MEMORANDUM



Executive Summary

HARP collaborators have intensively worked during more than  two years
to overcome challenges related to timescale, hardware,  data-taking,
software, detector performance, and now (finally)  physics
analysis. The preparation of the iDST is advancing with the  help of
all sub-detectors and will allow to distribute the analysis  effort
between the institutes.

Thus we can now expect relevant physics publications in a  reasonable
time scale --certainly before the end of the year--.  Those
forthcoming analyses will build on robust  software and analysis tools
(e.g, the iDST) and will proceed  through the obvious and necessary
steps: complete detector  calibration, correction of hardware related
issues (e.g, TPC  cross talk), Monte Carlo, validation of
reconstruction algorithms  and assessment of performance.

(In the HARP life-cycle there have been various iterations of the
software project, a large-angle analysis project, and three phases  of
the analysis project foreseen since its kick-off: for simplicity  they
will all be referred in the following under the term  "HARP
Software&Analysis").



Results and Strategy

The HARP software&analysis started in August 2000 and allowed a
successful data acquisition already for the Technical Run of end
September 2000. With the constant feedback and contribution of
detector experts, software&analysis have progressed  regularly as
documented in SPSC reports, allowing in  different cases: 
- to anticipate hardware-related problems  
  (such as beam intensity effects on NDCs),  
- to understand quantitatively efficiency problems  
  (such as the FTP paradox with the Monte-Carlo), 
- to reveal the existence of critical problems  
  (such as the X-talk in the TPC in the LA analysis project), 
- to produce the first preliminary physics results in Spring 2002   
  (successfully defending HARP in front of the CERN Research Board
  referees in view of the second year of data taking). 

The software&analysis have then turned toward a physics  publication
oriented strategy since Fall 2002, with a renewed  structure outlined
in the memorandum approved by the  Collaboration Board in July 2002.
Three phases have been defined at the kick-off meeting, including: 
- calibration of all sub-detectors 
- detector performance assessment  
- physics analysis.

The first phase is well under way. Calibration ntuples are being
produced and are being analyzed by detector experts.  Detector
calibration --an essential precondition for any serious physics
analysis-- will, therefore, be completed within the next very 
few months.

The next essential precondition for a sound physics analysis is a  set
of tools, such as the iDST and Harp Analysis Package. These  will
allow distributed analysis in all HARP institutes, and will permit  a
very fast turn-around of results. We note that every high energy
physics experiment has based the analysis on tools such as DSTs  and
programs running on them.  The production of analysis tools is also
well advanced and starts now  to involve a relevant fraction of
institutes, thus, accelerating the process.  As an example, since the
proposal of the iDST, less than two months  ago, the iDST concept has
been prototyped, its contents defined, and  we are now entering the
implementation phase, with the participation of CERN, Sofia, Geneve,
Oxford, Ral, Roma and the MiniBoone group.  In addition to its
relevance for the analysis, the current work is proving to be very
useful to improve the distributed-responsibility philosophy  on which
we must base our (present and future) efforts.
 
In order to produce reliable papers, it is also mandatory to
understand,  correct (and quantitatively assess the residual effects
after the correction)  our hardware-related problems, most relevant of
which is the TPC cross  talk. This in turn, requires a detailed Monte
Carlo modeling of those  effects (which also serves to quantify
the detector response, e.g, what  exactly is the uncertainty in the
position measurement introduced by the  cross talk w.r.t. to other
effects).  Obviously, these Monte Carlo studies, need to be
complemented and  expanded with the study of selected data sets (e.g.,
in the case of the  TPC cross-talk, Krypton data, cosmic-ray data,
etc.). 

The next unavoidable step is the correct assessment of detector
performance (for example reconstruction efficiency as a function of
angle, multiplicity  and momentum) and the selection of the best
available algorithms (e.g. for pattern recognition in the forward
region).

The program outlined above requires a minimum time which can be
optimized with sufficient manpower and proper management. With  the
current resources we estimate that sound physics results can be
achieved in a time scale of nine months. However, with sufficient
involvement from all HARP institutes, this time can likely be reduced
to about six months, provided that an effective management process  is
established. This is consistent with the expectations of our clients
(e.g. MiniBoone and K2K collaborations, which need a 5% precision
result in the timescale of one year).  It appears certainly possible
to produce a result for the first SPSC after the summer, if required.

Of course, another style is possible, and can lead to results.  If one
would put as requirement that the first result would be available as
fast as possible, modifications to this strategy are needed.  However,
rushing to a fast publication, will prevent us from developing
analysis tools that make a broad participation possible, and will
discontinue a delicate and sound program to obtain robust and
reproducible results. In exchange we will only achieve a mediocre
result, the motivation and usefulness of which is highly contestable.

Note that the core team of the analysis and software projects could
take on such a fast publication program, given their expertise on the
existing reconstruction and analysis tools. However, the price to pay
would be to a) interrupt or seriously disrupt the calibration process,
which would then be very hard to re-start at a later date, b) delay
the availability of the tools permitting broad participation in the
analysis. In our opinion, this is a too heavy toll to pay, in order to
anticipate publication just by a couple of months.

Along these lines we believe it is essential to set up a mechanism
which  assesses the level of reliability of the papers to be submitted
for  publication such as for instance an editorial board whose
composition  and mission should be discussed by the Collaboration
Board.

Finally, we believe HARP can produce excellent quality physics
publications without falling in the trap of a "panic-management" style.




Addendum (on publication time scale) 
---------- 
It has been pointed out that the software and analysis effort is too
slow.  In order to shed some light onto this statement, it is
illustrative to consider the cases of two recent experiments, whose
characteristics remind to some extent those of HARP.

NOMAD 
----- 
This was an experiment searching for (numu(nue) -->nutau)
neutrino oscillations. The experiment was approved in '91, originally
for a two year period, that was later extended by two more years. Data
taking started in '94 and extended up to '98. We note the following
facts: a) There was a period of three years between approval and start
of data taking that allowed, among other things, for extensive
software development.  b) During the "active" (data taking) life of
the experiment there was time to correct for hardware problems (e.g,
bubbles in the drift chambers) and  compensate with more data.  c) The
long running period, also allowed for extensive detector calibration
and further software and analysis refinements.  However, the first
physics papers, were only published in 1998. Notice, though, that
since that date there has been more than twenty NOMAD papers.  One
lesson to draw is that investing some extra time, in preparing a good
analysis, results in numerous high-quality publications. Another
interesting conclusion would be that a Harp publication by the end of
the year would be a very fast result on that scale.

CHORUS 
------ 
This experiment was also designed to search for neutrino
oscillations.  The experiment was approved in '91, and took first data
in '94.  Data taking (with emulsions) ended in '97.  First physics
results were published in '97, while the most prolific year in terms
of physics publications up to now was 2002.  The expectation is to
double the number of publications in the coming 15 months.


In summary, for NOMAD the first physics paper was published 7 years
after the approval and 4 years after the start of data taking.
Similarly, for CHORUS the first physics paper was published 6 years
after the approval and 3 years after the start of data taking.  If
HARP will publish the first physics paper by the end of this year,
this would be just 3 and a half years from approval and 2 years from
the start of data taking.


Signatures:
---------------
---------------
---------------
---------------