You are on page 1of 31

Expected limits for

HH to bbZZ to 2b 2l 2nu
Rami Kamalieddin, Ilya Kravchenko (UNL)
Michele de Gruttola (CERN)
Lesya Shchutska (ETH)

UNLHEP meeting, 8 Aug 2017


2

OUTLINE
Updated theory introduction
Analysis strategy and goals
Analysis overview and updates
Mass boundary for low/high mass splitting
Optimization of selection BDT cut
Distributions of BDT and HH_mt in SR and CRs
Datacard with some systematics
From r to limits
Conclusions
Next steps

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


3

Updated theory introduction


Current HH searches focus in two areas:

BSM resonant production:


- Many different BSM models predict HH
resonances
- Solution for analysts: work on model independent results
on narrow width resonances
- Two classes of particles: spin-0 (Radion) and spin-2 (Graviton)
- Mass range 250 -1000 GeV

From R. de Lima

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


4

Theory, cont.

Current HH searches focus in two areas:

BSM resonant production


non-resonant production:
- HH is the only direct test of the Higgs self-coupling!
- SM and BSM non-resonant production by introducing
6th order EFT terms in the Lagrangian and
testing the standard BSM benchmarks for that

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


5

Theory, cont. 2
Current HH searches focus in two areas:

BSM resonant production


non-resonant production

If k equals 10, then BSM is 10 times bigger than SM!


Meaning will have a discovery sooner than one reaches the
sensitivity to the SM values (HHSM= 33.45fb @ 13 TeV)

We are interested in all HH activities J SM

From A. Carvalho

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


6

Analysis strategy

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


7

Recap and final decision

Object Select Prepare input Limits


selection Candidate to Combine

Requirements BDT with nine Simple shape Derive the


on leptons, jets variables of HH_mt best limits
and met we can

Our objective is:


Set 95% CL limit on (pp X hh) BR(hh bbZZ bbllmet ) vs mass of X
(where X is the BulkGraviton as the first analysis target)

The method we will use to achieve it:


Binned shape analysis using Higgs Combination Tool

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


8

Analysis goals
We will work on spin-2 resonance first

At the advanced stage (~pre-approval level at least), we will consider


adding spin-0 and non-resonant case

Already preparing for the future: producing non-res samples, discussing


the methodology, strategy, etc

HH group will be making grand combination with the paper in mind:


- first non-res combination of matured and approved analyses
- then combination of resonant results
- we aim to be in!

We already have AN-17-198:


- Will be working on the documentation throughout the analysis timeline
- For the pre-approval we may need to write a PAS (as of now, may
change with the new convener)

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


9

Analysis overview

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


10

2b 2l 2 signature
MET from the off shell Z* boson

H ZZ*

2 leptons (eles, muons) from on shell Z

H bb
2 b-jets from the Higgs bb

In our analysis we have HH decay through:


H to bb to give the highest BR
H to ZZ*, with Z to ll to give important handles,
which improves sensitivity
* Pic based on the plot from bbWW team
Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017
11

Recent analysis updates

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


12

Analysis updates
Solved the issue of orthogonality with bbWW, inverting Zll cut and keeping only
the upper side-band

Need to update many parts of the analysis to the most recent POG
recommendations/corrections available:
- regressed jet quantities to shrink the resolution and decrease scale bias
- most recent lepton SFs
- maybe CMVAv2 for b-tagging

We got large samples of all types and currently running on them:


- crab3 is almost working (95% success rate J)
- possibly one wrong non-res sample, we spot it fast, but production people are on vacation untill Aug 20

Will have to train new MVA discriminants (perhaps again BDT) using new big
statistics samples

Incorporate into the framework shape uncertainties

Obtain limits_v2 and start working on the combination with the other channels

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


13

Mass boundary for low/high mass


splitting of the 250 1000 GeV range

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


14

Test of several boundary option


To increase low statistics/group similar samples, we stitch signal samples
with the weight 1, separating 2501000 GeV range into low/high mass regions.

Then we test four mass splitting cases:

Mass boundary low mass region ROC high mass region ROC

300 GeV 0.9138 0.9805


350 GeV 0.9037 0.988
400 GeV 0.9074 0.9928
450 GeV 0.9086 0.9957

We chose the fourth option.

For the analysis, doing the low/high


mass BDT training, we use the same
set of nine variables:

Rami Kamalieddin (UNL) HHbbZZ2b2l2 8 Aug 2017


15

Optimization of the BDT cut

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


16

Limit vs cut on BDT


For the 1st version of the expected limits, we pick 260 and 400 GeV
samples in the low mass region
For each BDT cut, produce HH_mt shape and derive expected limit.
Scan the whole BDT range to find the best value of the limit.
bbVV, Xsec 1pb, mass = 260GeV bbVV, Xsec 1pb, mass = 400GeV
45
'r' from 'Combine'

'r' from 'Combine'


10
40

35
8

30

25 6

20
4
15

10
2
5

0 0
1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8
Cut on BDT value Cut on BDT value

After cut at 0.5, limit growths for 260 GeV case with BDT cut value
For simplicity, for both low mass region samples we pick cut at 0.5
Of course, when new samples arrive, will optimize BDT cut for
each mass point separately
Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017
17

Limit vs cut on BDT, cont.


In the high mass region, we pick 650 and 900 GeV samples:

bbVV, Xsec 1pb, mass = 600GeV bbVV, Xsec 1pb, mass = 900GeV
'r' from 'Combine'

'r' from 'Combine'


10 10

8 8

6 6

4 4

2 2

0 0
1 0.5 0 0.5 1 1 0.5 0 0.5 1
Cut on BDT value Cut on BDT value

Both samples show the best limit at the BDT cut of 0.95

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


18

Distributions of BDT and HH_mt for SR/CRs

Hbb mass
150
90

76 106 Zll mass

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


19

Distributions of BDT and HH_mt, 400 GeV

SR CRDY CRTT

Low statistics in CRs, need bigger samples to draw any conclusions


Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017
20

Datacard with some systematics

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


21

400 GeV, signal region

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


22

From r to limits

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


23

Derivation of the expected limits

As of now, do not split bbVV into ZZ and WW components,


keep both and manipulate with cuts to be orthogonal to bbWW

For bbVV signal samples we used xsec of 1pb

We run simple command:

combine -M Asymptotic -t -1 m massValue --run blind comb_card_massValue.txt

Obtained value of r we multiply by 1000 to get 95% CL limit on (pp X hh)


BR(hh bbVV bbllmet ) in fb, with the emphasis on the ZZ selection

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


24

Fit results from combine

Mass DY_norm TT_norm

260 GeV 1.0295 +/- 0.00492473 1.03318 +/- 0.0136259

400 GeV 1.03649 +/- 0.00479558 1.0215 +/- 0.0130207

650 GeV 1.04855 +/- 0.00586312 1.05314 +/- 0.0207581

900 GeV 1.04189 +/- 0.00506078 1.04724 +/- 0.0180156

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


25

Expected limits

260 GeV 400 GeV 650 GeV 900 GeV


Expected 2.5%: r < 1668.5 Expected 2.5%: r < 1262.2 Expected 2.5%: r < 128.4 Expected 2.5%: r < 60.8
Expected 16.0%: r < 2237.8 Expected 16.0%: r < 1685.2 Expected 16.0%: r < 174.1 Expected 16.0%: r < 82.7
Expected 50.0%: r < 3140.6 Expected 50.0%: r < 2367.2 Expected 50.0%: r < 249.0 Expected 50.0%: r < 120.6
Expected 84.0%: r < 4480.1 Expected 84.0%: r < 3339.1 Expected 84.0%: r < 365.2 Expected 84.0%: r < 178.8
Expected 97.5%: r < 6163.5 Expected 97.5%: r < 4560.1 Expected 97.5%: r < 513.7 Expected 97.5%: r < 255.9

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


26

Possible improvements
+ Regressed jet quantities

+ Add EGamma data

+ New more powerful MVA discriminant

Some shape uncertainties and corrections


can eat up the limit

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


27

Conclusions
Analysis strategy is finalized

Optimization of the BDT cut is shown

Machinery for the derivation of the limits is in place:


suggestions to improve limits will be be taken into account
missing shape systematics will be added
reoptimization will be done once new big samples arrive

Will use new specific big bbZZ signal samples

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


28

Double Higgs is not easy,


but it is doable!
If it sounds complicated, you are not the only one

...the screenshot that I


took from the real HH talk!

Even Joe Incandela looks


double confused...

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


29

BACK UP for
HH bbZZ 2b 2l 2nu

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


30

Loose selection for BDT training

Apply almost no cuts to give BDT as much info as possible.


Few cuts though had to be applied to avoid NaNs

HH common sense preselection:


Dilepton (ee/) mass > 50 GeV
2 Jets: pt > 30 GeV and || < 2.4

75 GeV < Hbb mass < 175 GeV

56 GeV < Dilepton mass < 126

Transverse mass of HH > 100 GeV

Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017


31

HH candidate and SR/CRs


Loose selection for BDT (previous slide)
Cut on BDT 0.5 or 0.95 (signal sample or > 450 GeV)

SR: Dilepton mass = [76, 106], Hbb mass = [90, 150]


CRDY: Dilepton mass = [76, 106], Hbb mass = ]90-150[
CRTT: Dilepton mass >106, Hbb mass = [90, 150]
Hbb
150
90

76 106 Zll mass


Rami Kamalieddin (UNL) HHbbZZ2b2l2 UNLHEP meeting 8 Aug 2017

You might also like