------------------------------------------------------------------
            Calorimeter Task Force - Summary Report
                   February 15, 2002
------------------------------------------------------------------

The primary findings of the Calorimeter Task Force are:
 
 - Despite significant progress in understanding of effects adversely
   affecting calorimeter data and reconstructed objects, a number of 
   major issues remains unresolved. We recommend that a broad effort 
   in the areas of calorimeter hardware, data quality monitoring, 
   calibration, reconstruction and simulation software, and data
   analysis tools continues for an extended period.
 - A group dedicated to calorimeter issues should be established
   with sufficient manpower and strong connections to the related
   detector, ID, and physics analysis efforts, to continue progress 
   on the already identified problems and suggested solutions, and 
   to provide a long-term support of calorimeter-related needs.
 - In near term, the conservative settings of the zero suppression
   thresholds should remain in force, until appropriate algorithmic
   improvements are implemented and certified, at which time
   the suppression parameters should be revisited.
 
A brief discussion of the CTF activities and recommendations follows;
an expanded documentation of CTF studies will be provided separately.
 
    The Calorimeter Task Force was established in August 2002 to
investigate a wide range of issues in the performance of the
calorimeter hardware and software, in response to a significant
deterioration of the properties of calorimeter-based reconstructed
objects, particularly jets and missing ET, observed in D0 data after
the online zero-suppression threshold for calorimeter data acquisition
was lowered (to 1.5\sigma) in July 2002. Most strikingly, as a
consequence, the average number of jets reconstructed per event more
than tripled, and the average missing ET almost doubled.

    While these most obvious symptoms could be diminished by
raising the cell-level zero-suppression threshold (back to the initial
value of 2.5\sigma), it became apparent that more subtle, but
significant, effects, induced by a combination of high values of the
electronic noise and of the vulnerability of the reconstruction
software, remained, and adversely affected the physics quality of the
data. To preserve the possibility of making use of the low-energy
cells for improving the energy resolutions and shape discrimination
for calorimeter-based physics objects (known to be significantly worse
in Run II reconstruction so far than achieved in Run I), the approach
was taken to maintain the low online suppression, and introduce an
offline suppression at a higher threshold at the unpacking stage of
the reconstruction and in Level-3 trigger software.

     Aiming at the optimization of the zero-suppression procedure to
improve the properties of EM and jet reconstruction, resolution of
missing ET, and detection efficiencies of the calorimetric signatures
of taus and muons, the CTF embarked on a detailed review of the
performance of the calorimeter hardware, online calibration
procedures, offline reconstruction of calorimeter-based objects, and
of the simulation software. The deficiencies in the quality of
calorimeter objects reconstructed in the low-suppression data have
been traced, in large part, to the high level of noise present in the
coarse-hadronic and endcap massless gap layers (with noise values
exceeding 300 MeV for many cells in these layers). It has been
verified that, for individual cells, the noise exhibits a Gaussian
behavior, however with additional offset components that result in
"warm zones" of the detector, exacerbating the impact of noise
fluctuations on the jet and MET reconstruction. Pedestal calibration
and the impact of the SCA readout components, have been scrutinized
and understood in more detail than before, leading to proposed
improvements in the calibration procedures and in the front-end code.
These efforts should continue under the auspices of the Calorimeter
hardware and software groups, but will require significant additional
personpower in both groups to complete the necessary studies and
tasks.

    A particularly painful impact on the jet reconstruction
results from the sensitivity of the current (cone) algorithms to jet
"seeds" created by the high-noise cells, leading to significant
contamination by "fake" reconstructed jet objects. To alleviate this
problem, the CTF has begun studies of modifications to this aspect of
jet reconstruction; the current indications of the resulting
performance seem very promising. Another ongoing algorithmic
development attempts to retain the relatively low-energy cells,
provided that these cells are adjacent to cells of sufficiently high
significance (the exact thresholds to be determined). Schemes in
which the (offline) zero-suppression threshold is varied for different
calorimeter layers are also under investigation. These approaches are
expected to lead to a coherent treatment of calorimeter cells for
object reconstruction and to help improve the energy response and
resolutions of the reconstructed physics objects while minimizing the
undesirable effects from the detector noise.

    Not surprisingly, comparisons between, and optimizations of
such approaches could greatly benefit if detailed and realistic
simulation of the calorimeter performance were achieved. The
verification of the existing simulation was undertaken by the CTF as
one of the first projects, and became a continuing major effort
throughout the past 6 months, which is still ongoing. Initially, as
the simulation of the calorimeter noise and zero-suppression aspects
was found to be clearly not adequate for the CTF purposes, the
respective code packages were redesigned and rewritten, and the input
data for both the noise simulation and the zero-suppression were
carefully re-derived from special calibration studies. In the process
of scrutinizing the calorimeter code, a number of other aspects have
been corrected or improved as well, including the implementation of
additional energy weighting factors to reflect the actual status of
the electronics, aspects of the non-linearity corrections and
simulation, and of the offline zero-suppression.

    This broad effort resulted in a much improved understanding
of the low-level aspects and a more correct reconstruction of data,
and a significantly improved performance of the calorimeter simulation,
as compared to data. It provides a solid basis, and a performance
reference point, for further development of the already initiated new
reconstruction algorithms, and additional improvements to the quality
of the simulations, which are expected to be taken over and continued
by the appropriate ID and software groups after the termination of the
CTF effort. The interaction and feedback between the various detector, 
ID, and physics groups has been key to this understanding and
a mechanism for continuing such communication in the future should
be put in place. 

    The current zero suppression scheme has minimal impact on data
taking efficiency and data storage needs while maximizing the
effectiveness of the collected data for future studies and
re-reconstruction. Therefore, the Calorimeter Task Force is
recommending that these conservative settings for the zero-suppression
thresholds be preserved, namely a loose suppression at the online
electronic readout with the tighter offline suppression at the Level-3
trigger and in data reconstruction, while the algorithmic developments
and studies are in progress, so that the detailed future solutions
based on this work could be retroactively applied to the data
collected until the time of the implementation of the final choices.

    The Task Force members believe that perhaps the major outcome
of the CTF effort has been to underscore and bring into focus the wide
range of effort still needed to take full advantage of the calorimeter
detector sybsystem, at the level necessary to achieve the ambitious
physics goals of the experiment in Run II. To this end, the
calorimeter software group should be significantly strengthened over
the next several months, with manpower sufficient to complete the work
initiated by the CTF, and to provide a long-term dedicated detector
support and software development team. Ideally, this group would also
be in charge of further improving the calorimeter-related simulations
and of developing and maintaining data-quality monitoring tools. In
the immediate future, support and manpower for the Level-3 trigger
studies are urgently needed, and rapid feedback from offline analyses
of the calorimeter data to online and harware experts is essential for
maximizing the quality of the vast amount of data D0 is expected to
accumulate. We hope that providing the resources necessary to achieve
the goals outined above will be given full and careful consideration
and support by the D0 collaboration.