Liquid Argon Calorimeter

The Liquid Argon (LAr) calorimeter is the most important detector component for the present analysis. It provides an identification and measurement of the scattered electron at high Q² and (together with the tracker and SpaCal) the measurement of the hadronic final state. The LAr calorimeter covers the polar angle range of 4^◦ . θ . 153^◦ and is housed in a single cryostat. The superconducting solenoid is located outside of the cryostat to minimise the amount of inactive material in front of the calorimeter.

The LAr technique offers the advantages of good stability, ease of electronic cali-bration, good homogeneity of the response and fine granularity. These properties allow for the identification of electrons and the precise measurement of their energies and positions as well as the accurate measurement of the hadronic energy flow. The structure of the LAr calorimeter and measurement procedure of the electromagnetic and hadronic energy are described below. More details can be found in [65].

Structure of the LAr Calorimeter

The LAr calorimeter is a sampling calorimeter which consists of an inner, fine gran-ulated electromagnetic section followed by a hadronic part with coarser granulation.

Figure 3.11 shows a vertical cut along the beam axis of the LAr calorimeter. It is divided along the z-direction into eight self-supporting wheels, named according to their position w.r.t. nominal interaction point: Backward Barrel (BBE), Central Barrel (CB1, CB2, CB3), Forward Barrel(FB1, FB2), Outer Forward (OF) and In-ner Forward(IF). The BBE consists of an electromagnetic section only, the OF only of two hadronic sections. In ϕ-direction, each wheel is segmented into from eight identical units, the so-called octants. Figure 3.12 shows a transverse cross section of a central barrel wheel with the typical octagon structure. In the BBE for a better approximation of the circle, the structure of the octants has 16-fold polygonal sur-face of the calorimeter front. The insensitive areas between the modules are called

“z-cracks” (between wheels) and “ϕ-cracks” (between octants).

The LAr calorimeter is built up of absorber plates, the space between the plates is filled with liquid argon as active medium supplemented by high voltage and read-out electrodes. To obtain a uniform energy resolution, the orientation of the plates is ar-ranged such that the angle of incidence of particles originating form theepinteraction point is always larger than 45^◦. The electromagnetic section consists of 2.4 mm lead absorber plates. The LAr active gap thickness is 2.35 mm on average. The absorp-tion length of the electromagnetic part varies between 22 and 30 radiaabsorp-tion lengths (X0) in the central and forward directions, respectively. The absorber material in the hadronic section consists of 19 mm thick stainless steel plates with an active double gap of 2.4 mm filled with liquid argon. The total amount of absorbing material of the calorimeter corresponds to about 5 to 8 hadronic interaction lengths (λ).

The LAr calorimeter is segmented into about 45000 read-out cells to enable a good spatial resolution of the deposited energies. The segmentation is coarse in the

back-Figure 3.11: Longitudinal cross section of the LAr calorimeter. The upper half shows the sampling structure with the orientation of absorber plates. The read-out cell structure is indicated in the lower part. “WWP” denotes the nominal interaction point.

Figure 3.12: Transverse cross section of a central barrel (CB2) wheel of the LAr calorimeter, viewed along the proton beam direction.

ward part and becomes finer towards the forward direction (cf. bottom part of figure 3.11). As viewed from the ep interaction point, the number of layers increases from three to six in the electromagnetic section, and from four to six in the hadronic section. In terms of the Moliere radius⁶,RM, which is a measure of the transverse ex-tension of electromagnetic showers, the typical size of the cells varies between 2.5RM

in the backward region and 1.0 RM in the forward part. The fine granularity allows for both a precise position measurement of electromagnetically interacting particles and a clean separation of electromagnetic and hadronic showers. The latter provides the basis for an efficient electron identification.

Energy Measurement in the LAr Calorimeter

The energies deposited by incident particles in the electromagnetic and hadronic cells are reconstructed in several steps by the LAr reconstruction software.

Input to LAr calorimeter reconstruction are charges collected with charge sensitive amplifiers from the read-out pads. During data taking, depending on the cell loca-tion only cells with absolute value of the collected charge above 2-3 sigmas of the electronic noise are recorded (so-called “zero suppression”). The calorimeter recon-struction program converts charges to energies in the calorimeter for both hadronic and electromagnetic showers, corrected for the effects of dead material, eliminates electronic noise and forms clusters from groups of cells.

The conversion from charge to energy (electromagnetic scale) involves a charge to energy calibration factor (determined for each stack geometry in calibration runs at CERN [104]), a correction for the charge collection efficiency for operating at 1 500 V (derived from HV curves obtained with cosmic muons) and correction factors for local variations of gap and absorber thickness (measured during stack construction). The calibration of electronics is performed during special pulse runs [65], which are taken once every few weeks.

An important first step of the reconstruction program is noise suppression. The electronic noise is measured for each channel during electronic calibration [65]. It varies between 15 and 30 MeV equivalent energy depending on the calorimeter re-gion. In events recorded with a random trigger, 1100 cells out of a total 45000 cells pass a +2σ noise threshold on average. Adding up this energy of the full calorimeter yields an average value of 48 GeV with a standard deviation of 3 GeV [65, 68]. The basic idea of the noise suppression algorithm is to keep a localised energy deposit several standard deviations above the noise level together with all neighbouring cells.

This rejects single noisy cells as well as noise adding up from small contributions of a large number of cells. If to keep cells with energy above +2σ and below −2σ, the residual noise contribution after noise suppression is 0.1 GeV with aσ = 0.5 GeV.

For Monte Carlo simulations (see chapter 4), noise is included for each cell by using events recorded with random triggers in special runs with no zero suppression. This noise is added on top of the simulated energy deposit and than the full noise

sup-6The Moliere radius,RM, is a characteristic constant of a material describing its electromagnetic properties, and is related to the radiation length byRM =X0Es/Ec, with the radiation lengthX0, a scale energyEs≈21 MeV, and the critical energyEc [67].

pression procedure is applied as for the data.

Neighbouring cells which have not been rejected as noise are associated to clusters, i.e. groups of cells which are likely to contain the shower of the same incident particle.

The clustering algorithm works quite well for compact showers induced by electrons and photons. It was found [69], that 95-97% (depending on the energy of the incident particle) of electromagnetic showers in the LAr are reconstructed as a single cluster, while simultaneously resolving pairs of electrons into two separate clusters down to opening angles of about 2^◦ (IF) - 5^◦ (FB, CB) between the two electromagnetic showers [69]. The clustering algorithm does not work so well for hadronic showers, however. As a result of the broader and more fluctuating shower shape, hadronic showers induced by single hadrons are often reconstructed as several clusters.

The clusters found are then classified as either belonging to electromagnetic or hadronic showers, depending on the compactness of the cluster and on the position at which the shower started [64, 68]. An early shower start in the first electromagnetic layer of the calorimeter indicates that the shower is induced by a photon or electron. Addition-ally, the cell energies are corrected for energy loss in the cracks and the dead-material in front of the calorimeter (between 1-2 X₀, varying with polar angle [65]).

Identified hadronic objects are subjected to an energy weighting algorithm [62, 70], which has been developed to equalise the response of the LAr calorimeter to electro-magnetic and hadronic showers. The fine granularity of the LAr makes it possible to detect the electromagnetic components of hadronic showers, which are induced by neutral pions and indicated by high local energy densities. Note that the energy con-tained in the electromagnetic components of showers need no correction. By applying individual energy correction factors only to those cells in the hadronic shower which are not associated with electromagnetic subshowers, the influence of the reconstructed energy on variations in the number of neutral pions in the hadronic shower is reduced.

Final corrections to the reconstructed energies are applied by a calibration on the particle level. Both the electromagnetic energy scale and the hadronic energy scale are calibrated using neutral current DIS events (see sections 6.6 and 6.7). The energy resolution of the LAr calorimeter has been determined in CERN tests to be

σ_E^em

E = 0.12 pE [GeV], for electrons [65] and

σ^had_E

E = 0.50 pE [GeV], for charged pions [66].