Introduction Dipole Showers Results ˇ
Color corrections in parton showers
Malin Sjödahl
In collaboration with Simon Plätzer (Vienna) and Johan Thorén (Lund)
arXiv: 1808.00332 October 17, 2018
Malin Sjödahl
Introduction Dipole Showers Results ˇ
Section 1
Introduction
Introduction Dipole Showers Results ˇ
Parton shower basics
I A parton shower starts from a hard matrix element for some scattering process and dresses it up with additional radiation (mostly gluons)
Hard part
Radiation added by shower
Malin Sjödahl
Introduction Dipole Showers Results ˇ
In a leading color parton shower things are easy ...
I In standard parton showers where QCD is treated as if it had infinitely many colors, the colors are described by color lines and only color connected partons (sharing a line) can radiate coherently→ ∼ Npartondipoles
Introduction Dipole Showers Results ˇ
In real QCD with three colors
I For Nc= 3 radiation from almost any pair of partons can interfere→∼ Npartons2 possibilities, suppressed by (1/Nc), 1/Nc2,...
Malin Sjödahl
Introduction Dipole Showers Results ˇ
Why investigate N
c= 3 color corrections?
I Expect that color suppressed terms become very important for many partons
I The colored initial state and the higher energy at the LHC gives rise to many colored partons and hence many color suppressed terms
I Needed for exact matching of matrix elements to parton showers
I Needed for Nc = 3 hadronization
Introduction Dipole Showers Results ˇ
Section 2 Dipole Showers
Malin Sjödahl
Introduction Dipole Showers Results ˇ
Dipole Factorization
I Parton showers work under the approximation that the next parton to be emitted is soft or collinear to one of the
existing partons
I Dipole factorization gives, whenever i and j become collinear or one of them soft:
|Mn+1(..., pi, ..., pj, ..., pk, ...)|2= X
k6=i,j
1
2pi· pjhMn(p˜ij, p˜k, ...)|Vij,k(pi, pj, pk)| Mn(p˜ij, p˜k, ...)i An emitter ˜ijsplits into two partons i and j, with the spectator ˜kabsorbing the momentum to keep all partons (before and after) on-shell. (Catani, Seymour
hep-ph/9605323)
ij˜ i
j
Color corrections in parton showers 8
Malin Sjödahl
In collaboration with Simon Plätzer (Vienna) and Johan Thorén (Lund)arXiv: 1808.00332
Introduction Dipole Showers Results ˇ
The spin averaged splitting kernel is
Vij,k(pi, pj, pk) =−8παsVij,k(pi, pj, pk)T˜ij· Tk
T2˜ij
Where, for example, for a final-final dipole configuration, we have
Vq→qg,k(pi, pj, pk) = CF
2(1− z)
(1− z)2+ p2⊥/sijk − (1 + z)
Malin Sjödahl
Introduction Dipole Showers Results ˇ
Emission probability
For a leading Nc shower, the emission probability is dPij,k(p2⊥, z) = Vij,k(p2⊥, z)dφn+1(p2⊥, z)
dφn ×δ(˜ij, ˜k color connected) 1+ δ˜ij g
Including subleading emissions, instead gives dPij,k(p2⊥, z) = Vij,k(p2⊥, z)dφn+1(p2⊥, z)
dφn ×−1 T2˜
ij
hMn|T˜ij· T˜k|Mni
|M|2
Introduction Dipole Showers Results ˇ
Overall Picture
UsingHerwig’s dipole shower
I Instead of only allowing color connected emitter-spectator pairs to radiate, all possible pairs can radiate
I All pairs may radiate in proportion to(for the first emission) ωnik= −1
T2˜
ij
hMn|T˜ij· T˜k|Mni
|M|2
I Reweighting to encompass negative contributions
I The full color structure is evolved to be able to evaluate the above factor for the next emission
I Color structure is calculated usingColorFull(MS 1412.3967)
I Nc = 3 shower for a number of emissions, then standard leading Nc shower
Malin Sjödahl
Introduction Dipole Showers Results ˇ
Color structure
I A major challange is the SU(3) color structure of QCD
I The color structure can be decomposed in color bases
|Mni =
dn
X
α=1
cn,α|αni ↔ Mn= (cn,1, ..., cn,dn)T
and for this project we use trace bases
I ... but these standard “bases” are non-orthogonal and overcomplete, with a dimension scaling∼ (Ng+ Nq¯q)!→ (Ng+ Nq¯q)!2terms when squaring
I See next talk by Johan Thorén for better bases
Introduction Dipole Showers Results ˇ
New Features
Compared to our previous e+e−results(SP, MS 1206.0180), we have added
I The g→ q¯q splitting
I Hadronic initial state, meaning initial state radiation
I Full compatibility with all of the additional functionality in Herwig 7.1. (So we can run any process now, in particular LHC events)
I Subsequent standard leading Ncshowering after the Nc = 3 shower
Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
Section 3
Results
Dipole Showersˆ Results Current Status and Future Work
Full Color Shower Reaching Soft Scales
Since a limited number of Nc = 3 emissions are kept, up to 3 for LHC and 5 for LEP, we check the pT of the last corrected
emission
5th Nc=3 emission
0 1 2 3 4p⟂
0.00 0.05 0.10 0.15
LEP
3rd Nc=3 emission
5 10 15 20 25 p⟂
0.01 0.02 0.03 0.04 0.05 0.06
LHC 50 GeV
I → We go far down in pTcompared to relevant jet scales, at LEP close to the hadronization scale
Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
LEP Preliminary Results
For most e+e−observables we find small corrections, at the percent level. However, some observables (thrust, out-of-plane p⊥, hemisphere masses, aplanarity, jet multiplicities for many jets) are corrected by∼ 5%.
Leading Nc 1 Nc=3 emission 3 Nc=3 emissions 5 Nc=3 emissions
10−5 10−4 10−3 10−2 10−1
Jets
2 3 4 5 7
0.9 0.95 1.0 1.05
Ratio
Leading Nc 1 Nc=3 emission 3 Nc=3 emissions 5 Nc=3 emissions
10−5 10−4 10−3 10−2 10−1 1 101
Aplanarity
1/σdσ/dA
0 0.05 0.1 0.15 0.2
0.9 0.95 1.0 1.05
Ratio
Dipole Showersˆ Results Current Status and Future Work
LHC Preliminary Results
For LHC observables, corrections are typically of order a few percent, but some observables show corrections of 10− 20%
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions 105
106
Rapidity of first jet
dσ/dy(jet1)[pb]
-4 -2 0 2 4
0.8 0.9 1.0 1.1 1.2
y(jet 1)
Ratio
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions 104
105 106
Rapidity of second jet
dσ/dy(jet2)[pb]
-4 -2 0 2 4
0.8 0.9 1.0 1.1 1.2
y(jet 2)
Ratio
Figure:Rapidity of hardest and second hardest jetusing a 50GeV analysis cut Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
If we could study quark-gluon scattering, we would find large corrections
Leading Nc 3 Nc=3 emissions
102 103 104 105 106
Rapidity of first jet
dσ/dy(jet1)[pb]
-4 -2 0 2 4
0.8 0.9 1.0 1.1 1.2
y(jet 1)
Ratio
Leading Nc 3 Nc=3 emissions
101 102 103 104 105
Rapidity of second jet
dσ/dy(jet2)[pb]
-4 -2 0 2 4
0.8 0.9 1.0 1.1 1.2
y(jet 2)
Ratio
Figure:Rapidity distribution of the hardest and second hardest jet while considering only qg→ qg scattering and a 50 GeV analysis cut.
Dipole Showersˆ Results Current Status and Future Work
Requiring one forward (quark dominated) and one central (gluon dominated) jet we find sizable corrections for many observables
Leading Nc 3 Nc=3 em.
1 101 102 103 104
Rapidity of second jet
dσ/dy(jet2)[pb]
-4 -2 0 2 4
0.85 0.9 0.95 1.0 1.05 1.1
y(jet 2)
Ratio
Leading Nc 3 Nc=3 emissions
102 103
Azimuthal separation between jets
dσ/d∆φ(jet1,jet3)[pb]
0 0.5 1 1.5 2 2.5 3
0.85 0.9 0.95 1.0 1.05 1.1
∆φ(jet 1, jet 3)
Ratio
Figure:Rapidity and∆φ1,3for the central/forward case
(400< M12<600 GeV, 3.8 <|y1+ y2| < 5.2, 1.5 < |y2− y1| < 3.5)
Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
We have compared to LHC data for a wide range of
observables. In general we find small corrections and no overall visible change in data description.
b b b b b b b b b b b b b
Data
b
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Central Transv. Thrust, 90 GeV<pjet 1⊥ <125 GeV,√s=7 TeV
1/NdN/dln(1−TC)
-12 -10 -8 -6 -4 -2
0.6 0.8 1 1.2 1.4
ln(1−TC)
MC/Data b b b b b b b b b b b b b
Data
b
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
0.4Central Transv. Minor, 90 GeV<pjet 1⊥ <125 GeV,√s=7 TeV
1/NdN/dln(Tm,C)
-6 -5 -4 -3 -2 -1
0.6 0.8 1 1.2 1.4
ln(Tm,C)
MC/Data
Figure:Central transverse thrust and thrust minor for√s= 7GeV, CMS 1102.0068, TC= maxˆnT
P
i|p⊥,i·ˆnT|
Pp , Tm,C=
P
i|p⊥,i׈nT|
Pp for jet i, with
Dipole Showersˆ Results Current Status and Future Work
Conclusion, Hard Perturbative Region
I We have considered a wide range of observables at LEP and LHC and compared to data
I Overall the data description does not change
I As long as soft scales/observables with very many jets are not considered, the matrix element correction type of corrections are accurately described by correcting the first few emissions
I In general, percent level corrections are found at LEP, for some observables (thrust, out-of-plane p⊥, hemisphere masses, aplanarity, jet multiplicities for many jets) effects of around 5%
I At the LHC, corrections are often a few percent, for some observables (mostly rapidity) corrections around 10-20%
Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
Going Soft/Very Many Colored Partons
For soft QCD, where we cannot expect reliable results due to the need of more color suppressed terms, resummation, hadronization and MPI, we find larger corrections in many cases, (jet resolution scales, cluster masses in Herwig, number of very soft jets at LEP, charged multiplicity distribution,
individual hadron multiplicities), indicating that subleading Nc
effects probably play an important role for soft(ish) QCD
Leading Nc 1 Nc=3 emission 3 Nc=3 emissions 5 Nc=3 emissions 10−5
10−4 10−3 10−2 10−1 Jets
2 4 6 8 10 12 14
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
Jets
Ratio bbbbbbbbbbbbbbbbbbbbbbbbbb
Data
b
Leading Nc 3 Nc=3 emissions 5 Nc=3 emissions
10−5 10−4 10−3 10−2 10−1
Charged multiplicity distribution
1/NdN/dNch
10 20 30 40 50
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
Nch
MC/Data
Figure:Examples of large corrections: first clustermass in Herwig,
Color corrections in parton showers 22
Malin Sjödahl
In collaboration with Simon Plätzer (Vienna) and Johan Thorén (Lund)arXiv: 1808.00332
Dipole Showersˆ Results Current Status and Future Work
Conclusion, Soft Region
In the soft region/region of many colored partons:
I In this region, we cannot claim accurate results, however,
I we often find large corrections of several ten percent
I This affects the state going into the hadronization
I meaning that we can expect a significant effect on the tune
I Subleading Nc effects can therefore be hidden in the tune
I Need to retune
Malin Sjödahl
Dipole Showersˆ Results Current Status and Future Work
Section 4
Current Status and Future Work
Dipole Showersˆ Results Current Status and Future Work
Current Status and Future Work
I We have a fully functional Nc = 3 parton shower for any LEP or LHC process
I Tuning should be performed before a reliable comparison to standard showers can be done
I We still miss virtual corrections, which rearrange the color structure without any real emissions. These are important for gap-survival observables
I In the more distant future, an update of hadronization models to an Nc = 3 final state would be and interesting research task
Thank you!
Malin Sjödahl
Section 5
Backup Slides
Weight distribution
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions 4 Nc=3 emissions 5 Nc=3 emissions
-10 -5 0 5 10
10−5 10−4 10−3 10−2 10−1 1
Weight distribution
w N(w)/Ntot
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions
-10 -5 0 5 10
10−5 10−4 10−3 10−2 10−1 1
Weight distribution
w N(w)/Ntot
Figure:Weight distribution for e+e−(left) and pp collisions (right) depending on the number of Nc= 3 emissions allowed. All generated events are used in these plots, i.e., no further analysis cut is applied.
Malin Sjödahl
N
c= 3 Shower Reaching Soft Scales
4th Nc=3 emission 5th Nc=3 emission 6th Nc=3 emission
0 1 2 3 4p⟂
0.00 0.05 0.10 0.15 0.20
LEP
3rd Nc=3 emission 4th Nc=3 emission
5 10 15 20 25 p⟂
0.02 0.04 0.06 0.08
LHC 50 GeV
More LEP Observables
bbbbbbbbb b b b b b b b b
Data
b
Leading Nc 3 Nc=3 emissions 5 Nc=3 emissions
10−2 10−1 1 101
102Out-of-plane p⊥in GeV w.r.t. thrust axes
Ndσ/dpout⊥
0 0.5 1 1.5 2 2.5 3 3.5
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
pout⊥/ GeV
MC/Data b b b b b b b b b
Data
b
Leading Nc 3 Nc=3 emissions 5 Nc=3 emissions
10−1 1 101
Light hemisphere masses, M2l/Evis2
Ndσ/dM
2 l/E 2 vis
0 0.02 0.04 0.06 0.08 0.1
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
M2l/E2vis
MC/Data bbbbbbbbbbbbbbbbbbbbbbbbbb
Data
b
Leading Nc 3 Nc=3 emissions 5 Nc=3 emissions
10−5 10−4 10−3 10−2 10−1
Charged multiplicity distribution
1/NdN/dNch
10 20 30 40 50
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
Nch
MC/Data
Figure:Out-of-plane p⊥w.r.t. the thrust and thrust major axes (left), light hemisphere mass (middle) and fraction of events containing Nch
charged particles.DELPHI, ALEPH
Malin Sjödahl
Top at LHC
b b b b b b b b b b b b b b b b b b
Data
b
Leading Nc 3 Nc=3 emissions
0.7 0.75 0.8 0.85 0.9 0.95 1.0
Gap fraction vs. Q0for veto region:|y| <0.8 fgap
50 100 150 200 250 300
0.94 0.96 0.98 1.0 1.02 1.04
Q0[GeV]
MC/Data bbbbb b b b b b b b b b b b b b
Data
b
Leading Nc 3 Nc=3 emissions
0.86 0.88 0.9 0.92 0.94 0.96 0.98 1.0
Gap fraction vs. Qsumfor veto region:|y| <0.8 fgap
50 100 150 200 250 300 350 400
0.94 0.96 0.98 1.0 1.02 1.04
Qsum[GeV]
MC/Data
Figure:Fraction of events having no additional jet with p⊥ above Q0
within a rapidity interval|y| < 0.8 (left) and fraction of events where the scalar sum of transverse momenta within√ |y| < 0.8 does not
QCD “Coherence” observable
b b b b b b b b b b b b b b b b b b
Data
b
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions 3 Nc=3 emissions
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
CMS,√s=7 TeV, central jet 2–3 correlation,|η2| <0.8 Fη2(β)
0.5 1 1.5 2 2.5 3
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
β
MC/Data b b b b b b b b b b b b b b b b b b
Data
b
Leading Nc 1 Nc=3 emission 2 Nc=3 emissions
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
CMS,√s=7 TeV, central jet 2–3 correlation,|η2| <0.8 Fη2(β)
0.5 1 1.5 2 2.5 3
0.50.6 0.70.8 0.91.0 1.11.2 1.31.4
β
MC/Data
Figure:The angleβ, tan β =sign|φ(η3−φ2)(η23|−η2), using (left) an underlying 2→ 2 hard process and (right) an underlying 2 → 3 hard process.
CMS 1102.0068
Malin Sjödahl
Density Operator
We can write the amplitude as a vector in some basis (trace, multiplet, etc.),
|Mni =
dn
X
α=1
cn,α|αni ↔ Mn = (cn,1, ..., cn,dn)T (1)
and construct a “density operator” Mn =MnM†n, that we evolve by
Mn+1=−X
i6=j
X
k6=i,j
4παs
pi· pj
Vij,k(pi, pj, pk) T2˜
ij
T˜k,nMnT˜†
ij,n (2)
where
Vij,k = T2˜ijpi· pk
pj· pk. (3)
Color Matrix Element Corrections
Evolving the density operator, we can calculate the color matrix element corrections for any number of emissions
ωnik= −1 T2˜
ij
Tr
Sn+1× T˜k,nMnT˜†
ij,n
Tr(Sn× Mn) (4)
I Note thatωikn can be negative, this is included through the weighted Sudakov algorithm(Bellm, SP, Richardson, Siodmok, Webster, 1605.08256)
I This initially resulted in very large weights. Modifications to the weighted Sudakov veto algorithm drastically reduced the weights.
Malin Sjödahl
Standard veto algorithm
Standard veto algorithm: we want to generate a scale q and additional splitting variables x (e.g. z andφ) according to a distribution dSP.
dSP(µ, xµ|q, x|Q)
= dqddx(∆P(µ|Q)δ(q − µ)δ(x − xµ)
+P(q, x)θ(Q− q)θ(q − µ)∆P(q|Q)) Where∆Pis the Sudakov form factor,
∆P(q|Q) = exp
− Z Q
q
dk Z
ddzP(k, z)
To do this we use an overestimate of the distribution (with nicer analytical properties) dSR (change P→ R in the above eqs.).
Standard veto algorithm
Standard veto algorithm: we want to generate a scale q and additional splitting variables x (e.g. z andφ) according to a distribution dSP.
dSP(µ, xµ|q, x|Q)
= dqddx(∆P(µ|Q)δ(q − µ)δ(x − xµ)
+P(q, x)θ(Q− q)θ(q − µ)∆P(q|Q)) Where∆Pis the Sudakov form factor,
∆P(q|Q) = exp
− Z Q
q
dk Z
ddzP(k, z)
To do this we use an overestimate of the distribution (with nicer analytical properties) dSR (change P→ R in the above eqs.).
Where we require R(q, x)Malin Sjödahl≥ P(q, x) for all q, x.
Standard veto algorithm
P(q, x) > 0 and R(q, x)≥ P(q, x). Set k = Q
1. Generate q and x according to SR(µ, xµ|q, x|k).
2. If q= µ, there is no emission above the cutoff scale.
3. Else, accept the emission with the probability P(q, x)
R(q, x).
4. If the emission was vetoed, set k= q and go back to 1.
Weighted veto algorithm
Introduce an acceptance probability 0≤ (q, x|k, y) < 1 and a weightω. Set k = Q, ω = 1.
1. Generate q and x according to SR(µ, xµ|q, x|k).
2. If q= µ, there is no emission above the cutoff scale.
3. Accept the emission with the probability(q, x|k, y), update the weight
ω → ω ×1
×P R 4. Otherwise update the weight to
ω→ ω × 1 1− ×
1−P
R
and start over at 1 with k= q.
Malin Sjödahl
Example of 1 /N
csuppressed terms
Leading color structure:
2
=
= TR = TR2(Nc2− 1) ∝ Nc2.
Example of 1 /N
csuppressed terms
Leading color structure:
2
∝ Nc2. Interference term:
∗
=
= TR −TR
Nc
= 0− TR2
Nc2− 1 Nc ∝ Nc.
Malin Sjödahl
Example of 1 /N
csuppressed terms
∗
=
= TR
| {z }
∝N2c
−TR
Nc
| {z }
∝Nc2