1. ppZZ4μ background process
Les Houches Workshop, May '05
LO vs. NLO dynamics in
ppH4μ signal and
ppZZ4μ background process
Salavat Abdullin (FNAL)
Darin Acosta (UF)
Paolo Bartalini (UF)
Alexey Drozdetskiy (UF)
Andrey Korytov (UF)
Guenakh Mitselmakher (UF)
Alexander Sherstnev (MSU)
May '05
Alexey Drozdetskiy, University of Florida

2. Alexey Drozdetskiy, University of Florida
Samples preparation
Signal (ppHZZ4μ, mH = 150, 250, 500)
LO: HERWIG
NLO:  HERWIG
real NLO vs. LO comparison
Background (ppZZ4μ)
LO: MadGraph/PYTHIA
NLO: MadGraph/PYTHIA and MCFM (for σNLO calculation)
effective NLO vs. LO comparison
no PHOTOS, default PYTHIA FSR(QED) instead
for both…
only μ from the original Z/γ* considered
Events selection cuts
μ PTbarrel(|η|<1.1) >7 GeV, Pendcap(1.1<|η|<2.4)>9 GeV
inv. mass of the first muon pair (closest to mass Z0) and second muon pair (highest PT muons from the rest excluding muons from the first pair): 12 < Minv < GeV
May '05
Alexey Drozdetskiy, University of Florida

3. Alexey Drozdetskiy, University of Florida
Higgs: diagrams
LO:
NLO:
May '05
Alexey Drozdetskiy, University of Florida

4. Higgs: cross sections, K-factors
Generation conditions:
MZ = GeV, MW = GeV
WH: 16.7 MeV (MH=150 GeV), 4.0 GeV (MH=250 GeV), 66.9 GeV (MH=500GeV)
PDF: CTEQ 5M1 (NLO), CTEQ 5L (LO)
QQCD: Minv(4m)
Cuts: no cuts in
Cross sections and K-factors (no cuts)
mh150: sNLO/sLO=23.48 pb/9.76 pb = 2.40 (KSpira=2.32)
mh250: sNLO/sLO = 8.79 pb/3.95 pb = 2.22 (KSpira=2.17)
mh500: sNLO/sLO = 3.01 pb/1.53 pb = 1.98 (KSpira=1.95)
May '05
Alexey Drozdetskiy, University of Florida

5. Higgs: inv. mass, m(H) = 250 GeV
Normalized to the number of events for 30 fb-1
NLO: Nevent(selection)=58.7;
LO: Nevent(selection)=25.7
KNLO/LO = 2.22 (before selection, previous slide)
KNLO/LO = 2.28 (after selection)
Conclusion: no
significant difference
Blue: NLO
Red: K*LO
Analysis cuts:
243 < Minv(H) < 257 GeV
May '05
Alexey Drozdetskiy, University of Florida

6. Higgs: PT of all four selected muons, m(H)=250 GeV
Normalized to the number of events for 30 fb-1
analysis PT(m) cut
Blue: NLO
Red: KNLO/LO*LO
May '05
Alexey Drozdetskiy, University of Florida

7. Higgs: Higgs PT, m(H)=250 GeV
Some difference
cut on Higgs PT may be useful for high Higgs masses
Blue: NLO
Red: KNLO/LO*LO
May '05
Alexey Drozdetskiy, University of Florida

8. High mass higgs: m(H) = 500 GeV
Wide Breit-Wigner resonance: WH=67 GeV
Problems:
All NLO tools are zero-width approximation with artificial Breit-Wigner imitation...
HERWIG and are not consistent for high mass Higgs yet...
Would like to have a running BR (because of large WH)
discussing with and HERWIG authors
Conclusion: further investigation needed
artefact
Blue: NLO
Red: KNLO/LO*LO
May '05
Alexey Drozdetskiy, University of Florida

9. Alexey Drozdetskiy, University of Florida
ZZ: diagrams
LO:
NLO:
May '05
Alexey Drozdetskiy, University of Florida

10. ZZ: s-channel diagram problem
PYTHIA doesn't include the s-channel diagram (manual, p.141)
General LO generators (CompHEP, MadGraph) allow to generate events taking into account both diagrams
Before analysis cuts it's contribution is significant in mH<200 region
After a la analysis cuts with generated statistics in each bin we have difference at the level of stat. error
still CompHEP with all diagrams systematically bigger at the level 20-30% with respect to PYTHIA (t-chan. only)
may clarify with bigger statistics
peak in 4 muon invariant mass around 90 GeV may be useful for calibration from data
NB!: s-channel suppressed for a particular set of cuts
μ PT > 10 GeV (14, 10, 10, 7 GeV), |η|<2.4
inv. masses: 60 < Z1 < 110 GeV, 12 < mH, Z2, Z3, Z4 (where Z3, Z4 – other permutations of muon pairs, different to selected as Z1, Z2)
at the moment have bigger issues (σ calculations with γ* taken into account)  if know them to a better than 10% precision, need to return to s-channel
May '05
Alexey Drozdetskiy, University of Florida

11. ZZ: s-channel diagram problem (selection cuts)
Minv(Z1)
what we have with selection cuts only…
Minv(Z2)
GeV
GeV
Minv(4mu)
GeV
May '05
Alexey Drozdetskiy, University of Florida

12. ZZ: s-channel diagram problem (after cuts)
4 muons invariant mass, N events for 30 fb-1
All analysis cuts applied:
CompHEP s,t-channel diagrams
PYTHIA
CompHEP t-channel diagram only
stat errors here
are at the level
of events
GeV
May '05
Alexey Drozdetskiy, University of Florida

13. ZZ: effective NLO approximation
Main idea:
NLO QCD corrections: virtual (loops) and real (extra parton). Take into account the virtual correction by a K-factor and the real correction generated correctly (by a LO Monte-Carlo).
Method description:
Generate 2Z/g* and 2Z/g*+j samples with some cuts (on Minv(Z/g*) and PT(jet)) by MadGraph.
Pass 2Z/g* events through PYTHIA (ISR switched on) and 2Z/g*+j events (ISR switched off)
Consider the extra parton Pt (in ZZ – ISR jet). Take 2Z/g* events with PT(jISR) < P0 (with a k-factor) and 2Z/g*+j events with PT(j) > P0.
Set 2 conditions on the PT(j) distribution:
The full cross section has to be equal NLO cross section and
The jet Pt distribution has to be smooth.
The method is being applied by D0 in real single top analyses (D0 note-4722)
see also CMS Note 2000/065 and hep-ph/
May '05
Alexey Drozdetskiy, University of Florida

14. Alexey Drozdetskiy, University of Florida
ZZ: cross sections
Generation conditions:
MZ = GeV,
PDF: CTEQ 5L,
QQCD=Minv(4m)
Cuts: Minv(2m)>12 GeV, PT(j)>1 GeV, |(j)|< 10 in MadGraph
LO cross sections in MadGraph:
sLO(2Z/g* 4μ ) = fb
sLO(2Z/g*+j 4μ +j) = fb
definitely out of LO approximation area in 2Z/g*+j
NLO cross section:
The last ingradient for our method is sNLO(pp 2Z/g* 4μ )
sNLO(pp 2Z/g*)=23.20 pb by MCFM-4.1 (with PDF: cteq5m1)
It corresponds to sNLO(pp 2Z/g* 4μ ) =35.67 fb
May '05
Alexey Drozdetskiy, University of Florida

15. Alexey Drozdetskiy, University of Florida
ZZ: 4μ invariant mass
Normalized on the number of events for 30 fb-1
NLO/LO: Nevent(selection)=356/245
KNLO/LO = 1.38 (before selection)
KNLO/LO = 1.45 (after selection)
Small difference in PT(ZZ)
Analysis cuts (mh250):
243 < Minv(H)< 257 GeV
Blue: NLO
Red: KNLO/LO*LO
May '05
Alexey Drozdetskiy, University of Florida

16. ZZ: PT of all 4 selected muons
Normalized on the number of events for 30 fb-1
Blue: NLO
Red: KNLO/LO*LO
analysis PT(m) cut
May '05
Alexey Drozdetskiy, University of Florida

17. Alexey Drozdetskiy, University of Florida
Summary & Plans
NLO/LO cross section (K-factors):
We received K-factors close to KSpira(a bit higher: 2-3%) for Higgs
K-factors are slightly affected by selection cuts (effect ~3-6%)
K-factor depends on mH cuts
NLO/LO distribution shapes:
No large difference for light mass Higgs and ZZ for currently used in the analysis cuts
ZZ background:
s-channel diagram: the found effect is at the level of 20-30% (with a la analysis cuts applied) for mH < 180 GeV (need to verify with bigger statistics)
uses 0-width approximation for the process, effective NLO approximation allows to exclude this drawback
Near future plans:
Note with description of the results
More reasonable approach to the high mass Higgs case
May '05
Alexey Drozdetskiy, University of Florida

18. Alexey Drozdetskiy, University of Florida
May '05
Alexey Drozdetskiy, University of Florida