The nucleon mass in N_f=2 lattice QCD: finite size effects from chiral perturbation theory

The thing worth knowing: this paper derives the finite-volume correction to the nucleon mass in relativistic SU(2) baryon ChPT through O(p⁴) (Eqs. 15, 23, 24), and the derivation is correct as far as I can check. The sum-integral replacement is standard Hasenfratz-Leutwyler, the Bessel reductions are explicit, and the Appendix cross-check against Lüscher's dispersive form is a real independent test rather than decoration. That check is also the most interesting part: they find their leading exponential term is twice Lüscher's, and they say so plainly, attributing the difference to nucleon exchange diagrams in his treatment. I would want to see that resolved by someone, because they note the exponential and the integral nearly cancel in Lüscher's form, so a factor of two is not cosmetic. But they flag it themselves, which is the right instinct. The infinite-volume formula is quoted from Ref. [1]; the finite-volume extension is theirs and is new. The data compilation across UKQCD/QCDSF, CP-PACS and JLQCD is genuinely useful, and the observation that different actions land on one curve is a nice sanity check on lattice artefacts. On the soft spots: the curves in Figs. 3-5 are normalised so that mN(L) matches the Monte Carlo at the largest volume, so what is tested is the shape of the L-dependence, not the absolute prediction. That is a legitimate test but weaker than 'parameter-free prediction' implies, and the agreement is visual, with no error band from the LECs. The stress-test note is right that the curves are never recomputed with Fit 3/Fit 4 or with m0, c1 varied inside their errors. Fit 3 and Fit 4 do show the LECs are reasonably stable, which blunts the concern, but propagating the uncertainty into the curves is a half-day's work and would settle it. The ten-point selection is post hoc, and the r0 = 0.5 fm scale setting is a known weak point that the authors acknowledge. None of this breaks the central claim. Who should read it: anyone doing chiral extrapolations of baryon masses who needs to decide whether small-volume points can be corrected rather than discarded. It deserves referees, and I would cite the formulae.

The finite-size formula Eq. (24) depends on m0 and c1 through both Delta_a (Eq. 15) and Delta_b (Eq. 23), and these are fixed by Fit 1 to ten large-volume points with m_pi up to about 800 MeV (Table 4). The paper's own Fit 3 (four smallest-mass points, m0 = 0.88 GeV, c1 = -0.93 GeV^-1) and Fit 4 (m0 = 0.87 GeV, c1 = -1.11 GeV^-1) show the LECs are reasonably stable, which is partial mitigation. But the curves in Figs. 3-5 are computed only with Fit 1 and are never recomputed with Fit 3/Fit 4 or with m0 and c1 varied within their quoted errors (m0 = 0.89(6) GeV, c1 = -0.93(5) GeV^-1). Because each curve is anchored to the largest-volume Monte Carlo point, only the shape of the L-dependence is tested, and the shape's sensitivity to m0 and c1 is not quantified. If the LECs shift within their errors, the predicted curves could move by amounts comparable to the data errors (point 31 carries +/-82 MeV), which would weaken the 'remarkable agreement' of abstract:s3. The paper acknowledges the convergence question in Section 6 but does not close this specific loop by propagating the LEC uncertainty into the finite-size curves.

The derivation is transparent and the Appendix cross-check against Luescher's formula is an honest independent test, including the reported factor-of-two discrepancy in the leading exponential term. That discrepancy is numerically irrelevant for the data comparison because at m_pi L ~ 4-8 the exponential term is negligible and the agreement relies on the |n| > 1 terms, as the paper itself notes in Section 6. The central claim is supported by Figs. 3-5, where the central values of the smaller-volume points align well with the anchored curves. The remaining soft spot is the bridge from the Fit 1 LECs to the finite-size curves: the paper shows LEC stability via Fit 3/Fit 4 but does not propagate it to the curves, and the visual agreement is not quantified with an error band. This is a real but not fatal gap, and the reader's CONDITIONAL verdict is appropriate. I would keep the verdict unchanged and note that the paper's Fit 3/Fit 4 provide more mitigation than the reader's weakest_assumption credits, since the LECs are already shown to be stable under a stricter mass cut; what remains undone is the final propagation step to the finite-size curves.

The calculation is a genuine first-principles derivation: the finite-volume formula introduces no new parameters beyond those already fixed from the infinite-volume chiral fit, and the L-dependence is therefore independent structure rather than a fitted shape. The Appendix cross-check against Lüscher's formula is a real independent consistency test, and the paper reports the factor-of-two discrepancy honestly rather than suppressing it. The comparison spans four collaborations and several lattice actions, which is a meaningful consistency check, and the prediction is falsifiable by future simulations at smaller pion mass. These are substantive strengths.

Lattice simulations of QCD place the nucleon inside a finite box, and the measured mass drifts with box size because pions wrap around the periodic boundary. This paper computes that drift analytically in relativistic SU(2) baryon chiral perturbation theory through order $p^4$, obtaining a formula (Eq. 24) whose only inputs are low-energy constants already fixed from the pion-mass dependence of large-volume nucleon masses plus a few phenomenological couplings. The result is a parameter-free prediction: once the large-volume curve is anchored, the volume dependence at smaller $L$ is determined. The authors then confront this prediction with N_f=2 dynamical data from UKQCD, QCDSF, CP-PACS and JLQCD, using matched sets of simulations that differ only in lattice size, and find the predicted curvature tracks the data at pion masses of roughly 545, 717 and 732 MeV. If this holds up, finite-volume effects in nucleon-mass calculations become a computable correction rather than an extrapolation uncertainty.

What carries the argument. The engine is the standard sum-minus-integral identity for a massive propagator in a periodic box (Eq. 7, following Hasenfratz and Leutwyler), which converts $L^{-3}sum_{p} - (2pi)^{-3}int d^3p$ of $(p^2+M^2)^{-r}$ into a sum over integer winding vectors $n$ of modified Bessel functions $K_{r-3/2}(L|n|M)$. Applied to the pion-nucleon loop after Feynman parametrisation and a Wick rotation, this turns the finite-volume shift into a rapidly converging Bessel series whose terms are labelled by how many times the pion circumnavigates the box. The same identity, applied to the $O(p^4)$ counterterm graph, gives the $K_1$ and $K_2$ terms weighted by $2c_1-c_3$ and $c_2$. The appendix re-expresses the $|n|=1$ pieces in the dispersive form used by Lüscher, which relates them to the pion-proton forward scattering amplitude, and finds the leading exponential term carries a coefficient twice Lüscher's.

Core claim. The paper's claim is that the volume dependence of the nucleon mass is under control in chiral effective field theory. Two graphs contribute at the stated order: a one-loop pion-nucleon graph giving the $O(p^3)$ piece $m_N(L)-m_N(0) = 3g_A^2 m_0 m_2^2/(16pi^2 f_pi^2) int_0^infty dx sum'_n K_0(L|n| sqrt(m_0^2 x^2 + m_pi^2(1-x)))$ (Eq. 15), and a counterterm graph giving an $O(p^4)$ piece built from $c_1, c_2, c_3$ and Bessel functions $K_1, K_2$ of $L|n|m_pi$ (Eq. 23). Neither piece introduces a new low-energy constant. Fixing $m_0$, $c_1$ and $e_1^r$ from a fit to ten large-volume, fine-lattice points with $m_pi L > 5$ and $m_pi < 800$ MeV (Fit 1: $m_0 = 0.89(6)$ GeV, $c_1 = -0.93(5)$ GeV$^{-1}$, $e_1^r = 2.8(4)$ GeV$^{-3}$), the authors overlay the resulting curves on three same-coupling, different-volume data sets and report agreement. They also state that truncating the winding sum to $|n|=1$ would spoil that agreement, so pions crossing the box more than once matter numerically.

The premise it rests on. The low-energy constants fitted to infinite-volume nucleon masses at pion masses up to about 800 MeV are assumed to be the same constants that control the finite-volume formula, so any higher-order contamination of that fit passes straight into the supposedly parameter-free prediction.

Finite-volume corrections to the nucleon mass at a given pion mass and box size become a calculable number rather than an extrapolation ansatz, provided the low-energy constants are known.

The same constants that describe the pion-mass dependence of $m_N$ also describe its volume dependence, so lattice groups can cross-check their chiral fits against their finite-size systematics.

Because the winding sum must include $|n| > 1$ to match the data, analyses that keep only pions crossing the box once will underestimate the effect.

The comparison with Lüscher's formula isolates a factor-of-two difference in the leading term, attributed by the authors to nucleon-exchange diagrams that their chiral loop does not include.

If the chiral finite-volume series is trustworthy at these pion masses, the same machinery should predict volume shifts for other baryon masses and for matrix elements such as $g_A$; agreement or disagreement there would be an independent stress test of the constants extracted here.

The factor-of-two mismatch with Lüscher's leading term is a sharp, checkable discrepancy rather than a rounding issue, since the authors note the exponential and the integral nearly cancel in Lüscher's bookkeeping; a dedicated lattice study at very large $m_pi L$ where only the leading term survives could discriminate between the two bookkeepings.

The paper's reliance on $r_0 = 0.5$ fm to set the scale is flagged by the authors themselves as a weak point; recasting the analysis in dimensionless ratios, for instance $m_N/f_pi$ versus $m_pi/f_pi$, would remove that dependence and could be done with the same data tables.

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