CDF Logo Global Search for New Physics in 2.0 fb-1 of p-pbar Collisions at sqrt(s)=1.96 TeV
Public Page
Exotics Logo

Data collected in Run II of the Fermilab Tevatron are searched for indications of new electroweak scale physics.
Rather than focusing on particular new physics scenarios, CDF data are analyzed for discrepancies with the Standard Model prediction.
A model-independent approach (Vista) considers gross features of the data, and is sensitive to new large cross section physics.
A quasi-model-independent approach (Sleuth) emphasizes the high-pT tails, and is particularly sensitive to new electroweak scale physics.
An algorithm has been developed to search invariant mass distributions for 'bumps' that could indicate resonant production of new particles.
This combined global search for new physics in 2.0 fb^-1 of pbarp collisions at sqrt(s)=1.96 TeV reveals no indication of physics beyond the Standard Model.



Blessed Plots and Tables:

Vista plots and tables:

Sleuth plots and tables:

Bump Hunter plots and tables:

Vista Summary Information

Vista summary of final state populations
EPS GIF
Top ten Vista final states
The calculation of &sigmat includes the trials factor
EPS GIF
Graphical summary of Vista final state population discrepancies
No final state shows a significant population discrepancy, after accounting for the trials factor
EPS GIF
Graphical summary of Vista kinematic variable distribution discrepancies
Interest is focused on the outliers, representing distributions with significant discrepancy
However, after analysis, none of these discrepant distributions motivate a new physics claim
EPS GIF

Vista Final State Examples

Vista electron-jet final state: M(e,j)
EPS GIF
Vista muon-jet final state: muon pT
EPS GIF
Vista photon-jet final state: photon pT
EPS GIF
Vista Z production: e+ pT
EPS GIF
Vista W production: W pT
EPS GIF
Vista b-jet final state: b pT
EPS GIF
Vista tau-jet final state: tau pT
EPS GIF
Vista electron-photon final state: e pT
EPS GIF
Vista electron-photon final state: photon detector Eta
EPS GIF
Vista final state with largest deficit of data:
b-electron-pmiss: missing pT distribution
EPS GIF
Vista final state with largest excess of data:
photon-tau: tau pT
EPS GIF

Vista Discrepancies

Below we show a few representative plots for the "three jet" discrepancy. The most likely hypothesis for this effect is that our parton shower model doesn't model the Tevatron data well. This discrepancy remains currently unresolved.
Vista 3 jet low sumPt final state: deltaR(j2,j3)
EPS GIF
Vista 3 jet final state: mass (j1)
EPS GIF
Vista 1 electron, 2 jet final state: deltaR(j1,j2)
EPS GIF
Vista 1 b-jet, 1 electron, 1 jet final state: deltaR(j,b)
EPS GIF
Vista 1 b-jet, 2 jet low sumPt final state: deltaR(j2,b)
EPS GIF
Vista 1 b-jet, 2 jet low sumPt final state: deltaR(j2,b)
EPS GIF
Another cause of shape discrepancy in Vista
is difficulty modeling intrinsic kT of the colliding system.
This is illustrated by this plot of unclustered energy for the 2j final state
EPS GIF

Vista Plots Relevant for Sleuth's Most Discrepant Final States

Below are some Vista distributions relevent to the top Sleuth final states, all involving same sign electron muon final states.
Vista 1e+ 1mu+ final state: e+ detector eta
EPS GIF
Vista 1e+ 1mu+ 1pmiss final state: e+ detector eta
EPS GIF
Vista 1e+ 1mu+ final state: e+ detector eta
EPS GIF
Vista 1e+ 1mu+ 1pmiss final state: e+ detector eta
EPS GIF
Below we show analogous distribution for opposite sign electron muon final states. They do not display significant discrepancies indicating that the issue is unlikely to be due to a misestimate of the jet to electron fake rates.
Vista 1e+ 1mu- final state: detector eta.
EPS GIF
Vista 1e+ 1mu- 1pmiss final state: e+ detector eta
EPS GIF
Vista 1e+ 1jet 1mu- final state: e+ detector eta.
EPS GIF
Vista 1e+ 2jet 1mu- 1pmiss final state: e+ detector eta
EPS GIF
Vista electron-jet final state: electron detector eta. This plot similarly shows that the jet to electron fake rate is well constrained.
EPS GIF
Below we show the muon pT distributions for some Vista final states relevent to the top Sleuth final states. The excesses predominantly appear at large muon transverse momentum.
Vista 1e+ 1mu+ final state: mu+ pT
EPS GIF
Vista 1e+ 1mu+ 1pmiss final state: mu+ pT
EPS GIF
Vista 1e+ 1jet 1mu+ final state: mu+ pT
EPS GIF
Vista 1e+ 2jet 1mu+ 1pmiss final state: mu+ pT
EPS GIF

Vista Tables

Vista correction factor definitions
Values and errors obtained from global fit to data
These are only applicable within the context of the Vista correction model
EPS GIF

Sleuth Summary Information

TildeScriptP, which is the probability that a final state with a scriptP as small as the smallest scriptP observed would appear in pseudo-data.
EPS GIF
Observed distribution of scriptP from all Sleuth final states (data)
The deviation from a uniform distribution illustrates the known remaining imperfections in the correction model used to obtain the SM estimate
Conglomeration around 0 (1) indicates probable under(over)-estimation of the SM background in these final states
EPS GIF
Observed distribution of scriptP from all Sleuth final states (data)
Same as above, but in a Gaussian mapping to reveal values closer to 0 and 1
EPS GIF
Illustration of the Sleuth &Sigma pT variable in the top quark lepton+jets channel with at least 1 b-tag
EPS GIF

Sleuth's Most Discrepant Final States

ScriptP values for Sleuth's top 5 most discrepant final states
EPS GIF
Sleuth #1 discrepant final state
EPS GIF
Sleuth #2 discrepant final state
EPS GIF
Sleuth #3 discrepant final state
EPS GIF
Sleuth #4 discrepant final state
EPS GIF
Sleuth #5 discrepant final state
EPS GIF

Sleuth Sensitivity

Sleuth Result for the 1e+1e-1mu+1pmiss final state with the standard model WZ background removed. This shows the observed sensitivity for WZ production using Sleuth. After removing the WZ background, Sleuth finds a scriptP value of 5.7e-05, and a tildeScriptP of 0.01.
EPS GIF

Bump Hunter Summary Information

Overview of significances of most interesting bumps in all mass variables. Each entry corresponds to one mass distribution that was scanned, and was found to contain at least one bump satisfying quality criteria. The quantity distributed is Pa, the probability for the most significant bump in a distribution to be surpassed in significance by a bump in pseudo-data pulled from the same distribution. These probabilities are transformed to units of standard deviation, using the formula Pa=0.5*[1 - erf(&sigma/sqrt(2))]. The values corresponding to less than 4.5&sigma have been estimated approximately. When applied on psudo-data pulled from the mass distributions, the approximate method of estimation of Pa returns values distributed according to the black curve. Values corresponding to more than 4.5&sigma are estimated with a more accurate method. When that method is applied on pseudo-data pulled from the mass distributions, the resulting values of Pa are distributed according to a Normal distribution. Therefore, values of Pa corresponding to more than 4.5&sigma can be translated directly to significance, since their expected distribution is a Normal distribution. About 5000 mass distributions are considered, which means that to have an effect of significance 3&sigma after all trials factor, it needs to have a significance of 5&sigma or more in this scale of Pa.
EPS GIF

Bump Hunter Results


The only significant bump found, which is in the final state with 4 jets and SumPt < 400 GeV. Its significance is 4.1&sigma after accounting for the trials factor.
EPS GIF

Zoom in at the significant bump found in 4j_sumPt0-400.
EPS GIF

p-values of all bump windows accross mass(j1,j2,j3,j4) in 4j_sumPt0-400.
EPS GIF

Final state 4j_sumPt0-400 is affected by soft final state radiation, which is not modelled well by Pythia, as this analysis has demonstrated first in the 3j_sumPt0-400 final state.
EPS GIF

The "3-jet" effect, appearing as an excess of data with the third leading jet (j3) at low pT.
EPS GIF

The "3-jet" effect, appearing as an excess of data with the fourth leading jet (j4) at low pT.
EPS GIF

The "3-jet" effect, appearing in mass(j1,j2,j3) in the 3j_sumPt0-400 final state. The visible excess does not qualify as a bump, because it is too wide and sidebands are discrepant.
EPS GIF

The "3-jet" effect, appearing in mass(j1,j2,j3,j4,j5) in the 5j final state. The most interesting bump, indicated by the blue lines, corresponds to a 1.5&sigma effect after trials factor.
EPS GIF

Bump Hunter Additional Material


The most interesting bump found in the 1e+1e- final state. This represents a slight imperfection in modelling the Z-peak, which appears in the Bump Hunter because statistics are very enhanced, therefore even a tiny mismodelling appears significant. Still, the "bump" found here has significance only 2.7&sigma after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in 1e+1e-. This is clearly not new physics, but a tiny mismodelling with no effect, other than appearing in the results of the Bump Hunter.
EPS GIF

p-values of all bump windows accross mass(e+,e-) in 1e+1e-. Apart from the uninteresting feature at the Z-peak, nothing else is significant in this mass spectrum.
EPS GIF

The most interesting bump found in the 1mu+1mu- final state. This represents a slight imperfection in modelling the Z-peak, which appears in the Bump Hunter because statistics are very enhanced, therefore even a tiny mismodelling appears significant. Still, the "bump" found here has probability 74% to appear from a random fluctuation, is therefore statistically insignificant.
EPS GIF

Zoom in at the most interesting bump found in 1mu+1mu-. This is clearly a non-significant excess, though it is the most significant one within this mass variable.
EPS GIF

p-values of all bump windows accross mass(mu+,mu-) in 1mu+1mu-. Apart from the uninteresting feature at the Z-peak, nothing else is significant in this mass spectrum.
EPS GIF

The most interesting bump found in the high-SumPt di-jet final state. This "bump" is extremely insignificant, with probability over 99% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the high-SumPt 2j final state.
EPS GIF

p-values of all bump windows accross mass(j1,j2) in 2j high-SumPt final state.
EPS GIF

The most interesting bump found in the final state 1b1e+3j1pmiss, which is dominated by top quark production, where one of the two b-quark jets is not tagged. The scanned variable is the mass of the reconstructed top pair. This "bump" is extremely insignificant, with probability almost 100% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the 1b1e+3j1pmiss final state.
EPS GIF

p-values of all bump windows accross mass(ttbar) in final state 1b1e+3j1pmiss.
EPS GIF

The most interesting bump found in the final state 1e+1mu-, with an electron and muon of opposite charge. The scanned variable is the mass of the lepton pair. This "bump" is extremely insignificant, with probability almost 100% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the 1e+1mu- final state.
EPS GIF

p-values of all bump windows accross mass(e+,mu-) in final state 1e+1mu-.
EPS GIF

The most interesting bump found in the final state 1e+1mu+, with an electron and muon of same charge. The scanned variable is the mass of the lepton pair. This "bump" is extremely insignificant, with probability almost 100% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the 1e+1mu+ final state.
EPS GIF

p-values of all bump windows accross mass(e+,mu+) in final state 1e+1mu+.
EPS GIF

The most interesting bump found in the final state 1tau+1tau-, with two opposite charge hadronically decaying 1-prong taus. The interval of this bump, indicated by the blue lines, is only partly included in the drawn interval, which includes the bulk of the distribution, but in another plot we show exactly the region of the bump. The scanned variable is the mass of the lepton pair. This "bump" is very insignificant, with probability about 28% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the 1tau+1tau- final state.
EPS GIF

p-values of all bump windows accross mass(tau+,tau-) in final state 1tau+1tau-.
EPS GIF

The most interesting bump found in the final state 1e+1tau-, with an electron and hadronically decaying (1-prong) tau of opposite charge. The scanned variable is the mass of the lepton pair. This "bump" is has probability about 3% to appear due to a random fluctuation, after accounting for the trials factor.
EPS GIF

Zoom in at the most interesting bump found in the 1e+1tau- final state.
EPS GIF

p-values of all bump windows accross mass(e+,tau-) in final state 1e+1tau-.
EPS GIF

Bump Hunter Sensitivity Tests


Most significant bump after injecting 700 Z' events [Z'250 GeV&rarr charged leptons] on top of pseudo-data following the Standard Model background. That adds 55 signal events to the 1e+1e- final state, causing this bump of significance 3.74&sigma after trials factor. This amount of signal is slightly more than needed to reach the 3&sigma threshold. The actual discovery cost is between 600 and 700 events.
EPS GIF

Most significant bump after injecting 7000 Higgs events [WH(&rarr l&nu+b+bbar)] on top of pseudo-data following the Standard Model background. That adds 15 events to the 2b1e+1pmiss final state, which causes this bump of significance 3.4&sigma after trials factor. This amount of signal is slightly more than needed to reach the 3&sigma threshold. In another attempt, the required signal was 6000 events, resulting in a bump at the same region, with significance 3.1&sigma .
EPS GIF

Most significant bump after injecting 4600 Z'(500 GeV)&rarr ttbar events on top of pseudo-data following the Standard Model background. That adds 41 signal events to the 1b1e+3j1pmiss final state, causing this bump of significance 3.3&sigma after trials factor.
EPS GIF

Most significant bump after injecting 5000 Z'(500 GeV)&rarr ttbar events on top of pseudo-data following the Standard Model background. That adds 47 signal events to the 1b1e+3j1pmiss final state, causing this bump of significance 3.2&sigma after trials factor.
EPS GIF