Calorimeter response and energy calibration

In a local linear model, write measured signal S as pedestal plus gain times deposited or incident reference energy, with additional fluctuations. Subtract the pedestal and divide by gain to reconstruct energy. The gain is a calibration coefficient in signal units per GeV, not a universal material constant. Channel response, temperature, radiation damage, electronics, and reconstruction choices can all affect it. Nonlinearity or saturation requires a more complete response function than one constant slope.

An electromagnetic shower is largely governed by bremsstrahlung and pair production, while hadronic showers also involve nuclear breakup, invisible energy, and fluctuating electromagnetic fractions. A calibration derived for electrons does not automatically give an unbiased energy for every hadron or jet. Sampling calorimeters observe only part of the shower directly, and leakage or dead material can create asymmetric losses. Particle-flow methods combine detector information, adding correlations that a single isolated-calorimeter formula does not capture.

A commonly fitted relative-resolution form combines a stochastic term, a noise term, and a constant term in quadrature. Let e denote the numerical energy in GeV so all coefficients in that expression are dimensionless. Independent contributions with variance proportional to energy give inverse-square-root relative scaling. Approximately fixed additive electronic noise gives inverse-energy relative scaling. Spatial nonuniformity and other effects can approach a constant relative contribution at high energy.

Quadrature addition requires that the modeled contributions be sufficiently independent in the specified measurement. It is a useful parameterization over a stated range, not a theorem that every detector response is Gaussian or exactly follows three powers. Tails and thresholds can change selection probabilities even when the central resolution looks satisfactory. A quoted standard deviation must also specify whether it describes the full distribution, a fitted core, or a restricted effective interval. Those definitions need not agree for non-Gaussian showers.

Test beams, electronic injections, standard resonances, and momentum-to-energy comparisons constrain different pieces of a calibration chain. They cover particular particle types, positions, and energies, so transferring a calibration requires checks in the analysis region. Scale and resolution can be fitted simultaneously, but their nuisance parameters should remain conceptually distinct. Smearing a simulated response broadens its distribution; rescaling changes its central value. One operation cannot generally substitute for the other.

A common scale uncertainty moves many energies coherently. Averaging independent shower fluctuations reduces their random contribution, but a shared gain error survives that average and correlates mass bins and channels. Near a steeply falling spectrum or a selection boundary, either scale shifts or resolution changes can move events between bins. Propagating the modified response through selection and the likelihood is more faithful than attaching the same percentage independently to every final histogram entry.