QCD MEETING
March 25, 1999 at 3 PM
Black Hole
NEWS/ANNOUNCEMENTS
1. General news and announcements/Run II news
UPDATES
2. Min Bias Jets: The Evolution of Jets from 0.5 to 50 GeV
Rick Field 30 min
3. Update on differential dijet comparisons to theory
Frank Chlebana 20 min
4. Update on inclusive jet cross section comparisons to theory
Brenna Flaugher/Anwar Bhatti 20 min
5. Chi**2 and probability issues for the dijet invariant mass cross section
Eve Kovacs 20 min
1. News/announcements
1.0 QCD meeting time/day to be changed
There is a plan that will be put into effect in the
near future to concentrate on upgrade-related activities
on Thursdays and to have the physics meetings on Fridays.
Tentatively, the QCD meeting would start at 1 PM on Fridays.
Please send Anwar and Joey any comments that you may have.
Also, the series of talks on new accelerators seems to conflict
with the QCD meetings. We may try to rearrange the QCD meeting
schedule when that happens to allow Brenna (and others) to
attend the accelerator talks.
1.1 EPS abstracts due on May 3, 1999; note the recent change of date
Possible abstracts include:
-inclusive jet cross sections at 1800 and 630 GeV
-inclusive photon cross sections at 1800 and 630 GeV
-differential dijet cross sections
-dijet mass cross sections
-photon + muon cross section
-NLO 3 jet comparisons
-W + jets (including R10, differential jet results)
-MLLA comparisons to jet fragmentation
-diffractive dijet production at 1800 and 630 GeV
-diffractive heavy flavor and J/psi production
-your topic here...
1.2 Paper of the week
hep-ph/9903436, Sudakov resummation effects in prompt photon hadroproduction,
Michelangelo Mangano and company; companion paper to their previous effort
in which they do some phenomenological comparisons to fixed target
data from E706 and UA6. They consider threshold resummation effects
which reduce the scale dependence over all xT and cause an enhancement
to the cross section at high xT.
1.3 Run II news
As onset of blessings is mostly over and physics analyses are winding down,
we will be trying over the course of the rest of the year to step up
reviews of Run II triggers/software during the QCD meetings.
UPDATES
2. Min Bias Jets: The Evolution of Jets from 0.5 to 50 GeV Rick Field
Rick presented a plethora of plots from work that he is doing with
Dave Stuart plus the Bologna group (N. Moggi, F. Rimondi and S. Zucchelli).
The transparencies are available at
http://www.phys.ufl.edu/~rfield/cdf/QCD_Talk2.html.
The plan is to study the CDF minimum bias data with the goal of
finding a Monte Carlo event generator that will fit the data.
Rick would like to describe (approximately) all the features of the entire
inelastic ("hard-core") cross section at both low pT and high pT.
Many observables in the data are compared to Herwig, Isajet and Pythia.
For now only charged particles with pT>0.5 GeV/c with abs(eta) >1 are
considered.
The total ppbar cross section can be written as the sum of the elastic
cross section and the inelastic cross section. Furthermore, the inelastic cross
section can be written as the sum of the single diffractive and the double
diffractive cross section, plus the "hard core" cross section. The sigma_HC
is equal to the total inelastic cross section -single diffractive cross section
-double diffractive cross section=60 mb - 9 mb 1 mb = 50 mb. We want to describe
the entire "hard core" part of the inelastic cross section, sigma_HC, which has
a QCD "hard scattering" component that becomes infinite as pT_hard becomes
small, and which may also have a "soft collision" component that is not
calculable from perturbation theory.
There are two possible approaches:
1. two component model: sigma_inelastic=sigma("hard perturbative",pT>pt_min)
+ sigma("soft", everything else)
2. one component model: sigma_inelastic=sigma("hard perturbative", all pT
but remove the divergences)
Both Herwig and Isajet provide a soft collision generator for
simulating minimum bias events. They produce roughly 4 charged particles
per unit eta and have no correlations except resonances.
Herwig QCD 2->2 parton-parton hard scattering with pT_hard>3 Gev/c has a
cross section of 19.3 mb
ISAJET QCD 2->2 parton-parton hard scattering with pT_hard>3 GeV has a cross
section of 31.1 mb
ISAJET QCD 2->2 parton-parton hard scattering with pT_hard>3 GeV has a cross
section of 55.1 mb
The reasons for the differences in cross section predictions are not
known but may be related to scale choice, pdf's etc. The above numbers are to be
compared to the hard-core cross section of 50 mb.
Rick defines pT_max to be the highest pT charged particle in the event
satisfying the conditions given above and then looks for correlations in phi.
A plot of Ncharg vs phi shows that the level in the data is higher than the
level in either ("soft MB") Herwig or ("soft MB") Isajet. In addition, there
is structure in the data that is not present in the Monte Carlos.
For his comparisons, Rick uses a simple jet algorithm in which one
examines "circular regions" in eta-phi space of radius 0.7. In this jet
algorithm, one orders charged particles in pT(pT>0.5 GeV/c and abs(eta)<1),
starts with the highest pT particles and includes in the jet all particles
within a radius R of 0.7, then goes onto the next highest pT particle not
already in the previous jet and makes a new jet, etc.
There can be a considerable overlap between two jets. The maximum
possible number of jets in the central eta region is on the order of 16.
The pT of the jet is the sum of the pT values of all of the charged particles
in the jet cone. Rick showed distributions of the jet pseudo-cross section
plotted as a function of the jet ET, using both the min bias and JET20
data sets. He normalizes the JET20 and min bias data sets where they overlap,
around 25 GeV/c. Herwig seems to reproduce the combined min bias-JET20 shape
to the order of 1 GeV/c.
Rick also plotted the average number of charged particles vs the pT
of the jet. There is an initial steep rise with a point of inflection around
10 GeV/c followed by a continuing rise with a smaller slope. One can get out
to jet ET values of 100 GeV/c by also including JET50 and JET70. Rick compared
these distributions to the 3 Monte Carlos. Herwig may do the best job of
describing the data. There are many other distributions that can be compared to,
for example the radius for the jet containing 80% of the jet particles, of the
jet pT, the number of charged particles witin the jet, the radial pT flow
in the jet, and the fragmentation function in the jet. All give reasonable
agreement with the Monte Carlo predictions.
3. Update on differential dijet cross section comparisons to theory
Frank Chlebana
Frank presented some results from comparing the differential dijet cross
section results to theory predictions, fitting the needed systematic error
deviations to the residuals between data and theory. For example, the CTEQ4HJ
pdf needs smaller systematic shifts than does the MRST3 pdf, as expected. For
the latter case, the high pT response has to shift by 1.7 sigma, the resolution
by 2.47 sigma, and the normalization by 1.95 sigma. The chisquare can be
characterized by a sum of statistical and systematic terms. Some results are
given below (note that there are 51 points and 8 systematic errors):
PDF chisquare(stat) chisquare(sys)
CTEQ4HJ 135.85 0.64
CTEQ4M 318.81 4.81
MRST1 254.82 5.21
MRST2 379.72 3.27
MRST3 252.99 13.15
Again, one can note that the systematic deviations are smallest for
CTEQ4HJ and largest for MRST3.
Note that even for the case of CTEQ4HJ, the statistical chisquare
isn't so great. When only one eta range is used in the fit, a better
chisquare can be obtained (but at the ultimate expense of the other
chisquare ranges). So far, no systematic effects between eta ranges have
been included.
Frank presented a series of plots showing the residuals before and
after the systematic error shifts for the pdf's mentioned above. The biggest
problem with most pdf's seems to be the two higher eta bins. It was
suggested to Frank that he try fitting the first eta bin, the first two eta
bins, and the first 3 eta bins to see how the chisquare accumulates. Also,
Frank will try removing the data points in the last two eta bins were the
theoretical K-factor (NLO/LO) is large. This is a regime where NLO theory
may not be adequate, and one would not expect agreement between data and
theory. JH will provide Frank with the relevant kinematic range to exclude.
4. Update on inclusive jet cross section comparisons to theory
Brenna Flaugher
Brenna presented only 1 transparency due to a breakdown of her computer,
but what a transparency. At the request of the inclusive jet godparents, she
plotted the residuals (data-standard)/uncertainty on data. One expects a
Gaussian with an rms of 1. The resulting distribution does look very
Gaussian-like but with an rms of 1.2. The two "outlying" points that Brenna
has pointed out before actually do not look out of place in this distribution.
They have looked in depth for reasons to scale the statistical uncertainty
but have found none. The probability for this to happen (rms of 1.2) is on
the order of 5%.
There was a discussion of the best course of action to follow. The
sentiment was in favor of giving relative chisquares (relative to the standard
curve) rather than rescaling the errors so that the standard curve had a
chisquare of 1.
5. Chi**2 and probability issues for the dijet invariant mass cross section
Eve Kovacs
Eve reviewed the two techniques that have been used for calculating
chisquares, the covariance matrix technique and the parametric fit. In the
latter, the chisquare is obtained by fitting the fraction (alpha_k)
of each of the 8 systematic errors that gives the best description of the
residuals of the data with respect to a particular theory. Minuit is
used to find the alpha_k that minimize the chisquare.
In a previous QCD meeting, Eve presented a proof that the two methods
gave equivalent results. She points out however that care must be taken to
insure that the treatment of the systematic errors be the same in each case.
jet analyses have sed a variety of different choices in implementing the
statistical and systematic uncertainties; for some of these choices, it is not
possible to treat the errors identically for the two chisquare methods. In these
Cases, the results do not agree.
Systematic errors are expressed as (f_i^k), fractional errors (%). These
must be converted to absolute errors.
Definition 1: sigma_i^k= f_i^k*standard_fit(i); in this case the systematic
errors follow a smooth curve
Definition 2: sigma_i^k=f_i^k*data(i); in this case, the systematic errors
contain the statistical fluctuations in the data
->this is the source of Peele's Pertinent Puzzle; a
bias is introduced into the fit causing the fit
to be lowered
Definition 3: sigma_i^k=f_i^k*data(i) but the statistical errors are
rescaled; this was proposed by D'Agostini to fix
the problem with definition 2
Definitions 1 and 2 give identical results for both chisquare methods
provided the treatment is consistent. Definition 3 cannot be implemented for
the covariance matrix method. The statistical errors are involved in the
minimization procedure.
Eve argues for the use of definition 1.
There followed a discussion on probability and confidence levels along
with the critique of a technique that Steve Kuhlmann is advocating. There will
be further discussion of this at the next QCD meeting.
Joey Huston - March 30, 1999