Localized 4σ and 5σ dijet mass excesses in ALEPH LEP2 four-jet events

Abstract

We investigate an excess observed in hadronic events in the archived LEP2 ALEPH data. This excess was observed at preselection level during data-MC comparisons of four-jet events when no search was being performed. The events are clustered into four jets and paired such that the mass difference between the two dijet systems is minimized. The excess occurs in the region M 1 + M 2 ∼ 110 GeV; about half of the excess is concentrated in the region M 1 ∼ 80 GeV, M 2 ∼ 25 GeV, with a local significance between 4.7σ and 5.5σ, depending on assumptions about hadronization uncertainties. The other half of the events are in a broad excess near M 1 ∼ M 2 ∼ 55 GeV; these display a local significance of 4.1 − 4.5σ. We investigate the effects of changing the SM QCD Monte Carlo sample, the jet-clustering algorithm, and the jet rescaling method. We find that the excess is remarkably robust under these changes, and we find no source of systematic uncertainty that can explain the excess. No analogue of the excess is seen at LEP1. than either of the masses measured separately; such a plot also displays a nice separation of QCD and W + W − events, useful for data-MC comparisons.Unlike what was typical during the LEP era, we plotted M 1 + M 2 using the entire ALEPH LEP2 data sample (∼ 735pb −1 ) simultaneously; with our very loose cuts, this amounts to O(17, 000) events. In doing so, we noticed an excess of about 200 events (∼ 3σ) in the dijet mass sum in the region M 1 + M 2 ∼ 110 GeV. Interestingly, this is the same region in which ALEPH observed an excess in the search for hA final states [2] with a much smaller dataset (5.7pb −1 ) and after analysis cuts; the excess was later dismissed as a statistical fluctuation [3] 3 . This excess brought with it a set of unusual challenges. The first of these is that we had no model on which to base selection cuts for an analysis. Second, as this excess was found during data-MC comparisons at preselection level when no search was being conducted, we had to develop the machinery of the analysis after knowing of the excess; the analysis is unavoidably unblind.Were the experiment still running, a simple solution to these challenges would be to explore the features of the excess region, design a set of cuts which efficiently selects the excess, and then see if such cuts also yield an excess in future data. Unfortunately, this is not possible in our case. However, there are three additional LEP data sets which could potentially be used to confirm or refute our findings. With this in mind, we follow the philosophy of Ref.[2] and set about cataloging the features in the excess region to develop analyses which can be used by the three other LEP experiments. In this paper, we will concentrate on establishing the location of the excess, its significance, and its dependence upon MC samples, jet-clustering algorithm, etc. We will reserve further details, including distributions of many observables in the excess region, for future work.At the time that we initially noticed the excess near M 1 + M 2 ∼ 110 G...

Searching in CMS open data for dimuon resonances with substantial transverse momentum

“…We repeat the procedure, increasing or decreasing the smearing by an amount that is independent of p T and η, to obtain other choices of resolution that we need later in our statistical analysis. 10 Our generated statistics are high enough that we may use the smeared MC as our prediction. As a check, we studied smoothing our prediction by fitting it with a single-or double-shouldered Crystal Ball function.…”

“…[6][7][8] for machine-learning studies on detector-simulated CMS samples, Refs. [9][10][11] for QCD studies on archival ALEPH data, and Ref. [12] for a diphoton analysis with public LHCb data.…”

Studying QCD modeling of uncertainties in particle spectra from dark-matter annihilation into jets

“…It was found that different MC event generators have excellent agreement in both the peak as well as in the bulk of the spectra while they can have very large differences in the tails [19]. We stress out that the excellent level of agreement between the different MC event generators is simply due to the fact that they are tuned to the same set of data comprising of LEP measurements at the Z-boson pole [20,21,22,23,24,25,26,27,28,29]. To assess this, we perform a comprehensive analysis of QCD uncertainties on both the fragmentation function parameters and the perturbative shower model using an alternative method and within the same model wherein the default Monash 2013 tune [25] of the Pythia version 8.2.35 event generator [30] is used as our baseline.…”

Particle spectra from dark matter annihilation: physics modelling and QCD uncertainties

“…Photons from π 0 → γγ dominate bulk (and peak) of spectra most-populated regions of particle spectra may be interpreted as due to the fact that the different MC generators tend to be tuned to roughly the same set of data mostly coming from LEP measurements at the Z-boson pole [25][26][27][28][29][30][31][32][33][34]. Therefore, the envelope spanned by the predictions of the different MC models cannot represent the true estimate of the uncertainty on the predicted spectra.…”