Distillation at High-Momentum

You should know this paper has one genuinely useful idea and one half that doesn't yet land. The idea: in distillation, you can momentum-smear simply by multiplying precomputed Laplacian eigenvectors by a single phase factor e^{i zeta·z}. The authors show this is the only phase choice among their candidates that preserves translational invariance, and that's a clean, correct argument. On the a094m358 ensemble, they get ground-state nucleon energies out to |p| ~ 3 GeV with a variational basis, and the dispersion relation is consistent. That part is convincing and reproducible in structure.

The soft spot is the matrix-element section, exactly where the abstract makes its second claim. For apz = 4, the phased three-point functions do have dramatically better signal than unphased ones, but the extracted scalar charge is 25% below the rest-frame value, and the authors themselves call it 'dubious.' The axial ratios don't plateau and the gamma3gamma5 and gamma4gamma5 channels trend apart. The Vz determination is constrained by only two source-sink separations. This traces to the two-exponential ansatz (Eqs. 14-15) fit over a narrow T window without three-state tests. The stress-test note is right: the phasing construction isn't the risk; the completeness of the two-state model at high momentum is.

To be fair, the authors flag these limitations in the text. They don't oversell the charges as final numbers; they present them as evidence the method can be pushed. But the abstract's phrase 'matrix elements featuring a large momentum dependence can be resolved' is stronger than the data show. It should be toned down to something like 'can be accessed with improved signal, pending control of excited-state contamination.'

Citation pattern is fine: they build on Bali et al. and Peardon et al., and the relevant prior work is cited. No self-citation inflation.

For lattice practitioners working on hadron structure and boosted-frame observables, the energy-side result is worth citing; the charge determinations are not yet citable as numbers. I'd send it to a serious referee—it's a methods paper with a real advance and honest reporting—and ask specifically for scrutiny of the matrix-element fits. With more statistics and three-state fits, the method could deliver what the abstract claims.

Load-bearing for the abstract's matrix-element claim is the assumption that the two-exponential parametrizations (14)–(15) fully capture excited-state contamination in the phased apz = 4(2π/L) three-point functions. This is not established. The phased high-momentum extraction uses T_max_fit = 12, and the authors state the vector Vz determination is 'constrained by only two values of T' (Sec. V.B). Within this data set, Eq. (15) has amplitudes A, B, C plus ΔE; with only two source–sink separations, a missed third state can be absorbed into A, biasing gΓ. The paper itself reports the phased scalar charge differs by 25% from the apz = 0,1 values and calls the extraction 'dubious' (Sec. V.A), while the axial ratios do not plateau and the γ3γ5 and γ4γ5 channels trend apart (Sec. V.C). The Conclusions also note that the phasing 'improves overlap onto unwanted single- and multi-particle excited states.' These in-paper admissions locate the central risk: not the phasing construction, but the completeness of the two-state ansatz for highly boosted matrix elements.

I agree with the reader's weakest-assumption: the two-exponential ansatz is the fragile point for the matrix-element claim. The phasing idea itself has independent support: the type-1 translational-invariance calculation is clean, the dispersion relation is consistent out to roughly 3 GeV with two operator bases, and the V4 check (ZV g_V4 = 0.982(18) in the phased frame) is a useful control. None of that, however, validates the charge extraction in the highly boosted frame, where the paper's own caveats are explicit and the 25% scalar-charge discrepancy is unresolved. A three-state reanalysis with additional T values is the decisive check: if the scalar and axial charges shift significantly, the abstract's claim that matrix elements with large momentum dependence can be resolved must be narrowed to energies and low-momentum matrix elements; if they do not shift, the current conditional acceptance stands, though the axial plateau issue should be revisited with more statistics. I therefore keep the reader's CONDITIONAL verdict unchanged.

If the energy-sector result holds, this is a genuinely useful technical advance: it allows the variational operator bases and reusable perambulators of distillation to be combined with high-momentum kinematics, which is important for LaMET, pseudo-PDF, and lattice-cross-section calculations of hadron structure. The derivation of translational invariance for the single-phase construction (Section II.A) is explicit and correct, and the phase factors are chosen rather than fitted, so the circularity burden is low. The energy extraction is validated against external expectations, including the continuum dispersion relation, consistency between two operator bases, and, for the matrix elements, the Z_V g_V = 1 normalization at zero momentum. However, the abstract's claim that 'matrix elements featuring a large momentum dependence can be resolved' is not supported by the present three-point analysis; the high-momentum charge determinations exhibit unresolved excited-state contamination and, in the scalar and axial channels, frame dependence that the authors themselves call dubious or superficial.

This paper claims that momentum smearing can be realized inside distillation—the lattice-QCD smearing scheme built from eigenvectors of the gauge-covariant Laplacian—by multiplying those eigenvectors by a single coordinate-dependent phase factor, and that only this single-phase variant preserves translational invariance. It reports that with one and two units of allowed lattice momentum applied to the eigenvectors, a variational analysis with an expanded nucleon operator basis reliably extracts ground-state nucleon energies up to roughly 3 GeV on a $32^{3}$ x 64 lattice with pion mass 358 MeV, whereas unphased distillation loses the signal by lattice momentum $ap_z=4(2\pi/L)$. It further shows that the same phased framework yields usable nucleon isovector charges at momentum $ap_z=4(2\pi/L)$, while unphased three-point correlators at that momentum are noise-dominated. A reader should care because this lifts the low-momentum ceiling on distillation-based hadron-structure calculations, which need matrix elements over a wide range of momenta.

What carries the argument. The central object is the phased eigenvector basis $\tilde{\xi}^{(k)}_a(\vec z,t)=e^{i\vec\zeta\cdot\vec z}\xi^{(k)}_a(\vec z,t)$, obtained by multiplying each of the $R_D$ gauge-covariant Laplacian eigenvectors by a single plane-wave phase. The perambulators built from these phased vectors remain translationally invariant, because the phase at the sink cancels the conjugate phase at the source; the paper proves this cancellation and shows that the other proposed phasings—opposing phases, identity plus opposing phases, and multiple unidirectional phases—do not survive translation. This mechanism shifts the momentum-space overlap of the distillation kernel without changing the group-theoretic construction of the interpolating operators.

Core claim. The central claim is that attaching a single phase factor to the precomputed Laplacian eigenvectors realizes momentum smearing inside distillation, and that this is sufficient to push the practical momentum ceiling of the method far beyond where unphased distillation fails. With one unit of the allowed lattice momentum, the phased variational principal correlators are notably better determined than unphased ones; with two units, ground-state nucleon energies can be resolved up to $ap_z=6(2\pi/L)$, corresponding to $p_z\simeq 3$ GeV, and the dispersion relation is consistent with continuum expectations over that range. The same construction is then applied to nucleon three-point functions: at $ap_z=4(2\pi/L)$ unphased correlators are noise-dominated, while phased correlators yield renormalized isovector vector and tensor charges consistent with rest-frame values. The scalar charge in the phased high-momentum frame comes out about 25 percent lower than the rest-frame value, which the authors leave as an open systematic question rather than a resolved discrepancy.

The premise it rests on. The results assume that the two-exponential fits used for the boosted-frame energies and charges describe all excited-state contamination in the fitted time range, so the extracted ground-state values are unbiased.

Distillation can now be used for hadron-structure calculations at momenta up to about 3 GeV, more than doubling the roughly 1.75 GeV ceiling of unphased distillation on the same lattice.

A phased basis with one unit of allowed momentum covers a broad momentum window rather than a single value, so a few recomputed perambulators serve many boosted frames.

The variational operator analysis remains valid after phasing; the principal correlators show single-state dominance, and two different operator bases give consistent dispersion relations.

Isovector vector and tensor charges extracted in the phased $ap_z=4(2\pi/L)$ frame agree with rest-frame values, whereas the unphased three-point functions carry no usable signal.

The scalar charge at that same momentum differs from rest-frame values by about 25 percent and the axial effective-charge ratios do not plateau, so the method does not by itself remove excited-state contamination at high momentum.

If the single-phase recipe transfers to mesons and to non-forward matrix elements, one perambulator set could support form-factor and distribution calculations over a range of momenta rather than a separate calculation for each frame.

The translational-invariance failure of the multi-phase variants suggests a route for future work: realize simultaneous multi-momentum overlap through a gauge-covariant construction rather than fixed plane-wave phases, avoiding the cost of a separate basis per momentum.

The 25 percent scalar-charge deficit is the sharpest stress test of the method; a dedicated excited-state analysis at several more source–sink separations would show whether high-momentum ground-state matrix elements can be trusted without frame-by-frame form-factor control.

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