LHC Physics an experimentalists perspective Claudio Campagnari UC

LHC Physics: an experimentalist’s perspective Claudio Campagnari UC Santa Barbara 1

What are these lectures about? • For HEP theory graduate students – Unknown (to me) mix of formal/model builders/phenomenology • About 1 year before start of LHC physics – A new energy regime hopefully will lead to historic discoveries • But with no guarantees. . • Give you a flavor of how we (experimentalists) go about this business • Help you in understanding experimental issues – Help you understand appreciate results as they come along – Even as a theorist need some intuition if you want to be “in the game” • Point out issues on theoretical side • Personal slant • Based on talks (2006) at the West Coast LHC Theory Network and at the LHC Olympics 2

What this talk is not about • A status report on LHC machine and detectors • A collection of pretty pictures – Although the detectors and machine pictures are pretty • A collection of results from physics reach studies, eg, if there is SUSY with this set of parameters, we’ll have a 5 signal in XX fb-1, or we’ll exclude it at 95% CL in YY fb-1 – – (Perhaps) interesting and important but …. . boring (to me) New Physics models are guesses anyway I don’t believe these studies all that much You will not learn very much. IMHO, more important to think in terms of general principles – I will, however, show some such results to illustrate the principles 3

Disclaimer • Most (all) of the LHC related pictures/plots that I will show to illustrate my points are from CMS • Not because I do not like Atlas • But because I am on CMS – easier for me to find CMS material – and I understand it better • Much of what I'll talk about will be quite general and applies to Atlas also 4

Outline 1. Detectors 101 • • How do these detectors work in broad terms, capabilities, limitations Physics Objects 2. Physics “landscape” at the LHC • Cross sections, rates for SM processes and generic NP 3. Searches for NP • • • How are they done, what are the ingredients Issues on theoretical side Case studies Obviously, can only scratch the surface. Give rules of thumb to help you think about this stuff 5

Detectors 101 • Detectors for high PT at colliders are designed to identify and measure the "objects" that are used to do physics – – – electrons muons taus photons quark and gluons as jets • including b-quark jets – neutrinos (and dark matter, etc) • as missing energy • A "generic" detector is a cylinder (with endplugs) 6 with concentric layers of detector elements

Detector Slice muon hadronic calorimeter Tracking in solenoidal B-field to measure PT tracking EM cal. e , K, p. . Artwork by Corinne Mills 7

A more realistic slice (CMS) 8

And what it actually looks like (late April) 9

Physics Objects • Go through the physics objects one-by-one – Not many details, but general picture • • How are they detected? How well are they measured? How are they misidentified? Will conclude with score card on objects • A little understanding of how this works useful to 1. Appreciate the forthcoming exp. results 2. If you want to participate in the LHC program 10

Electron signature • Track in the inner detector • Shower and complete energy deposition in EM calorimeter – electron bremsstrahlung – e+e- pair production http: //www. irs. inms. nrc. ca/EGSnrc/pirs 701/node 22. html http: //www. irs. inms. nrc. ca/EGSnrc/pirs 701/img 12. png 11

Electron Signature (2) www-zeus. physik. uni-bonn. de/~brock/ http: //student. physik. uni-mainz. de/~reiffert/atlas/em-shower. jpg X 0 = radiation length Pb. W 04: 0. 9 cm, Pb: 0. 6 cm, Cu: 1. 4 cm 12

Photons • Just like electron, but no track • Resolution of EM calorimeters very good, eg, CMS (E in Ge. V) • Gets better with increasing E • Question: where do all these terms come from? • Answers – 1 st term: shower statistics (fluctuations of number of particles in shower) – 2 nd term: mostly module-to-module calibration – 3 rd term: noise, pileup, etc 13

Hadrons • Track in inner detector (unless neutral, eg, n) • Hadronic interaction – Some energy deposition in EM calorimeter – Energy deposition in HAD calorimeter Interaction length Pb. WO 4: 22 cm Pb: 17 cm Cu: 15 cm 14 http: //student. physik. uni-mainz. de/~reiffert/atlas/hardron-shower. jpg

EM vs HAD showers • The pretty pictures look similar, but the physics is different – with important consequences X 0 << 1. • • Longitudinal (and transverse) evolutions quite different e/ on average shower 1 st and stop 1 st use it to separate e from This is a good thing 2. Hadronic shower fluctuations large • • Energy resolution poor Response often not linear with E This is not a good thing 15

Calorimeter response to (CMS) Poor resolution compared to e/ Non-linear response 16

Jets • Traditionally reconstructed by summing the energy in nearby calorimeter towers • Limitations in the had energy measurement leads to poor resolution 17

Aside on rapidity • In hadron collider not very convenient to use angular cylindrical coordin ates ( , ) – even if the geometry is cylindrical • Basically because the CM is boosted • Rapidity y = ½ log(E+PZ)/(E-PZ) • d 3 p/(2 E) = ½ d. PT 2 dy d – Phase space: flat in dy • Longitudinal boost (eg takes from LAB to CM frame): – y y + y – = tanh y (same for all particles) • Pseudorapidity = log(cot( /2)) = y in the limit E>>m – Jets of given PT have constant ~ at all 18

Muons • Measured in the inner tracker, go through the calorimeter, measured again outside CMS • Unlike e case, resolution gets worse at high energy. Why? 19

Neutrinos (or dark matter) • Sum up the momenta of everything, what is left to get back to zero (missing energy) is the neutrino(s) • Longitudinal information is lost down the beampipe can only do in transverse plane – Missing transverse energy (MET) • If > 1 , you only infer the sum of the trans. momenta 20

Missing Transverse Energy • Fake MET mostly from jets, resolutions and tails 1 min bias event contribution to MET component in a given direction ~ 6 Ge. V • Also from missed muons • Also from "underlying event" CMS And the tails don't come without some work. . D 0 21

Taus • Detect decay products • 25% as e or • Remainder: narrow jet with 1 or 3 charged tracks High momentum Boosted daughters • Energy resolution intrinsically poor (miss ) 22

B-jets • Wouldn’t it be nice to be able to say: this jet comes from a gluon. Or an up-quark. Or a down-quark. Or…. • Dream on! • Fortunately we can do it for b-jets – And c-jets, to some extent – Because (b) ~ 1. 5 psec for P=40 Ge. V mean decay length = c ~ 4 mm Find evidence of “long” lived particles inside a jet to “tag” a b-jet • Efficiency typically ~ 50% per jet • And can use b c (ce ) also 23

Instrumental backgrounds (fakes) • In a given analysis you select the objects (e, , , b-jets, etc) for the specific physics channel that you are interested in • But how do you know that what you called an electron (say) isn’t really something else? • And does it really matter? – Interestingly enough, in some cases you don’t really care that much (will see some examples later) • Although if you don’t know, it is guaranteed that you will find someone in the collaboration that will make your life miserable when you try to publish – But in most cases it does matter, and in some cases it is the most important/difficult piece of the analysis • e. g. , many searches for rare NP processes 24

Fakes 1. Where do they come from 2. How can we deal with them 25

Fake electrons (1) Electron shower Hadron shower Vs • Sometimes a hadron shower looks like an EM shower • Or 0 , e+e- conversion in material • Or …. EM shower High momentum 0 e+ e- 26

Fake electrons (2) • Usually we are interested in high PT, isolated electrons • High momentum hadrons that can fake the electron signature are in jets – And there a lot of jets, as we will see later • These hadrons are usually not isolated – So requiring “isolation” helps – But occasionally jet fragments to a leading particle • Typical probability of “jet faking isolated electron” ~ 10 -410 -5. • Watch out! Some jets can have real electrons, e. g. , b ce 27

Fake photon • Same picture as before, high momentum EM shower 0 High momentum 0 • Can look just like single • Again: high momentum 0 from jets • Probability of “jet faking isolated ” ~ 10 -4 28

Fake muons • A hadron can occasionally “sail through” without interacting – It will then look exactly like a muon • and K decays in flight • Again: from high PT pions and kaons in jets • Typical probability of “jet faking isolated ” 10 -4 -10 -5 29

Fake taus High momentum Boosted daughters • Sometimes a jet will fragment in a small number of well collimated tracks, and look like a • Typical “probability of jet faking a ” ~ 10 -2 -10 -3 • Much worse than for e and • A real shame – 3 rd generation in many NP scenarios is special – Heavier than e and , higher coupling to Higgs, window on EWSB 30

Fake b-tagged jets • Best b-tag: evidence of long lived particles inside a jet – Typical flight lengths ~ O(mm) • Fake b-tags – Track reconstruction errors, resolution tails – Residual long-lived particles inside a jet that have nothing to do with b-quarks, eg, KS + -, p – Particles directly originating from collision, eg, conversions, nuclear interactions in beam pipe – b-quark creation as part of jet evolution. Gluon splitting g bb • Typical “probability of jet faking b-tag” ~ 1% 31

Fakes 1. Where do they come from 2. How can we deal with them 32

Dealing with fakes • Clearly: try to do the best job you can and be as smart as you can to reduce them • But you will never get rid of them. And you need to understand them. • In general you do not get these from simulation – You are looking at improbable occurrences in the tail of distributions • Jet fragmentation • Detector response • Two general approaches 33

1 st approach: use fake probabilities • Suppose you are looking for a process with X+Y+b-tag (where X and Y are, eg, electrons or muons, or taus). • You are worried about background from X+Y+jet with jet faking a b-tag • You figure out your rate of X+Y+jet – Either from theory (Monte Carlo) or data (better) – You measure the fake probability somewhere else • This is not always easy! – You apply it to X+Y+jet to predict the background from X+Y+(fake b-tag) 34

Some fake probabilities D 0 Lauer Ph. D Thesis Iowa State Jets as Jeans (Rome, CDF) LHC Symposium 05 Btag fake rate 35

2 nd approach to fakes • Relax your requirements, let backgrounds in • By studying how much background comes in as a function of your requirements, try to estimate how much background is left after your final requirements 36

Final Comments on fakes • Jets can fake lepton/b signature – not very probable – but "jets are everywhere" • we'll come back to this later • A major component of the physics program at a collider is to develop robust criteria for lepton/b identification – efficiency vs rejection • Generally not emphasized much when you hear a talk at a conference – experimentalists take it "as a given", (some) theorists are oblivious to it • Some appreciation of this required if you want to participate to the LHC program – A paper suggesting to look at a signature for NP which is swamped by instrumental backgrounds is (nearly) useless 37 • Nearly: do not underestimate the cleverness of your exp. when challenged

Score. Card on objects Object Notes Typical eff Jet fake rate e Excellent resolution Improves with E ~ 90% ~ 10 -4 -10 -5 Excellent resolution Degrades with E ~ 90% ~ 10 -4 -10 -5 So-so resolution ~ 50% ~ 10 -2 - 10 -3 Excellent resolution Improves with E ~ 90% ~ 10 -3 -10 -4 Jet Poor resolution (> 10%) ~ 100% if above threshold - ~ 50% ~ 1% Btag MET Depends on everything else in the event! 38

Outline 1. Detectors 101 • • How do these detectors work in broad terms, capabilities, limitations Physics objects 2. Physics “landscape” at the LHC • Cross sections, rates for SM processes and generic NP 3. Searches for NP • • • How are they done, what are the ingredients Issues on theoretical side Case studies 39

Hadron Collider LHC opens up new energy regime – 14 Te. V in CM vs 2 Te. V at the Tevatron – With higher luminosity – I will not belabor this point… Hadron collider = collisions of two broadband beams of partons (q, q, and gluons) 40 http: //www. physics. nmt. edu/~raymond/classes/ph 13 xbook/img 2047. png

Parton-parton luminosities • A way to think about this and develop a semi-quantitative intuition: parton-parton luminosities • Reminder: luminosity is a measure of the intensity and brightness of the beam: N=L • Define "effective luminosity" for parton-parton collisions as a function of the ECM of the parton-parton system • Parton-parton x-section, i+j X: • pp (or pp) x-section, pp X or pp X: (the sum is over all the i's and j's that result in X) 41

Parton Distribution functions fi(x) = probability of parton i having momentum fraction x cannot be calculated from first principles Go to http: //durpdg. dur. ac. uk/HEPDATA/PDF click around and make your own plots. It’s pretty cool! The LHC is mostly a gluon-gluon collider! 42

Parton-parton luminosities (2) • We had: pp (or pp) x-section, pp X or pp X: • Rewrite it as: Luminosity for parton-parton collisions as a function of parton-parton ECM EHLQ RMP 56 579 (1984) 43

Parton-Parton luminosities (3) gg luminosity @ LHC qq luminosity @ LHC gg luminosity @ Tevatron qq luminosity @ Tevatron 44

Zooming-in on the < 1 Te. V region gg luminosity @ LHC qq luminosity @ LHC gg luminosity @ Tevatron qq luminosity @ Tevatron 45

Ratio of LHC and Tevatron parton luminosities LHC vs Tevatron gg qq 1 st (simplistic) rule of thumb: – For 1 Te. V gg processes, 1 fb-1 at FNAL is like 1 nb-1 at LHC – For 1 Te. V qq processes, 1 fb-1 at FNAL is like 1 pb-1 at LHC 46

Another rule of thumb: d. L/d falls steeply with ECM • In multi-Te. V region, ~ by factor 10 every 600 Ge. V • New states produced near threshold • Suppose you have a limit on some pair-produced object, M > 1 Te. V • How does your sensitivity improve with more data? LHC gg qq Answer: by ~ (600/2)=300 Ge. V = 30% for 10 times more lumi 47 Improving sensitivity with lumi is tough. . but you might turn a hint into discovery

Cross Sections T. Han Tev 4 LHC Good to keep these in mind when thinking about NP 48

Total x-section: pp anything • Three components 1. pp pp (elastic) 2. pp p. X (diffractive) 3. pp X (inelastic) • Don’t know to 10 -20% – – • Order 100 mb Will measure it (TOTEM) Interaction rate: – N=L¢ • For L = 1033 cm-2 sec-1 : N = 100 MHz 49

What does the average event look like? Minimum Bias events: events selected with a minimal interaction trigger Trigger when “something happened”, ie, the protons broke up d 3 p/(2 E) = ½ d. PT 2 dy d About 6 charged particles per unit rapidity (+ neutral) Of order 90 particles in detector acceptance Plots stolen from: Czech. Jour. of Phys. , Vol. 54 (2004), Suppl. A, p 221 50

More min bias One min-bias event in CMS HCAL http: //home. fnal. gov/~sceno/jpg/minutes/aug 121999/dg_doc. pdf Lots of soft particles 51

Moving on to harder scatters: jets 52

Jets Jet rates are enormous…. Compared to anything else that is interesting 53

A two jet event from D 0 (di-jet) Two jets back-to-back in Note: 45 Ge. V of MET Also: looks like there is more stuff in the event (underlying event) 54

More on jets • At lowest order: 2 back-to-back jets • Higher order diagrams: multijets – Suppressed by ~ s per additional jet • Because of the high x-section for QCD jet processes it is very difficult to look for NP in final states with jets only, eg, H bb – Not entirely true. Interesting to look for deviations from QCD. Some great examples later – Need to key off e, , , MET, etc • A real limitation of hadron colliders • Perhaps the main reason for ILC 55

Underlying event & Pileup • The D 0 di-jet event looked to have additional stuff besides the two jets • This is what you expect. The remnant of the protons after the hard scatter must end up somewhere! They recombine into hadrons and the show up in the detector • A nuisance: underlying event – Looks like min bias (sort of) • Also: pileup – At LHC luminosity there will be on average several interactions per beam crossing (up to ~ 20 at full luminosity) – Even more of a nuisance! • eg high track multiplicity degrades tracking performance 56

Pileup Missing Et resolution degraded by pileup 1 min bias event contribution to MET component in a given direction ~ 6 Ge. V CMS 57

Further down in x-section…. (bb, high PT) ~ 1 b (W l ) ~ 60 nb (tt) ~ 1 nb (WW) ~ 200 pb (tt) ~ 1 nb Te. V scale SUSY ~ pb 58

tt • (tt) huge compared to the Tevatron – 1 nb vs 10 pb • Coupled with the high luminosity, LHC is a top factory – This is a great thing! • On the other hand: tt leptons + jets + MET – This is similar to many possible NP signatures tt is background to many NP searches Yesterday’s hard won discovery is today’s background! 59

Outline 1. Detectors 101 • • How do these detectors work in broad terms, capabilities, limitations Physics objects 2. Physics “landscape” at the LHC • Cross sections, rates for SM processes and generic NP 3. Searches for NP • • • How are they done, what are the ingredients Issues on theoretical side Case studies 60

NP discoveries at the LHC 3 + 1 ingredients 0. Detector and machine: If they don't work, forget it • How fast will they come up? Don't know, but probably not fast. 1. Trigger: If you didn't trigger on it, it never happened • See following discussion 2. Backgrounds: It's the background, stupid • Need to understand SM and instrumental backgrounds • • • Instrumental BG: us (experimentalists, mostly) Physics BG: you (theorists, mostly) There are exceptions. . 3. Searches: If you look for something, you may not find it. But if you don't look, you will never find it 61 • Model independent vs model dependent searches

Trigger (1) • Inelastic cross section O(100 mb) • For low luminosity ~ 1033 cm-2 sec-1, event rate ~ 100 MHz • Data Acquisition Capability: ~ 100 Hz • Most of the events are thrown away! 62

Trigger (2) • The decision on what to trigger on has enormous impact on the physics that we can do • Trigger selects objects (e, , jets. . ) or combinations thereof • All kinematical distributions fall steeply with PT trigger selects objects above a threshold • It is always a compromise • A balance between competing priorities • A source of great debate in the collaboration • If you, as a theorist have a great idea for NP…. . 1. Check that your events have been triggered on 2. If not, try to convince people to devote bandwidth to your theory And your argument better be good… 63

Example of possible CMS 33 trigger menu (L=2 x 10 ) 64

Comments on the trigger menu • Will grow to be much more complicated – More lower threshold prescaled triggers – More triggers with combinations of objects aimed at specific channels – ……. . D 0 W e • Thresholds for di-objects < thresholds for single objects – eg 2 e: ET>17 Ge. V 1 e: ET>29 Ge. V • Some of these thresholds are already pretty high! 65

New Physics discoveries @ LHC Broadly speaking, three possibilities 1. Self Calibrating • e. g. , a mass peak 2. Counting experiments • The number of observed events of some type is >> than the SM prediction 3. Distributions • The distribution of some kinematical quantity is inconsistent with the SM prediction NB: the distinction is not always clean, but still 66 useful to think in these terms

Self Calibrating Signals (SCS) • A NP signal that obviously stands out – where you do not need to know the SM BG very precisely • or do you? • watch out for irrational exuberance • For example: – A mass peak – A huge distortion to some kinematical distribution 67

Reminder: Drell-Yan e- or e+ or + SM Drell-Yan Di-jet background WW background 68

SCS example: Z' What a 100 pb-1 expt might look like Cousins, Mumford, Valuev UCLA 69

Another SCS example: di-jet resonances e. g. , excited quarks, axigluons, E 6 di-quarks, Z', W', . . . Rules of thumb: • If produced strongly about same cross section as QCD at same mass, fairly easy to see • If produced weakly, tougher CMS 70

Di-jet resonances (cont. ) 95% CL Sensitivity to Dijet Resonances CMS Published Exclusion (Dijets) CMS 100 pb-1 CMS 1 fb-1 CMS 10 fb-1 E 6 Diquark Excited Quark Axigluon or Coloron Color Octet Technirho W’ RS Graviton Esen and Harris (FNAL) Gumus and Akchurin (Texas Tech) Z’ 0 1 2 3 4 5 6 Mass (Te. V) 71

(Yet) Another SCS example: di-jet mass distribution • Distorts angular distributions • More scatters at high angles – More jets at high PT – More di-jets at high mass d / d. M or d /d. PT • Like Rutherford scattering, but with quarks! Quark Compositeness New Interactions q q M~L q q Dijet Mass << L Quark Contact Interaction q q L QCD Background q q QCD + Signal Dijet Mass or jet PT If the "edge" is low enough, this could be a relatively easy discovery (Self-calibrating variety) 72

Di-jet mass distribution distortion • Ratio of events at high-low is a sensitive variable that eliminates many syst uncertaintes CMS Left-Handed Quark Contact Interaction + for 100 pb-1 (Te. V) + for 1 fb-1 (Te. V) + for 10 fb-1 (Te. V) 95% CL Exclusion 6. 2 10. 4 14. 8 5σ Discovery 4. 7 7. 8 12. 0 Esen and Harris (FNAL) Gumus and Akchurin (Texas Tech) 73

SCS: Edges 10 fb-1 M(l+l-) 74

Not all that glitters is gold Pentaquarks z(8. 3) Leptoquarks 40 Ge. V top Buyer beware. Especially in the tails of distributions 75

An aside Tail of the jet ET distribution. Definitely not a self calibrating signal (SCS) CDF PRL 77 438 (1996) • Data in the tail not consistent with QCD + (then) existing sets of parton distribution functions (PDFs) • Looks like contact term ~ 1. 6 Te. V • Further PDF analysis found that the discrepancy could be absorbed by modifying gluon distribution – without conflicting with other data – even though all existing PDF fits were "low" • Modern PDFs include uncertainties • A great step forward Example of careful, not-so-glamorous, phenomenological work that has a major impact 76

PDFs with error analysis http: //durpdg. dur. ac. uk/HEPDATA/PDF 77

Counting experiments, distributions • Not all NP signals are as dramatic as a mass peak • Counting experiment – Select events passing some requirements – Count them. Is the number of events consistent with the SM prediction? • Kinematical distributions – Select events passing some requirements – Look at kinematical distributions, eg, the PT of the jets – Are these distributions consistent with the SM prediction? • Of course: not mutually exclusive methods – Ideally: do both Need the SM prediction 78

SM prediction • In some cases the SM prediction can come entirely from the experiment (data driven) • Robust • In other cases the SM prediction relies heavily on theory • Not so robust • And of course the SM prediction must include instrumental backgrounds. . . (fakes) • A couple of examples to understand typical issues 79

Example 1: CDF search for NP in lep + + MET • www-cdf. fnal. gov/physics/exotic/r 2 a/20050714. loginov_Lep. Photon. X/ • A fairly simple final state • Motivated by few weird events in Run 1 (before 1996) • Run 2 (>2002): select events, then compare with SM – both number of events and kinematical distributions • Requires careful accounting of SM sources • A lot of work! – typical for this type of searches – painstaking accounting of many BG sources – you don't just "run the Monte Carlo" 80

SM contributions to lep+ +MET (1) • pp W+jet, W lep , jet fakes – estimated from observed rate of W+jet and measured probability for jet to fake a – difficult (100% uncertainty), but data driven • Drell-Yan e+e- pairs with hard brehmstrahlung, where the electron is lost and looks like a and the MET fluctuates high – estimated from observed rate of Z ee and Z e" " and observed MET distribution – data driven • pp jets, jets fake leptons – estimated from data by relaxing the lepton quality requirements, and extrapolating 81

SM contributions to lep+ +MET (2) • pp W or Z – This turns out to be the main background – Need theoretical input – Tools are: • LO parton level event generators, interfaced to Pythia – yes: more than one • NLO calculation – Good case because the NLO calculation exists • Often it doesn't • The NLO/LO K-factor is ~ 1. 5, but it varies across phase space • The LO MC is then "fudged" to account for that 82

NLO V+ (V=W or Z) 83

Baur, Han, and Ohnemus. PRD 57 (1998) 2823 NLO changes shape of distributions 84

Results of CDF lep + + MET search Decent agreement in shape and normalization Without NLO, SM prediction ~ 26 ± ? What would you have concluded? 85

Results recently updated Stolen from H. Frisch’s talk at Gordy Kane’s Symposium (Jan 2007) 86

Example 2: UA 2 W tb search (1989) • Ancient, but an example of a search based on a shape analysis that is independent of theoretical assumptions – yes, sometimes this happens! Z. Phys. C 46, 179 1990 • Signal is W tb, t e b – M(e ) < MW • BG is W+jets, W e – M(e ) = MW • Cannot reconstruct M(e ) – Because PZ( ) unknown • Next best thing: transv. Mass • Shape: kinematics + MET resolution 87

Example 3: CDF tt evidence (1994) • Also ancient, but example of counting expt independent of theoretical assumptions • Signal: lep + MET + ≥ 3 jets (≥ 1 of them b-tagged) • Background: W+jets (fake b-tag), or Wbb (real b-tag) many more diagrams. . 88

tt search strategy • Count number of events with lep + MET + ≥ 3 jets (≥ 1 of them b-tagged) • Compare with W + jet prediction • If excess signal ! • But where does the prediction come from? • Measure probability of tagging ordinary jets in QCD jet events. Sources are: – Fakes. Should be the same in W + jets events – Real b-quarks. Expect fraction of b-quarks/jet to be larger in QCD jet events than in W+jets • because QCD jet events are mostly gluons and g bb BG prediction overestimated BUT entirely from data! 89

CDF tt evidence results • Good agreement in events with N=1 or 2 jets where tt events are expected to be (almost) absent • Excess in N=3 and 4 jets, where tt is expected to show up • Also: calculate BG from stateof-the-art theory of Wbb (method 2) – Confirm BG calculation from data (method 1) is pessimistic! PRD 50 2966 1994 90

Example 4 From Joe Lykken: Is particle physics ready for the LHC? 91

92

93

94

What was the problem? • Incorrect use of theoretical tools • The general purpose Monte Carlos only have leading order diagrams, eg, q q Z • If you run these event generators you will get events with Z + jets – When Z these are a source of jets + MET events • The jets come from parton showers from the initial state • For Z + 1 moderate PT jet, you get about the right answer • For Z + many high PT jets you get the wrong answer 95

Is everything lost? • No. People have calculated, eg, Z + multijet. • Leading order matrix element calculation only significant uncertainty • And in order for exp to use this calculation, need to interface to shower MC • Lots of tricky issues. Need to be very careful • As an experimentalist, relying on these tools to make a discovery makes me a bit queasy – Many of my colleagues are more optimistic than I am. . . 96

Comments • Often purely data driven BG estimates do not work • SM BG to LO have large normalization uncertainties – Makes counting experiments difficult • SM LO event generators can have large shape uncertainties – Makes shape analyses difficult • What are the uncertainties at LO? at NLO? – Often can get handle from data, e. g. , W+jets vs Z+jets • Where is the smoking gun? – As an experimentalist, more comfortable if uncertainties are under my control – As a theorist, you might feel differently – Don't ask how sausages are made 97

What can theorists do for us? Slides from Z. Bern at LBNL LHC West Coast Theory Network meeting And don't forget to implement them in a MC so that we can actually use them Now that we are about to get data, nuts-and-bolts contributions can be more useful than suggestions for another beyond the SM theory 98

Model dependent vs. model independent searches • Can search for generic NP signatures e. g. , the lep + + MET CDF search described earlier • Or, for very specific, complicated signatures e. g. , pp TT, T t. Z T b. W, t e b Z , W • Because we do not know what the NP is, generic searches are very powerful • But in a generic search worry about missing complicated signature • With O(1000) physicists both approaches will be pursued 99

Ba. Bar Palano (Bari) Ds. J(2317) DS 0 PRL 90 242001 (2003) cs meson. Was expected to have mass ~ 2. 5 Ge. V, decay strongly to DK, and therefore have broad width So: noone bothered to look! This huge signal had been in various data sets for many years!! – What is hiding in the Tevatron data sets? – What was missed by the Tevatron triggers? 100

A case study: tt at the Tevatron • The high PT discovery at Tevatron • Not NP, the ultimate known unknown • Complicated signature, search narrowly focused on expected SM properties Would it have been seen in generic search? • In the high statistics lep+jets channel probably not for a long time – Lots of BG, theoretical tools (W+multijet & Wbb calculations/MC) developed specifically for the search • In the dilepton channel would have slowly emerged as excess of events with jets (and eventually, b-jets) Power of multi-lepton searches 101

If we see NP, can we tell what it is? • Great question – Supersymmetry and the LHC inverse problem (hep-ph/0512190) • Ian Hinchliffe: I’ll discover it first, I’ll think about it on the way to Stockholm, and I’ll tell you on the way back • It does not come as a surprise that figuring out precisely what is going on at a hadron collider is not trivial – Motivation for ILC. . • Not something that keeps me awake at night – Should we be so lucky! 102

NP Interpretation • Emphasis shifts to "Given that you see X, if the NP is Y, you should see Z" – suggestions with Z experimentally impossible not very useful • but do not underestimate your experimental colleagues! – a well developed feel for experimental issues could make a difference 103

Conclusions • LHC may trigger a revolution in particle physics • I hope to have given you a flavor of how searches for NP at the LHC can be carried out • You are beginning your career at an exciting time • You have a golden opportunity to be part of this adventure. Seize it! 104