Combination of CDF and DO results on the mass of the top quark using up to 8.7 fb^-1 at the Tevatron

This is the March 2013 Tevatron top-mass average, and it does what a combination note should do: it puts every input's uncertainty into fourteen named categories (Table 1), publishes the resulting 12×12 correlation matrix (Table 2), and states each correlation rule in plain sentences in Section 4. You can check the arithmetic by hand, which is more than most combinations allow. The genuinely new content is narrow but real: the CDF lepton+jets input updated to 8.7 fb⁻¹, the CDF MEt input updated to the full Run II dataset, and a refined correlation treatment. The result, 173.20 ± 0.51 (stat) ± 0.71 (syst) GeV/c², is a modest improvement on the July 2012 average and is not in any prior reference. The cross-checks are the right ones: a 50%-correlation variation, a four-channel split with pairwise chi-squares, and separate CDF and DØ averages with χ²(CDF−DØ) = 2.25/1. That is honest self-testing, and I take it seriously. On the stress-test concern about the Signal category: the paper does say in Section 3 that colour reconnection was never evaluated for Run I, and then takes Signal as 100% correlated across all inputs in Section 4. That is a real asymmetry, but the 50% cross-check bounds the shift at +0.19 GeV, and overstating correlation in BLUE inflates rather than deflates the quoted error, so the direction is conservative. I would not call this load-bearing. Two things I would actually ask them to fix. First, the CDF Run II dilepton input reads 170.28 ± 1.95 (stat) in Table 1 and 170.56 ± 2.19 (stat) in Figure 1, with the same total; eleven of twelve entries match, so it is a transcription slip, and with a weight of −0.3% it changes nothing, but it should be reconciled. Second, the summary attributes the precision limit to the jet energy scale, while Table 3 shows Signal modelling at 0.52 GeV is the largest single systematic, ahead of in-situ JES at 0.36 GeV. That framing is worth correcting. Who should read it: anyone quoting a Tevatron top mass, and anyone building a correlated combination who wants a clean template for how to document category-by-category correlations. It deserves a referee, though the internal review by two collaborations already does much of that work.

Section 3 states that the colour-reconnection (CR) subcomponent of the Signal category 'was not evaluated in Run I' and that Run I measurements 'do not include this source of systematic uncertainty.' Section 4 nevertheless takes Signal to be 100% correlated 'among all measurements,' including Run I–Run II pairs. Signal is the largest single systematic in the combination (0.52 GeV, Table 3), so the correlation model overstates the Run I–Run II Signal correlation. Two facts bound the concern. First, the paper's own 50%-correlation cross-check (Section 5) shifts the mass by only +0.19 GeV and changes the uncertainty negligibly, so the maximum effect of any single category is well below that. Second, overstating correlation in BLUE generally increases the coherent sum (Σ w_i σ_i)² relative to the quadrature sum Σ w_i² σ_i², so the reported 0.87 GeV is conservative on this axis. A hand estimate using the Table 4 weights and Table 1 Signal uncertainties gives a coherent sum of ~0.52 GeV (matching Table 3) and an uncorrelated sum of ~0.44 GeV, so the true Signal contribution lies between 0.44 and 0.52 GeV. The concern is real but does not threaten the headline. Separately, the CDF-II dilepton input appears as 170.28 ± 1.95 in Table 1 and 170.56 ± 2.19 in Figure 1; this is a genuine internal inconsistency, but its weight is -0.3%, so it is negligible for the combination.

The paper's central claim is a BLUE combination of twelve top-mass measurements, and the load-bearing assumption is the correlation model in Section 4. The reader correctly identifies the Signal-category 100% correlation as the weakest link, given the explicit statement in Section 3 that the CR subcomponent was never evaluated in Run I. I agree with that identification. My refinement is about direction: overstating correlation in BLUE increases the coherent sum for a category with predominantly positive weights, so the reported uncertainty is conservative on this axis, not anti-conservative. A hand calculation using the Table 4 weights and Table 1 Signal uncertainties confirms that the 0.52 GeV Signal contribution in Table 3 is the coherent sum, and that the uncorrelated sum would be ~0.44 GeV, bounding the effect to ~0.08 GeV. The paper's 50% cross-check (+0.19 GeV shift, negligible uncertainty change) provides an independent bound. The Table 1/Figure 1 discrepancy for CDF-II dilepton is a real internal inconsistency but carries weight -0.3%, so it cannot move the headline. The paper also acknowledges that most measurements are not yet on the full datasets, which is a limitation on future precision, not on the current claim. The CDF-DØ split (172.72 ± 0.93 vs 174.89 ± 1.42, χ² = 2.25/1, p = 13%) is a mild tension but not a concern for the combined number. Overall, the central claim holds up; the reader's CONDITIONAL verdict is appropriate, and my concern does not move it.

The result is the canonical Tevatron average of the top-quark mass and is a reference number for the field. The paper's methodological strength is transparency: every input uncertainty is broken into the same fourteen categories, the correlation rules are stated one by one in §4, the full correlation matrix is published in Table 2, and two mathematically equivalent combination methods are shown to agree. The per-channel consistency tests (χ² between l+jets, ll, alljets and MEt, all with probabilities between 25% and 79%) and the separate CDF and DØ averages are exactly the right diagnostics. The 50% correlation stress test in §5 is a genuine, if blunt, robustness check. I verified that the quadrature sums in Table 1 reproduce the quoted totals for eleven of the twelve inputs, that Table 2 is symmetric, that the Table 4 weights sum to 100%, and that the Table 3 category uncertainties reproduce the quoted 0.71 GeV systematic. The central value is sound; my major comments concern the documentation and robustness of the correlation model, not the arithmetic.

This paper asks whether the twelve separate measurements of the top-quark mass made by the CDF and DØ experiments between 1992 and 2011 actually agree, and if they do, what single number should stand for them. The answer given is $M_t = 173.20$ GeV/$c^2$, with a statistical uncertainty of $0.51$ and a systematic uncertainty of $0.71$, for a total of $0.87$ GeV/$c^2$ and a relative precision of $0.50$ percent. The agreement check is explicit: $8.5$ for $11$ degrees of freedom, a $67$ percent probability, so no input is badly out of line with the rest. What makes the number worth having is not the central value but the bookkeeping behind it, because the two collaborations first agreed on which uncertainties are shared between channels, runs and detectors and which are not, so that the combination neither double-counts a common systematic nor pretends a shared one is independent. The result is the Tevatron's best single statement of a quantity that sets the scale for electroweak fits and for the calibration of simulated top-quark events.

What carries the argument. The engine is the best linear unbiased estimator (BLUE), run alongside an independent numerical chi-squared minimisation and checked to give identical results. Its input is not just the twelve masses and errors but a correlation matrix: each measurement's systematic uncertainty is decomposed into the fourteen categories defined in the text (six separate components of the jet energy scale, plus signal modelling, background from theory, background from data, lepton modelling, jet modelling, calibration method, offset, and multiple-interaction modelling), and each category is then assigned a correlation rule across the twelve inputs. Statistical and calibration-method errors are taken uncorrelated; the b-jet model, out-of-cone and signal-modelling categories are taken fully correlated across everything; the remaining categories are correlated within an experiment, within a run period, or within a decay channel as the physics dictates. This decomposition is what allows the combination to assign a $62$ percent weight to the single most precise CDF lepton-plus-jets input while still giving small or even negative weights to less precise measurements that carry correlated systematics.

Core claim. The paper's claim is that five published Run I measurements and seven Run II measurements (six published, one preliminary), spanning all-jets, lepton-plus-jets, dilepton and a lepton-free missing-transverse-energy channel plus a tracking-based decay-length measurement, can be combined into one top-quark mass with a defensible total uncertainty. The quoted result is $M_t = 173.20$ GeV/$c^2$, with the uncertainty budget broken into fourteen named categories; the largest single entry is signal modelling at $0.52$ GeV, followed by the in-situ light-jet calibration at $0.36$ GeV, and the whole systematic side sums to $0.71$ GeV. The paper also reports the combination performed channel by channel, giving $M_t^{alljets} = 172.7$, $M_t^{lepton+jets} = 173.2$, $M_t^{dilepton} = 171.0$ and $M_t^{MEt} = 173.8$ GeV/$c^2$ with pairwise chi-squared probabilities between $25$ and $79$ percent, and separate CDF and DØ averages of $172.72$ and $174.89$ GeV/$c^2$ whose difference has a chi-squared probability of $13$ percent. The central value sits $0.02$ GeV/$c^2$ above the previous published Tevatron average, while the total uncertainty improves by $8$ percent.

The premise it rests on. The combination assumes that the signal-modelling uncertainty is $100$ percent correlated across all twelve inputs, even though the colour-reconnection part of that uncertainty was never evaluated for the Run I measurements, and signal modelling is the largest single systematic in the final budget at $0.52$ GeV.

The top-quark mass is fixed to $0.50$ percent relative precision, so the dominant residual uncertainty is no longer statistics but the jet energy scale and the modelling of the simulated signal.

Because the four decay channels give consistent masses, the mass extraction methods used in each channel can be treated as mutually trustworthy and cross-calibrated against one another.

The correlation model provides a template that any future combination, including one that adds LHC measurements, must reproduce or explicitly supersede in order to claim a genuine improvement in precision.

The result constrains the top mass entering electroweak fits, where the top mass and the W mass together predict the Higgs boson mass.

The paper states that further improvement awaits the finalisation of all CDF and DØ channels on the full Run II data set, so the quoted uncertainty is a waypoint rather than an endpoint.

The paper's own cross-check, in which all non-diagonal correlations are cut from $100$ percent to $50$ percent and the central value moves by $+0.19$ GeV/$c^2$, suggests the correlation model itself carries an uncertainty of roughly a fifth of a GeV that is not included in the quoted error bar.

The negative weights assigned to several inputs mean the Tevatron average is not a weighted mean of twelve numbers; a reader who averages the central values naively will get a different answer, which is a useful test of whether the correlation treatment is doing real work.

Because the colour-reconnection component of signal modelling was never evaluated for the Run I measurements, a modern re-analysis of those old data sets with a generator that can vary the reconnection model would directly test the one correlation assumption the paper cannot check.

The $13$ percent probability for the CDF-versus-DØ difference is low enough that a future combination with more precise inputs could turn that mild tension into a real discrepancy worth investigating.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review