A number of methods are used to calibrate the electromagnetic
calorimeter (ECAL) of ALEPH [5]. The ECAL gain is
directly monitored by looking at an Fe55 source
(which ages [11]). The amplitude
of the gain variation is
2 - 3% over one year, which
after correction is stable to better than 0.3%.
For a range of electron energies, the ratio of ECAL energies
to electron track momenta can be measured from various processes
at the Z-peak:
e+e- e+e-e+e- yields electrons
in the 1-10 GeV range and
Z
(
e
) electrons in the 10-30 GeV
range. Also the
Z
e+e- and Bhabha processes
produce electrons at the beam energy (45.6 GeV).
For high energy LEP runs, one can measure EECAL/Ebeam for Bhabha events. With large data samples, the high energy runs can also be split into smaller runs to estimate the time dependence of ECAL variations.
The typical net ECAL energy calibration uncertainty is
(0.7 - 0.9)%.
The hadronic calorimeter (HCAL) of ALEPH also uses physics
processes for calibration [5]. The idea is to
constrain the peak
of the muon energy distribution in HCAL to its expected
position ( 3.7 GeV for muons crossing the calorimeter) which
is measured in beam test. Then at the start of every data-taking period,
Z
and
Z
q
events
are used to calibrate. The use of hadronic Z decays provides
a much more statistically powerful sample, which can be used
because the ratio between the average energy released by hadronic
Z decays and an isolated muon in an HCAL module is well
known from data. This technique gives a `time 0' uncertainty
of
1%.
For high energy running it is possible to compare data and MC
for
events, yielding muons
in the energy range 2.5-10 GeV. The energy distributions agree
at the 1.5% level.
The typical net uncertainty is thus 2%.
The effects of the calorimeter uncertainties on the mW measurements are evaluated by changing, in Monte Carlo samples, the calibration of the subdetectors by the net uncertainties described above, and performing mass fits before and after such a change.