Unlocking the Structure of New Physics at the LHC · Structure of New Physics at the LHC Natalia...

Unlocking the Structure of New Physics

at the LHCNatalia Toro

hep-ph/0703088: Arkani-Hamed et al0810.3921: Alwall, Schuster, NTwork in progress: UCSB CMS group (special thanks: S.A. Koay)

Hadron Collider 101x1E1 x2E2

A

B

x1, x2: fraction of beam energy carried by each parton

d!inc

dVars=

!dx1

x1

dx2

x2x1fg(x1, Q)x2fq(x2, Q)

d!(qg ! AB)dVars

Hadron Collider 101x1E1 x2E2

A

B

x1, x2: fraction of beam energy carried by each parton

d!inc

dVars=

!dx1

x1

dx2


d!(qg ! AB)dVars

CM-frame boost ⇒multi-particle Lorentz invariants and pT’s

!

parton E2cm

CM boost

=ds

sdy

uu

gg

uu/ug

y (CM boost)

arbit. scale

gg

uu/ug

uu

y (CM boost)

arbit. scale

(300 GeV)(1 TeV)

Multi-Particle Mass Invariants

To construct invariants, must pair/group particles. To pair, must know decay topology. Not known a priori.

What can be learned from simpler pT’s? (and lower statistics)

Edge/endpoint:

q !02 !0

1!

q ! !Full reconstruction and mT2:

Many more variables:– precision mass measurement at hadron colliders!

...100 fb-1

& counts are search variables → understood early. They suffice to build good hypotheses for mass spectra, cascades, then isolate decay modes for precision mass measurement.

Using Transverse MomentaUseful combinations: HT =

!|pT |, !ET =

!!pT

1) HT bump ~ 1–2 x produced particle mass:

2) Locations of pT bumps ~ relative mass scales

(depends on decay chain, LSP mass)

pT , HT , !ET

[picture]

Lepton pT Leading jet pT

+lepton

+jet

HT for Models with M=650–700 GeV

(after cuts)

HT (GeV)

Outline1. Hadron Collider Observables and Ambiguities

- Goal: “Basis of Parameters” for new physics to model most relevant observables and address (subset of) theoretical questions.

- pT in Pair Production (mostly independent of M.E!)

- pT’s and counts insensitive to complex decay chains

2. Designing Robust and General New-Physics Searches(results from UCSB CMS group)

3. Building up from very simple description of new physics

d!inc

dVars=

!dx1

x1

dx2


d!(qg ! AB)dVars

pT Distributions

! !!PDF’s parton cross-section

→parton luminosity

Simple and instructive to calculate pT distribution for 2→2 product with general matrix element:

parton E2cm CM boost

=ds

sdy

s0 = 2M2( : threshold s)

!

!dy

s20

d!

dtds=

1s

s20

s2"(s, Q2)

!s2 d!

dt

"

(q~1–1.5)!(s, Q2) ! (s/Stot)!q

uu

uguu

gg

! = s/Stot

!(s/Stot, Q2)

Q2 = (500 GeV)2

(1 TeV)(300 GeV)

! (1" x)px!q

18!|M(s, t)|2

t = !12

[s(1! !)! s0]

CM-frame Lorentz invariants: or or

pT Distributions

s & t s & p2T

related by:

“pure angular” variable linearly related to → good variable for M.E. expansion

s & !

! ! " cos #CM :

p2T =

tu!M4

s

1s

! dp2T ds = !dtds

s20

d!

dtds=

s20

s2"(s, Q2)|M|2 !(s, s0) ! A(s/Stot)!q

t = !12

[s(1! !)! s0]


pT Distributions

s & t s & p2T

related by:


s & !

! ! " cos #CM :

p2T =

tu!M4

s

s20

d!

dp2T

=1"

!ds

!1!

ds

s

! dp2T ds = !dtds

s0 + 4p2T s0 + 4p2

T

StotStot

s20

d!

dtds=

s20

s2"(s, Q2)|M|2 !(s, s0) ! A(s/Stot)!q

t = !12

[s(1! !)! s0]


pT Distributions

s & t s & p2T

related by:


Expand near threshold (usually dominated by low m, n)

s & !

! ! " cos #CM :

p2T =

tu!M4

s

s20

d!

dp2T

=1"

!ds

!1!

ds

s

|M|2 =!

Cm,n(s/s0)m!n

! dp2T ds = !dtds

s0 + 4p2T s0 + 4p2

T

StotStot

s20

d!

dtds=

s20

s2"(s, Q2)|M|2 !(s, s0) ! A(s/Stot)!q

s0 + 4p2T

Stot

s20

d!

dp2T

=!

s0

Stot

"!q #

m,n

Cm,n

$ds

"s(s/s0)m!q!2"n

t = !12

[s(1! !)! s0]


pT Distributions

s & t s & p2T

related by:



s & !

! ! " cos #CM :

p2T =

tu!M4

s

s20

d!

dp2T

=1"

!ds

!1!

ds

s

|M|2 =!

Cm,n(s/s0)m!n

! dp2T ds = !dtds

s0 + 4p2T s0 + 4p2

T

StotStot

s20

d!

dtds=

s20

s2"(s, Q2)|M|2 !(s, s0) ! A(s/Stot)!q

s0 + 4p2T

Stot

s20

d!

dp2T

=!

s0

Stot

"!q #

m,n

Cm,n

$ds

"s(s/s0)m!q!2"n

s/s0 =1 + 4p2

T /s0

1! !2

0

! 1

=!

s0

Stot

"!q #

m,n

Cm,n

$2d!

1! !2(1! !2)!m+q+2!n " (1 + 4p2

T /s0)m!q!2

t = !12

[s(1! !)! s0]


pT Distributions

s & t s & p2T

related by:



s & !

! ! " cos #CM :

p2T =

tu!M4

s

s20

d!

dp2T

=1"

!ds

!1!

ds

s

|M|2 =!

Cm,n(s/s0)m!n

! dp2T ds = !dtds

s0 + 4p2T s0 + 4p2

T

StotStot

s20

d!

dtds=

s20

s2"(s, Q2)|M|2 !(s, s0) ! A(s/Stot)!q

s0 + 4p2T

Stot

s20

d!

dp2T

=!

s0

Stot

"!q #

m,n

Cm,n

$ds

"s(s/s0)m!q!2"n

shape independent of n

!

Euler B-function

s/s0 =1 + 4p2

T /s0

1! !2

0

! 1

=!

s0

Stot

"!q #

m,n

Cm,n

$2d!

1! !2(1! !2)!m+q+2!n " (1 + 4p2

T /s0)m!q!2

pT Universality

“Shape invariance” Arkani-Hamed et al, hep-ph/0703....

pT variables are useful because they are simple, single-particle Lorentz invariants and insensitive to production matrix element!

• Not completely universal- Depends on m (different for p-wave and contact operators)- Depends on q (sensitive to init. state)- Observable pT’s depend on decay M.E.

• But easy to get similar effects (after cuts) by changing s0 – simple analysis can’t distinguish

• Similarly, η distribution indep. of m – even different n convolved with y distribution have similar shape

|M|2 ! (s/s0)m!n, "(s) ! s!qford!

dp2T

! (1 + p2T /M2)m!q!2

Typical pT~0.5 M

–

Why bother?• Shape invariance: a clear guide to information that

can be stripped out & still do meaningful analysis

• Why understand these?- Important (approximate) ambiguities to be aware of in

any description of positive signal at LHC

- Allows predictions, MC generation, simulation of detector response w/o full knowledge of model Lagrangian

- Suggest search/interpretation strategies with wide reach compared to no. of parameters

How much do you need to say about model to predict LHC signals?

• Masses and quantum numbers of produced particles• Production cross-sections (and near-threshold behavior)• Branching fractions to different final states• To predict invariant mass distributions, also need to

know intermediate spins.

Specialize to models like SUSY – pair production, no fully-reconstructed decays

First three: On-Shell Effective Theory – hep-ph/0703088

Much less detail than full Lagrangian – but even at this level data can be ambiguous...

Squarks + Gluino ExampleM

ass

!01

!02

+2j+2j

+jet

g

qL,R

!qq,!qg ! !gg

!01

!02

+2j+2j

+jetg

qL,R

mq !mg " mg

!

can ignore squarks

mq !mg " mg

!

can ignore squarks

jet from squark decay very soft

Extreme spectra well described by fewer particles –> can’t

resolve squark mass in these cases

Simplified Models of Lepton Cascades

From gluon partner:

q

q

q

q q

W/Z(!)

G

G

G G G

!G

q

*!

!

qG

q

*

! or "! or "

BW /BZBLSP B!! B!"

!ET

!ET !ET !ET

MI

(ML)MLSP

MG

Masses

From quark partner:

q q q

*

Q

Q

Q Q Q

W/Z(!)

q

*

! or "! or "!

!

BW /BZBLSP B!! B!"!Q

Q

!ET !ET !ET

!ET

Masses

MI

(ML)MLSP

MQ

*on or off-shell


From gluon partner:

q

q

q

q q

W/Z(!)

G

G

G G G

!G

q

*!

!

qG

q

*

! or "! or "

BW /BZBLSP B!! B!"

!ET

!ET !ET !ET

MI

(ML)MLSP

MG

Masses

From quark partner:

q q q

*

Q

Q

Q Q Q

W/Z(!)

q

*

! or "! or "!

!


Q

!ET !ET !ET

!ET

Masses

MI

(ML)MLSP

MQ

*on or off-shell

q

q

Two decay modes populate 0, 2, 4 leptons,

flavor correlation

Just 2 flavor-uncorrelated leptons

distinguishable


From gluon partner:

q

q

q

q q

W/Z(!)

G

G

G G G

!G

q

*!

!

qG

q

*

! or "! or "

BW /BZBLSP B!! B!"

!ET

!ET !ET !ET

MI

(ML)MLSP

MG

Masses

From quark partner:

q q q

*

Q

Q

Q Q Q

W/Z(!)

q

*

! or "! or "!

!


Q

!ET !ET !ET

!ET

Masses

MI

(ML)MLSP

MQ

*on or off-shell


From gluon partner:

q

q

q

q q

W/Z(!)

G

G

G G G

!G

q

*!

!

qG

q

*

! or "! or "

BW /BZBLSP B!! B!"

!ET

!ET !ET !ET

MI

(ML)MLSP

MG

Masses

From quark partner:

q q q

*

Q

Q

Q Q Q

W/Z(!)

q

*

! or "! or "!

!


Q

!ET !ET !ET

!ET

Masses

MI

(ML)MLSP

MQ

*on or off-shell

!!

Many handles: frequency of n-lepton events, flavor & sign correlations.

Overlapping Lepton Sources

but....

(Not) Resolving Leptonic Decays:An Example

Counts:

Kinematic distributions:

Points:Model with very complicated cascades:

!/"

500 GeV

380 GeV

140 GeV115 GeV

!/"

W 0,±

h!soft

W (Z) !/"

!/"

Red/Green:One-stage fit(2!, W, Z, prompt)

Summary• If we’re agnostic about sparticle orderings (even

assume SUSY!):- Determining production matrix elements is hard

(Excellent approximation: info. erased by PDF integration)

- Determining spectrum and decay modes isn’t easy(Overlapping processes)

• This is a covenient misfortune!- Artificially simple few-parameter models mimic wide

range of SUSY (etc.) models well (in pT’s, some m’s)- Search and first-pass characterization that is simple,

broadly applicable, and transparent* - Precise starting point for building evidence of

complex production/decay modes


From gluon partner:

q

q

q

q q

W/Z(!)

G

G

G G G

!G

q

*!

!

qG

q

*

! or "! or "

BW /BZBLSP B!! B!"

!ET

!ET !ET !ET

MI

(ML)MLSP

MG

Masses

From quark partner:

q q q

*

Q

Q

Q Q Q

W/Z(!)

q

*

! or "! or "!

!


Q

!ET !ET !ET

!ET

Masses

MI

(ML)MLSP

MQ

*on or off-shell

[Alwall, Schuster, Toro 0810.3921]

Heavy Flavor Models

From quark partner:

From gluon partner:

tt

G

G

q

qG b

bG G

Bbb BttBqq!G

!ET!ET!ET MLSP

MG

Masses

q bT

!Q !B !T

T

TQ

Q Q

B

B B

!ET!ET!ET

MLSP

MassesMQ/T/B

Different structures / different patterns of b-tag multiplicity[Alwall, Schuster, Toro 0810.3921]

1) Which colored particles dominate production?

2) What color-singlet decay channels are present, and in what fractions?

Models with one produced species, one-stage cascade decay (produced species either G or Q).

3) How b-rich are the events?

G: Produce gluon partners that decay to qq, bb, or tt +LSP

Q: Pair-produce parters of q12, b, and t

What Can We Learn Using Simplified Models?

Quark partnerQ

Gluon partnerG

_ _ _

Either or

[Alwall, Schuster, Toro 0810.3921]

Surprising Success!

FIG. 24: Left: Spectrum cartoon for the model used in Example 2 (parameters in Appendix B 3b).

Right: Spectra for SUSY models o!B and onC used in the comparison in sec. VI D2 (parameters

in Appendix B3 c)

Second, provided the basic jet and lepton kinematics are well-modeled, we expect that

the simplified model fits can be simulated in a crude detector simulator (with approximately

similar features as the experimental environment, such as cone size, and overall geometry),

and then used as a target for vetting models that any particular theorist has in mind. Where

the simplified model fully describes the data — the HT distribution and lepton and jet counts

in the case of Lep(G), lepton-inclusive b-tag counts and b kinematics in Btag(G) — it can

be used as a target for full models. We are not saying that strict exclusions can be derived

from comparisons to the fits, but certainly the approximate consistent regions of parameter

space can be identified, and others broadly ruled out.

Third, sources of tension in the fits, such as the soft lepton deficits in this example, can

be used quite readily in the comparisons. As can be seen in figures 22 and 23, qualita-

tive di!erences from the simplified models can be seen to agree with those in the models

considered. For instance, all of the models have softer leptons than in the Lep(G) model

(because they have several light states with small splittings), as does the signal, and similar

67

0 100 200 300 400 500 600

# E

vts

/Bin

0

50

100

150

200

250

300

350pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

One Lepton Region Leading Lepton Pt (in 1-lepton region)0 100 200 300 400 500 600

0

0.5

1

1.5

2 0 100 200 300 400 500 600

# E

vts

/Bin

0

50

100

150

200

250

300

350

pseudoData

Lep(G) B_W=0,B_nul/B_lnu=1 (on-shell)

Lep(G) B_W=0 (off-shell)

Lep(G) B_W=0,B_nul/B_lnu=0 (on-shell)

One Lepton Region Leading Lepton Pt (in 1-lepton region)0 100 200 300 400 500 600

0

0.5

1

1.5

2

0 50 100 150 200 250 300 350 400 450 500

# E

vts

/Bin

0

10

20

30

40

50pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

OSSF Lepton Invariant Mass (in 2-lepton region)0 50 100 150 200 250 300 350 400 450 500

0

0.5

1

1.5

2 0 50 100 150 200 250 300 350 400 450 500

# E

vts

/Bin

0

5

10

15

20

25

30

35 pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

OF lepton Invariant Mass (in 2-lepton region)0 50 100 150 200 250 300 350 400 450 500

0

0.5

1

1.5

2

FIG. 19: Example 2: Representative lepton signatures where the Lep(Q/G) fits exhibit tension

accounting for the data.

3. Sources of Tension and Kinematics

Before studying heavy flavor sources in the Btag(Q/G) models, we comment on a few

persistent sources of tension with the Lep(Q/G) fits. The most dramatic source of tension

is with the lepton kinematics. In figure 19, we show both the lepton pT distribution in the 1

lepton region, and the opposite and same flavor di-lepton mass distributions in the 2 lepton

region. We see that there is a deficit of leptons below pT ! 75 GeV, and that in general

the lepton pT distribution is too hard. This problem persists for both on- and o!- shell

kinematics in the leptonic models. While not justified in detail here, varying the masses

in the Lep(Q/G) models does not appreciably help this structural problem. For opposite

flavor events, the Lep(Q/G) fits give rise to harder than observed leptons. This is reflected

in the bulge of events at an invariant mass of ! 30 " 100 GeV relative to either simplified

model. Again, these structural problems cannot be completely resolved within the simplified

models. We should note that the signatures shown in figure 19 are representative. We do

58

0 500 1000 1500 2000 2500 3000

# E

vts

/Bin

0

50

100

150

200

250 pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

HT (scalar sum Et of 4jets+leptons+met) (in lepton-veto region)0 500 1000 1500 2000 2500 3000

0

0.5

1

1.5

2

# E

vts

/Bin

0

500

1000

1500

2000

2500 pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

No. of Events in Each Signal Region

0Lep3Jet 1Lep3Jet 2Lep2Jet 3Lep 4Lep0

0.5

1

1.5

2

0 50 100 150 200 250 300 350 400 450 500

# E

vts

/Bin

0

10

20

30

40

50pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

OSSF Lepton Invariant Mass (in 2-lepton region)0 50 100 150 200 250 300 350 400 450 500

0

0.5

1

1.5

2

# E

vts

/Bin

50

100

150

200

250pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

2Lep2Jet_counts

OSOF OSSF OSSF_ZCand SSOF SSSF0

0.5

1

1.5

2

0 2 4 6 8 10

# E

vts

/Bin

0

100

200

300

400

500

600

700

800pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

Number of Jets (pT>30 GeV) (in lepton-veto region)0 2 4 6 8 10

0

0.5

1

1.5

2 0 2 4 6 8 10

# E

vts

/Bin

0

20

40

60

80

100

120

140 pseudoData

Lep(G) B_lnu=0

Lep(Q) B_lnu=0

Number of Jets (pT>30 GeV) (in 2-lepton region)0 2 4 6 8 10

0

0.5

1

1.5

2

FIG. 16: Example 2: A subset of signatures as described by the Lep(G) (Bl! = 0) and Lep(Q)

(Bl! = 0) fits. Jet counts and kinematics are well-approxmated by the Lep(G) fits with W . All fits

have di!culties modeling the di-lepton correlations, such as the opposite sign same flavor di-lepton

invariant mass shown here. We will comment on other sources of tension in subsection VI B3.

good fit is the Lep(G) assuming Bl! = 0 (W boson rich), with a lower bound mass estimate

of MG ! 700 GeV, MI ! 440 GeV, and MLSP = 100 GeV. Another decent fit is the Lep(Q)

assuming no primary l! decay mode with masses of MQ ! 650 GeV, MI ! 440 GeV, and

MLSP = 100 GeV. Also shown in table VI are on-shell variants of these fits. The fit cross

section are in the range of ! 11 " 14 pb for Lep(G) fits, and ! 45 pb for Lep(Q) fits

54

0 2 4 6 8 10

# E

vts

/Bin

0

100

200

300

400

500

600

700

800

900pseudoData

Lep(Q) B_W=0

Lep(Q) B_lnu=0


0

0.5

1

1.5

2 0 2 4 6 8 10

# E

vts

/Bin

0

20

40

60

80

100

120

140pseudoData

Lep(Q) B_W=0

Lep(Q) B_lnu=0


0

0.5

1

1.5

2

0 2 4 6 8 10

# E

vts

/Bin

0

100

200

300

400

500

600

700pseudoData

Lep(G) B_lnu=0

Lep(G) B_W=0


0

0.5

1

1.5

2 0 2 4 6 8 10

# E

vts

/Bin

0

20

40

60

80

100

120pseudoData

Lep(G) B_lnu=0

Lep(G) B_W=0


0

0.5

1

1.5

2

FIG. 17: Example 2: Jet count distributions of jets with pT ! 30 GeV. The 0 lepton region is

shown on the right, while the 2 lepton region is shown on the left. This comparison is meant to

highlight any jet-lepton correlations that exist in the data or the fits to leptonic models. The top

row shows Lep(Q) fits, while the bottom row shows Lep(G) fits.

decays is clearly the most consistent, while the BW = 0 fit give slightly too few jets. This

general trend remains true, even as the jet pT threshold is increased, though mild tension

accounting for the highest multiplicity (5, 6, or 7 jet) bins is apparent as the threshold is

increased. This is mostly above the trigger threshold, so we do not expect significant trigger

bias systematics in this case.

The correlation of jet counts and lepton counts, shown here by comparing the jet counts

in 0 and 2 lepton regions, again appears most consistent with a W hypothesis for Lep(G)

with the statistics available.

56

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

200

400

600

800

1000

1200

1400

1600

1800 pseudoData

Btag(Q) (lep inc)

Btag(G) (lep inc)

Number of B Jets (pT>30 GeV) (in lepton-inclusive region)-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

0.5

1

1.5

2 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

200

400

600

800

1000

1200

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)

Number of B Jets (pT>30 GeV) (in lepton-veto region)-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

0.5

1

1.5

2

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

100

200

300

400

500

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)

Number of B Jets (pT>30 GeV) (in 1-lepton region)-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

0.5

1

1.5

2 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

50

100

150

200

250

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)


0.5

1

1.5

2

FIG. 20: Example 2: b-jet count distributions for the lepton inclusive, 0, 1, and 2 lepton regions.

Note that Btag(G) provides a better overall description of these signatures. Also note that the

lepton exclusive fits, in which W ’s from top-channels are used to account for leptons, fail to account

for all the lepton in the 0 b-jet regions. This is strong evidence for lepton channels beyond those

that may accompany any third generation channels.

Figure 21 shows several other useful comparisons of the b-tag fits. For the range of masses

considered, both b-tag models describe the bb invariant mass signature quite well when it

is dominated by top decay modes, as is the case with the lepton exclusive fits. Moreover,

the kinematics of the b-jets themselves are better modeled by the lepton exclusive fits, in

which the primary source of b-jets is from top. These comparisons do not directly imply

a preponderance of top decays (an attempt at direct top reconstruction might be a better

source of evidence for this), but they certainly support a top-rich hypothesis.

The data is globally well described by a subset of limits of the four simplified models,

as presented above, despite the simplicity of these models. As we’ll see, the qualitative

and quantitative information from the above fits is good enough to directly motivate model-

building. However, due to tensions in the fits, more precise information about the underlying

60

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

200

400

600

800

1000

1200

1400

1600

1800 pseudoData

Btag(Q) (lep inc)

Btag(G) (lep inc)

Number of B Jets (pT>30 GeV) (in lepton-inclusive region)-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

0.5

1

1.5

2 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

200

400

600

800

1000

1200

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)

Number of B Jets (pT>30 GeV) (in lepton-veto region)-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.50

0.5

1

1.5

2

-0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

100

200

300

400

500

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)


0.5

1

1.5

2 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5

# E

vts

/Bin

0

50

100

150

200

250

pseudoData

Btag(Q) (lep exc)

Btag(G) (lep exc)


0.5

1

1.5

2

FIG. 20: Example 2: b-jet count distributions for the lepton inclusive, 0, 1, and 2 lepton regions.

Note that Btag(G) provides a better overall description of these signatures. Also note that the

lepton exclusive fits, in which W ’s from top-channels are used to account for leptons, fail to account

for all the lepton in the 0 b-jet regions. This is strong evidence for lepton channels beyond those

that may accompany any third generation channels.

Figure 21 shows several other useful comparisons of the b-tag fits. For the range of masses

considered, both b-tag models describe the bb invariant mass signature quite well when it

is dominated by top decay modes, as is the case with the lepton exclusive fits. Moreover,

the kinematics of the b-jets themselves are better modeled by the lepton exclusive fits, in

which the primary source of b-jets is from top. These comparisons do not directly imply

a preponderance of top decays (an attempt at direct top reconstruction might be a better

source of evidence for this), but they certainly support a top-rich hypothesis.

The data is globally well described by a subset of limits of the four simplified models,

as presented above, despite the simplicity of these models. As we’ll see, the qualitative

and quantitative information from the above fits is good enough to directly motivate model-

building. However, due to tensions in the fits, more precise information about the underlying

60

Lept

onic

Hea

vy

flavo

r

Good agreement in many, not all

distributions & well-defined best-fit parameters –

Discrepancies hint at (specific!) additional

structure, but extensions can’t be fully constrained

• Current study: hadronic searches (leptonic search study underway)

• First step: Validation of mSUGRA benchmark points LM*- Make sure LM* distributions are reproduced- Then topology-based searches guaranteed to be sensitive to

LM* – “first, do no harm”

Simplified Searches- Optimize sensitivity to general models

- Present results for general models

Design searches around individual topologies, with more softer or few harder jets.

(work in progress by UCSB CMS group)

24

!!"#

!$%#

!&'!&(

#)

!!%#

*$#

*!#

+$#

+!#

,-%

,,$

,.%

,/-

01#

21#

31#

41#

05#25#35#45#

,/-

,!%

.%&

,'%

-%/

/-%/-/

/..

(-6.

!/%#

!.%#

!$"#

78!%9:;<=9>;?+@A2

-!B"#$$

'/B"#CDEDFDG

-!B89"#%#&

!"#

"$#

"%#"&#

./B89"#'#&

.%B

!&B

'6/B

!.B989"#(#&

/%B89"#%#&

!.B

H1H1

,B

H1H5

!!B

H5H5

!!B

:H1

$/B

:H5

/-B

::9

!%B'()*+'*,-./0),1

('B89"#%#&

/.B89"#%#&

'6&B

!BI.B89"#%#&

!/B

'-B9@J9K>>9FK0L@AM4

?L@034+M@A9MA91N!94@A2M2+29

@J9O2MP?>;Q90;4KR94FKMA26

SFM29PKT;29M+9?KL+M43>KL>R9

KP;AK*>;9+@9*;MA)9

!""#$%&'!()*+,)--+UM+F9K9

/I?KL+M4>;97VWS6

e.g. Production modes in the LM1 Benchmark:(after hadronic search cuts: lepton veto, 3 or more jets)

–S.A. Koay

It Works!Fit gg, ug, and uu production fractions (and masses, by eye) from HT, jet pT

~~ ~~ ~~

“Do no harm” : search optimized for this topology can discover LM1 as well as an LM1-optimized search

(generator-level comparison)

17

!"

!!#$

%&!

!%#

'&!

'#!

(&&

(#)

($*

+,!

-,!

.,!

/,!

($0

1)0

&$2

()$

($3

##1##(

0$4)

+5!-5!.5!/5!

!!#6

!&$!

)0789"#$#%

((789"#&#%

3*789"#'#%

)*7:0(789"#'#%

1#7:1)789"#'#%

#$$7989"#$#%

1(789"#$#%

($789"#(#%

#3789"#&#%

)789"#$#%

2079"#;;&(79"#!)

*#79"#;;*79"#!!##79"))

8<=> 8?=>

@A5

&)7

@A,

&&7BB

#17

@@

#&7A,A5

37

A5A5

27

A,A,

)7

CC

#7

Extreme case: LM0 (significant stop production and cascade decays)

–S.A. Koay

20

Works again! (Look at blue vs. black)MET/HT very sensitive to cascade shape, most discrepant

Affects efficiency of search cuts, but minor impact on distributions after cuts

Topology-Driven SearchesDesign cuts for sensitivity to processes with more/fewer jets, wide range of spectra.

“Three hard jets” “~5 hadronic jets”(Effect of cascade

depends on C+ mass)

Even more/softer jets(not visible – ignore for now)

Fixed cuts: lepton veto, 3 jetsOptimize Jet pT, HT, MET cuts for sensitivity to A/B topologies over wide mass rangeLeptonic search effort underway...

Search Generally Present Generally

Meaningful steps beyond mSUGRA

Sensitivity (and eventually exclusion) can be quoted in terms of all relevant parameters: cross-section, mg, mu, and mC+, mLSP

Models with similar topologies don’t require separate searches. If topology is dissimilar, motivation to search for it is clear.

Ensure sensitivity to multiple topologies

~ ~ ~

Applying deltaPhi cuts to every jet makes search insensitive to longer cascades – dangerous if they dominate!

And for wide range of mass splittings!

If new physics is seen in “SUSY” search, What Next?

Crude “Simplified Models” from earlier are general starting point for analysis.

Example:– what do they tell us?– how do we move beyond them?

– what do we learn from simplified model fits “inside,” but not outside theorists’ analysis of published data?

Branching Ratios

OSOF(e+µ-)

OSSF(e+e-)

ZCand SSOF(e+µ+)

SSSF(e+e+)

5 params and 3 independent counts in 2-lepton data (under-constrained)

Additional constraint from 0-, 1- or 3-lepton data

AMBIGUITY:W goes to 1 lepton (30%) or 0 leptons (70%).

Hard to distinguish W’s from combination of direct and one-lepton cascade

Branching Ratios(Best Fits)

Parameters that fit counts, HT, pT(lepton):

ambiguity –affects conclusions!

big syst. effect on masses, xsec

some branching ratios more stable than others

Theorist on the outside can estimate these from 1,2-lepton data...but given large systematics, we’re likely to make mistakes combining channels reliably

What the best fits look likeCounts, jet kinematics reproduced well!

(also jet pT plots, MET...)

What the best fits look like(2-lepton plots)(1-lepton plots)

Cannot reproduce the data with these models(or with tops). Robustly demonstrating this is hard, but provides STRONG EVIDENCE for more complex source of soft, flavor-uncorrelated leptons.

Lepton pTOSSF (e+e-) invariant mass

Opposite-flavor (eµ) invariant mass

Q/G

weak

LSP+leptons/W/Z

+jets

Q/G

weak

LSP

weak’

(only believable if studied by experimentalists)

Interim Conclusions and Questions

• Data consistent with squark and/or gluino production

• Need two-stage cascades to explain data

• Large rate of single-lepton cascade (+ precise numbers)

• To reproduce the 2-lepton counts (trial & error) ...on-shell slepton and charginos. See if this can be confirmed

from kinematics – dilepton invariant mass should have

an EDGE (this is sub-dominant source

of 2-lepton events, edge didn’t jump out but this motivates

looking harder)

Q/G

weak

LSP

slepton

Q/G

weak

LSPslepton

or ?

I can find SUSY models with both hierarchies, see if any of them are consistent with larger set of distributions in data...

More conclusions fromb-jet studies

• Gluon-partner with ~60% branching fraction to heavy flavor works well. Not flavor-universal!

• Lepton-rich events have fewer b-jets

G

weak

LSP

slepton

+ light flavor+ heavyflavor

(G decay could haveintermediate on-shell Q’s)

More conclusions fromb-jet studies

• Gluon-partner with ~60% branching fraction to heavy flavor works well. Not flavor-universal!

• Lepton-rich events have fewer b-jets

G

weak

LSP

slepton

+ light flavor+ heavyflavor

(G decay could haveintermediate on-shell Q’s)

three SUSY ideas

gluino

weak

LSPslepton

stop

squarks

top dominates because stop is lighter

gluino

weak

H LSPslepton

stop &squarks

top dominates because it has biggest coupling

~

gluino

weak

H LSPslepton

top dominates because stop is lighter

~

stopsquarks

ConclusionsHadron colliders swallow a lot of information! Sharpen the question: “What can be probed?”

Two natural classes of simplification:– insensitivity to production matrix element– smearing-together of decay chains

Used at CMS to generalize some SUSY searches

Basis for observable properties of new physics will assist in making sense of a discovery

Unlocking the Structure of New Physics at the LHC · Structure of New Physics at the LHC Natalia...

Documents

Transcript of Unlocking the Structure of New Physics at the LHC · Structure of New Physics at the LHC Natalia...