Семинар «Использование современных информационных технологий для решения современных задач физики частиц» в московском офисе Яндекса, 3 июля 2012 Mitesh Patel, Imperial College

1. Searching for new physics
with the LHCb experiment
Mitesh Patel (Imperial College London)
Yandex, Moscow 3rd July 2012

2. What is Particle Physics?
Particle physics is the study of the basic constituents
of matter and the forces that act between them
2

3. Subatomic structure
• The protons and neutrons that make-up ordinary matter are not
fundamental – they are made of quarks
3

4. Subatomic structure
• The protons and neutrons that make-up ordinary matter are not
fundamental – they are made of quarks
• proton
– up up down
• neutron
– up down down
4

5. The Standard Model
Gauge Bosons
5

6. The Standard Model
• Mathematical description (not a
classification!) of particle
interactions
– Quantitative and predictive theory
– Agrees with the results of virtually
all experiments …
• Incredibly successful theory –
describes virtually all known
phenomena with amazing accuracy
• Incomplete…
– The Higgs Boson supposed to give
mass to other particles
– Theory doesn’t describe gravity
– Number of other open questions…
6

7. Problems with the SM Higgs
• Even if the Higgs boson is found at CERN’s Large Hadron Collider,
our problems aren’t over…
• If we compute the Higgs mass find contributions from processes
like,
f
H
H
f
→ Higgs mass blows-up to ∞ (aside which we will ignore: ‘or
incredible fine-tuning’)
• This can’t be the case …
7

8. Problems with the SM (cont’d)
Two colliding clusters of galaxies
• Observations of the stars → much more
mass that visible
– “Dark Matter” – 23% of the mass/energy
of the Universe is missing!
– SM has no Dark Matter candidate!
• Observations also indicate that the
Universe is expanding at an
accelerating rate
– “Dark energy” – 73% of mass/energy in
Universe is missing!
– Try to compute this from the SM - find
something 1054 times too big !
8

9. Problems with the SM (cont’d)
• Whole host of other open questions:
– Why are there so many types of matter particles?
• Mixing of different flavours of quarks and leptons
• Observed matter-antimatter difference
– Are fundamental forces unified?
• Do all the forces unify at some higher energy scale?
– What is quantum theory of gravity?
• String theory?
– …
• → Expect to find new phenomena (“new physics”) at experiments
at CERN’s latest accelerator, the Large Hadron Collider !
• Solving the problems of the Standard Model:
– (Super-)partners to all existing particles
– Extra spatial dimensions
– …
9

10. Supersymmetry… ?
• Supersymmetric theory (SUSY) postulates that every particle we
observe has a partner with spin different by 1/2
– denoted by adding tildes (~) to the symbols for the SM particles
→ squarks, sleptons, gauginos
f
~
+
f
H
H
H
H
f
~
f
10

11. Supersymmetry… ?
• The symmetry must be “broken” – the partners must have higher
masses than the SM particles or we would have seen them!
• Superpartners stablise the Higgs mass
f
~
+
f
H
H
H
H
f
~
f
• In order to make this cancellation the superpartners cannot be too
heavy
• Lightest supersymmetric partner good candidate for dark matter
11

12. The Large Hadron Collider
• CERN’s Large Hadron Collider (LHC) will
explore the physics beyond the SM
– world's largest and highest-energy particle
accelerator
– contained in a circular tunnel, 27km
around, at a depth ~100m underground
– two adjacent parallel beam pipes that
intersect at points where expts are placed
– 1600 superconducting magnets bend two
proton beams into circular trajectory
– ~96 tonnes of liquid helium used to keep
the magnets at their operating temperature
of 1.9K (−271.25 °C)
– beams accelerated to 0.99999999 of the
speed of light
– Beam energy
• Channel tunnel train at 150km/h
• Admiral Kuznetsov cruiser @ 8 knots
• 77kg of TNT, your car at ~1000 mph
… within the width of a human hair 12

13. Searching for new particles
• Two ways of searching for new particles, X
– Try and produce X directly from pp interactions (and Direct production
detect its subsequent decay into known particles) p
p
Designed to X
study pp
interactions
– Look for the effect of X as an intermediary in decay
of well known particles
• So called ‘loop’ decays of B particles particularly Loop decay
interesting
Designed to B
study B decays X
Have to integrate over
• Uncertainty principle means that, provided it exists all possible momenta of
only for a very short time, X can be much heavier than intermediate partlcles
allowed by energy conservation
13

14. The LHCb Experiment
• LHCb is used to study a wide range of “golden decays” where we
have precise theory predictions
• Perhaps, the highest profile measurement is the search for the
decay Bs0→µ+µ- 14

15. The decay Bs0→µ+µ-
• The decay Bs0→µ+µ- is very sensitive to contributions from new
particles e.g. Higgs boson A0
• The decay is very suppressed in the SM but the rate expected from
SM processes can be computed precisely,
– B(Bs0→µ+µ-) = (3.5±0.2)×10-9
– → 1 Bs0→µ+µ- decay in every 285 Million Bs0 decays…
– … but only get 1 Bs0 in every 2000 pp interactions, some of which can
fake a Bs0→µ+µ- decay → few events in >> 285 Million decays
… and rate can be substantially modified in presence of e.g. Higgs
boson, A0
• Rely on combination of all event properties: Multivariate Analysis
15

16. Multivariate Analysis
• This, and pretty much all other analyses at LHCb, use the package
Toolkit for MultiVariate Analysis (TMVA)
• Boosted Decision Tree (BDT) seems to be best performing method
• Not clear how optimal this is :
– Most people just use default boosting procedure (AdaBoost), choice of
depth, number of nodes etc.
– Notable feature of our problems: not enough training data
• From analysis side application of MVA is the problem in extracting
particle physics results :
– Acquiring the data is extremely time consuming and expensive
– Anything that allows you to get more “power” out of same data is
therefore vitally important
→ Something with a demonstrable advantage would be used
everywhere, very quickly
16

17. Latest Experimental Results
2011 data (5fb-1)
• Profile is such that search made
at all three LHC experiments
• Intense rivalry to see the first
signal events
• Will then need to make a precise
measurement of the decay rate
Part of 2011 data (2.4fb-1)
3.5
Events/60 MeV
2011 data (1fb-1)
3
ATLAS
| |max< 1
s = 7 TeV
-1 Data
2.5 Ldt = 2.4 fb Bs µ+µ- MC (10×)
2
1.5
1
0.5
0
4800 5000 5200 5400 5600 5800
mµ µ [MeV]
17
3.5
0 MeV
ATLAS
3 | |max< 1.5

18. The Future of Bs0→µ+µ-
12
B(Bs → µ + µ -) [10 -9]
11 LHCb
10 Projection from 1 fb-1
9
8
7
0
6
5
4 time integrated SM
3 (arXiv:1204.1737)
2
1
1.5 2 2.5 3 3.5 4 4.5 5
Luminosity [fb-1]
18

19. Other Analyses relying on MVA
• Whole host of other analyses rely on MVA…
500
Events / (0.5 a. u.)
Bd → K*0 µµ Background
500
2011 Bd → J/ ψ K* 0 Signal
0
450
2010 Bd → J/ ψ K* Signal
400 2011 Bd → K* 0
µ+µ- Background 400
0
2010 Bd → K* µ+µ- Background
350
300 300
250
200 200
150
100 100
50
0 0
-1 -0.5 0 0.5 1 0 100 200 300
BDT Response (a. u.) Bd → K*0 µµ Signal
`
19

20. Triggering
• Accelerator collides particles 40Million times / second
• Cannot process or store all of events from collisions and look
afterwards for events we are interested in – have to chose which
events to keep for further study
L0 “high pT” signals in calorimeter
Hardware and muon systems
HLT1 Partial reconstruction, selection
Software based on one or two (dimuon)
displaced tracks, muon ID
HLT2 Global reconstruction (very close
Software to offline) dominantly inclusive
signatures – use BDT
20

21. HLT2
• LHCb uses a BDT in the second level of the High Level Trigger,
HLT2
– selects N-dimensional regions of parameter space to keep by learning
from training samples
– Have to ensure that selected regions are not so small relative to the
resolution and/or stability of the detector st they could cause the signal
events to oscillate in and out of the kept regions (→ less efficiency, or
a trigger that is impossible to understand the efficiency of)
– Only allow decision tree to split at certain pre-defined points in the
parameter space
• e.g. know that the track quality of a particle discriminates between signal and
background – requirement of χ2 < 4 or χ2 < 9 are sensible, effect of χ2
<1.000045 might vary between data-taking period
– Triggering is one of the major challenges for the experiment – any
advantage that could get from new methods would make a tremendous
difference
21

22. Conclusions
• Our knowledge of particle physics is embodied by a mathematical
description of particle interactions, ‘The Standard Model’
• The model is tremendously successful but has some significant
problems – latest experiments may find new phenomena!
• LHCb experiment searching for signatures of new phenomena by
probing certain rare B particle decay modes such as Bs0→µ+µ-
• In this and in many other analyses, and in other aspects of the
experiment, searching for small signal over large backgrounds –
multivariate analysis a key requirement
• Any improvement in MVA would be hugely beneficial and sought
after by everyone working in this field, and in other fields
22

23. Backup
23

24. Extra Dimensions
• Which is weaker: –
Gravity or Electromagnetism?
• Alternatively, which is more
powerful: –
The gravitational pull of the entire
earth
or
The boy with his
magnet?
• Gravity is extremely weak! Why?
24

25. Extra Dimensions
• Electromagnetism is confined
to our usual three dimensions
of space
• Maybe gravity is special: –
maybe gravity sees other
dimensions of space … ? Gravity
• As the force is spread out, it is
weakened
• How can there be extra
dimensions of space?!
25

26. 26

27. Black Holes
• Microscopic Black Holes! Not like astronomical Black Holes!
• If matter is sufficiently compressed, its gravity becomes so strong
that it carves out a region of space from which nothing can escape
• Size you have to compress to depends on the mass -> smaller hole,
greater amount of compression required
• Gravity weak -> amount of compression required way beyond
accelerators… but with extra-dimensions maybe gravity is strong on
small enough scales… -> microscopic black holes at the LHC?
• Hawking radiation -> black holes shrink
• Quantum effects -> microscopic black holes “evaporate” -> produce
lots of particles

28. Cosmic rays are continuously bombarding Earth's atmosphere with far
more energy than protons will have at the LHC, so cosmic rays would
produce everything LHC can produce
They have done so throughout the 4.5 billion years of the Earth's
existence, and the Earth is still here!
The LHC just lets us see these processes in the lab (though at a
much, much lower energies than some cosmic rays)
So, there is no danger at all!

29. Pair Production and Annihilation
• Picture shows pair-production:
γ + γ -> e+ + e-
• Observe that particle and antiparticle are
always created in pairs
• Annihilation also occurs in pairs:
e+ + e- -> γ + γ
• Hence,
Particles − Antiparticles = 0
p.29/41

30. The History of the Universe
• t = 13.7×109 yrs
• All energy in Universe confined in a tiny
region -> extremely hot and dense
• ‘Soup’ of basic particles
• Only later, as Universe expanded and
cooled, temperature became low enough to
form neutrons and protons, nuclei, atoms…
• t=0 s ????
p.30/41

31. Where did the antimatter go?
• Shortly after the Big Bang (extremely
dense/hot) -> equal amounts of matter
and antimatter were created from the
available energy
• Where did the antimatter go?
• Particle Physics – smallest of scales
Big Bang – largest of scales
p.31/41

32. A matter-antimatter asymmetry
• We have found a small difference between matter and antimatter
that could generate such an asymmetry
• Some processes generate slightly more matter than antimatter
• Such processes violate a symmetry known as “CP-symmetry”
– A process obeys CP-symmetry if its results are identical after changing
all particle positions to a mirror image and changing all particles to their
antiparticles [… next slides…]
– Processes that don’t obey CP-symmetry said to be “CP-violating” – can
produce an excess of matter over antimatter as they treat particles and
antiparticles differently
p.32/41

33. CP Violation
Parity
P
Inversion
Spatial
mirror
p.33/41

34. CP Violation
Charge Inversion
C
P
Particle-antiparticle
C
CP
mirror
Parity
P
Inversion
Spatial
mirror
p.34/41

35. CP Violation
Charge Inversion
C
P
Particle-antiparticle
C
CP
mirror
Parity
P
Inversion
Spatial
mirror
p.35/41

36. CP Violation
CP
• We have found that matter and
antimatter behave differently after the
C and P mirrors: “CP violation”
• Allows for some reactions to proceed
more easily that their CP-opposites
p.36/41

37. A matter-antimatter asymmetry
• While CP violation could
generate a matter-antimatter
asymmetry the effect we see is
tiny – much too small to explain
the matter-antimatter
asymmetry in the Universe
• Expect there are additional
sources of CP violation -> hope
to see evidence of these in the
collisions at CERNs Large
Hadron Collider (LHC)
p.37/41

38. The Large Hadron Collider
• CERN’s Large Hadron Collider (LHC) will
explore the physics beyond the Standard
Model
– world's largest and highest-energy particle
accelerator
– contained in a circular tunnel, 27km around,
at a depth ~100m underground
– two adjacent parallel beam pipes that
intersect at four points where experiments
are placed
– 1600 superconducting magnets bend
protons into circular trajectory
– ~96 tonnes of liquid helium used to keep the
magnets at their operating temperature of
1.9K (−271.25 °C)
– beams accelerated to 0.99999999 of the
speed of light
– Beam energy
• Channel tunnel train at 150km/h
• Aircraft carriers HMS invisible and HMS
Illustrious (combined) at 6.0 m/s
• 77kg of TNT, your car at ~1000 mph
… within the width of a human hair
38

39. The Large Hadron Collider
• CERN’s Large Hadron Collider (LHC) will
explore the physics beyond the Standard
Model
– world's largest and highest-energy particle
accelerator
– contained in a circular tunnel, 27km around,
at a depth ~100m underground
– two adjacent parallel beam pipes that
intersect at four points where experiments
are placed
– 1600 superconducting magnets bend
protons into circular trajectory
– ~96 tonnes of liquid helium used to keep the
magnets at their operating temperature of
1.9K (−271.25 °C)
– beams accelerated to 0.99999999 of the
speed of light
– Beam energy
• Channel tunnel train at 150km/h
• Aircraft carriers HMS invisible and HMS
Illustrious (combined) at 6.0 m/s
• 77kg of TNT, your car at ~1000 mph
… within the width of a human hair
39

40. The Higgs Boson H?
• One of main problems of Standard Model – in its simplest form the
mathematical structure of theory does not allow the introduction of
mass for the particles!
• The Higgs Boson, through the Higgs mechanism, is the particle that
‘gives’ particles mass …
– How can a particle give mass to other particles?!
– Don’t particles just have mass?
40

41. The Higgs
Mechanism
• Imagine a cocktail party of political party
workers who are uniformly distributed
across the floor, all talking to their
nearest neighbours
• A certain ex-Prime-Minister enters and
crosses the room. All of the workers in
her neighbourhood are strongly attracted
to her and cluster round her. As she
moves she attracts the people she
comes close to, while the ones she has
left return to their even spacing
• Because of the knot of people always
clustered around her she acquires a
greater mass than normal, that is, she
has more momentum for the same
speed of movement across the room.
Once moving she is harder to stop, and
once stopped she is harder to get
moving again because the clustering
process has to be restarted
A
quasi-­‐poli?cal
Explana?on
of
the
Higgs
Boson;
for
Mr
Waldegrave,
41
UK
Science
Minister
1993
(David
J.
Miller,
UCL)

42. The Higgs
Boson
• Now consider a rumour passing through
our room full of uniformly spread political
workers. Those near the door hear of it
first and cluster together to get the
details, then they turn and move closer
to their next neighbours who want to
know about it too
• A wave of clustering passes through the
room. It may spread out to all the
corners, or it may form a compact bunch
which carries the news along a line of
workers from the door to some dignitary
at the other side of the room
• Since the information is carried by
clusters of people, and since it was
clustering which gave extra mass to the
ex-Prime Minister, then the rumour-
carrying clusters also have mass
• The Higgs boson is predicted to be just
such a clustering in the Higgs field
A
quasi-­‐poli?cal
Explana?on
of
the
Higgs
Boson;
for
Mr
Waldegrave,
42
UK
Science
Minister
1993
(David
J.
Miller,
UCL)

43. The Higgs
Boson
• Now consider a rumour passing through
our room full of uniformly spread political
workers. Those near the door hear of it
first and cluster together to get the
details, then they turn and move closer
The Higgs field pervades all
to their next neighbours who want to
know about it too space, the Higgs boson is
• A wave of clustering passes through the
like the clustering in that
room. It may spread out to all the field. It is the interactions of
corners, or it may form a compact bunch particles with the Higgs
which carries the news along a line of boson that give particles
workers from the door to some dignitary mass.
at the other side of the room
• Since the information is carried by
clusters of people, and since it was
clustering which gave extra mass to the
ex-Prime Minister, then the rumour-
carrying clusters also have mass
• The Higgs boson is predicted to be just
such a clustering in the Higgs field
A
quasi-­‐poli?cal
Explana?on
of
the
Higgs
Boson;
for
Mr
Waldegrave,
43
UK
Science
Minister
1993
(David
J.
Miller,
UCL)

44. The Higgs Boson H?
• Existing measurements tell us that the Higgs Boson, or some other
phenomena, must appear at energies accessible at CERN’s LHC
e+e-→W+W-
Only if we put the Higgs in
with the couplings predicted in
the SM do we get a theory
related to interaction prediction (the turquoise line)
that agrees with the
probability: must be measurements (green points)
less than ~17 May not be the Higgs boson
but something is doing the
job!
Energy of e e collision
• The simplest theories predict only one boson, but others say there
+ -
might be several
44