QCD MEETING
May 20, 1999 at 3 PM
Black Hole
NEWS/ANNOUNCEMENTS
(Note: the qcd meeting from May 6 had the video cut off for most of the time.)
1. General news and announcements/Run II news
1.1 Run II
-report from photon and weak boson working group and the
jet clustering meeting Joey Huston
1.2 Report on calorimeter reconstruction
issues for Run 2 John Mayer 30 min
UPDATES
2. What have we been doing with the dijet mass cross sections?
John Mayer 20 min
3. NLO 3 jet analysis
Alex Brandl/Sally Seidel 30 min
PREBLESSINGS
4. Diffractive dijets at 1800 GeV Kerstin Borras 30 min
5. DPE dijets Koji Terashi 30 min
6. Systematic error analysis for MLLA studies Alexei Safonov 30 min
1. News/announcements/Run II news
1.01 13 EPS abstracts submitted; see http://www.pa.msu.edu/~huston/eps_abstracts
1.02 Change of QCD meeting schedule
To emphasize the growing concentration of the
collaboration on Run II preparations, there is a proposal to have Thursdays
devoted to these topics with the physics meetings on Fridays.
The QCD meeting would be from 1-4 PM on Friday.
This would take place starting at the beginning of June.
There may also be some restructuring of the physics groups with the QCD group
taking on more responsibility.
1.03 QCD Baby of the Week
hep-ex/8 lbs 0 ounces; Ali Ahmad Anwar Bhatti; authors: Shabnamad and Anwar
Bhatti
1.1 Report from photon and weak boson working group and the jet clustering
meeting Joey Huston
In the context of the Run 2 workshop, several meetings have been held
that may be of interest to the QCD group. In particular, there was a meeting
held on discussing physics possibilities for photons and W/Z's in Run 2 and
another meeting to discuss a common (and improved jet algorithm) that could
be used by both CDF and D0 in Run 2. There will be followup meetings on these
and other Run 2 subjects at the next workshop date (June 3,4).
A core dump of ideas that were presented for photons is reproduced
(redumped below). Jeff Owens is making up many of the plots/calculations
discussed. If you would like to become involved, please contact Uli Baur or
JH.
Some of the plots/studies/etc may already exist, but it would still be
helpful to collect them in one place, i.e. the Run 2
document.
* Kinematic reach for 2 fb-1, 4fb-1; produce a plot showing the reach in ET for
inclusive single photon production with statistical error bars at each point
* Subprocess plot;indicate the fraction of photon production due to gluon-quark
scattering, qqbar annihilation, fragmentation processes (with isolation cut)
as a function of photon ET
* What is really known about the production mechanisms for direct photon
production? What evidence is there for the contributions of fragmentation
processes to isolated direct photon production? (The isolation energy in a cone
around the direct photons seems similar to that observed in MB.) How
important are the NLO fragmentation contributions for the isolated photon
cross sections?
* Calculation of isolated photon cross sections at the Tevatron? Can anything
definitive be said about the "safeness" of the cross sections at the Tevatron?
* What is the "K-factor" (NLO/LO) for a scale of ET/2, to give an idea
of the importance of NLO corrections in the region of measurement?
* Sources of background to single photon production: discuss fragmentation of
jets/what is known about jet fragmentation/what we can learn: a detailed
study will help to normalize the Monte Carlo calculations that are being
used to predict the gg backgrounds at the LHC
* Photon purity: what is the photon purity (true photon/(photon+background)) as
a function of photon ET for given isolation cuts; the imposition of a tight
isolation cut can improve the signal/background ratio by several orders of
magnitude; these calculations can be compared to CDF (and D0?) results for Run 1
to show that the formalism is reasonable. One of the implications is that any
isolated photon candidate at high ET is almost certainly a photon and not a pio
This has implications for QCD photon measurements at high ET as well as for
exotic uses of photons.
Measurement techniques for Run 2; regions of applicability and
limitations; for example, in CDF, we can use the shower profile technique only
until about 40 GeV/c; the conversion technique has no explicit ET limitation
but "splashback" becomes increasingly more probable as the ET of the photon
increases
* Photon + jet(s) cross sections; triple differential cross sections, photon +
jet mass, etc
* Photon + heavy flavor cross sections (g+m, g+D*, etc)
* Constraints on the gluon distribution; can we get out to high enough x to
shed some light on the high x gluon, a la CTEQ4HJ, perhaps using photon + jet
to get to the right kinematic region; how low in x can we go? What sort of
overlap with HERA?
* Comparisons of the inclusive photon cross section to NLO, NLO+parton showers,
NLO+kT predictions and an understanding of the role of soft gluon radiation at
low ET; can looking at the ratio of forward to central photons say anything
about the kinematic dependence of kT? Can direct photon data be described
without kT, by steeper pdf's, by more fragmentation contributions?
* What observables can be helpful in describing kT for photon production?
Pout for photon + jet? Recoil jet z distribution?
What is the proper formulation for a resummation calculation for collider
direct photon cross sections? How does this fit in with resummation
Most of the above items can be duplicated for the case of
diphoton production; this has particular implications for
diphoton production at the LHC and the backgrounds to
low mass Higgs productions.The statistics of Run 2 will
allow diphoton production to be studied in detail.
Overlaps:
-with pdf group: determination of the gluon distribution, especially at
high x; need for resummed cross sections
-with jet group; is kT present at the same level in jet events as in
direct photon and diphoton events?
The jet clustering meeting was less "dumpy" as the attendees (both
theorists and experimentalists) were struggling to arrive at jet algorithms
(using both the cone and kT formalisms) that could be
common (as much as possible) between the two experiments
and could be easily compared to a variety of pQCD calculations, including
(N)NLO and resummed. The primary worry was the sensitivity
of the jet clustering algorithm to seed cells between two nearby jets.
This problem may be fixed by requiring jets (in both the
data and theory) to be a certain distance apart; Steve Ellis also had a
suggestion in the past to fix any possible problem by always using the
cell between two nearby jets as a seed cell
regardless of whether the energy in the cell is above the seed threshold. My
brain hurt after the meeting. The discussion will continue June 3 and 4.
1.2 Report on calorimeter reconstruction issues for Run 2 John Mayer
John gave a report on the jet reconstruction efforts for Run 2. He
started with a review of what currently exists. There are currently two modules:
1. Wrappered fortran-Run 1 cone algorithm (JETCLU)
->copies TOWE, SETA banks onto an array that the old JETCLU routines
can access
->at the end of an event, TOWE, SETA, CLSL, CALL and JETS banks are
copied onto the TRYBOS record.
This module is written primarily to check the validity of newer code written in
C++. It is not intended to be used in production.
2. C++ cone algorithm (JETCLU)
->a rewrite of the Run 1 code into C++
->data access has been re-written to use TOWE_BANK_DRIVER and
ANNULUS_LOCATION classes to determine the ET of the EM and HAD
towers. An array of CAL_TOWERS is formed which contains, for each
tower
-EM, HAD and TOTAL energies
-ET, eta and phi
-sin(theta)
CAL_TOWER is then used by the clustering algorithms to make jets. At
present, there is only one "calculator" (Snowmass). Running on
HEPG has not been fully implemented.
The calorimetry reconstruction software was reviewed in March (but not
JETS). Many structural changes are presently being made and will
be completed in the 1-2 week time frame. The JETS modules
must then be modified to reflect these changes (completed in mid to late June).
The most important change is that "clusterable" objects are now
defined. These may be different than simple reconstructed calorimeter
towers (CAL_TOWERS). For instance, clusters of towers may be considered
clusterable, a feature which may be useful in implementing kT-type algorithms.
Cluster algorithms should work on calorimeter towers, HEPG particles
and tracks. The user will be able to "talk-to" each module within
the AC++ framework to adjust the parameters.
The flexibility and quality of the jet package will only be as good as
dictated by the users. Feedback is imperative. The physics groups must:
->define the algorithms they want
->define the flexibility
->define the parameter sets for the algorithms
->test, test, validate and test
Some algorithm hints so far:
-a backward compatible cone algorithm (exists)
-a kT algorithm based on calorimter towers
-a new and improved cone algorithm, a la the QCD workshop
-a high resolution algorithm using energies corrected with preshower
and shower max information
UPDATES
2. What have we been doing with the dijet mass analysis? John Mayer
John then switched hats and discussed the dijet mass analysis, which he has
been worrying about since Bjoern's graduation. The key question, as in
the inclusive jet and differential dijet analyses, is the interpretation of
the comparison of the data cross section to the theoretical predictions.
Five techniques were summarized: a covariant matrix method and 4 best fit
methods.
Fitting Method #1: This is Eve's prescription and adjusts the data in the
fit by an additive term calculated as the sum of the absolute systematic
uncertainties used in the fit. This method is equivalent to the covariant
matrix method as long as the systematic and statistical uncertainties are
calculated in a consistent manner. For both of the uncertainties, either the
standard curve or the data can be used.
Fitting Method #2: This method adjusts the data by a multiplicative factor
calculated from the product of the minimized fractional systematic
uncertainties. For statistical uncertainties calculated from the data, MRST
is the favored PDF.
Fitting Method #3: This method is described in the talk 2 weeks ago by
Brenna. Theory points are adjusted by a factor calculated from the
minimized fractional systematic uncertainties. In this case the fit favors
CTEQ4HJ.
Fitting Method #4: Fitting methods #1-#3 all assume Gaussian statistics in
each fitted bin. Method #4 is one where the traditional Gaussian chi**2
formula (data-theory)**2/sigma**2 is replaced by an equivalent one based on
Poisson statistics. Here the factor which adjusts the theory is similar to
that of Method #2. The favored PDF is CTEQ4HJ.
A plot was also shown which graphs the total chi**2 for each of the methods
versus the PDF. This plot showed that for the most part the results from
the different methods track each other and are very similar. The only
variation is that Methods #3 and #4 do not favour CTEQ4M (mu=0.25) whereas
Methods #1 and #2 do.
3. NLO 3 jet analysis Alex Brandl
The goal of the analysis is to obtain a measurement of alpha_s by
comparing the energy partitioning in multi-jet events to the predictions
of NLO QCD (the new calculation by Walter and Bill Kilgore).
The data was extracted from the CDF sum ET Stream B data banks. The
analysis considers systems of 3 or more massless jets. The three leading
jets in the laboratory frame are used as the basis of transformation
into the three-jet rest frame. The variables of interest are the mass of the
three-jet system (m3J), the dimensionless Dalitz variables x3 and x4
(xi=2Ei/m3J) and the two angular variables cos(theta3*) and
psi*. (See CDF note 3076 for the previous 3 jet analysis and comparison to
leading order predictions.)
So far, 18000 events have been used from the data to tune the cuts. The
Giele-Kilgore Monte Carlo (hep-ph/9903361) is the event generator for the three
jet production. It's the first program to calculate all parton sub-processes to
NLO in
perturbation theory. This includes the one-loop 2->3 subprocesses as well as the
tree-level 2->4 subprocesses. The Monte Carlo outputs the cross section in two
parts, a hard emission with all partons well-resolved and an infrared part with
one
or more partons in the soft or collinear limit. The two parts are added
algebraically.
For the analyses, 6 template Monte Carlo data sets were generated, each
with
a different value of alpha_s (using the appropriate CTEQ3 pdf corresponding to
that alpha_s). The Monte Carlo data sets were then compared to the data.
To remove beam halo, calorimeter malfunctions and cosmic rays, events
were excluded if:
-significant energy out of time (HATFLT)
-total energy greater than 2000 GeV
-missing transverse energy significance greater than 6
-listed by module BADRUN
A requirement was also made that the primary reconstructed vertex has
abs(z)<60 cm.
Jet corrections were made with JTC96S (underlying event
correction applied, out-of-cone corrections not used), jet ET>20 GeV
abs(eta)<2.0 for all jets, and a requirement was made that the 2 jet
separation be greater than 1.0 (delta-R). Data were excluded from
those regions of phase space (high cos(theta*)) where the acceptance is
less than 95%. The vertex was also redefined as the one containing
the maximum momentum (VXPRIM).
The application of all of the cuts reduces the number of events
from 18000 to 3952.
For comparing the data to Monte Carlo predictions, the data are binned
in x3-x4 space (bin size is 0.02 by 0.02), the binned Monte Carlo cross
sections are multiplied by the luminosity expected for the given data set
to obtain the expected number of events in each bin. The data are compared to
Monte Carlo using the method of maximum likelihood. The data have not
yet been unsmeared for this comparison. The chisquare is then calculated for
the different template samples using a log likelihood technique.
No cuts were made on x3 and x4 but the expectation is that the
theoretical prediction will be unstable at the edge of phase space. This
is borne out by the high chisquares obtained for large values of x3 and
x4. In the future, those regions will be explicitly cut out.
The next steps:
-unsmear the data for direct comparison to predictions
-use all of the sum ET data stored on tape (from an integrated
luminosity of 86 pb**-1)
-investigate cuts in the Dalitz plane
-analyze systematic errors
-optimize kinematic cuts
-investigate backgrounds (cosmic rays, beam halo,...)
-as a check on the result, calculate the three-jet to two-jet
ratio and compare to NLO QCD
PREBLESSINGS
4. Diffractive dijets at 1800 GeV Kerstin Borras
The idea is to measure the diffractive structure relative to the
proton structure (SD/ND) using dijet production, as a function of
x_pbar. The measured quantities are xi=p_pomeron/p_pbar and
beta=p_gluon,quark/p_pomeron=x_pbar*xi.
Kerstin presented the outline for a PRL:
a.) view on the Roman pot data with Roman pot acceptance and the ratios
of SD dijets to SD inclusive as a function of the relevant quantities t
and xi
b.) properties of the diffractive dijets compared to non-diffractive dijets
c.) ratio of SD dijets to ND dijets as a function of x_pbar
d.) unfolding of the diffractive structure from the ratio as a function of
beta and xi without any MC (this is a new result)
Kerstin showed a plot (for preblessing) giving the rate for
dijet/inclusive production as a function of xi (increasing with increasing
xi) and t (flat with t). She also showed a plot showing the ET distribution
for diffractive and non-diffractive dijet production. The slope of the two
is similar. The rapidity of the dijet system is shifted for diffractive
production, as expected, and the delta-phi distribution seems more
peaked back-to-back for the case of diffractive production.
The ratio of diffractive and non-diffractive dijet production was
calculated (for ET_jet>7 GeV/c) and 0.04<xi<0.1 and t<t_min (for the
diffractive production). The advantage of doing this is that all detector
effects cancel out, since jets at the same value of x_pbar have similar
eta values and jet ET and the influence of the minimum required x_pbar for
dijet production cancels out in the ratio. The interpretation of the plot
is that the ratio is the diffractive structure relative to the proton
structure.
A power law is observed in the kinematic safe region of 10^-3<x_pbar<
0.5<xi>. One question that remains is whether the structure is the same at all
xi values. The answer is yes for <xi> values from 0.04 to 0.09, if the fit
region from 10^-3<x_pbar<0.5<xi> is used.
The ratio that is measured is equal to the diffractive structure
function over the non-diffractive structure function. One can use the
single effective subprocess matrix element in LO, and turn the ratio
around and calculate the diffractive structure function as
F_D = Ratio(x_pbar,xi) X F_ND(x_pbar)
where F_ND is calculated from proton pdf's. One expects a power law,
R(x,xi) ~ A/x^m, F_ND ~ B/x^r, so F_D ~ C/beta^n X 1/xi^k.
Everything is independent of any assumption for the diffractive
process. There is no mention of the Pomeron at all and no MC dependence.
The main topics of the analysis are about to be finished and a draft
for a PRL is well-underway.
5. DPE dijets Koji Terashi
Since the last blessed results, a new Roman pot position
database has been used and an improved underlying event correction has
been applied in JTC96X. The DPE/SD/ND cross sections have been measured
and new results of x_p have been obtained for DPE events.
As a reminder, in single diffractive dijet production we can
measure xi_pbar = P_pomeron/P_pbar, where xi_pbar is measured with
the Roman pots. In double Pomeron exchange, we can measure xi_p=P_pomeron/P_p
=M_X**2/(xi_pbar*s) where M_X is the DPE system mass.
The ND dijet cross section is calculated using data from the MB trigger
at low luminosity. The ND dijet cross section is based on the measurement of
the BBC effective cross section of 51.2+/-1.7 (syst) mb. An underlying event
energy subtraction of 1.16 GeV is applied. The dijet cross section for ET_min>7
GeV/c is 5.17+/-0.04 mb and for ET_min>10 GeV is 1.67+/-0.02 mb.
For SD (DPE) dijets, an underlying event energy of 0.54 (0.37) GeV is
subtracted. The SD dijet cross section is calculated based on the measurement
of the Roman pot cross section (sigma_SD_inclusive=0.73 mb (0.04<xi<0.095,
abs(t)<1.0 GeV**2)). The single diffractive cross section is 16.1+/-0.1
microbarns for ET_min>7 GeV/c and 2.8+/-0.1 microbarns for ET_min>10 GeV/c
(errors are statistical only).
The DPE dijet search is based on finding a rapidity gap due to
Pomeron exchange from the proton. For DPE exchange, we expect low
multiplicity on the east (proton) side. The data signal is zero
multiplicity in the BBC and forward calorimeter. A clear rapidity gap
signal is observed in the (0,0) bin corresponding to a gap in the
rapidity interval from 2.4 to 5.9.
The following ratios are found:
ET_min 7 GeV/c 10 GeV/c
R_(DPE/SD) 0.39+/-0.07% 0.28+/-0.09%
R_(SD/ND) 0.31+/-0.004% 0.17+/-0.004%
R_(DPE/ND) 1.2+/-0.2X10**-5 0.5+/-0.2X10**05
Sigma_DPE_jj 62.5+/-10.6 nb 7.8+/-2.5 nb (statistical errors only)
One can calculate the momentum fractions x_p and x_pbar from the
ET's and eta's of the dijets. One can calculate the ratio of DPE to SD
dijet production as a function of x_p, in a similar manner to what has
been done for the ratio of SD to ND (plotted as a function of x_pbar).
The DPE/SD ratio increases as x_p decreases and has a similar slope to the
ratio of SD to ND (as a function of x_pbar).
6. Systematic error analysis for MLLA studies Alexei Korytov
Alexei discussed the comparison of CDF jet fragmentation data to
modified leading logarithm approximation (MLLA) predictions, with a
concentration on the systematic errors for the comparison. As a brief
review, within the MLLA framework, jet fragmentation can be described
all the way from the hardest scale in the collision down to a
perturbative cutoff (Q_eff) on the order of lambda_QCD.
The only two parameters
in the description are the cutoff and and the number of hadrons per
parton. Most of the distributions come out as functions of Ejet*theta/
Q_eff, where theta is an opening angle around the jet axis. The scaling
in this variable was explicitly tested for the first time in this
CDF measurement. The material is his talk can also be found in CDF note 4996.
To compare experimental results with MLLA, one needs to take
into account that the relative amount of quark and gluon jets change
from about 0.7 to 0.25 for dijet masses from 80 to 600 GeV. Within
MLLA, Nq (the number of partons in quark jets) differs from Ng (the number of
partons in gluon jets) by a simple factor 1/r=C_F/C_A=4/9, i.e.
Nq=Ng/r. There have been attempts to calculate this ratio taking into
account next-to-MLLA and next-to-next-to-MLLA terms. The value of r at
next-to-MLLA is on the order of 2.05 while two results from NNMLLA give
2.0 and 1.85. A few attempts to measure Ng/Nq have lead to smaller values,
in the range from 1.0-1.5. In this analysis, the comparisons for
different values of r were made to the data, with the fraction of
quark and gluon jets being determined from Herwig.
The data was taken from the 30, 50, 70 and 100 GeV jet triggers in
Run 1B. The data were plotted for 9 different dijet mass bins starting
at 72 GeV and ending at 740 GeV, and 5 different cone sizes. The track
parameters were determined using only CTC information. In addition
to the particles from the jet fragmentation, there will also be particles
present in the jet cones from the underlying event. This background
can be determined using cones separated in phi from the jets by
90 degrees. An important additional
background comes from track contamination from electrons from gamma
conversions, Ks decay products etc. To investigate these latter
effects, studies using Herwig + QFL were performed.
The major sources of systematic errors include:
-CTC inefficiency
-choice of track vertex cuts
-jet direction definition
-jet energy determination
-jet energy determination
-choice of fitting range for for track momenta
For the 45 (9X5) data samples, the parameters Qeff,
K(=Ntrk/Nparton) and peak position xi_o were determined.
For every reported value, uncertainties were
evaluated corresponding to the five major systematic error
sources listed above and added in quadrature. For the fitted
parameters, all nine mass jet data samples corresponding to
a particular cone were considered as one ensemble and the
maximum systematic error among them was picked as an overall
conservative estimate.
CTC inefficiency and a poor knowledge of CTC related
effects can result in signficant errors. A new parameterization
of the CTC correction function was given in CDF note 4883. For
all observables, three possible parameterizations of CTC
efficiency were considered: nominal, optimistic and
pessimistic.
The uncertainties in K and Qeff have a significant dependence on
variations in fit boundaries. This is due to the fact that the
shape of the MLLA limiting spectrum does not quite fit the data. If
the spectrum is forced to follow closely the left side of the
distribution, it is off on the right side and vice versa. To
evaluate this systematic uncertainty, separate fits were performed to
the left-hand side of the curve (including the peak) and the right-hand
side (including the peak). Again, for every cone size the largest
deviation for each of the dijet mass samples was assumed for all of the
mass samples.
The inclusive multiplicity distribution has a total error (from
the above sources) of about 10% for each cone size. A comparison of the
jet multiplicity observed in data to MLLA predictions indicates best
agreement with the MLLA predictions with a value of r of between 1.4
and 1.6. The data is also in agreement with Herwig 5.6, if the Herwig
multiplicity is decreased by 0.934.
A fit to the peak position as a function of M_jj*sin(theta) indicates
good scaling for all dijet masses and jet opening angles and results in a
value of Q_eff of 235+/- 40 MeV. The parameter K should decrease as the
dijet mass increases (and a larger fraction of the jets are quark jets) and
it does. It is also observed that K increases as the opening angle of
the jets increases, a feature which should not be present in the MLLA
framework.
Some conclusions:
-MLLA seems to give a reasonable description of inclusive momentum
distributions in a wide range of jet energies and opening angles. The
cutoff scale was found to be 240+/-40 MeV. Qeff tends to become smaller
for larger jet energies, although the difference is not large and remains
within the systematic errors.
-the high statistics data samples show some discrepancies in shape,
such as a noticeable excess to the left from the peak
-the evolution of the mean track multiplicit and the MLLA
parameter K indicates that the ratio r is larger than 1 and can be as
large as 1.4-2.0. However, the fact that this value may be energy
dependent does not allow for more definitive conclusions.
-energy evolution of the peak of the inclusive momentum distribution
is free from spectrum normalization issues and depends only Qeff.
-as a byproduct of the comparisons, it was found that Herwig
5.6 gives a very reasonable description of the distribution shapes,
although the charged particle distribution is about 7% high.
Joey Huston - July 13, 1999