Multi-jet physics with BlackHat Kemal Ozeren UCLA Work in collaboration with: Z. Bern, G. Diana, L. Dixon, F. Febres Cordero, S. Hoeche, H. Ita, D. Kosower, D. Maitre 1 [1106.1423], [1112,3940], [1206.6064], [forthcoming] Tuesday, September 4, 12
Multi-jet physics with BlackHat
Kemal OzerenUCLA
Work in collaboration with:
Z. Bern, G. Diana, L. Dixon, F. Febres Cordero, S. Hoeche, H. Ita, D. Kosower, D. Maitre
1
[1106.1423], [1112,3940], [1206.6064], [forthcoming]
Tuesday, September 4, 12
Outline• quest for precision: NLO and BlackHat
• Precision QCD used in CMS jets+MET search
• new result: W+5jets @ NLO
• distributing results: ROOT ntuples
• Fernando’s talk: First NLO prediction of 4-jet
production
2
Tuesday, September 4, 12
Many hard jets...num
ber
of events
10
210
310
410
510
610
data
(MadGraph)νµ → W
top
other backgrounds
CMS preliminary
= 7 TeVs at -136 pb
> 30 GeVjet
TE
exclusive jet multiplicity0 1 2 3 4 5 6
da
ta/M
C
0.5
1
1.5
To get the most from the LHC, we need to understand multi-jet final states
3
Tuesday, September 4, 12
A few words on QCD Predictions• LHC workhorses for full event
simulation: Herwig, Sherpa, Pythia
• ME+PS matching when there are
many hard jets - big improvement
• But need NLO for reliable first
principles QCD prediction,
including correct normalization
• Recent exciting progress in matching NLO/PS MC@NLO [Frixione, Webber; Sherpa]
POWHEG [Nason; Frixione, Nason, Oleari]
• These tools still require the one-loop amplitude as input...
[p
b]
tot
σP
roduction
Cro
ss S
ection,
-110
1
10
210
310
410
510
CMS
W
1j≥
2j≥
3j≥
4j≥
Z
1j≥
2j≥
3j≥
4j≥
> 30 GeV jet
TE
| < 2.4 jetη|
γW
> 10 GeVγ
TE
,l) > 0.7γR(Δ
γZ
WWWZ
ZZ
ZZ→
(127)H
-136 pb -136 pb -11.1 fb -14.7 fb
JHEP10(2011)132
CMS-PAS-EWK-10-012PLB701(2011)535 CMS-PAS-EWK-11-010 CMS-PAS-HIG-11-025
theory prediction
syst)⊕CMS measurement (stat
CMS 95%CL limit
4
Tuesday, September 4, 12
BlackHat
• Efficient evaluation of one-loop QCD amplitudes (traditionally the hardest aspect of NLO predictions)
• Implementation of modern generalized unitarity method
• Evaluates coefficients of integrals
• Tree amplitudes in → one-loop amplitudes out
•Opens the door to precise predictions for multi-jets, used in conjunction
with SHERPA
•Well tested and battle-hardened:
W+1,2,3,4,(5) jets, Z+1,2,3,4
• similar codes available: GoSam, HELAC-NLO, Madloop, NGluon, Rocket ...5
Tuesday, September 4, 12
6
Jets+MET at CMS
Tuesday, September 4, 12
7
Data-Driven Background Estimation
•New physics search in jets + MET channel
•CMS estimates Z + 3 jets background by measuring
photon + 3 jets events
�(pp � Z(� ��)) = �(pp � �) � RZ/�
background to NP measure this theory input
So what is R ? Let’s calculate it at NLO QCD...
Tuesday, September 4, 12
8
Photon Isolation: theory vs experiment
!0
!
"
#
•We are interested in isolated photons
• “Fixed cone” used by experiments is not IR-safe.
Described using fragmentation functions on the
theory side.
•We employ Frixione isolation [Frixione arxiv:9801442]
(no fragmentation piece)�
i
EiT �(� � Ri�) � H(�)
Transverse Energy in Cone vanishes as δ→0
• This is ideal for theorists, problematic for experiments (detector is granular, but isolation condition is continuous)
Tuesday, September 4, 12
9
Z/gamma ratio for CMS•We calculate Z+3j and γ+3j to NLO using BlackHat +Sherpa
• Critical variables:
HT =�
jets
EjetT
• Jet cut: 50 GeV
•Many different regions of interest:
study them all [1106.1423], [1206.6064]
����MET = �
�
jets
�pT,jet
Tuesday, September 4, 12
10
Z/gamma ratio for CMSHow to estimate the error on a ratio?
• Correlated scale variation leads to tiny errors. Should we trust this?
•We study NLO - MEPS and take this as a guide to the uncertainty,
finding around 5-10% error
• Other issues: large QCD logs, large EW logs [0508253]. What is their
impact?
Can use jet ratios as a diagnostic for
large logs
/ [GeV]TH0 100 200 300 400 500 600 700 800
- M
ET /
[GeV
]TH
-400
-200
0
200
400
600
800
1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1, R = 0.5 [anti-kt]2
TH’ = µ
Z+3j / Z+2j (NLO)
BlackHat+Sherpa
LHC 7 TeV
Z+3j/Z+2j
Tuesday, September 4, 12
11
Outcome
• We worked closely with groups from
CMS
• Fruitful cross-talk between theory and
experiment
• This search was very constraining...
CMS [1207.1898]/ GeV
0 100 200 300 400 500 600 700 800 900 10000
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
LONLOME+PS
, R = 0.5 [anti-kt]2TH'
= µ
+3jγZ+3j /
Set 1 cuts
BlackHat+SherpaLHC 7 TeV
[GeV]0m500 1000 1500 2000 2500 3000
[GeV
]1/
2m
100
200
300
400
500
600
700
800
900
1000
± l~ LEP2 ± 1χ∼ LEP2 No EWSB
= L
SPτ∼
Non-Convergent RGE's) = 500g~m(
) = 1000g~m(
) = 1500g~m(
) = 2000g~m(
) = 1000q~m(
) = 1500q~m(
) = 2000
q~m(
) = 2500
q~m()=10βtan(
= 0 GeV0A > 0µ = 173.2 GeVtm = 7 TeVs, -1 L dt = 4.98 fb∫CMS,
Observed signal theoryσ1±Observed exp.σ1±Expected
-1CMS, 36 pb
LM5
Good example of utility of high-precision
theory: ratio = input to data-driven
method
See [1106.1423] and [1206.6064] for
many plots and numerical results
Tuesday, September 4, 12
W+5j Production at NLO
this talk: preliminary results
12
Tuesday, September 4, 12
W + 5jets Production at the LHCFirst 2→6 NLO calculation at a
hadron collider
Preliminary
T T �; FSWSR� �� �
PITXSR + 1)8 + �Nq
g
l
!q
g
g
g
g
sample Feynman diagram (octogon!)
For searches: background for various NP signatures
(also to top pair production)
For measurements: check theory vs experiment
leading colour approx. in the virtual part only: expect 2-3%13
Tuesday, September 4, 12
Theoretical Tools
•BlackHat for virtual part [Bern, Dixon, Febres Cordero, Hoeche, Ita,
Kosower, Maitre, KJO]
•COMIX (part of Sherpa) for real emission [Hoeche]
•Sherpa - organizational framework [Hoeche, Hoeth, Krauss,
Schoenherr, Schumann, Siegert, Winter]
•LHAPDF for parton distributions [Whalley, Bourilkov, Group]
•FASTJET for jet clustering [Cacciari, Salam, Soyez]
•ROOT for analysis and storing events [see later]14
Tuesday, September 4, 12
W+5jets: reduced scale variation at NLO
• Plot shows effect of varying μ up and down by a factor of 2
Strong reduction in scale
uncertainty
µ* = µ6
15
/ [GeV]T
First jet p0 50 100 150 200 250 300 350 400
[pb
/ GeV
]T
/ d
pσd
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014 LO
NLO
PRELIMINARYBlackHat+Sherpa
+ 5 jets-W, R = 0.5 [anti-kt]2
TH' = µ
Tuesday, September 4, 12
Jet Ratios: key observables
W+3j / W+2j
W+4j / W+3j
W+5j / W+4j
dash=LOsolid=NLO
• Note - they are not constant!• Both theory and experimental errors are minimized in the ratio• Can be important input to data-driven methods for backgrounds (compare Z/γ + 3jets used in SUSY search )
/ GeVTH0 100 200 300 400 500 600 700 800
+(n-
1)j
-+n
j / W
-W
0
0.2
0.4
0.6
0.8
1
1.2NLO32NLO43NLO54
BlackHat+SherpaPreliminary
LHC 7 TeV
LO
NLO
16
Tuesday, September 4, 12
/ GeVT
lead jet p50 100 150 200 250
+nj
-+n
j / W
+W
0
0.20.40.60.8
11.21.41.6
1.82
2.2
NLO 2-jetNLO 3-jet x 0.8NLO 4-jet x 0.6NLO 5-jet x 0.4
More Ratios: W+ / W-
W+ / W- + 2jet
W+ / W- + 3jet
W+ / W- + 4jet
dash=LOsolid=NLO
x 0.8
x 0.6
NOTE scalings for visual clarity
x 0.4
W+ / W- + 5jetBlackHat+SherpaPreliminary
LHC 7 TeV
17
Tuesday, September 4, 12
18
100 200 300 400 500
0.0001
0.001
0.01
0.1
1
10
100
dσ /
dpT
[ pb
/ G
eV ]
LONLO
100 200 300 400 500First Jet pT [ GeV ]
0.5
1
1.5
2
2.5
3
3.5 LO / NLO
100 200 300 400 500
100 200 300 400 500Second Jet pT [ GeV ]
100 200 300 400 500
100 200 300 400 500Third Jet pT [ GeV ]
100 200 300 400 500
100 200 300 400 500Fourth Jet pT [ GeV ]
100 200 300 400 500
0.0001
0.001
0.01
0.1
1
10
100
100 200 300 400 500Fifth Jet pT [ GeV ]
0.5
1
1.5
2
2.5
3
3.5
NLO scale dependence
BlackHat+Sherpa
LO scale dependence
pTjet > 25 GeV, | ηjet | < 3
ETe > 20 GeV, | ηe | < 2.5
ETν > 20 GeV, MT
W > 20 GeV
R = 0.4 [anti-kT]
√%s = 7 TeV
µR = µF = HT^ ’ / 2
W- + 5 jets + X
nth jet pT in events with at least 5 jets [preliminary]
1st jet 2nd 3rd 4th 5th
Preliminary
Tuesday, September 4, 12
Distributing the Results - ROOT ntuples
• NLO calculations often very computationally intensive
• don’t want to run again and again with different cuts etc.
• solution: store events and apply analysis cuts later
• ROOT ntuple files ideal for this purpose
• store coefficients of ln μ. For example, for the virtual piece:
1PSST = %+ & ln µ + ' ln� µ
• Major benefit 1: can change scales, PDF, add new observables...
• Major benefit 2: can hand over the ntuples to experimentalists
We are working towards release of ntuples, including a library of code for their analysis
19
Tuesday, September 4, 12
20
Using the Ntuples
• events stored in ntuple are parton level
• user should perform clustering into jets [FASTJET]
• at generation level, event selection is such as to permit only certain jet algorithms (we choose siscone, kt, anti-kt R=0.4, 0.5, 0.6, 0.7)
• four parts to be added [c.f. Catani-Seymour subtraction]:
Born Virtual Int Sub Real-Sub� �� � � �� �
n-parton events (n+1) and n-parton events
+ + +
• cancellations within RS piece - need to be careful when evaluating statistical error
Tuesday, September 4, 12
ROOT ntuples in Action ATLAS W+jets [1201.1276]
• ntuples created with BlackHat+Sherpa [1009.2338]
• experimenters perform their own analysis of the NLO results
• we are currently working also with CMS people
... we look forward to similar comparisons of our W+5jet NLO results with data!
21
Tuesday, September 4, 12
Summary
• I stressed the importance of NLO precision in achieving our physics goals at the LHC
• Application of BlackHat+Sherpa to Jets+MET new physics search
• new result: W+5jets → highest multiplicity NLO calculation at a hadron collider
• described how our results are passed to experimenters and compared to data: ROOT ntuples
22
Tuesday, September 4, 12