optimizing the w resonance in dijet mass

Optimizing the W resonance in dijet mass

Daniel AbercrombiePennsylvania State University

8 August 2013

Advisors: Phil Harris and Andreas Hinzmann

The Goal of the Project• Compare jet cone sizes and algorithms

• Identify the algorithm and parameters that givesa stable W mass and narrowest resonance

• Results will be used in talks with ATLAS to determine a common set of parameters for jet reconstruction between the experiments

Daniel Abercrombie 2

The Event

Daniel Abercrombie 3

Characterizing the W peak

Searching for stable mean and smallest fractional width

Daniel Abercrombie 4

200 GeV < pT < 225 GeV

Comparing cone sizes

Daniel Abercrombie 5

• Using the anti-kT algorithm gives the most conic shape and is resistant to soft radiation

• Scanned through cone sizes from ΔR = 0.4 to ΔR = 0.8 with a resolution of 0.1

Comparing cone sizes

Daniel Abercrombie 6

• Jump in larger cones probably due pT cut for single jets

Comparing cone sizes

Daniel Abercrombie 7

• ΔR = 0.4 gives narrowest width

Comparing cone sizes

Daniel Abercrombie 8

• Reasonably constant responses from each cone size

Comparing cone sizes

Daniel Abercrombie 9

• Again, ΔR = 0.4 gives the narrowest width

Comparing cone sizes

Daniel Abercrombie 10

• Again, ΔR = 0.4 gives the narrowest width

Comparing algorithms

Daniel Abercrombie 11

Comparing algorithms

Daniel Abercrombie 12

• Grooming keeps mass relatively constant compared to anti-kT

ΔR = 0.5

Comparing algorithms

Daniel Abercrombie 13

ΔR = 0.5

• Trimming and filtering compete for best resolution

Comparing algorithms

Daniel Abercrombie 14

• Pruning may be too aggressive at low pileup

ΔR = 0.5

Comparing algorithms

Daniel Abercrombie 15

ΔR = 0.5

• Trimming and filtering compete for best resolution

Conclusions• Smaller cone sizes give the best mass resolution with

a reasonably small response

• Pruning looks like it might be too aggressive

• Current plots should be improved by finding ways to increase the efficiency of picking the correct jets

Daniel Abercrombie 16

Future work• Explore additional parameter space of the algorithms

• Look at the effects of jet reconstruction onthe top quark mass

• Work on selection cuts and parameters to increase the efficiency of selecting the correct jet

Daniel Abercrombie 17

Thank you!

Daniel Abercrombie 18

Thank you!

Daniel Abercrombie 19

Backup Slides

Daniel Abercrombie 20

Selection criteria jets• Events must have at least two b tagged jets

and one isolated muon with pT > 10 GeV and |η| < 2.4

• Two jets with pT > 20 GeV and the highest combined secondary vertex values were selected as the b jets

• Other jets were in the opposite hemisphere from the muon, MET, and b tagged jet closer to the muon

i.e.

Daniel Abercrombie 21

Selection criteria jets (cont.)

• Single jets were picked with the following cuts:p > 200 GeV; mass > 60 GeV; MET > 30 GeV– MET cut helps ensure boosted tops

• If there were no single jets, the dijet system with the highest pT jets with a invariant mass of 30 GeV < m < 250 GeV is picked

Daniel Abercrombie 22

Comparing algorithms

Daniel Abercrombie 23

• Pruningtight: nsubjets=2, zcut=0.1, dcut factor=0.5, algo = CAloose: nsubjets=2, zcut=0.1, dcut factor=0.2, algo = CA

• Filteringtight: rfilt=0.2, nfilt=3, algo = CA loose: rfilt=0.3, nfilt=3, algo = CA

• Trimmingtight: rtrim=0.2, pTfrac=0.05, algo = CA loose: rtrim=0.2, pTfrac=0.03, algo = CA

Other measures of efficiency

Daniel Abercrombie 24

ΔR = 0.5

• All of the lines for each algorithm fall well withinthe uncertainties

Other measures of efficiency

Daniel Abercrombie 25

ΔR = 0.5

• All of the lines for each algorithm fall well withinthe uncertainties

Effects of PU

Daniel Abercrombie 26

ΔR = 0.4

• Pileup decreases efficiency• This is more prominent using larger cone sizes

Effects of PU

Daniel Abercrombie 27

ΔR = 0.5

• Pileup decreases efficiency• This is more prominent using larger cone sizes

Effects of PU

Daniel Abercrombie 28

ΔR = 0.7

• Pileup decreases efficiency• This is more prominent using larger cone sizes

Effects of PU

Daniel Abercrombie 29

ΔR = 0.9

• Pileup decreases efficiency• This is more prominent using larger cone sizes

PU jets simulation

Daniel Abercrombie 30

𝑑𝜎𝑑𝑝𝑇

∝𝑝𝑇❑− 5 ;𝑝𝑇>3GeV

𝑑𝜎𝑑𝑝𝑇

=𝑚𝑝𝑇+𝑏 ;0GeV<𝑝𝑇<3GeV

Weighting:

𝑤 (𝑁𝑃𝑈 ,𝑛 𝑗𝑒𝑡𝑠 )= 𝑁𝑃𝑈 !(𝑁𝑃𝑈−𝑛 𝑗𝑒𝑡𝑠 ) !𝑛 𝑗𝑒𝑡𝑠 !

(0.0125 )𝑛 𝑗𝑒𝑡𝑠 (0.9875 )𝑁𝑃𝑈 −𝑛 𝑗𝑒𝑡𝑠

𝐴 𝑗𝑒𝑡

𝐴𝐶𝑀𝑆≈0.0125

PU jets simulation

Daniel Abercrombie 31

NPU = 10

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 32

NPU = 15

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 33

NPU = 20

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 34

NPU = 25

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 35

NPU = 30

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 36

NPU = 35

• Everything above 20 GeV can be mistakenfor a quark jet

PU jets simulation

Daniel Abercrombie 37

NPU = 40

• Everything above 20 GeV can be mistakenfor a quark jet

Daniel Abercrombie 38

Daniel Abercrombie 39

Daniel Abercrombie 40

Daniel Abercrombie 41

Daniel Abercrombie 42

ΔR = 0.3

Daniel Abercrombie 43

ΔR = 0.4

Daniel Abercrombie 44

ΔR = 0.5

Daniel Abercrombie 45

ΔR = 0.6

Daniel Abercrombie 46

ΔR = 0.7

Daniel Abercrombie 47

ΔR = 0.8

Daniel Abercrombie 48

ΔR = 0.9

Daniel Abercrombie 49

ΔR = 1.0

Daniel Abercrombie 50

ΔR = 0.7

175 GeV < pT < 200 GeV

Daniel Abercrombie 51

ΔR = 0.7

200 GeV < pT < 225 GeV

Daniel Abercrombie 52

ΔR = 0.7

225 GeV < pT < 250 GeV

Daniel Abercrombie 53

ΔR = 0.7

250 GeV < pT < 275 GeV

Daniel Abercrombie 54

ΔR = 0.7

275 GeV < pT < 300 GeV

Daniel Abercrombie 55

Daniel Abercrombie 56

Daniel Abercrombie 57

Daniel Abercrombie 58

Cacciari, M., et al. JHEP04(2008)063

Daniel Abercrombie 59

Comparing algorithms

Daniel Abercrombie 60

ΔR = 0.5

• Grooming keeps mass relatively constant compared to anti-kT

Comparing algorithms

Daniel Abercrombie 61

ΔR = 0.5

• Anti-kT seems to have the smallest width

Comparing algorithms

Daniel Abercrombie 62

ΔR = 0.5

• Pruning may be too aggressive at low pileup

Comparing algorithms

Daniel Abercrombie 63

ΔR = 0.5

• Again, anti-kT has narrowest width

Daniel Abercrombie 64

Top Mass

Daniel Abercrombie 65

Top Mass Width