studies of pdf uncertainties for the measurement of the ... · october 21, 2014 stefano camarda 2....
TRANSCRIPT
Studies of PDF uncertainties for the measurementof the mass of the W boson at the LHC
Stefano Camarda (DESY)
October 21, 2014
October 21, 2014 Stefano Camarda 1
PDF uncertainties on mW
ATL-PHYS-PUB-2014-015“Studies of theoretical uncertainties for the measurement ofthe mass of the W boson at the LHC”
IntroductionTheory predictions and PDF uncertaintiesEvent selection and methodologyW polarization and PDF uncertaintiesCharm-initiated W production and PDF uncertaintiesDetector effects and summary of PDF uncertaintiesParton shower uncertaintiesConclusions
Points of discussion for a coherent treatment of PDFuncertainties between ATLAS and CMS
October 21, 2014 Stefano Camarda 2
Introduction
The extraction of mW from the p`T spectrum, is likely to belimited by theoretical uncertainties.
This study addresses PDF uncertainties, and non-perturbativeQCD uncertainties related to the parton shower model
We need not only a precise estimation of PDF and othertheoretical uncertainties, but also a roadmap to control andreduce them, by mean of precise experimental measurementsof alternative observables
The idea is to perform a breakdown of the physicalmechanisms behind the PDF uncertainties, and estimatewhich are the most relevant sources of uncertainties.
By pointing out the largest uncertainties, the idea is toprovide a pattern to reduce them, rather than a preciseestimation to be used in our measurement
October 21, 2014 Stefano Camarda 3
Introduction
PDF uncertainties for the extraction MW from p`T at theTevatron are 9 MeV (CDF) and 11 MeV (D0)
It has been suggested in Eur. Phys. J. C 69 (2010) 379397[Krasny, Dydak, Fayette, Placzek, Siodmok, ’10] that PDFuncertainties in proton-proton collisions could be larger thanin proton-antiproton collisions for
1st quark generation effect: u and d PDF uncertainties on Wboson polarisation2nd quark generation effect: strange-quark PDF uncertaintyon charm-initiated W boson production3rd quark generation effect: bottom quark mass uncertainty inthe extraction of non pertubative parameters from pZT
The idea of this study is to estimate such effects withstandard tools for theory predictions and Monte Carlo, andget feedback from the theory community
Are we missing something important?Do we need better, additional theoretical predictions toestimate these effects?
October 21, 2014 Stefano Camarda 4
Introduction - disclaimer
The emphasis is on tracking the physical sources of theuncertainties, we are not estimating the uncertainties to beused in the measurementThis is not the ATLAS final word on PDF and PSuncertainties for the W mass, and on theory uncertaintiesWe are considering in this study:
u and d valence and sea PDF uncertaintiesStrange PDF uncertaintiesPS uncertainties, assuming they can be extrapolated from themeasurement of pZT to the modelling of pWTDetector effects on the muon momentum resolution
We are not considering in this study:Gluon PDF uncertainties in all-order resummation (gluon PDFis varied only at NLO)Heavy flavour masses in the matrix-element calculationsDifferences in the heavy flavour content of W and Zproduction when propagating PS uncertainties from pZT to pWTAny QED FSR, and NLO EW effectDetector effects on the measurement of the hadronic recoil
October 21, 2014 Stefano Camarda 5
Theory predictions
EW schemeGµ-scheme: with GF , MZ , MW as inputs from PDG 2012, αand θW calculated at tree levelΓZ and ΓW measured value from PDG 2012CKM from PDG 2012, but Vtx = 0, no top in the initial state
Theory predictions and toolsMCFM
W+j production at LO O(αs), which is the real part of Winclusive calculation at NLOfinite width, leptonic decay, spin correlations
CuTe
differential W pT at NNLL with matching corrections atO(αs) (NLO+NNLL)zero width approximationno decay of the W
APPLGRID: Fast PDF convolution
Need to combine the two codes, MCFM and CuTe, to get arealistic prediction of the lepton pT spectrum
October 21, 2014 Stefano Camarda 6
CuTe
Infrared Safety from the Collinear Anomaly [Becher, Neubert,Wilhelm ’11]
The factorisation scale is set to µ = q∗ + qT , withq∗ ∼ 1.82 GeV
The non-perturbative scale q∗ ∼ e−C/αs(mV ) protects theprocesses from receiving large long-distance hadroniccontributions
Allows to calculate the derivative of pW ,ZT for pT → 0 with
perturbative QCD
One additional non perturbative parameter ΛNP = 0.6 GeVintroduce a gaussian smearing for hadronic non-pQCD effects
Public C++ code, very fast
October 21, 2014 Stefano Camarda 7
Theory predictions - benchmark
0 20 40
[p
b/G
eV
] +
W T/d
pσ
d
10
210
310
410
510
610 = 7 TeV s; +
W→pp
NLO (fixed order)CuTe CT10nloMCFM CT10nlo
[GeV] +
W
T p
0 20 40
Ra
tio
0.98
1
1.02 0 20 40
[p
b/G
eV
]
W T/d
pσ
d
10
210
310
410
510
610 = 7 TeV s;
W→pp
NLO (fixed order)CuTe CT10nloMCFM CT10nlo
[GeV]
W
T p
0 20 40
Ra
tio
0.98
1
1.02
Perfect agreement between the two codes, at fixed order O(αs),zero width, no decay.
October 21, 2014 Stefano Camarda 8
Theory predictions - combination
Reweighting function defined as r(pT ) = NLO+NNLLNLO
The reweighing is applied, in the range 0.1 < qT < 150 GeV,outside this range the weight is set to 0
[GeV]+
W
Tp
0 20 40 60 80 100 120 140
[p
b/G
eV
] T
/dp
σd
10
210
310
410
510
+ W→CuTe; pp
NLO (fixed order)NLO+NNLL
[GeV]
W
Tp
0 20 40 60 80 100 120 140
[p
b/G
eV
] T
/dp
σd
10
210
310
410
510
W→CuTe; pp
NLO (fixed order)NLO+NNLL
[GeV]+
W
Tp
0 20 40 60 80 100 120 140
Ra
tio
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6 + W→CuTe; pp
NLONLO+NNLL
[GeV]
W
Tp
0 20 40 60 80 100 120 140
Ra
tio
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
W→CuTe; pp
NLONLO+NNLL
October 21, 2014 Stefano Camarda 9
CKM decomposition of the reweighting function
The NLO+NNLL/NLO ratio has a significant dependence onthe flavour of the quarks initiating the W -boson productionprocess: heavy quarks result in a harder pWT spectrum and aharder ratio between resummed and fixed order predictionsThe reweighting function is decomposed in terms of the CKMmatrix: 6× 2 functions are the NLO+NNLL/NLO ratiosevaluated by setting to 0 all the CKM matrix except the Vxy
term, separately for W+ and W−
[GeV]+
W
Tp
0 20 40 60 80 100 120 140
Rew
eig
hting function
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
+ W→CuTe; pp
udV
usV
ubV
cdV
csV
cbV
[GeV]
W
Tp
0 20 40 60 80 100 120 140
Rew
eig
hting function
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
W→CuTe; pp
udV
usV
ubV
cdV
csV
cbV
October 21, 2014 Stefano Camarda 10
Theory predictions - benchmark
0 20 40
[p
b/G
eV
] +
W T/d
pσ
d
210
310
410 = 7 TeV s;
+ W→pp
NLO+NNLL
CuTeMCFM+CuTe
[GeV] +
W
T p
0 20 40
Ra
tio
0.95
1
1.05 0 20 40
[p
b/G
eV
]
W T/d
pσ
d
210
310
410 = 7 TeV s;
W→pp
NLO+NNLL
CuTeMCFM+CuTe
[GeV]
W
T p
0 20 40
Ra
tio
0.95
1
1.05
Good agreement at NLO+NNLL after reweighting.PDF uncertainties reproduced at high pT , at low pT CuTe haslarger PDF uncertainties.In CuTe the factorisation scale is set to µ = q∗ + qT , withq∗ ∼ 1.82 GeV, → CuTe is sensitive to larger PDF uncertaintiesassociated with the low x region. In MCFM prediction, thefactorisation scale is set to 80 GeV.
October 21, 2014 Stefano Camarda 11
Dedicated PDF set
A dedicated PDF set has been produced to study the PDFuncertainties for MW extraction from plTSimple setup which allows breakdown of uncertainties, notintended for final estimate of PDF uncertainties
NLO Fit to HERA I data
Starting scale Q20 = 1.7 GeV2
charm mass mc = 1.38 GeV
bottom mass mb = 4.75 GeV
top mass mt = 3.5 TeV → 5 flavour
strange fraction rs = s/d = 1
13p parametrisation
26 hessian variations
4 model variations: mc = 1.32, 1.44, rs = 0.72, 1.25
Total of 30 variations
October 21, 2014 Stefano Camarda 12
PDF at the starting scale
x 410
310 210 110 1
)2
(x,Q
V x
u
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
(x,Q
V x
d
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
xg
(x,Q
1
0.5
0
0.5
1
1.5
2
2.5
3
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
(x,Q
Σ x
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
(x,Q
u x
0
0.1
0.2
0.3
0.4
0.5
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
(x,Q
d x
0
0.1
0.2
0.3
0.4
0.5
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
xs(x
,Q
0
0.1
0.2
0.3
0.4
0.5
0.6
2
= 1.7 GeV2QMWNLO exp+modMWNLO exp
x 410
310 210 110 1
)2
)(x,Q
d+
u)/
(s
x(s
+
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
2 = 1.7 GeV2Q
MWNLO exp+modMWNLO exp
Blue band: experimental (hessian) uncertainties
Red band: experimental uncertainties plus model variations, rsand mc
October 21, 2014 Stefano Camarda 13
PDF uncertainties at the scale of MW
x 410
310 210 110 1
ref
)2
(x,Q
V)/
xu
2(x
,QV
xu
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
(x,Q
V)/
xd
2(x
,QV
xd
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
)/xg
(x,Q
2 x
g(x
,Q
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
(x,Q
Σ)/
x2
(x,Q
Σ x
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
(x,Q
u)/
x2
(x,Q
u x
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
(x,Q
d)/
x2
(x,Q
d x
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
)/xs(x
,Q2
xs(x
,Q
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
x 410
310 210 110 1
ref
)2
)(x,Q
d+
u)/
(s
)/x(s
+2
)(x,Q
d+
u)/
(s
x(s
+
0.8
1
1.2
2 = 6464 GeV2Q
MWNLO exp+modMWNLO exp
Valence PDF dominated by experimental uncertainties fromthe HERA I data
u, d and strange PDF dominated by model variations, rs andmc
October 21, 2014 Stefano Camarda 14
Methodology and event selection
Event selection|ηl | < 2.4pνT > 30 GeVMT > 60 GeV30 < plT < 50 GeV
Generate a test sample with MW = 80.385Estimate only statistical experimental uncertainty in 5 fb−1,assuming no efficiency lossConsider normalised plT distribution, in bins of 0.5 GeVCalculate a χ2 profile as a function of MW , in the region±100 MeV (80.286 < MW < 80.484), in steps of 2 MeVFit the χ2 profile with a parabolic function to find minimumand sigma at ∆χ2 = 1Compute the minimum for each PDF variation of a given PDFset, calculate PDF uncertainty as the difference between theminimum for the central PDF and each PDF variationRepeat the analysis for W+ only, W− only and both W+ andW−
October 21, 2014 Stefano Camarda 15
Methodology and event selection - Example of χ2 profile
[MeV]WM80300 80350 80400 80450 80500
2χ
200
0
200
400
600
800
+W
W
and W
+W
5.0±80385.2 5.8±80385.1
3.8±80385.2
The procedure gives us a rough estimate of the statisticaluncertainty expected from the 2011 data set at 7 TeV.Detector resolution effect are not yet included in this plot
The statistic of the MC sample is much higher, the statisticalbias on the central value of MW = 80.385 is below 0.2 MeV
October 21, 2014 Stefano Camarda 16
W polarization and PDF uncertainties
The production of W by u and d quarks at LO, at a given rapidityy can be decomposed in two terms corresponding to λ = +1,−1helicity states:
σW+(y) ∝ u(x1) · d(x2) + d(x1) · u(x2) (1)
σW−(y) ∝ d(x1) · u(x2) + u(x1) · d(x2) (2)
x1,2 = MW√s· e±y
At central rapidity y = 0, x1 = x2, the two terms are equal, →unpolarised W
For y 6= 0, the two terms are different, the W is polarised onaverage
The uncertainty in the u and d valence and sea PDF determinesan uncertainty in the average polarisation of the W , which in turnspropagates into an uncertainty on the plT spectrum.
October 21, 2014 Stefano Camarda 17
W polarization and PDF uncertainties
•
λ = −1
zlab
`+
(WRF)
ν(WRF)
θ∗
•λ = 0
zlab
`+
(WRF)
ν(WRF)
θ∗
•
λ = +1
zlab
`+
(WRF)
ν(WRF)
θ∗
1
)*
θcos(
-1 -0.8-0.6-0.4-0.2 0 0.2 0.4 0.6 0.8 10
0.5
1
1.5
2
2.5
3
3.5
4=-1λ=+1λ
=0λ=-1,0,+1λ=-1,1λ
The λ = +1,−1, and 0 helicity states correspond to differentdistribution of the polar angle θ∗ between the direction of themomentum of the incoming quark in the laboratory frame,assumed to be parallel to the beam axis, and the direction of theleptons momenta in the W -boson rest frame
October 21, 2014 Stefano Camarda 18
W polarization and PDF uncertainties
The uncertainty on the u and d valence and sea PDFtranslates into an uncertainty in the average W polarisation,which affects the plT distribution
Simple setup to disentagle the effect of W polarisation
Keep only Vud 6= 0 and set all the other terms of the CKMmatrix to 0
Apply a random rotation to the decay angle of the leptons inthe W rest frame → No spin correlations
Apply a sign flip to the lepton momentum in W rest frame →λ = ±1 symmetric
Switch on spin correlations and compare PDF uncertainties
Thanks to Paolo Nason for the idea of randomising the leptonmomenta to switch off spin correlations
October 21, 2014 Stefano Camarda 19
W polarization and PDF uncertainties
*) θ cos(1 0.5 0 0.5 1
*) [
pb
]θ
/dco
s(
σ d
1000
2000
3000
4000
5000
6000
7000
= 7 TeV s; +
W→pp
MCFM+CuTe
No spin correlations
1 symmetric± = λ
Spin correlations
*) θ cos(1 0.5 0 0.5 1
*) [
pb
]θ
/dco
s(
σ d
500
1000
1500
2000
2500
3000
3500
4000 = 7 TeV s;
W→pp
MCFM+CuTe
No spin correlations
1 symmetric± = λ
Spin correlations
cos θ∗ distributions in the 3 samples
The dashed bands show PDF uncertainties, only in the samplewith spin correlations the PDF uncertainties affect theaverage W polarisation
October 21, 2014 Stefano Camarda 20
W polarization and PDF uncertainties
30 35 40 45 50
[p
b /
Ge
V]
l T/d
pσ
d⋅ σ
1/
210
110 = 7 TeV s;
+ W→pp
Spin correlationsNo spin correlations
[GeV/c] l
T p
30 35 40 45 50
Ra
tio
0.995
1
1.005 30 35 40 45 50
[p
b /
Ge
V]
l T/d
pσ
d⋅ σ
1/
210
110 = 7 TeV s;
W→pp
Spin correlationsNo spin correlations
[GeV/c] l
T p
30 35 40 45 50
Ra
tio
0.995
1
1.005
Dramatic shrink of the PDF uncertainties on p`T distributionwhen spin correlations are switched off
October 21, 2014 Stefano Camarda 21
W polarization and PDF uncertainties
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80360
80370
80380
80390
80400
80384.8 +3.1 3.0
80386.2 +3.1 3.3
80385 +17 21
No spin correlations+
W
1 symmetric± = λ +
W
Spin correlations+
W
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80365
80370
80375
80380
80385
80390
80395
80400
80405
80384.5 +5.6 5.5
80386.4 +6.8 6.1
80385 +28 27
No spin correlations
W
1 symmetric± = λ
W
Spin correlations
W
Sets 1-26, hessian variations
Set 27 mc = 1.32, Set 28 mc = 1.44
Set 29 rs = 0.72, Set 30 rs = 1.25
∼ 20 MeV effect in W+, mostly due rs variations
∼ 25 MeV effect in W−, spread across eigenvector variations
October 21, 2014 Stefano Camarda 22
W polarization and PDF uncertainties
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80365
80370
80375
80380
80385
80390
80395
80400
80384.7 +3.4 3.5
80386.3 +3.8 3.7
80385 +15 17
No spin correlations±
W
1 symmetric± = λ ±
W
Spin correlations±
W
Since PDF variations are different between W+ and W−,when W+ and W− spectra are used simultaneously, theuncertainty is reduced to ∼ 15 MeV
October 21, 2014 Stefano Camarda 23
W polarization and PDF uncertainties
Why rs variations affects the W -boson polarisation, when onlyu- and d- initiated W production is considered?
x 410
310 210 110 1
ref
)2
)(x
,Qd /
V)/
x(d
2)(
x,Q
d/V
x(d
0.8
1
1.2
2 = 6464 GeV2QMWNLO exp+modMWNLO exp
varsMWNLO r
DIS data constrains the sum of d and strange PDFIn the medium and high x region, variations of the ratio of thestrange-quark PDF over the d PDF, corresponding tors = 0.72 and rs = 1.25, give a significant contribution to thetotal PDF uncertainty of the dV /d ratio.
October 21, 2014 Stefano Camarda 24
Charm quark in the initial state and strange PDF
A charm in the initial state is expected to alter the kinematicof the W productions
Roughly, a kick of the order of mc = 1.4 GeV in the pWTspectrum is expected
Randomise decay angle of the leptons in the W rest frame →unpolarised W
Switch between a setup with only Vud term in the CKMmatrix, and a setup with Vud and Vcs terms
However, the amount of charm initiated W production willalso alter the balance between valence quark initiated and seaquark initiated production, which in turns affect the Wpolarisation and the W plT spectrum
October 21, 2014 Stefano Camarda 25
Charm mass effect
0 20 40
] 1
[G
eV
+W T
/dp
σ d⋅
σ 1
/
310
210
110 = 7 TeV s; +
W→pp
CuTe NLO+NNLL
udV
csV
cs and VudV
[GeV] +
W
T p
0 20 40
Ra
tio
0.60.8
11.2 0 20 40
] 1
[G
eV
W T
/dp
σ d⋅
σ 1
/
310
210
110 = 7 TeV s;
W→pp
CuTe NLO+NNLL
udV
csV
cs and VudV
[GeV]
W
T p
0 20 40
Ra
tio
0.60.8
11.2
As expected, the pWT spectrum acquires a kink fromcharm-initiated production
Notice that the charm is treated as massless in the MEcalculation, the effect of the charm mass we see is encoded inthe PDF, as a threshold between 3 and 4 flavours
October 21, 2014 Stefano Camarda 26
Charm quark in the initial state and strange PDF
30 35 40 45 50
[p
b / G
eV
] l T
/dp
σ d⋅
σ 1
/
210
110 = 7 TeV s;
+ W→pp
cs and VudV
udV
[GeV/c] l
T p
30 35 40 45 50
Ra
tio
0.995
1
1.005 30 35 40 45 50
[p
b / G
eV
] l T
/dp
σ d⋅
σ 1
/
210
110 = 7 TeV s;
W→pp
cs and VudV
udV
[GeV/c] l
T p
30 35 40 45 50
Ra
tio
0.995
1
1.005
Significant shrink of the PDF uncertainties on p`T distributionwhen charm-initiated W production is switched off
October 21, 2014 Stefano Camarda 27
Charm quark in the initial state and strange PDF
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80376
80378
80380
80382
80384
80386
80388
80390
80392
80394
80384.8 +3.1 3.0
80384.9 +8.4 6.9
ud V+
W
cs and Vud V+
W
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80375
80380
80385
80390
80395
80384.5 +5.6 5.5
80384.6 +11.6 9.5
ud V
W
cs and Vud V
W
Sets 1-26, hessian variations
Set 27 mc = 1.32, Set 28 mc = 1.44
Set 29 rs = 0.72, Set 30 rs = 1.25
∼ 5 MeV effect in W+ and W− due to rs variations
October 21, 2014 Stefano Camarda 28
Charm quark in the initial state and strange PDF
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80375
80380
80385
80390
80395
80384.7 +3.4 3.5
80384.8 +9.4 7.7ud V
pmW
cs and Vud V±
W
Since the rs variations are correlated between W+ and W−,there is no significant reduction of the uncertainty when W+
and W− p`T spectra are used simultaneously
October 21, 2014 Stefano Camarda 29
Total PDF uncertainties
Consider altogether polarisation and charm-initiated effects,by using normal spin correlations and SM CKM matrix
Observed a partial cancellation of the two effects
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80375
80380
80385
80390
80395
80385 +12 11
80385 +20 19
80385 +11 10
+W
W
and W
+W
October 21, 2014 Stefano Camarda 30
Detector effects - Muon pT smearing
Implemented detector pT smearing for muons fromarXiv:1404.4562, to asses the impact of detector smearing on PDFuncertainties
October 21, 2014 Stefano Camarda 31
Detector effects - Muon pT smearing
[MeV]Wm80300 80350 80400 80450 80500
2χ
0
100
200
300
400
+W
with detector effects+
W
5.0±80385.2
5.7±80384.9
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80375
80380
80385
80390
80395
80385 +12 11
80385 +13 12
+W
with detector effects+
W
[MeV]Wm80300 80350 80400 80450 80500
2χ
50
0
50
100
150
200
250
300
350
W
with detector effects
W
5.8±80385.1
6.5±80385.2
PDF member
1 6 11
16 21
26
[M
eV
]W
m
80375
80380
80385
80390
80395
80400
80385 +20 19
80385 +22 21
W
with detector effects
W
Small effect, increase of the PDF uncertainties by ∼ 10%
However, detector effects on the hadronic recoil are notconsidered
October 21, 2014 Stefano Camarda 32
Summary of PDF uncertainties
MW-NLO CT10nlo MSTW2008CPdeutnlo NNPDF30 nlo as 118
W+ +13 -12 +18 -22 +11 -10 +8 -10
W− +22 -22 +18 -23 +11 -10 +8 -9
W± +11 -11 +14 -18 +7 -7 +6 -5
CT10nlo scaled to 68 cl
The dedicate PDF set used for the study has a reasonableuncertainty compared to the global PDF set
The larger W− uncertainty is due to missing constraints fromadditional dataset like Tevatron (and LHC) W asymmetry
October 21, 2014 Stefano Camarda 33
Summary of PDF uncertainties
Difference between the input value of mW = 80385 MeV and theextracted value of mW , when using CT10nlo as PDF for thereference p`T spectrum, and another PDF for the test spectra withvarious values of mW
MW-NLO CT10nlo MSTW2008CPdeutnlo NNPDF30 nlo as 118
W+ -9 -0.1 -20 -1.2
W− +48 +0.2 +13 +12
W± +16 0.0 -6 +5
Large, unrealistic shift for the dedicated PDF set MW-NLOfor W− production
Known issue related to the limited number of parameters (13parameters), and not enough constraints from HERA data onthe d valence PDF
October 21, 2014 Stefano Camarda 34
Low pWT modelling uncertainty
Exploit the pZT data to constrain the non perturbative QCDparameters at low pT of Pythia8: Tune AZ
Propagate the data uncertainty to the pWT by mean ofeigentunes hessian uncertainties
[GeV] ZT
p1 10 210
Pre
dict
ion/
Dat
a
0.8
0.9
1
1.1
Data uncertainty
PYTHIA8 4C
PYTHIA8 AZ
ATLAS
-1 Ldt = 4.7 fb∫ = 7 TeV; s
[GeV] ZT
p1 10 210
Pre
dict
ion/
Dat
a
0.8
0.9
1
1.1
Data uncertainty
POWHEG+PYTHIA8 4C
POWHEG+PYTHIA8 AZNLO
ATLAS
-1 Ldt = 4.7 fb∫ = 7 TeV; s
Tune AZ AZNLO 4C
Primordial kT [GeV] 1.71 ± 0.03 1.75 ± 0.03 2.0
αISRs (mZ ) 0.1237 ± 0.0002 0.118 (fixed) 0.137
ISR cut-off [GeV] 0.59 ± 0.08 1.92 ± 0.12 2.0
χ2min/dof 45.4/32 46.0/33 -
October 21, 2014 Stefano Camarda 35
Low pWT modelling uncertainty
[GeV]µ
Tp
30 32 34 36 38 40 42 44 46 48 50
Rat
io to
AZ
0.995
0.996
0.997
0.998
0.999
1
1.001
1.002
1.003
1.004
1.005Central tune
Variation 1+
Variation 2+
Variation 3+
[GeV]µ
Tp
30 32 34 36 38 40 42 44 46 48 50
Rat
io to
AZ
0.995
0.996
0.997
0.998
0.999
1
1.001
1.002
1.003
1.004
1.005Central tune
Variation 1-
Variation 2-
Variation 3-
Tune Variation Positive Negative
1± 3 −42± 3 −43± 3 −4
Total 5 7
October 21, 2014 Stefano Camarda 36
Low pWT modelling uncertainty
[GeV]+
W
Tp
0 10 20 30 40 50
]1
[G
eV
T/d
pσ
d⋅ σ
1/
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07CuTe NLO+NNLL
+ W→pp
udV
usV
ubV
cdV
csV
cbV
[GeV]
W
Tp
0 10 20 30 40 50
]1
[G
eV
T/d
pσ
d⋅ σ
1/
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07CuTe NLO+NNLL
W→pp
udV
usV
ubV
cdV
csV
cbV
[GeV]Z
Tp
0 10 20 30 40 50
]1
[G
eV
T/d
pσ
d⋅ σ
1/
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07 CuTe NLO+NNLL Z→pp
uu
ddss
cc
bb
Heavy flavour production results in different distorsions of theW and Z pT spectrum
Differences in the heavy flavour content of W and Zproduction when propagating PS uncertainties from pZT to pWTare not accounted
October 21, 2014 Stefano Camarda 37
Summary and conclusions
Performed a breakdown of most important PDF uncertaintiesfor the MW extraction from p`T , by tracking the physicalmechanism which give rise to PDF uncertainties
Large contribution of polarisation, of the order of 10− 20MeV, can be improved by better knowledge of u, d valenceand sea PDF
Strange-quark PDF uncertainty due to charm-initiated Wproduction is at the level of 5 MeV, but partially cancel theeffect of strange-quark PDF on polarisation
Preliminary and partial estimation of parton showeruncertainties at the level of 5− 7 MeV
October 21, 2014 Stefano Camarda 38
POINTS FOR DISCUSSION
With inputs from Luca Perozzi and Maria Rosaria D’Alfonso
October 21, 2014 Stefano Camarda 39
Coherent treatment of PDF uncertainties between ATLASand CMS
Various theory predictions can be used to evaluate PDFuncertainties: MCFM, RESBOS, DYRES, POWHEG,[email protected] we expect such tools to provide the same estimation ofPDF uncertainties? Should we agree on which prediction(s)can be safely used?
Various PDF set are available on the market, at least CT10,MSTW, NNPDF, but also ABM, JR, HERAPDF, and recentlyalso ATLAS and CMS PDF sets. Do we want to agree on acommon PDF set? We should at least be sure that ATLASand CMS results can be combined accounting for thecorrelation of PDF uncertainties
Prescription for evaluating PDF uncertainties: single set?envelope as in PDF4LHC prescription? META-PDF?
October 21, 2014 Stefano Camarda 40
How to reduce PDF uncertainties for mW
Is the measurement of mW at different collider energies, 7, 8,13 TeV, beneficial for the reduction of PDF uncertainties?
Which LHC measurements can further constrain the PDFuncertainties for mW ?
October 21, 2014 Stefano Camarda 41
BACKUP
October 21, 2014 Stefano Camarda 42
Choice of ptsqmin 4.0 in POWHEG+PYTHIA8
ptsqmin 0.2ptsqmin 0.8ptsqmin 2.0
0 1 2 3 4 50
10
20
30
40
50
60
70
80
PZ
T [GeV]
dσfid./dPZ T[pb/GeV
]
October 21, 2014 Stefano Camarda 43