W mass Oxford - University of Oxford · Oxford Jan 2017 gigi.rolandi@cern.ch 1 ... W Pt spectrum....

Post on 15-Jul-2018

247 views 0 download

transcript

gigi.rolandi@cern.ch Oxford Jan 2017 1

Precise measurement of W mass at LHC with the CMS detector

M. Bachtis*, M.R. d’ Alfonso, L. Perrozzi +GR+… CERN

O. Cerri, N. Foppiani, E. Manca, A. Stacchiotti Students @ Scuola Normale - Pisa

* now UCLA

arXi

v:16

08.0

1509

GeV

gigi.rolandi@cern.ch Oxford Jan 2017 3

Measurement of the TOP quark mass

Classical method

gigi.rolandi@cern.ch Oxford Jan 2017 4

Top mass

Alternative methods

gigi.rolandi@cern.ch Oxford Jan 2017

5

Past (and present) W mass measurements

mW = 80.370 ± 19 MeV

New ATLAS Measurement december 2016

8-98-105-3

9-14 2-13

6-12

ATLAS

11

7

14

7

ATLA

S N

umbe

rs

Gig

i’s B

reak

dow

n

gigi.rolandi@cern.ch Oxford Jan 2017 6

Measure μ pT : you need the prediction of the W pt distribution to measure the W mass.

Observable sensitive to Mw : muon transverse momentum

Mean 0.01145± 10.56

Std Dev 0.008099± 7.433

W Pt [GeV]0 5 10 15 20 25 30

Nor

mal

ized

ent

ries

/ 0.5

GeV

0.005

0.01

0.015

0.02

0.025

0.03 Mean 0.01145± 10.56

Std Dev 0.008099± 7.433

W Pt spectrum

gigi.rolandi@cern.ch Oxford Jan 2017 7

How to extract the W mass information from a distribution ?

1- Compare one measured distribution (eg pT of the muon) with several simulated distributions generated with different mass assumptions and compute the likelihood ratio.

100 MeVThis technique implies a perfect mastering of the simulation : W production and measurement of the event properties (muon, recoil)

gigi.rolandi@cern.ch Oxford Jan 2017 8

Transverse mass: in principle a more robuste observableIn these formulae p== pT

The jacobian edge preserves the mass information

sensitive to production and

decay dynamics

PT

ηcm

gigi.rolandi@cern.ch Oxford Jan 2017 9

Small distortions due to ISR and muon resolution

Plot done with ideal measurement of the hadronic recoil

gigi.rolandi@cern.ch Oxford Jan 2017 10

…. however it is difficult to measure precisely the hadronic recoil !

With a resolution of 7 GeV the edge is lost. Still very small dependence on the Wpt distribution

Resolution in CDF ~6-8 GeV Resolution in ATLAS ~ 13 GeV

/55CERNSep. 2013 gigi.rolandi@cern.ch

11

muon pt ~ 39 GeV

W mass @ CMS

gigi.rolandi@cern.ch Oxford Jan 2017 12

All silicon tracker ~ 1500 pixel modules and 15000 strip modules providing ~ 14 measurements ~ 10-50 μm per track in a magnetic field of 3.8 T

gigi.rolandi@cern.ch Oxford Jan 2017 13

Tracker Resolution

gigi.rolandi@cern.ch Oxford Jan 2017 14

W mass measurement strategy

Calibration of the muon momentum scale

Calibration of the hadronic recoil

Precise definition of the production model

J/Psi - Y samples

Z samples

Ancillary measurements

gigi.rolandi@cern.ch Oxford Jan 2017 15

CMS Statistics on tape

Lumi(1/)) Pileup J/Psi Y Z W+ W-

7TeV 5 7 3.5 1 1.4 13 9

8TeV 22 22 5.3 3 6.2 61 45

13TeV 4+35 13+24 122 49 39 ~350 ~250

Using 8 TeV only the statistical error is <2 MeV

Million Events

gigi.rolandi@cern.ch Oxford Jan 2017 16

Calibration of the momentum scale

The 16500 modules of the tracker are aligned with collision data and cosmic rays. Local precision 2 μm thanks to overlaps of nearby modules. Alignment has weak modes - geometry is prone to global scale deformations

Tracker in a very uniform magnetic field Map measured when CMS was the surface assembly hall. 2D approx. map used in the reconstruction

Momentum biassed by Energy Loss How well do we know the tracker material ?

Kc=AK+ε sinθ K2 + qM

B field and radial length

K=curvature=1/p

material effects transverse alignment10 η bins 10 η bins20 η bins

gigi.rolandi@cern.ch Oxford Jan 2017 17

Fitting dimuon resonances: kalman filter with target mass mc = dimuon generator mass after FSR

Magnetic field in data about 0.05 % higher than in simulation expected because of more iron in the cavern

x 500

Radial deformations < 0.1% level

Michalis Bachtis

gigi.rolandi@cern.ch Oxford Jan 2017 18

Material effects < 10 MeV

Transverse alignment terms ~ few 10-5 GeV-1

Calibration performed only with J/ψ and ΥZ used in the closure plots

gigi.rolandi@cern.ch Oxford Jan 2017 19

Fit correlations

70% correlation between material terms and B-field terms

gigi.rolandi@cern.ch Oxford Jan 2017 20

In situ measurement of the Material Budget

Using ~ 1-2 GeV tracks compute “local sagitta s” using measurements in three consecutive layers

Elisabetta Manca

The variance measures the multiple scattering

2 out of the 39 measured tracker layers

100 s of data taking

gigi.rolandi@cern.ch Oxford Jan 2017 21

Calibration of the recoil

Montecarlo study : which particles to use to measure the W recoil ?

At fix muon momentum plot the component of the recoil parallel to the muon direction

+……

gigi.rolandi@cern.ch Oxford Jan 2017 22

gigi.rolandi@cern.ch Oxford Jan 2017 23

Hadronic recoil, definition and calibration

Ideal detector, no pileup tracks + clusters

Real detector, pileup tracks+clusters

Ideal detector, no pileup only charged tracks

Real detector, pileup only charged tracks from muon vertex

Maria Rosaria d’ Alfonso

gigi.rolandi@cern.ch Oxford Jan 2017 24

Hadronic recoil calibrated with Z μμ events

Good closure of the recoil

gigi.rolandi@cern.ch Oxford Jan 2017 25

Hadronic recoil calibrated with Z μμ events

Response ~ 50% Resolution ~ 10 GeV

Response CDF ~ 65%, ATLAS ~ 90%

Tracks

gigi.rolandi@cern.ch Oxford Jan 2017 26

March 2016

gigi.rolandi@cern.ch Oxford Jan 2017 27

Measurement of the Z mass “as W” CMS-SMP-14-007

Production model addressed reweighing the Z MC to Z data: This measurement is just a test of the experimental calibrations

gigi.rolandi@cern.ch Oxford Jan 2017 28

180 k Z , stat 36 MeV, muons scale 13 MeV , recoil 10 MeV

CDF 620 k W , stat 16 MeV, muons scale 7 MeV , recoil 8 MeV

gigi.rolandi@cern.ch Oxford Jan 2017 29

More on recoil

Nicolo Foppiani Olmo CerriExploiting correlations

Different muon pT bins

gigi.rolandi@cern.ch Oxford Jan 2017 30

1st results on MC resolution from 10 GeV —> 8 GeV

gigi.rolandi@cern.ch Oxford Jan 2017 31

Modeling W production

gigi.rolandi@cern.ch Oxford Jan 2017 32

Effect of the PDF uncertainties on W mass

PT

ηcm

ηcm

ηL

YWpT>30 GeV ~ |ηcm | <0.7

At a given ηL you sample different Y corresponding to different pT changing average pT . And also W polarization effects

PT cut

gigi.rolandi@cern.ch Oxford Jan 2017 33

PDF Effects

At LHC, differently from Tevatron, W+ and W- have different cross sections

and have ~ equal rate because at leading order the s and bar(s) content of the proton is the same.

Subtracting W+ and W- distributions selects “valence” “sea” combinations and the PDF’s of c and s quarks enter only at “Cabibbo suppressed” level.

ATLAS

gigi.rolandi@cern.ch Oxford Jan 2017 34

PDF Mitigation

Fitting transverse mass in different η bins helps in reducing PDF systematic error.

Fit the μ pT-η rate with fine binning together with the W mass

gigi.rolandi@cern.ch Oxford Jan 2017 35

W production modeling

gigi.rolandi@cern.ch Oxford Jan 2017 36

Angular coefficients

Perform the measurement at low pT ?ATLAS use fixed order NNLO prediction validating the model with Z measurement resulting in 5 MeV error.

gigi.rolandi@cern.ch Oxford Jan 2017 37

V Transverse Momentum Distribution

Fit prediction to Z data and apply to W

ATLAS considered Powheg+Pytia8 and Pytia8 standalone

After tuning 1% agreementprimordial kt αISR ISR cut-off

Exported to W

gigi.rolandi@cern.ch Oxford Jan 2017 38

Collinear gluon emission, factorization scale and quark masses

Gluon radiation has a cutoff depending on the quark mass

Depending how you vary the scale (W/Z correlations) , you get different results

gigi.rolandi@cern.ch Oxford Jan 2017 39

Resummation codes predict an harder pT W spectrum for a given measured pT Z spectrum

gigi.rolandi@cern.ch Oxford Jan 2017 40

Conclusions

CMS has calibrated the muon momentum scale to 0.02 % and there is room for further improvements

Measuring the W pt using charged tracks is possible and reasonably good resolution can be achieved. This allows to exploit the large luminosity collected at high pileup

The (non) agreement of the Z (W) pt spectrum with more advanced calculations based on resummation must be understood

We welcome very much the ATLAS measurement