Very recently, a combination of the precise measurements of the mass of the top quark obtained by the CDF and DZERO experiments at the Fermilab Tevatron collider with those produced by the ATLAS and CMS experiments at the CERN LHC collider has been produced, obtaining a result of 173.34 GeV, which surprised nobody -of course- with a very small total error bar: 0.76 GeV, a mere 760 MeV, not even a proton's mass.

That result was known to be bound to be revised soon, as the business of measuring with as much precision as is possible the mass of the sixth quark is definitely not over yet. The LHC experiments have in fact only started to do it a few years ago, and with the prospects of much larger datasets and higher center-of-mass energy (thus larger top-pair production rates), ATLAS and CMS look forward to getting to O(500 MeV) measurements of the top quark mass soon. You must realize that the issue is not only that of large datasets: the systematic uncertainties can only be reduced with painstaking work based on ancillary datasets, providing jet energy calibrations and lepton momentum scale, etcetera. That is why it takes a long time to squeeze the best results out of the data.

Anyway, DZERO surprisingly produced a few days ago a new single measurement of the top quark mass which is by itself as precise as the world average, and  a bit more precise than any of the inputs (the closest competitor being CMS, which found in 2012 a result with a total uncertainty of 0.82 GeV against DZERO's 0.76 GeV). This comes from the analysis of DZERO's full data sample of Run II proton-antiproton collisions, and the use of a matrix-element technique whereby the top quark mass and the jet energy scale (otherwise the largest single source of systematic measurement) are fit simultaneously, using the internal constraint of the W->jj decay.

In practice, the technique -by now over a decade old- consists in selecting single-lepton decays of top quark pairs, which arise from the decay of one top into lepton-neutrino and b-jet (the lepton-neutrino pair coming from the intermediate decay of the W boson emitted by the top) and the other top into three jets (two of which coming from the decay of the other intermediate W). The presence of the hadronically-decaying W boson, and the very precise knowledge of the W mass, allows to impose that those two jets make the W mass, constraining the mismatch between the jet energy that can be measured in real and simulated jets.

Nothing new here, and indeed the article is short, giving for granted that readers well know the subtleties of the measurement. In 2011 DZERO had performed a similar measurement only using a third of the data sample used today, obtaining a result which very closely matches the one now published, but with a twice as large total uncertainty. That result was M_top=174.94+-1.5 GeV, and the one obtained today is M_top=174.98+-0.76 GeV. What is noticeable is that both results have a central value which is a bit higher than what the other three experiments insist in finding: in fact, CMS recently quoted its experiment average as 172.2+-0.7+-0.1 GeV; ATLAS has two measurements of 172.31+-1.55 GeV and 173.09+-1.63 GeV; and CDF has several results in the 170 to 173 GeV range (see picture below, from the paper on the 2014 world average).

(Above, the top mass measurements used to extract the world average from CDF, DZERO, ATLAS, and CMS are listed, showing the breakdown of statistical, internal jet energy scale, and other systematic uncertainties as error bars of different colour (red, yellow, and blue respectively). At the bottom are the combination (in red) and the separate previous combinations of Tevatron and LHC measurements.)

So what is up with DZERO ? I am led to recall the 1995 discovery announcement of the top quark by CDF and DZERO, where DZERO had to rushedly put together a measurement to avoid losing the train to a joint claim of discovery, once CDF signalled to the Fermilab director that they had a paper ready for submission (there was a seven-day grace period in force on the matter). On that occasion DZERO pulled out a measurement of the top mass at 199 GeV, which was really far off. This was later traced to a mistake in the generation of the Monte Carlo simulations - a innocent-looking parameter had caused the generated transverse momentum distributions of top quarks higher than they should have.

In this case hurry is not a concern - the Tevatron experiments are producing their legacy papers at a comfortable pace. What one is led to question is whether we are in presence of a statistical fluctuation or some other effect, as I find it hard to believe that the other three experiments are right and DZERO is correct; the truth can be in the middle but it would be nice to know what is causing the discrepancy. I am too lazy to do the exact calculation, but I can easily eyeball that if we took out of the 2014 world average (M_top=173.32+-0.76 GeV) the two DZERO inputs (174.94+-1.50 and 174.00+-2.79) we would get a result which is almost two and a half sigma away from the latest DZERO result.

Note that such a "two sigma" effect is indeed significant - we are not dealing with particle bumps here, but with combining results which evaluate their error bars conservatively and where tables of systematic uncertainties are longer than your typical grocery list. It is, in my opinion, some sort of "notitia criminis" of something fishy going on.

So let us look at the breakdown of the uncertainties quoted by DZERO's latest paper. They have a 0.58 GeV statistical uncertainty, which includes the jet energy scale error. The list of systematic contributions is long as I said, and none of them is big enough to be a likely source. One is either led to consider the discrepancy a fluctuation, or to question the internal jet energy scale constraint of the DZERO result, which alone brings in a 0.41 GeV uncertainty to the final result.

Indeed, the DZERO fit returns a jet energy scale of 1.025+-0.005, which means that the fit "prefers" to scale up the jet energies by 2.5+-0.5% with respect to the default. Note that, since the weight of jet energies in determining the top quark is typically of 60%, the 2.5% corresponds to an over 2-GeV increase in the top mass. In other words, if the scale had been fit at 1.00, the DZERO result would be perfectly in line with those of the other experiments. In the paper it is explained (or so I understood) that DZERO is relying entirely, for the jet energy scale calibration in this measurement, to the W bosons in the data, without using the constraining 2% measurement of the scale coming from gamma-jet and dijet balancing as done in previous measurements. I wonder what would have happened if they had used it.

In any case, the new result ought to be soon incorporated in a new world average of the top mass, which at this point would probably end up getting closer to 174 GeV from the down side, with a uncertaintly of less than 700 MeV (the precise values require too much fiddling for my taste - they also depend on assumption on the correlation of different systematic uncertainties, so it is better to leave this job to those who have the macros to spin).

It is nice to observe that years after having stopped collecting data, DZERO (and CDF) are still producing world-class measurements. These collaborations are nowadays heavily undermanned (don't be deceived by the still longish author list - probably 80% of the physicists listed there have a stronger commitment elsewhere nowadays). So, kudos to DZERO today !