parton distribution functions at the lhc · 2013-02-27 · obtaining pdf sets – general...
TRANSCRIPT
Parton Distribution Functions at the LHC
Robert Thorne
February 15th, 2013
University College London
IPPP Research Associate
With contributions from Alan Martin, James Stirling, Graeme Watt
and Arnold Mathijssen and Ben Watt
UCL – February 2013
UCL – February 2013 1
UCL – February 2013 2
Obtaining PDF sets – General procedure.
Start parton evolution at low scale Q20 ∼ 1GeV2. In principle 11 different
partons to consider.
u, u, d, d, s, s, c, c, b, b, g
mc,mb � ΛQCD so heavy parton distributions determinedperturbatively. Leaves 7 independent combinations, or 6 if we assumes = s (just started not to).
uV = u− u, dV = d− d, sea = 2 ∗ (u + d + s), s + s d− u, g.
Input partons parameterised as, e.g. MSTW,
xf(x,Q20) = (1− x)η(1 + εx0.5 + γx)xδ.
Evolve partons upwards using LO, NLO (or increasingly NNLO) DGLAPequations.
dfi(x,Q2,αs(Q2))
d ln Q2 =∑
j Pij(x, αs(Q2))⊗ fj(x,Q2, αs(Q2))
UCL – February 2013 3
Fit data above ∼ 2GeV2. Need many types for full determination.
- Lepton-proton collider HERA – (DIS) → small-x quarks (best belowx ∼ 0.05). Also gluons from evolution (same x), and now FL(x,Q2).Also, jets → moderate-x gluon.Charged current data some limited infoon flavour separation. Heavy flavour structure functions – gluon andcharm, bottom distributions and masses.
- Fixed target DIS – higher x – leptons (BCDMS, NMC, . . .) → up quark(proton) or down quark (deuterium) and neutrinos (CHORUS, NuTeV,CCFR) → valence or singlet combinations.
- Di-muon production in neutrino DIS – strange quarks and neutrino-antineutrino comparison → asymmetry . Only for x > 0.01.
- Drell-Yan production of dileptons – quark-antiquark annihilation (E605,E866) – high-x sea quarks. Deuterium target – u/d asymmetry.
- High-pT jets at colliders (Tevatron) – high-x gluon distribution – x >0.01.
- W and Z production at colliders (Tevatron/LHC) – different quarkcontributions to DIS.
UCL – February 2013 4
This procedure is generally successful and is part of a large-scale,ongoing project. Results in partons of the form shown.
x-410 -310 -210 -110 1
)2xf
(x,Q
0
0.2
0.4
0.6
0.8
1
1.2
g/10
d
d
u
uss,
cc,
2 = 10 GeV2Q
x-410 -310 -210 -110 1
)2xf
(x,Q
0
0.2
0.4
0.6
0.8
1
1.2
x-410 -310 -210 -110 1
)2xf
(x,Q
0
0.2
0.4
0.6
0.8
1
1.2
g/10
d
d
u
u
ss,
cc,
bb,
2 GeV4 = 102Q
x-410 -310 -210 -110 1
)2xf
(x,Q
0
0.2
0.4
0.6
0.8
1
1.2
MSTW 2008 NLO PDFs (68% C.L.)
Various choices of PDF – MSTW, CTEQ, NNPDF, AB(K)M, HERA,Jimenez-Delgado et al etc.. All LHC cross-sections rely on ourunderstanding of these partons.
UCL – February 2013 5
|y|0 0.5 1 1.5 2 2.5 3
/dy
σ dσ
1/
0
0.1
0.2
0.3
Run II∅* rapidity shape distribution from DγZ/
MRST2006 NNLO PDFs
LO Vrap
NLO Vrap
NNLO Vrap
Run II∅* rapidity shape distribution from DγZ/Leads to very accurateand precise predictions.
Comparison of MSTWprediction for Z rapiditydistribution with data.
UCL – February 2013 6
Parton Fits and Uncertainties. Two main approaches.
Parton parameterization and Hessian (Error Matrix) approach first usedby H1 and ZEUS, and extended by CTEQ.
χ2 − χ2min ≡ ∆χ2 =
∑i,j Hij(ai − a
(0)i )(aj − a
(0)j )
The Hessian matrix H is related to the covariance matrix of theparameters by
Cij(a) = ∆χ2(H−1)ij.
We can then use the standard formula for linear error propagation.
(∆F )2 = ∆χ2∑
i,j∂F∂ai
(H)−1ij
∂F∂aj
,
This is the most common approach.
UCL – February 2013 7
Can find and rescale eigenvectors of H leading to diagonal form
∆χ2 =∑
i z2i
Implemented by CTEQ, then MRST/MSTW, HERAPDF. Uncertainty onphysical quantity then given by
(∆F )2 =∑
i
(F (S(+)
i )− F (S(−)i )
)2,
where S(+)i and S
(−)i are PDF “error sets” displaced along eigenvector
direction.
Must choose “correct” ∆χ2 given complication of errors in full fit andsometimes conflicting data sets.
UCL – February 2013 8
Neural Network group (Ball et al.) limit parameterization dependence.Leads to alternative approach to “best fit” and uncertainties.
First part of approach, no longer perturb about best fit.
Where r(k)p are random numbers following Gaussian distribution. Hence,
include information about measurements and errors in distribution ofO
art,(k)i,p .
Fit to the data replicas obtaining PDF replicas q(net)(k)i (follows Giele et
al.)
Mean µO and deviation σO of observable O then given by
µO = 1Nrep
∑Nrep1 O[q(net)(k)
i ], σ2O = 1
Nrep
∑Nrep1 (O[q(net)(k)
i ]− µO)2.
Eliminates parameterisation dependence by using a neural net whichundergoes a series of (mutations via genetic algorithm) to find the bestfit. In effect is a much larger sets of parameters – ∼ 37 per distribution.
UCL – February 2013 9
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
2 GeV4 = 102Up valence distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
2 GeV4 = 102Down valence distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
Showed equivalence of replicaapproach to Hessian study witheigenvectors for given ∆χ2.
However, can generate “random”PDF sets directly from parametersand variation from eigenvectors.
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
2 GeV4 = 102Up antiquark distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
2 GeV4 = 102Down antiquark distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
ai(Sk)=a0i +∑n
j=1eij
(±t±j
)|Rjk|
(k = 1, . . . , Npdf). Or fromeigenvectors directly (see LHCbstudy and De Lorenzi thesis).Far quicker.
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
2 GeV4 = 102Strange quark distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
2 GeV4 = 102Gluon distribution at Q
MSTW 2008 NLO (68% C.L.)Random params. (40)Random params. (40)Random PDFs (1000)
x-510 -410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
F (Sk)=F (S0)+∑
j
[F (S±j )−F (S0)
]|Rjk|
Can use in reweighting studiesas NNPDF, i.e. weight “random”set by χ2 of fit to new data set(s)to get new distribution.
UCL – February 2013 10
Speed of convergence of prediction for Z cross section.
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.94
0.96
0.98
1
1.02
1.04
1.06
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross section at the LHC (0NLO Z
pdfSame random numbers for each value of N
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.94
0.96
0.98
1
1.02
1.04
1.06
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross section at the LHC (0NLO Z
pdfDifferent random numbers for each value of N
Left, add a new random set to existing ones sequentially. Right,increasing numbers of independent random sets.
Very good with 40 sets. Excellent with 100 sets.
UCL – February 2013 11
Speed of convergence of prediction for W+/W− cross section ratio.
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.97
0.98
0.99
1
1.01
1.02
1.03
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross-section ratio at the LHC (-/W+NLO W
pdfSame random numbers for each value of N
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.97
0.98
0.99
1
1.01
1.02
1.03
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross-section ratio at the LHC (-/W+NLO W
pdfDifferent random numbers for each value of N
Left, add a new random set to existing ones sequentially. Right,increasing numbers of independent random sets.
Very good with 40 sets. Excellent with 100 sets.
UCL – February 2013 12
Speed of convergence of prediction for tt cross section.
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross section at the LHC (tNLO t
pdfSame random numbers for each value of N
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 7 TeV)s cross section at the LHC (tNLO t
pdfDifferent random numbers for each value of N
Left, add a new random set to existing ones sequentially. Right,increasing numbers of independent random sets.
Very good with 40 sets. Excellent with 100 sets.
UCL – February 2013 13
Speed of convergence of prediction for H cross section.
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 120 GeVH
= 7 TeV) for MsH at the LHC (→NLO gg
pdfSame random numbers for each value of N
pdfNumber of random predictions, N
10 210 310
Rat
io to
bes
t-fit
pre
dict
ion
0.95
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
1.05
MSTW 2008 NLO PDFs (68% C.L.)
predictionspdf
Average and s.d. over N
= 120 GeVH
= 7 TeV) for MsH at the LHC (→NLO gg
pdfDifferent random numbers for each value of N
Left, add a new random set to existing ones sequentially. Right,increasing numbers of independent random sets.
Very good with 40 sets. Excellent with 100 sets.
UCL – February 2013 14
Can combine PDF sets, e.g. comparison to PDF4LHC prescription.
) (
nb
)- l+ l
→ 0 B
(Z⋅ 0
Zσ
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1 = 7 TeV)s at the LHC (-l+ l→ 0NNLO Z
MSTW08) = 0.11712
Z(MSα
CT10) = 0.11802
Z(MSα
NNPDF2.3) = 0.11902
Z(MSα
Closed markers: average and s.d. over random predictions.Open markers: usual best-fit and 68% C.L. Hessian uncertainty.
Solid lines: envelope and midpoint.Dashed lines: statistical combination.
G. W
att
(
No
vem
ber
201
2)
-Wσ
/ +Wσ ≡ ±
R
1.4
1.41
1.42
1.43
1.44
1.45
1.46
1.47
1.48
1.49
1.5 = 7 TeV)s ratio at the LHC (-/W+NNLO W
MSTW08) = 0.11712
Z(MSα
CT10) = 0.11802
Z(MSα
NNPDF2.3) = 0.11902
Z(MSα
Closed markers: average and s.d. over random predictions.Open markers: usual best-fit and 68% C.L. Hessian uncertainty.
Solid lines: envelope and midpoint.Dashed lines: statistical combination.
G. W
att
(
No
vem
ber
201
2)
(p
b)
ttσ
145
150
155
160
165
= 7 TeV)s cross sections at the LHC (tNNLO+NNLL t
MSTW08) = 0.11712
Z(MSα
CT10) = 0.11802
Z(MSα
NNPDF2.3) = 0.11902
Z(MSα
Closed markers: average and s.d. over random predictions.Open markers: usual best-fit and 68% C.L. Hessian uncertainty.
Solid lines: envelope and midpoint.Dashed lines: statistical combination.
= 173.18 GeVpolet = m
Fµ =
RµTop++ (v1.4),
G. W
att
(
No
vem
ber
201
2)
(p
b)
Hσ
18.5
19
19.5
20
20.5
21
21.5
22
= 126 GeVH
= 8 TeV) for MsH at the LHC (→NNLO gg
MSTW08) = 0.11712
Z(MSα
CT10) = 0.11802
Z(MSα
NNPDF2.3) = 0.11902
Z(MSα
Closed markers: average and s.d. over random predictions.Open markers: usual best-fit and 68% C.L. Hessian uncertainty.
Solid lines: envelope and midpoint.Dashed lines: statistical combination.
/ 2H = MF
µ = R
µggh@nnlo (v1.4.1),
G. W
att
(
No
vem
ber
201
2)
Smaller uncertainty and shifted central value if disagreement betweenindividual predictions. (Plots by G. Watt at http://mstwpdf.hepforge.org/random/).
UCL – February 2013 15
Comparison to LHC data. (arXiv:1211.1215, A.D. Martin et. al.)
Start with ATLAS jets. Comparison clearly fine.
UCL – February 2013 16
Being more quantitative Using APPLGrid or FastNLO (B. Watt).
MSTW fit very good, see χ2 per point above. Always close to, andsometimes the best of any PDF set, particularly for R = 0.6. Not a hugevariation in PDFs across groups though.
0.96
0.97
0.98
0.99
1
1.01
1.02
1.03
1.04
0.0001 0.001 0.01 0.1
Rat
io to
MS
TW
Cen
tral
Val
ue
x
g(x) at q2=10000 (GeV)2
Before Reweighting (Npdf=1000)After Reweighting (Neff=774)Can see effect of data
on the gluon by usingthe reweighting technique,R = 0.6 (R = 0.4similar). Clearly little pullwith present data.
UCL – February 2013 17
Comparison of MSTW2008 to total W,Z excellent.
Ratio (CMS/Theory)0.6 0.8 1 1.2 1.4
= 7 TeVs at -136 pbCMS
B ( W )× σ th. 0.028± exp. 0.009±0.987
)+ B ( W× σ th. 0.030± exp. 0.009±0.982
)- B ( W× σ th. 0.029± exp. 0.010±0.993
B ( Z )× σ th. 0.032± exp. 0.010±1.002
W/ZR th. 0.015± exp. 0.010±0.981
+/-R th. 0.023± exp. 0.011±0.990
Ratio (CMS/Theory)0.6 0.8 1 1.2 1.4
lumi. uncertainty: 4%±
UCL – February 2013 18
Comparisons between different sets for Inclusive predictions.
) (nb)ν+ l→ + B(W⋅ +Wσ
5.6 5.8 6 6.2 6.4 6.6
) (
nb
)ν- l
→ - B
(W⋅ -
Wσ
3.9
4
4.1
4.2
4.3
4.4
4.5
= 7 TeV)s cross sections at the LHC (- and W+NNLO W
uncertaintiesSαPDF+
-1CMS, L = 36 pb
-1ATLAS, L = 33-36 pb
) = -/W+R(W 1.40
1.45
1.50
G. W
att
(
No
vem
ber
201
2)
68% C.L. PDF
MSTW08
CT10
NNPDF2.3 noLHC
NNPDF2.3
HERAPDF1.5
ABM11
JR09
) (nb)ν± l→ ± B(W⋅ ±Wσ
9.8 10 10.2 10.4 10.6 10.8 11
) (
nb
)- l+ l
→ 0 B
(Z⋅ 0 Zσ
0.88
0.9
0.92
0.94
0.96
0.98
1
1.02
= 7 TeV)sNNLO W and Z cross sections at the LHC (
uncertaintiesSαPDF+
-1CMS, L = 36 pb
-1ATLAS, L = 33-36 pb
R(W/Z) =
10.810.9
11.0
G. W
att
(
No
vem
ber
201
2)
68% C.L. PDF
MSTW08
CT10
NNPDF2.3 noLHC
NNPDF2.3
HERAPDF1.5
ABM11
JR09
(Plots by G. Watt at http://mstwpdf.hepforge.org/pdf4lhc/AR/).
UCL – February 2013 19
)2Z
(MSα0.112 0.113 0.114 0.115 0.116 0.117 0.118 0.119 0.12
(p
b)
ttσ
100
120
140
160
180
200
220
2012PDG
(prel., Sept. 2012)
-1L = 0.7-1.1 fbLHC combined,
= 7 TeV)s cross sections at the LHC (tNNLO+NNLL t
= 173.18 GeVpolet = m
Fµ =
RµTop++ (v1.4),
G. W
att
(
No
vem
ber
201
2)
68% C.L. PDF
MSTW08
CT10
NNPDF2.3 noLHC
NNPDF2.3
HERAPDF1.5
ABM11
JR09
)2Z
(MSα0.112 0.113 0.114 0.115 0.116 0.117 0.118 0.119 0.12
(p
b)
Hσ
17
17.5
18
18.5
19
19.5
20
20.5
21
21.5
22
2012PDG
= 126 GeVH
= 8 TeV) for MsH at the LHC (→NNLO gg
/ 2H = MF
µ = R
µggh@nnlo (v1.4.1),
G. W
att
(
No
vem
ber
201
2)
68% C.L. PDF
MSTW08
CT10
NNPDF2.3 noLHC
NNPDF2.3
HERAPDF1.5
ABM11
JR09
Note variety of values of αS(M2Z) used by different groups.
UCL – February 2013 20
Correlates with PDF luminosity plots
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2 = 8 TeV)s) luminosity at LHC (q(qqΣNNLO
WM ZM
MSTW 2008
CT10
NNPDF2.3 noLHC
NNPDF2.3
G. W
att
(
No
vem
ber
201
2)
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2 = 8 TeV)s) luminosity at LHC (q(qqΣNNLO
WM ZM
MSTW 2008
HERAPDF1.5
ABM11
JR09
G. W
att
(
No
vem
ber
201
2)
UCL – February 2013 21
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2 = 8 TeV)sNNLO gg luminosity at LHC (
top2mHiggsM
MSTW 2008
CT10
NNPDF2.3 noLHC
NNPDF2.3
G. W
att
(
No
vem
ber
201
2)
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2 = 8 TeV)sNNLO gg luminosity at LHC (
top2mHiggsM
MSTW 2008
HERAPDF1.5
ABM11
JR09
G. W
att
(
No
vem
ber
201
2)
Some pretty clear difference between some of the groups. More later.
UCL – February 2013 22
MSTW also pretty good for inclusive distributions. Except someproblems with asymmetry. Comparison gives χ2/Npts = 60/30
|Z
|y0 0.5 1 1.5 2 2.5 3 3.5
The
ory/
Dat
a
0.91
1.10 0.5 1 1.5 2 2.5 3 3.5
| [pb
]Z
/d|y
σd
20
40
60
80
100
120
140
160
= 7 TeV)sData 2010 (
MSTW08
HERAPDF1.5
ABKM09
JR09
-1 L dt = 33-36 pb∫ -l+ l→Z
Uncorr. uncertainty
Total uncertainty
ATLAS
|l
η|0 0.5 1 1.5 2 2.5
The
ory/
Dat
a
0.91
1.10 0.5 1 1.5 2 2.5
| [pb
]lη
/d|
σd
300
400
500
600
700
800
= 7 TeV)sData 2010 (
MSTW08
HERAPDF1.5ABKM09
JR09
-1 L dt = 33-36 pb∫ lν+ l→+W
Uncorr. uncertainty
Total uncertainty
ATLAS
|l
η|0 0.5 1 1.5 2 2.5
The
ory/
Dat
a
0.91
1.10 0.5 1 1.5 2 2.5
| [pb
]lη
/d|
σd
100
200
300
400
500
600
= 7 TeV)sData 2010 (
MSTW08
HERAPDF1.5ABKM09
JR09
-1 L dt = 33-36 pb∫ lν- l→-W
Uncorr. uncertainty
Total uncertainty
ATLAS
|l
η|0 0.5 1 1.5 2 2.5
lA
0.1
0.15
0.2
0.25
0.3
0.35 = 7 TeV)sData 2010 (
MSTW08HERAPDF1.5ABKM09JR09
-1 L dt = 33-36 pb∫
Stat. uncertainty
Total uncertainty
ATLAS
UCL – February 2013 23
MSTW pretty good for ATLAS W,Z distributions, except some problemswith asymmetry.
|ηElectron Pseudorapidity |0 1 2
Ele
ctro
n C
harg
e A
sym
met
ry
0.05
0.1
0.15
0.2
0.25
CMS = 7 TeVs at -1840 pb
ν e→W
MCFM:
theory bands: 68% CL
CT10 HERAPDF1.5 MSTW2008NLO NNPDF2.2 (NLO)
(e) > 35 GeVT
p
Similar for CMS data (will return to this later), though depends on pT
cut. Generally very good for LHCb.
µη0 1 2 3 4
µA
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
= 7 TeVsLHCb,
statData MSTW08
totData ABKM09
JR09
NNPDF21
HERA15
CTEQ6M (NLO)
> 20 GeV/cµT
p
UCL – February 2013 24
Asymmetry used (G. Watt, R.T.) in reweighting (JHEP 1208 (2012) 052),and moves uV −dV up near x = 0.01 - where parameterisation perhapsunderestimates uncertainty. (ATLAS left, CMS pT > 25GeV right).
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lep
ton
ch
arg
e as
ymm
etry
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3-1), L = 33-36 pbµATLAS (l = e+
MSTW08 NLO PDFs (68% C.L.)
/11 = 2.712χHessian,
/11 = 2.902χ = 1000, pdfN
/11 = 1.272χ = 264, effN
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lep
ton
ch
arg
e as
ymm
etry
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3 > 40 GeV
T > 25 GeV, M
νTE > 20 GeV, l
T = 7 TeV) with ps at the LHC (ν± l→ ±W
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lep
ton
ch
arg
e as
ymm
etry
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3
-1CMS (l = e), L = 36 pb-1), L = 36 pbµCMS (l =
MSTW08 NLO PDFs (68% C.L.)
/12 = 1.742χHessian,
/12 = 2.012χ = 1000, pdfN
/12 = 1.202χ = 485, effN
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lep
ton
ch
arg
e as
ymm
etry
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3
> 25 GeVl
T = 7 TeV) with ps at the LHC (ν± l→ ±W
x-410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
2 GeV4 = 102 distribution at Qv - dvu
MSTW08 (68% C.L.)
= 1000)pdf
MSTW08 (N
= 264)eff
MSTW08 (N
x-410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
x-410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
2 GeV4 = 102 distribution at Qv - dvu
MSTW08 (68% C.L.)
= 1000)pdf
MSTW08 (N
= 485)eff
MSTW08 (N
x-410 -310 -210 -110
Rat
io t
o M
ST
W 2
008
NL
O
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
UCL – February 2013 25
Investigation of Parameterisation Issues
In the light of Monte Carlo studies investigate parameterisationdependence, initially concentrating on valence quarks.
Decide to use Chebyshev polynomials (looked at other possibilities)
xf(x,Q20) = A(1− x)ηxδ(1 +
∑n
anTn(y))
i.e. keep high and low x limits. Choose y = 1− 2√
x.
0.001 0.01 0.1 1x
-0.5
0.0
0.5
1.0
TiHyHxL � 1 - 2 xL
0.001 0.01 0.1 1x
-0.5
0.0
0.5
1.0
TiJyHxL � 1 - 2 x N
0.001 0.01 0.1 1x
-0.5
0.0
0.5
1.0
TiIyHxL � 1 - 2 x0.25M
0.001 0.01 0.1 1x
-0.5
0.0
0.5
1.0
TiHyHxL � cosHΠ xLL
Same choice as in Pumplin study. Slightly different to Glazov, Moch andRadescu.
UCL – February 2013 26
It is an open questionhow many parametersare required to accuratelyfit a simple shape suchas that for xuV (x,Q2) orxdV (x,Q2).
Investigate by generatingpseudo-data for varioustypes of parton distributionusing a function withvery many parameters,then fitting with smallernumbers.
Fairly similar conclusionindependent of type ofparameterisation.
UCL – February 2013 27
Example, fit to pseudo-data for valence quark generated between x =0.01 and x = 0.7. For precise match to pseudo-data need 4 polynomials.
0.001 0.005 0.010 0.050 0.100 0.500 1.000x
0.001
0.01
0.1
Fractional deviation
0.001 0.005 0.010 0.050 0.100 0.500 1.000x
0.001
0.01
0.1
Fractional deviation
Look at χ2 distribution with increasing terms in polynomial.
0.02 0.05 0.10 0.20 0.50 1.00
0.001
0.1
10
0.02 0.05 0.10 0.20 0.50 1.00
0.01
0.1
1
10
0.02 0.05 0.10 0.20 0.50 1.00
10-4
0.01
1
0.02 0.05 0.10 0.20 0.50 1.00
10-5
0.001
0.1
10
Very good fits with 4 parameters. More tends to give over-fitting andpeculiarities outside of range of x fit.
UCL – February 2013 28
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of uV(x,Q2=1GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of uV(x,Q2=10000GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of dV(x,Q2=1GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of dV(x,Q2=10000GeV2)
Fits to same data as MSTW2008with 2 extra parameters forvalence quarks – 28 → 32params.
Also with 2 extra parametersfor valence quarks and sea –28 → 34 params. No extraparameters in the gluon foundto be necessary.
0.9
1
1.1
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of sea(x,Q2=1GeV2)
0.9
1
1.1
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of sea(x,Q2=10000GeV2)
0.9
1
1.1
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of g(x,Q2=5GeV2)
0.9
1
1.1
10-4
10-3
10-2
10-1
MSTWMSTWCPvMSTWCP
x
ratio of g(x,Q2=10000GeV2)
∆χ2 = −29, MSTW2008CPpdfs. Only real change in small-x uV , and sea at low Q2.
UCL – February 2013 29
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPMSTWCPdeut
x
ratio of uV(x,Q2=1GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPMSTWCPdeut
x
ratio of uV(x,Q2=10000GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPMSTWCPdeut
x
ratio of dV(x,Q2=1GeV2)
0.8
1
1.2
10-4
10-3
10-2
10-1
MSTWMSTWCPMSTWCPdeut
x
ratio of dV(x,Q2=10000GeV2)
Also do the same fits witha freedom in the deuteriumcorrections – more stablethan previous MSTW studies.MSTW2008CPdeut PDFs.
Now also get variation in dV (x) for higher x due to deuterium correction(seen before) and x ≤ 0.03 due to parameterization and corrections.
Fit to ATLAS W,Z rapidity data at NLO improves to 49/30 for MSTWCPand 46/30 for MSTWCPdeut.
UCL – February 2013 30
0
0.1
0.2
0.3
10-4
10-3
10-2
10-1
MSTW
MSTWCP
MSTWCPdeut
x
x(uV-dV)(x,Q2=10000GeV2)
-50
0
50
10-4
10-3
10-2
10-1
MSTWMSTWCPMSTWCPdeut
x
percentage difference at Q2=10000GeV2
Shown is change in central valueand uncertainty for uV (x)−dV (x)at Q2 = 10, 000 GeV2.
Biggest effect at lower x thanprobed at the LHC (yet).
Uncertainty sets have 23 eigenvectors(20 in MSTW2008).
Main effect in uncertainty anincrease in dV (x,Q2) due todeuterium correction uncertainties,and minor valence uncertaintyincrease from extra parameter.
UCL – February 2013 31
0 1 2 30.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
35
lepton asymmetry, variable p
Tl(min)
MSTW2008 MSTW2008CP MSTW2008CPdeut
0
20
10
30
A(y
l)
yl
LHC 7 TeVMSTW2008 NLO
Increases lepton asymmetry,but very preferentially forhigh pT cut. (Curves madehere with LO calculations).
Most of the effect alreadyobtained for parameterisationextension, but some fromdeuterium study.
UCL – February 2013 32
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lept
on c
harg
e as
ymm
etry
0
0.05
0.1
0.15
0.2
0.25
0.3
-1CMS (l = e), L = 840 pb
NLO PDFs (with NLO K-factors)
/11 = 5.342χMSTW08 (68% C.L.),
/11 = 1.532χMSTW08CP,
/11 = 0.952χMSTW08CPdeut,
> 35 GeVl
T = 7 TeV) with ps at the LHC (ν± l→ ±W
Prediction for pT > 35GeV CMS asymmetry data.
Note no change to data fit, just parameterisation and some fromdeuterium corrections.
Main deuterium effect absence of shadowing used in default fit.
UCL – February 2013 33
|l
ηLepton pseudorapidity, |0 0.5 1 1.5 2 2.5 3
Lept
on c
harg
e as
ymm
etry
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3-1), L = 33-36 pbµATLAS (l = e+
NLO PDFs (with NLO K-factors)
/11 = 2.712χMSTW08 (68% C.L.),
/11 = 1.382χMSTW08CP,
/11 = 0.862χMSTW08CPdeut,
> 40 GeVT
> 25 GeV, MνTE > 20 GeV, l
T = 7 TeV) with ps at the LHC (ν± l→ ±W
Prediction for ATLAS asymmetry data.
Note no change to data fit, just parameterisation and some fromdeuterium corrections.
UCL – February 2013 34
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1 = 8 TeV)s) luminosity at LHC (q(qqΣNLO
WM ZM
MSTW08
MSTW08CP
MSTW08CPdeut
Big change in high-pT cut asymmetry,but this is very specifically sensitive touV (x,Q2)− dV (x,Q2).
What about other quantities?
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1 = 8 TeV)sNLO gg luminosity at LHC (
top2mHiggsM
MSTW08
MSTW08CP
MSTW08CPdeut
Other PDFs changed little. αS free buttiny change. Expect little variation.
Seen clearly on luminosity plots (G.Watt).
(GeV)s10 100 1000
Rat
io t
o M
ST
W 2
008
(68%
C.L
.)
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
)q(q+q
Σ g + 4/9 ≡ = 8 TeV), G sNLO GG luminosity at LHC (
MSTW08
MSTW08CP
MSTW08CPdeut
A not totally insignificant change inthe high-x gluon luminosity for theMSTWCPdeut set . Due to softer dV
UCL – February 2013 35
The percentage change ofvarious cross sections dueto the modifications of theMSTW2008 PDFs. In orderto demonstrate the smallchanges in the cross sections,we also show, in the finalcolumn, the symmetrizedPDF + αS(M2
Z) percentageuncertainties for MSTW2008PDFs.
UCL – February 2013 36
6 7 8 9 10 11 12 13 14 15-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
_t t
gg→H(125)
Z0
W −
W +
δσ/σ
(
%)
√s (TeV)
% change in total (NLO) LHC cross sectionsfrom MSTW2008 to MSTW2008CPdeut
Can see variation of crosssections with energy.
UCL – February 2013 37
Choices for Heavy Flavours in DIS. (Extension of work in by RT inPhys.Rev. D86 (2012) 074017.)
Near threshold Q2 ∼ m2H massive quarks not partons. Created in final
state.
Described using Fixed Flavour Number Scheme (FFNS).
F (x,Q2) = CFF,nf
k (Q2/m2H)⊗ f
nf
k (Q2)
Does not sum αnS lnn Q2/m2
H terms in perturbative expansion. Usuallyachieved by definition of heavy flavour parton distributions and solutionof evolution equations.
Additional problem FFNS known up to NLO (Laenen et al.), but are notfully known at NNLO – α3
SCFF,32,Hi unknown.
Approximations based on some or all of threshold, low-x and high-Q2
limits can be derived, see Kawamura, et al.,, and are sometimes usedin fits, e.g. ABM11 and MSTW (at low Q2). Generally not large exceptat threshold and very low x.
UCL – February 2013 38
Variable Flavour - at high scales Q2 � m2H heavy quarks behave like
massless partons. Sum ln(Q2/m2H) terms via evolution. Zero Mass
Variable Flavour Number Scheme (ZM-VFNS). Ignores O(m2H/Q2)
corrections.
F (x,Q2) = CZM,nf
j ⊗ fnf
j (Q2).
Partons in different number regions related to each other perturbatively.
fnf+1
j (Q2) = Ajk(Q2/m2H)⊗ f
nf
k (Q2),
Perturbative matrix elements Ajk(Q2/m2H) (Buza et al.) containing
ln(Q2/m2H) terms relate f
nf
i (Q2) and fnf+1
i (Q2) → correct evolution forboth.
Want a General-Mass Variable Flavour Number Scheme (VFNS)taking one from the two well-defined limits of Q2 ≤ m2
H and Q2 � m2H.
UCL – February 2013 39
The GM-VFNS can be defined by demanding equivalence of the nf
light flavour and nf + 1 light flavour descriptions at all orders – abovetransition point nf → nf + 1
F (x,Q2)=CFF,nf
k (Q2/m2H)⊗f
nf
k (Q2)=CV F,nf+1
j (Q2/m2H)⊗f
nf+1
j (Q2)
≡ CV F,nf+1
j (Q2/m2H)⊗Ajk(Q2/m2
H)⊗ fnf
k (Q2).
Hence, the VFNS coefficient functions satisfy
CFF,nf
k (Q2/m2H) = C
V F,nf+1
j (Q2/m2H)⊗Ajk(Q2/m2
H),
which at O(αS) gives
CFF,nf ,(1)
2,Hg ( Q2
m2H
) = CV F,nf+1,(0)
2,HH ( Q2
m2H
)⊗P 0qg ln(Q2/m2
H)+CV F,nf+1,(1)
2,Hg ( Q2
m2H
),
The VFNS coefficient functions tend to massless limits as Q2/m2H →∞.
However, CV Fj (Q2/m2
H) only uniquely defined in this limit.
Can swapO(m2H/Q2) terms between CV F,0
2,HH(Q2/m2H) and CV F,1
2,g (Q2/m2H),
i.e. get scheme variations.
UCL – February 2013 40
Difference between FFNS and GM-VFNS
0
1
1 10 102
103
MSTW08MSTW08FFNS
NLO
x=0.0005
Q2
Fc 2(x,
Q2 )
0
0.25
0.5
1 10 102
103
MSTW08
MSTW08FFNS
x=0.005
Q2
Fc 2(x,
Q2 )
0
0.05
1 10 102
103
MSTW08
MSTW08FFNS
x=0.05
Q2
Fc 2(x,
Q2 )
0
1
1 10 102
103
MSTW08MSTW08FFNS
NNLO
x=0.0005
Q2
Fc 2(x,
Q2 )
0
0.25
0.5
1 10 102
103
MSTW08
MSTW08FFNS
x=0.005
Q2
Fc 2(x,
Q2 )
0
0.05
1 10 102
103
MSTW08
MSTW08FFNS
x=0.05
Q2Fc 2(
x,Q
2 )
As seen at higher Q2 charm structure function for FFNS nearly alwayslower than any GM-VFNS. NNLO uses O(α2
S) coefficient functions forF c
2 (x,Q2). Approx. O(α3S) raises slightly F c
2 (x,Q2) at higher x andlowers it at small x (except very small x at low Q2).
UCL – February 2013 41
µ=100 GeV
0.9
0.95
1
1.05
1.1
1.15
1.2
1.25
1.3
10-4
10-3
10-2
10-1
x
evol
ved/
fixe
d-or
der
c(2)(Nf=5)
b(2)(Nf=5)
Σ(2)(Nf=5)
G(2)(Nf=5)
Results for F c2 (x,Q2) in GM-VFNS compared to those for FFNS similar
to results for PDFs by Alekhin et al. in Phys.Rev. D81 (2010) 014032comparing NNLO evolution to the fixed order result up to O(α2
S). Detailsdepend on PDF set and αS(M2
Z) value used.
UCL – February 2013 42
0.96
1
1.04
10 102
103
104
MSTW08MSTW08FFNS
NNLO x=0.0005
Q2
F 2(x,
Q2 )/
MST
W08
0.955
1
1.045
10 102
103
104
x=0.005
Q2
F 2(x,
Q2 )/
MST
W08
0.96
1
1.04
10 102
103
104
x=0.05
Q2
F 2(x,
Q2 )/
MST
W08
Can lead to over 4% changes inthe total F2(x,Q2) if the sameinput PDFs are used in twoschemes.
At higher x mainly due toF c
2 (x,Q2).
At lower x there is a largecontribution from light quarksevolving slightly more slowly inFFNS.
At much higher x differencedies away. Charm componentbecomes very small and lightquark evolution not much different.(Light quarks slightly bigger atthe highest x.)
UCL – February 2013 43
0
0.5
1
1.5
2
1 10 102
103
σ~(x,Q2)HERA (F2(x,Q2)FixedTarget) + c
x=0.00016 (c=0.4)
x=0.0005 (c=0.35)
x=0.0013 (c=0.3)
x=0.0032 (c=0.25)
x=0.008 (c=0.2)
x=0.013 (c=0.15)
x=0.05 (c=0.1)
x=0.18 (c=0.05)
x=0.35 (c=0.0)
Q2(GeV2)
H1ZEUSNMCBCDMSSLAC
NNLO NLO LO
MSTW 2008This is certainly important giventhe precision and range of thedata on F2(x,Q2) included inPDF fits.
Mainly from HERA, but NMCdata also relevant.
UCL – February 2013 44
Performed a series of NLO fits using the FFNS scheme and NNLO withup to O(α2
S) heavy flavour coefficient functions.(Approximations to theO(α3
S) expressions change results very little).
Fit to only DIS and Drell-Yan data but also effectively fit to Tevatron Drell-Yan or Tevatron jet data, if necessary, in 5-flavour scheme as FFNScalculations do not exist.
Fits to DIS and Drell-Yan data usually at least a few tens of unitsworse than MSTW08 to same data (even without refitting MSTW08 torestricted data sets). Often slightly better for F c
2 (x,Q2), but flatter in Q2
for x ∼ 0.01 for inclusive structure function.
As well as (usually) a worse fit to DIS and Drell-Yan data only, in FFNSthe fit quality for the DIS and low-energy Drell Yan data deteriorates byin general ∼ 50 units when all jet data is included as opposed to < 10units when using a GM-VFNS.
UCL – February 2013 45
PDFs evolved up to Q2 = 10, 000GeV2 (using variable flavour evolutionfor consistent comparison) different in form to MSTW08.
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202FFNSDIS-0.1187FFNS+DY-0.1185FFNSjet-0.1222FFNSjetZ-0.1225FF
NS/
2008
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLOFFNSDISFFNS+DYFFNSjetFFNSjetZFF
NS/
2008
at N
LO
for
u(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171FFNSDIS-0.1144FFNS+DY-0.1152FFNSjet-0.1181FFNSjetZ-0.1184FF
NS/
2008
at N
NL
O f
or g
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLOFFNSDISFFNS+DYFFNSjetFFNSjetZFF
NS/
2008
at N
NL
O f
or u
(x,Q
2 )
In contrast in standard MSTW2008 fit PDFs usually within uncertaintiesif Tevatron jet data left out. Main effect loss of tight constraint onαS(M2
Z).
UCL – February 2013 46
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08GMVFNS1GMVFNS2GMVFNS3
GMVFNS4GMVFNS5GMVFNS6ZMVFNS
GMVFNSopt
GM
VFN
Sa/2
008
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08GMVFNS1GMVFNS2GMVFNS3
GMVFNS4GMVFNS5GMVFNS6ZMVFNS
GMVFNSopt
GM
VFN
Sa/2
008
at N
LO
for
u(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08NNLOGMVFNS1GMVFNS2GMVFNS3
GMVFNS4GMVFNS5GMVFNS6GMVFNSopt
GM
VFN
Sa/2
008
at N
NL
O f
or g
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08NNLOGMVFNS1GMVFNS2GMVFNS3
GMVFNS4GMVFNS5GMVFNS6GMVFNSopt
GM
VFN
Sa/2
008
at N
NL
O f
or u
(x,Q
2 )
Using FFNS leads to much larger changes than any choice of GM-VFNS mainly due to fitting high-Q2 DIS data.
UCL – February 2013 47
Low Q2 – Higher Twist.
Potentially large corrections at low Q2 and particularly low W 2. UsualMSTW cuts for
Q2cut - 2GeV2
W 2cut - 15GeV2
Have tried raising Q2 cut to 5GeV2 and 10GeV2 and W 2 to 20GeV2. Notmuch effect on PDFs or αS.
Can also lower W 2cut to 5GeV2 and try parameterising higher twist
contributions by
FHTi (x,Q2) = FLT
i (x,Q2)(
1 +Di(x)Q2
)
where i spans bins of x from x = 0.8− 0.9 down to x = 0− 0.0005.
Previously no evidence for much higher twist except at low W 2.
UCL – February 2013 48
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202
FFNSDIS-0.1187MSTW08HT-0.1189
FFNSHT-0.1188
PDF/
MST
W08
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO
FFNSDISMSTW08HT
FFNSHT
PDF/
MST
W08
at N
LO
for
u(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171
FFNSDIS-0.1144MSTW08HT-0.1175
FFNSHT-0.1152
PDF/
MST
W08
at N
NL
O f
or g
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO
FFNSDISMSTW08HT
FFNSHTPD
F/M
STW
08 a
t NN
LO
for
u(x
,Q2 )
Now more evidence for positive contribution also at very low x. Leads tolower input quarks, more gluon for evolution. Largely washes out quicklywith Q2. Similar effect using FFNS as for GM-VFNS.
UCL – February 2013 49
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202
FFNS+DYHT*-0.1165
FFNS+DY-0.1185
PDF/
MST
W08
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO
FFNS+DYHT*
FFNS+DY
PDF/
MST
W20
08 a
t NL
O f
or u
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171
FFNS+DYHT*-0.1136
FFNS+DY-0.1152
PDF/
MST
W08
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO
FFNS+DYHT*
FFNS+DY
PDF/
MST
W20
08 a
t NN
LO
for
u(x
,Q2 )
Restricting higher twist from lowest x value and omitting nuclear targetdata (except dimuon for strangeness) tends to keep values of αS lowerby ∼ 0.02.
UCL – February 2013 50
Explains some PDF differences? MSTW FFNS ratios and ABKM ratios.
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202
ABKM09-0.1179
PDF/
MST
W08
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO
ABKM09
PDF/
MST
W08
at N
LO
for
u(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171
ABKM09-0.1135
PDF/
MST
W08
at N
NL
O f
or g
(x,Q
2 )0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO
ABKM09
PDF/
MST
W08
at N
NL
O f
or u
(x,Q
2 )
General trend is very similar to fits on previous page.
UCL – February 2013 51
Conclusions
There is a reasonable amount of variation in comparisons todata/predictions at the LHC using different PDF sets.
The MSTWCP and MSTWCPdeut PDFs improve dramatically thepre(post)diction for lepton asymmetries from W bosons at the LHC.
This is due to a very localised effect in small-x valence quarkdifferences. There is extremely little change in parton luminosities orpredictions for a wide variety of other cross sections.
Lowering W 2cut and allowing for higher twist terms in structure functions
makes very little change to MSTW PDFs except at extremely high x
Performing fits using an FFNS leads to worse fits to DIS and low energyDrell-Yan data than GM-VFNS, and much more tension with jet data.
Light quarks in variable flavour number scheme are automatically largerin most regions for FFNS than for GM-VFNS. The gluon is smaller athigh x and larger at small x in FFNS, and αS(M2
Z) smaller. Seems tobe the likely explanation of some major differences in PDFs and LHCpredictions. (Similar conclusions in NNPDF studies.)
UCL – February 2013 52
Back-up
UCL – February 2013 53
Calculate χ2/N = 60/30 for ATLAS W,Z data again at NLO usingAPPLGrid. Not best, but fairly close to any other set except CT10 whichis best. Again look at eigenvectors.
Fit improves markedly in one direction with eigenvector 9, gluon, whichalters common shape and normalisation, and 14 and 18 which alterdV and uV , i.e. affect asymmetry. Not much variation in strangenormalisation.
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0 5 10 15 20
Χ2 p
er p
oint
Dev
iatio
n F
rom
Cen
tral
Val
ue
Eigenvector Number
MSTW Eigenvectors for WZ Fit
UCL – February 2013 54
Similar but slightly smaller effect on uV − dV than asymmetry alone.
Can also see the effect on the gluon. Slight raise near x = 0.01preferred. Improves overall shape of rapidity distribution. Afterreweighting χ2 = 48/60.
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
0.0001 0.001 0.01 0.1
Rat
io to
MS
TW
Cen
tral
Val
ue
x
g(x) at q2=10000 (GeV)2
Before Reweighting (Npdf=1000)After Reweighting (Neff=190)
UCL – February 2013 55
x0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Rat
io t
o M
ST
W 2
008
NL
O (
90%
C.L
.)
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
2 GeV4 = 102Gluon distribution at Q
No Tevatron jetsFit only Run I jets
II jets∅Fit CDFII(kT) + D/2JET
T = p
Fµ =
RµSame with
II jets∅Fit CDFII(Midpoint) + DII jets∅Fit Run I + CDFII(kT) + D
x0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Rat
io t
o M
ST
W 2
008
NL
O (
90%
C.L
.)
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
In contrast in MSTW2008 fit central gluon hardly changed if Tevatron jetdata left out, and only slight further rearrangement of quark flavours ifDrell-Yan data left out.
Main effect loss of tight constraint on αS(M2Z). Much the same at NNLO.
Similar results from various other groups.
UCL – February 2013 56
0.94
0.96
0.98
1
1.02
1.04
10-2
10-1
MSTWCP - 2 deuteron parameters
MSTWCP - 3 deuteron parameters
MSTWCP - 4 deuteron parameters
corr
ectio
n fa
ctor
Given previous relationship betweenTevatron asymmetry and deuteriumcorrections where partial successwas noted revisit with extendedparameterisation.
Default for MSTW some shadowingfor x < 0.01.
Previously big improvement in fit,but “unusual” corrections.
Now improvement again but muchmore stable, and sensible fordeuterium corrections. (Noshadowing favoured though.)
UCL – February 2013 57
The GM-VFNS can be defined by demanding equivalence of the nf
light flavour and nf + 1 light flavour descriptions at all orders – abovetransition point nf → nf + 1
F (x,Q2)=CFF,nf
k (Q2/m2H)⊗f
nf
k (Q2)=CV F,nf+1
j (Q2/m2H)⊗f
nf+1
j (Q2)
≡ CV F,nf+1
j (Q2/m2H)⊗Ajk(Q2/m2
H)⊗fnf
k (Q2).
Hence, the VFNS coefficient functions satisfy
CFF,nf
k (Q2/m2H) = C
V F,nf+1
j (Q2/m2H)⊗Ajk(Q2/m2
H),
which at O(αS) gives
CFF,nf ,(1)
2,Hg (Q2
m2H
) = CV F,nf+1,(0)
2,HH (Q2
m2H
)⊗P 0qg ln(Q2/m2
H)+CV F,nf+1,(1)
2,Hg (Q2
m2H
),
The VFNS coefficient functions tend to the m=0 limits as Q2/m2H →∞.
However, CV Fj (Q2/m2
H) only uniquely defined in this limit.
Can swapO(m2H/Q2) terms between CV F,0
2,HH(Q2/m2H) and CV F,1
2,g (Q2/m2H).
UCL – February 2013 58
0.96
1
1.04
10 102
103
104
MSTW08MSTW08nojet
FFNSDISFFNSjetZ
NNLO x=0.0005
Q2
F 2(x,
Q2 )/
MST
W08
0.96
1
1.04
10 102
103
104
x=0.005
Q2
F 2(x,
Q2 )/
MST
W08
0.96
1
1.04
10 102
103
104
x=0.05
Q2
F 2(x,
Q2 )/
MST
W08
The results for F2(x,Q2) whenrefits are performed.
As seen very little change whenusing GM-VFNS with no jets.
Much more tension and worsefits for FFNS.
UCL – February 2013 59
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
10-3
10-2
10-1
x
F c , Q2 = 20 GeV22
NNLO / NLO
NNLO approx.
x
F c , Q2 = 100 GeV22
NLO scale
NNLO scale
ABM11, m2 = 2 GeV2c
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
10-3
10-2
10-1
Approximate O(α3S) corrections to F c
2 (x,Q2) by Kawamura et al. inNucl.Phys. B864 (2012) 399-468.
Similar results for O(α3S) approximation used by MSTW at low Q2
extended to higher Q2.
UCL – February 2013 60
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08 NLOQ2=5GeV2, W2=20GeV2
Q2=10GeV2, W2=20GeV2
Preliminary
Q2=10,000GeV2
part
ons/
MST
W20
08 a
t NL
O f
or g
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08 NLOQ2=5GeV2, W2=20GeV2
Q2=10GeV2, W2=20GeV2
Preliminary
Q2=10,000GeV2
part
ons/
MST
W20
08 a
t NL
O f
or u
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08 NNLOQ2=5GeV2, W2=20GeV2
Q2=10GeV2, W2=20GeV2
Preliminary
part
ons/
MST
W20
08 a
t NN
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-5
10-4
10-3
10-2
10-1
MSTW08 NNLOQ2=5GeV2, W2=20GeV2
Q2=5GeV2, W2=20GeV2
Preliminarypa
rton
s/M
STW
2008
at N
NL
O f
or u
(x,Q
2 )
Similar to effect of higher twist, particularly at NNLO. Remember losedata at lowest x.
UCL – February 2013 61
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202FFNSHT*DIS-0.1162FFNS+DYHT*-0.1165FFNSjetHT*-0.1199FFNSjetZHT*-0.1215PD
F/M
STW
2008
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLOFFNSDISYHT* 2 LFFNS+DYHT*FFNSjetHT*FFNSjetZHT*PD
F/M
STW
08 a
t NL
O f
or u
(x,Q
2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171FFNSDISHT*-0.1132FFNS+DYHT*-0.1136FFNSjetHT*-0.1155FFNSjetZHT*-0.1174PD
F/M
STW
08 a
t NN
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLOFFNSDISHT*FFNS+DYHT*FFNSjetHT*FFNSjetZHT*PD
F/M
STW
08 a
t NN
LO
for
u(x
,Q2 )
Restricting higher twist from lowest x value and omitting nuclear targetdata (except dimuon for strangeness). Same trends as for standard fitsbut slightly lower αS
UCL – February 2013 62
Explains some PDF differences? MSTW FFNS ratios and ABKM ratios.
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO-0.1202
ABKM09-0.1179
ABM11-0.1180PDF/
MST
W08
at N
LO
for
g(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NLO
ABKM09
ABM11PDF/
MST
W08
at N
LO
for
u(x
,Q2 )
0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO-0.1171
ABKM09-0.1135
ABM11-0.1134PDF/
MST
W08
at N
NL
O f
or g
(x,Q
2 )0.9
0.95
1
1.05
1.1
10-4
10-3
10-2
10-1
MSTW08NNLO
ABKM09
ABM11PDF/
MST
W08
at N
NL
O f
or u
(x,Q
2 )
Better to compare to ABKM09 as mass scheme and data fit are moresimilar.
UCL – February 2013 63
0
20
40
60
80
100
10-5
10-4
10-3
10-2
10-1
1x
xg(x,Q2=10000GeV2)
-10
-5
0
5
10
15
10-5
10-4
10-3
10-2
10-1x
percentage difference at Q2=10000GeV2
0
2
4
6
10-5
10-4
10-3
10-2
10-1
1x
xu(x,Q2=10000GeV2)
-10
-5
0
5
10
15
10-5
10-4
10-3
10-2
10-1x
percentage difference at Q2=10000GeV2
Change in MSTW2008 NNLO PDFs when fitting HERA combined data.
UCL – February 2013 64
Comparison to LHC data.
Start with ATLAS jets. Use APPLGrid or FastNLO at NLO (Ben Watt)and correlated errors treated as in the formula,
χ2 =Npts.∑i=1
(Di − Ti
σuncorr.i
)2
+Ncorr.∑k=1
r2k,
where Di ≡ Di −∑Ncorr.
k=1 rk σcorr.k,i Di are the data points allowed to shift
by the systematic errors in order to give the best fit, and σcorr.k,i is a
fractional uncertainty. Normalisation is treated as the other correlateduncertainties.
MSTW fit very good (χ2 per point below left ), though numbers lower forinclusive data. Always close to, if not best, particularly for R = 0.6. Nothuge variation in PDFs though.
Scale pT/2 pT 2pTInclusive (R=0.4) 0.752 0.773 0.703Inclusive (R=0.6) 0.845 0.790 0.721Dijet (R=0.4) 2.53 2.24 2.20Dijet (R=0.6) 2.44 2.04 1.74
|rk| < 1 1 < |rk| < 2 2 < |rk| < 3 3 < |rk| < 4Inclusive (R=0.4) 85 2 1 0Inclusive (R=0.6) 87 1 0 0Dijet (R=0.4) 82 6 0 0Dijet (R=0.6) 74 12 2 0
UCL – February 2013 65