The pion mass in finite volume

The thing to take away: Table 2 and Figs. 6–7 are the deliverable, and they are worth having. Colangelo and Dürr take Lüscher's one-particle formula, feed it the NNLO two-flavor ππ forward amplitude from Bijnens et al., and produce the relative shift Mπ(L)/Mπ − 1 with an explicit error budget. The B2k integrals are done analytically (Eq. 26), the R integrals numerically, the LEC dependence is written out in Appendix A, and the correlation matrix in Table 3 is used properly in Eq. 37. That is a complete enough recipe to re-implement, which matters more here than any single number in the table. The comparison against the full one-loop Gasser–Leutwyler result is the right sanity check and they actually run it rather than gesturing at it. No circularity: the finite-volume target is built from infinite-volume inputs throughout. The genuinely valuable part is the negative result. For light pions the leading exponential in Lüscher's formula is not numerically dominant even at L = 4 fm, and they say so plainly instead of burying it. Telling the community that a full two-loop finite-volume calculation is needed below roughly 200 MeV is a real service. Where I would push back: the sentence 'for masses above 200 MeV and L ≥ 2 fm ... our evaluation of the finite size effects is reliable' is stronger than Table 2 supports. At 200 MeV, L = 2 fm the NNLO Lüscher entry is 2.39(27)×10⁻² and their own hybrid estimate is 3.31×10⁻², a 38% spread, right at the corner of the claimed reliable region. At 300 MeV, 2 fm it is 0.95(22) versus 1.12, about 18%. The magnitude claim of a few percent survives comfortably; the reliability claim does not, at the boundary. Related: Eq. 37 propagates only LEC errors. There is no estimate of the p8 truncation, and the r1–r4 resonance-saturation errors at 100% are flagged but never folded into a systematic. The domain where both expansions are demonstrably controlled is closer to MπL ≳ 4. None of this undermines the calculation. The hybrid column is explicitly labelled heuristic, and the authors are candid about the light-pion failure. It is a boundary-drawing quibble, not a flaw in the derivation. Who should read it: anyone quoting finite-volume corrections to lattice pion masses in full QCD, and anyone planning simulations in the 200–300 MeV, 2–2.5 fm window. Worth a serious referee. The revisions I would ask for are a softened reliability sentence and some estimate, even crude, of the truncation error.

The central claim is a tabulated, error-budgeted prediction, so the load-bearing question is whether the quoted uncertainty actually covers the dominant systematics. The error bar in Eq. (37) propagates only the LEC errors; it does not include (i) truncation of the chiral series at p6 (no estimate of the p8 term) or (ii) the neglected subleading exponentials in Lüscher's asymptotic formula (15). The paper's own Table 2 exposes (ii): at Mπ = 200 MeV, L = 2 fm the NNLO Lüscher entry is 2.39(27)×10⁻² while the hybrid 'best estimate' (last entry) is 3.31×10⁻², a 38% spread; at 300 MeV, L = 2 fm it is 0.95(22) vs 1.12×10⁻², roughly 18%. These cells lie inside the region the text calls reliable ('for masses above 200 MeV and L ≥ 2 fm ... our evaluation of the finite size effects is reliable'). The magnitude claim ('at most a few percent') survives, but the reliability claim does not: the domain where both expansions are demonstrably controlled is closer to MπL ≳ 4 (L ≳ 4 fm at 200 MeV). The authors acknowledge the light-pion failure and the heuristic nature of the hybrid column, but the tension between the stated reliable domain and the table's own spread is not resolved. Separately, the p6 constants r1–r4 rest on resonance saturation with 100% errors, so the NNLO term itself is soft, and the ξ-scaling of the p8 term is never bounded.

The reader's weakest_assumption correctly identifies the two truncations (p6 chiral order and leading exponential in Lüscher's formula) as the load-bearing premise, and the reader correctly notes the authors are candid about the light-pion failure. Where I part company is the localization: the reader treats the Mπ ≥ 200 MeV, L ≥ 2 fm region as safe, whereas the paper's own Table 2 shows the hybrid-vs-NNLO spread is 20–40% precisely in that corner. That is the same concern the reader raised, but it reaches further into the claimed-reliable domain than the reader allows. This is a refinement of the domain of validity, not a defect in the method: Lüscher's formula is correctly applied, the CHPT amplitude is the published NNLO result, and the finite-volume target is derived from infinite-volume inputs with no circularity. The paper is acceptable as it stands; the honest adjustment is to state the reliable domain more conservatively (MπL ≳ 4) and to fold chiral-truncation and subleading-exponential systematics into the quoted errors. Because the authors already flag the hybrid column as heuristic and call for a two-loop finite-volume calculation, the concern does not change the verdict, only the sharpness of the reliability statement. Hence UNCHANGED, with the caveat above recorded as a suggested tightening rather than a rejection ground.

The paper supplies something lattice practitioners need and rarely get: a tabulated, error-budgeted, explicitly derived finite-volume correction that can be applied directly to spectroscopy data, together with a candid statement of where it fails. The extension to next-to-next-to-leading order beyond the classic Gasser–Leutwyler one-loop result is the substantive new ingredient, and the use of the analyticity of the infinite-volume amplitude along the imaginary ν axis is a clean and legitimate application of Lüscher's theorem. The comparison with the full one-loop Gasser–Leutwyler expression is a genuine cross-check of the large-L expansion, and the paper is explicit that for Mπ = 100 MeV even L = 4 fm lies outside the formula's domain. The results are falsifiable and reproducible: every integral is either closed-form or defined by an explicit numerical prescription, and the low-energy constants and their correlations are listed. This is a useful reference paper.

Lattice QCD simulations put pions in a box of side $L$, so the mass measured there, $M_π(L)$, is not the infinite-volume mass $M_π$. The paper's aim is to turn that nuisance into a controlled, tabulated prediction: compute the relative shift $M_π(L)/M_π - 1$ for two-flavor QCD as a function of $M_π$ and $L$, with quantified uncertainty, so that lattice groups can subtract it rather than extrapolate blindly. The strategy is to feed the known infinite-volume $ππ$ forward scattering amplitude, evaluated to next-to-next-to-leading order in chiral perturbation theory, into Lüscher's one-particle finite-volume formula, which relates the mass shift to an integral of that amplitude along the imaginary axis. Adding a third chiral order is what makes the convergence check possible: comparing leading order, next-to-leading, and next-to-next-to-leading integrands shows whether the chiral series is under control for a given pion mass. The authors also confront Lüscher's formula with the full one-loop finite-volume chiral expression of Gasser and Leutwyler, which keeps the whole series of exponentials rather than only the leading one, and use the mismatch to mark where Lüscher's asymptotic formula stops being trustworthy. If the paper is right, finite-volume corrections to the pion mass are at most a few percent for $M_π$ above roughly 200 MeV and $L$ above roughly 2 fm, and below 0.1 percent for $M_π L$ around 4 at $L$ around 4 fm.

What carries the argument. The load-bearing device is Lüscher's one-particle formula, $M_π(L) - M_π = -3/(16π^2 M_π L) ∫ dy F(iy) e^{-√(M_π^2+y^2)L} + O(e^{-ML})$, which converts the finite-volume mass shift into an integral of the infinite-volume forward $ππ$ amplitude $F$ along the imaginary axis, where the amplitude is far from its cuts. Everything else is bookkeeping around it: decomposing $F$ into isospin amplitudes, expanding $F(iy) = 16π^2[ξF_2 + ξ^2F_4 + ξ^3F_6]$ in $ξ = M_π^2/(4πF_π)^2$, and reducing the integrals to Bessel functions $B_{2k}(λ)$ with $λ = M_π L$ plus a set of numerically evaluated integrals $R^k_i(λ)$ over the two-loop functions $K^{ππ}_i$. The comparison machinery is the leading-order finite-volume chiral series $M_π(L) = M_π[1 + 12N_f ξ g̃_1(λ) + …]$, whose Bessel sum $g̃_1(λ) = Σ_n 4m(n)/(√n λ) K_1(√n λ)$ retains every exponential and thus exposes what Lüscher's single leading exponential leaves out.

Core claim. The paper establishes a numerical map of the finite-volume pion mass shift in two-flavor QCD, obtained by inserting the order-$p^6$ $ππ$ forward scattering amplitude into Lüscher's formula. The result is Table 2 and Figs. 6 and 7: for $M_π$ between 200 and 500 MeV and $L$ between 2 and 4 fm, the relative shift $R_M(M_π,L)$ ranges from a few percent down to $10^{-5}$, with an uncertainty band from the low-energy constants that grows with the pion mass. The authors' own reading of the convergence pattern is that the chiral series behaves well for light pions and more slowly but still acceptably for heavier ones, and that going from next-to-leading to next-to-next-to-leading order changes the integrand by an amount that is not alarming even at $M_π = 500$ MeV. Equally part of the claim is a negative result stated plainly: for $M_π = 100$ MeV, even at $L = 4$ fm the leading exponential retained by Lüscher's formula is not numerically dominant, so the Lüscher-based numbers are not justified there and only a full two-loop finite-volume chiral calculation would settle the answer. The authors therefore recommend relying on their table only in the region where both expansions can be checked, and they give the full one-loop Gasser-Leutwyler column alongside as a cross-check, plus a combined entry that adds the large-$L$ suppressed exponentials missing from Lüscher's formula.

The premise it rests on. The load-bearing premise is that truncating the chiral amplitude at order $p^6$ and keeping only the leading exponential of Lüscher's asymptotic formula is accurate in the region the authors recommend, which they themselves show fails for $M_π = 100$ MeV even at $L = 4$ fm.

Lattice groups working with two dynamical flavors and $M_π$ above about 200 MeV at $L$ of 2 fm or more can read the finite-volume correction off Table 2 instead of extrapolating in volume.

For $M_π L$ around 4 with $L$ around 4 fm the predicted shift falls below 0.1 percent, so in that corner the lattice mass can be treated as the infinite-volume mass for practical purposes.

The largest corrections, of order one percent, sit at $M_π$ around 200 to 300 MeV with $L$ around 2 to 2.5 fm, which is precisely where precision lattice tests are heading.

The same machinery gives the finite-volume shift of $F_π$, which for two flavors is minus four times the fourth entry of each table cell.

Because Lüscher's formula rests on unitarity in the full theory, the numbers apply to full QCD and not to the quenched approximation.

The paper's negative result is arguably as useful as its table: it defines a boundary in the $(M_π, L)$ plane beyond which any single-exponential finite-volume estimate, however carefully computed, should be distrusted, and that boundary can be quoted independently of the specific chiral order used here.

Since the uncertainty band is dominated by the four order-$p^6$ constants estimated through resonance saturation with 100 percent errors, a modest improvement in those constants would tighten the heavy-pion entries more than any further loop order in the amplitude.

The tabulated shifts are a concrete target for a future two-loop finite-volume chiral calculation: agreement between that calculation and the combined column of Table 2 at $M_π$ around 200 MeV would validate the whole approach in the region where it currently has to be hedged.

The same combination of Lüscher's formula with a chiral amplitude could be applied to other hadron masses whose shift is generated by pion exchange, with the caveat that the parity argument removing the three-particle term is specific to the pion.

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