Proposed Revisions to MiniMax paper

Cyrus Taylor (cct@po.cwru.edu)
Fri, 27 Aug 1999 15:01:59 -0400

Fellow MiniMaxers-

I propose the following reply to PRD regading out paper.

(My apologies for taking so long. I was on vacation, then had to make
an additional unexpected trip...)

Please let me know
by next Wednesday (1 Sept) if you have any changes you would like to make
to my proposed modifications (which are in turn based on those Tom proposed
a few weeks ago).

After receiving your suggestions, I will make the final changes to the text,
which I hope to submit next week. Since the referee indicated

>The paper presents new results and it therefore should be published, in
>PRD.

these should be the final revisions.

I hope all is well,

Cyrus
***************
Dear Dr. Ripka,

This is in reply to the referees report on df7229 Brooks, T.,
Search for disoriented chiral condensate at the Fermilab Tevatron.

We would very much like to thank the referee for his/her comments. Members
of the collaboration uniformly noted that the referee was
someone sympathetic, conscientious, and very knowledgeable.

I'd now like to turn to the changes we have made in accordance with the
referees comments and suggestions.

>1. The authors begin by introducing the expected P(f) for a DCC state. It
>would be good if they could present the distribution they have found, or
>at least one that is consistent with their data. Can they not unfold this
>distribution? Is the neutral pion fraction indeed 1/3?

In order to completely unfold the distribution, one needs detailed knowledge
of the efficiencies, including the efficiencies for seeing two photons from
the decay of a single pi-zero. The point of the robust observables is that
they are sensitive to the presence or absence of DCC, without being sensitive
to such unknowns.

We conclude that the data are consistent with a binomial distribution. This
conclusion does not depend on detailed knowledge of the efficiencies.
Any P(f) of the form $\delta(f-f_0)$ , where
$0<f_0<1$, would be consistent with this conclusion.
However, as the referee notes in point 8, our Figure 8 indicates that the overall
ratio of neutral pions to total pions is roughly ("at the 10% level") that of Pythia,
that is, about 1/3. This conclusion, however, depends in detail on the Monte-Carlo
simulations so is a weaker than our overall conclusions.

We have addressed the referee's question by adding the following paragraph after the
sixth paragraph of section VI. CONCLUSIONS:

"These limits are consistent with generic, binomial-distribution partition of pions
into charged and neutral species. While the robust observables, being independent
of detection efficiencies, do not permit the determination of the neutral fraction,
we note that Figure 8 indicates that the normalization of
the observed inclusive measurements of
gammas and charged particles agree with Pythia/GEANT simulations at roughly the
10\% level."

>2. The authors should clarify a seeming discrepancy regarding event
>rates. They state (on page 6) that their trigger cross-section was 43 mb.
>The lowest luminosity they appear to have run at for the 6 day run seems
>to be of order 10**28 /cm**2/sec. This would imply a trigger rate of 430
>Hz so that they could collect all their 1.3M events is less than one hour.
>Even given modest dead-times, I don't see how they could have taken 6 days
>worth of data. The reported trigger rates don't seem to make sense either.

As noted in the paragraph before the one the author cites, the calculation of
the luminosity at C0 used the D0 luminosity (available real-time over the accelerator
network) as an input. The numbers quoted were these D0 luminosities. In order
to avoid confusion, we have replaced the sentence puzzling the referree with:

"The luminosity at the C0 collision point
was inferred from the D0 luminosity corrected for
differences in the magnetic architecture at the two points and the fact
that bunches that collide at C0 are not the same pairs that collide at
D0. The C0 luminosity
during these runs ranged from about 10^26 cm^-2 s^-1 to about 10^28 cm^-2
s^-1".

>3. A concern remains with this reader about how valid the results are,
>given that they have trouble simulating their backgrounds. Have they made
>any attempt at superposing additional hits on their simulated events and
>then testing whether their novel statistics change significantly? As I
>understand it, after cuts their simulation reproduces track and photon
>distributions [see comment 8 below] but it fails to reproduce lots of
>extra hits away from tracks. True? (If so, this is not uncommon or
>unexpected.)

This is an important issue. It is true that this is a common problem; we include
a reference to work by the ALICE collaboration reporting this problem with
chambers of similar geometry and gas mixture to those used in MiniMax. The
task of how to address the problem is, however, much more complicated. In
the absence of an understanding of the physical origin of the extra hits, attempts
to introduce extra hits in an ad-hoc fashion can easily give false senses of
(in)security.

Instead, MiniMax chose to address the problem in a data-driven
fashion. By being able to tag with opposite-side multiplicity (Section V.C), we
were able to study the robust observables while significantly changing the
mean multiplicity of the events. We observed no effect. Together with
the fact that Figure 8 indicates good overall agreement of observed particle
production with the simulations, we believe
(with the referee - see his point 8) that this goes far towards
valdating our results.

To make this point more clearly, we have added the following paragraph to the
conclusions, following the discussion of limits on DCC production in various
scenarios:

"Similar analyses (and conclusions)
are possible for the data subsets defined by the diffractive
and forward antinucleon tags, and for events with opposite side multiplicity tags.
Indeed, there is no evidence of a multiplicity dependence in the robust observables.
(See figure 9). We believe that this, together with the overall agreement between
data and experiment suggested by Figure 8, goes far towards validating our results."

>4. I don't understand the statement on page 14 that the deviations from
>unity for the higher order ratios in Table VII are not very significant.
>The reported values are FAR more significantly different from 1.0, using
>the quoted errors, than are those for the low order ratios. Perhaps when
>normalized to their monte-carlo, this effect goes away but the authors
>give no indication of that.

Perhaps by the phrase "The higher order ratios", the referee thought we
meant r(0,2), r(1,2), etc., which are, indeed, far more significantly
different from 1.0. These, however, involve more than one gamma being
observed, and are not robust, depending on a combination of the efficiencies
for observing both one and two gammas from the decay of a single pi-zero, as
well as the overall mean number of gammas. Since this point confused the
referee, we have added the following after the fourth sentence of V.A:

"While the $r_{i,1}$ are robust in the sense defined in section IV, this is
not true of the $r_{i,j}, j>1$ (see [6]). We have nevertheless tabulated these
results for completeness, though they are not useful for the present analysis."

We have also added the following sentence to the caption of Table VII:

"Only the $r_{i,1}$ are robust; the other quantities are tabulated for completeness."

>5. The authors claim that the low-order ratios for diffractive and
>forward tagged events are consistent with those from the total sample. Yet
>in every case, the numbers are greater in these samples than in the total
>sample; and this tendency continues for the higher order ratios as well.
>And by the way, does the simulation give the correct fractions of events
>for these tagged samples?

This is true, but not significant because the numbers in a given column of table
VII are not statistically independent. We have added the following sentence to
the caption of the table:

"Note also that the entries in a given column are not statistically independent."

>And by the way, does the simulation give the correct fractions of events
>for these tagged samples?

This simulations did not include the portions of the detector far downstream.
The numbers are consistent with hand-estimations. We have added a citation
to reference [2] to the sentence in II.B beginning "Detailed GEANT simulations
of the detector and its environment [2]...", since these are described in
detail in reference [2].

>6. At the bottom of table VII are found the values for the low order
>ratios using the alternative tracker. The event total just above the
>bottom three entries seems to be those found with the alternative tracker.
>If so, this should be stated somewhere; and then it needs to be explained
>why this tracker seems to be only 18% as efficient as the nominal one.
>This is doubly puzzling in that the other subsamples that are called out
>(diffractive, forward) have comparable event totals for the two trackers.

We have added a sentence to the caption of the table:
"# events refers to the number of raw events put through the respective
trackers."

>7. The technique of using ratios of factorial moments seems to allow
>extraction of the physics without the need to worry about things such as
>tracking efficiency, etc. But what information is lost? I.e., if one did
>fully understand efficiencies, could more be learned?

As we noted in our reply to point 1, the robust observables are sensitive
to DCC, but are not sufficient to unfold the parent multiplicity distribution
from the observed multiplicity distribution. This is the information which
is lost. To make this point more clearly, we have added the following
at the end of the final paragraph of IV.B:

"We thus use the robust observables as the basis for our analysis in the
remainder of this paper. It is important to note, however, that some
information is lost in this procedure. While we will be sensitive to the
presence of DCC, we will make no attempt to unfold the parent distribution
of charged and neutral pions, since this would require detailed knowledge
of the detection efficiencies for charged tracks and $\gamma$'s."

>8. I don't understand what conclusions are to be drawn from Figure 8.
>Is it that the simulation gets the normalization and distribution of
>charged and neutral particles correct at the 10% level? This would follow
>if the simulations include detector effects which they no doubt do. If
>so, the authors should state this as it gives important validation of
>their simulation procedure and of their understanding of their analysis
>and cuts. Also, the large shift between PYTHIA and GEANT for the photon
>distribution, no doubt having much to do with conversion probability,
>should be explained in the caption (or in the text).

We have re-written the caption of this figure to read:

"FIG. 8 Raw
distributions of the pseudorapidity
distributions of charged and neutral particles. The curve labelled
"PYTHIA" refers to simulated events produced by the PYTHIA event
generator. These events were then propagated through the GEANT
detector simulation and reconstruction algorithms. These results
are labelled "GEANT". The large shift between "PYTHIA" and "GEANT"
for the photon distribution is largely due to the conversion probability.
The observed data, uncorrected for
detection and trigger efficiencies are plotted as "DATA".
The close agreement between the "DATA" and
"GEANT" curves validates the simulation procedure and our understanding
of the analysis and cuts."

9. Finally, there is no clue as to why the experiment is called
"MiniMax."

MiniMax was a successor to an earlier unsuccessful proposal (Fermilab P-864)
for a "Maximum Acceptance Detector", MAX. Since MiniMax was much more
modest, the name was natural. This history has been described in many of
our earlier publications in reference [1], as well as in M. Convery's Ph. D. thesis,
reference [2]. We have thus not made any additions to the present paper to again repeat
this history.

In closing, we note that the referee commented:

>The paper presents new results and it therefore should be published, in
>PRD.

We hope that it can now be published expeditiously.

Thank you very much for your help,

Sincerely,

Cyrus Taylor
for the MiniMax collaboration

-----------------------------------
Cyrus Taylor (216) 368-3710
Armington Professor (216) 368-4671 (FAX)
Department of Physics cct@po.cwru.edu
Case Western Reserve University ctaylor@fnal.fnal.gov
Cleveland, OH 44106-7079
USA