 0.5$ T (left), $B = 0.2$ T (middle) and $B = 0$ T (right) for pp collisions at \sthirteen.

Figure 55

Left: Mass positions for \gls{PCM-EMC}(blue) and \gls{EMC} (green) after applying the nonlinearity correction obtained from \Sec{sec:TBNonLin} \Figure{fig:5-TB-EnergyNonLinearity}. The mass positions are normalized to the neutral pion rest mass. In the case of \gls{PCM-EMC} the data points represent the squared mass position. Right: Ratio of mass positions in data and \gls{MC} for both techniques with their corresponding fits according to \Eq{eq:convcalo}.

Figure 56

Correction functions ($f_{\rm corr}$) for the default analysis cuts (see \Tab{tab:photonbasiccuts}) for \gls{CRF} and \gls{CCRF}, which are applied in \gls{MC}, together with the test-beam correction for data and its variation from \Sec{sec:TBNonLin}.

Figure 57

Relative mass position difference between data and simulations for the neutral pion (left) and $\eta$ meson (right) for different reconstruction techniques in \pp\ collisions at \sthirteen.

Figure 58

Left: Fraction of clusters with 2 or more cells for data and \gls{MC} for clusters selected with \gls{PCM-EMC} tagging and \gls{EMC} tagging. Right: Ratio of fully corrected \piz\ meson spectra obtained with \gls{PCM-EMC} (top) and \gls{EMC} (bottom) with $N_{\rm{cell}} \geq 2$ to the $N_{\rm{cell}} \geq 1$ with and without applying the cluster-size correction.

Figure 59

Schematic view of the measured \gls{ADC} time distribution shapes considering the contribution of the signal and a baseline. The effect of a baseline modification on the final shape is shown.

Figure 60

Probability distributions of the shower shape parameter $\shshlo$ of neutral clusters in data and simulations The different panels show different neutral cluster energy intervals. All distributions are normalized to their integral Data are shown as black histograms and simulations (\gls{PYTHIA}6 events with two jets or a direct photon and a jet in the final state, with \gls{GEANT}3 default settings) in blue. For the red histograms the modelling of the cross-talk observed in the \gls{EMCal} electronics was included in the simulations.

Figure 61

Comparison of the probability distribution of the shower shape parameter, $\shshlo$, of neutral clusters with $n^{\rm w}_{\rm cell}>1$ (left) and $>4$ (right) for different fractions of induced energy in the cross talk model (see \Eq{eq:crosstalk}), in pp collisions at \sseven\ \gls{EMCal} triggered data are compared to \gls{PYTHIA}6 simulated events with a direct photon and a jet or two jets in the final state, where one jet is triggered by a decay $\gamma$ on \gls{EMCal} acceptance with \pt$>$ 3.5 GeV/$c$. Data and default \gls{MC} (untuned simulation) are the same as in \Fig{fig:SigmaLongDataMCxTalk}.

Figure 62

Fraction of clusters with $n^{\rm w}_{\rm cell}>4$ within the range $0.1<\sigma_{\rm long}^{2}<0.3$ per \gls{SM}. Left and middle: pp \sthirteen\ \gls{L1} $\gamma$ triggered data. Right: Clusters (enhanced merged decay population by few \%) in \gls{EMCal} for pp \sseven\ minimum bias and \gls{L0} trigger data (black marker), compared to simulations with 2 jets in the final state with different photon trigger thresholds, with (red marker) and without (blue marker) cross-talk tuning.

Figure 63

Left: $\shshlo$ as a function of the cluster energy. Data, pp collisions at \sthirteen\ triggered by the \gls{EMCal} and \gls{DCal} with \gls{L1}-$\gamma$ at $8.5$ GeV, V2 clusterizer Right: Number of cells in the cluster in the same T-Card as the highest energy cell as a function of the number of cells in a different T-Card, for clusters between 100 and 200 GeV.