Angular Momentum on the Lattice: The Case of Non-Zero Linear Momentum

Short version: this is a table paper, and the table is the deliverable. Moore and Fleming work out how continuum states of definite angular-momentum projection mj subduce onto the lattice irreps of the little group of a nonzero momentum vector — the (1,0,0), (1,1,0), (1,1,1) cases that moving-frame spectroscopy actually uses. Johnson and Basak et al. did k=0; Kilcup and Sharpe did nonzero momentum but for staggered fermions. Doing it for Wilson/overlap with parity, and printing the reduction table, is a real if modest contribution. The method is textbook — induce from translations, project with characters — and the outline is right. Nothing here is fitted or assumed; it's a character projection from group structure, so the circularity worry is basically nil. The one caveat is that the dicyclic character tables are taken from cited sources rather than derived in the paper, which is a reproducibility issue, not a circularity one. The harmonic-oscillator section is a toy and the authors say so; the inverse mapping is ambiguous past the lowest levels and they don't oversell it. Where it's soft: Table III is the whole paper, and you can't check it from the paper. Section II.A says the dicyclic character tables 'can be calculated' from the J^P = 1/2^- representation and then points at Lax's book and at ref. [12], which is listed as 'to be published (2006)'. That's an unpublished source carrying the input to the central table. Print Dic2, Dic3, Dic4 and the paper becomes reproducible; as it stands a reader has to trust the arithmetic. There's also a checkable oddity: in the C2 column, mj=0+ and mj=0- both get 'A', though their characters under the C2 rotation differ by sign. The same column gives 2A for mj=1 and 2B for mj=1/2, which doesn't look like what a cyclic double cover produces. Could be a typesetting slip in the arXiv version, but it's exactly the kind of thing an unprinted character table would hide, and the oscillator example only exercises C4v so it wouldn't catch it. On the finite-torus worry: Section III does assert rather than derive that the projection is N-independent. I'd call that secondary — the star sum factorizes into a momentum-conserving delta times a little-group sum, so N-independence is what you'd expect. Worth a sentence of derivation, not a rewrite. Who should read it: anyone doing moving-frame spectroscopy who needs the mj-to-irrep lookup. It deserves a referee, mainly to get the character tables into print and to clear up the C2 column.

The paper's deliverable is Table III: the reduction of O(2) (double cover) onto each little group, which is what makes the inverse mapping in Section IV possible. The reader flags the finite-torus size-independence claim in Section III as the weakest assumption. That is a genuine soft spot, but it is not the most load-bearing one, because the size-independence statement is a standard consequence of the semidirect structure: the character sum over the star factorizes into a translation sum times a little-group character sum, and the translation sum is a Kronecker delta on momentum conservation that is exact on any torus. So the tables are plausibly N-independent even without the exhibited derivation. The more load-bearing gap is upstream: Table III is computed from dicyclic character tables (Dic2, Dic3, Dic4) that the paper never prints. Section II.A says only that 'the character tables for the double covers of the little groups can be calculated using the J^P = 1/2^- representation of SU(2) x Z2 to determine the multiplication tables' and cites Lax's book plus a reference listed as 'to be published (2006)' (ref. [12], Jahn, Negele, Sigaev). A reader cannot reproduce a single row of Table III from the paper alone. This matters because the reader already spotted a checkable inconsistency in the C2 column, where mj=0+ and mj=0- are both assigned 'A' despite having opposite characters under the C2 rotation; that is exactly the kind of error that an unprinted character table would hide. If the dicyclic character tables are wrong or the reduction was done by hand without them, individual entries in Table III could be wrong in ways the harmonic-oscillator example (which only exercises C4v) would not catch. The central claim 'continuum mj decomposes into lattice irrep alpha as tabulated' is only as good as that table, and the table is asserted, not derived in print.

The reader's verdict of CONDITIONAL is defensible and I would not move it. I agree with the reader that the paper is competent standard group theory whose deliverable is a reusable lookup table, and that the harmonic-oscillator ambiguity is the authors' own acknowledged limitation (Section IV says the inverse mapping is 'increasingly difficult' for higher states and 'relies heavily upon knowledge of the continuum states'), so it is not a new finding. Where I part company is on which gap is most load-bearing. The reader puts the finite-torus size-independence assertion (Section III) first. That assertion is stated as a finding rather than derived, which is a fair complaint, but the underlying factorization is standard and the translation sum is an exact momentum-conservation delta on any torus, so I do not think finite-volume admixtures are the place the central claim is most likely to break. The more load-bearing gap is that Table III, the actual product, is computed from dicyclic character tables that are never printed; the paper points to Lax's book and to ref. [12] listed as 'to be published (2006)'. Per the review rules, a cited published source is support, but an unpublished 'to be published' reference is not, and here the character tables are the input to the central table. The reader's own observation of the C2 column inconsistency (mj=0+ and mj=0- both labelled 'A' despite opposite C2 characters) is a concrete symptom of exactly this: an unprinted intermediate that could conceal a real error. That is a checkable, located concern about the central table, and it is the one I would press. It does not rise to a load-bearing error that overturns the method, so UNCHANGED is right; but the concrete test should be a from-scratch recomputation of Table III rather than a finite-N study.

If correct, Table III is a practical lookup for lattice practitioners doing moving-frame spectroscopy: it maps a measured lattice irrep at momentum k to the continuum m_j values that can contribute in the continuum limit. The method (induce from translations, project with characters, Eq. (3)) is standard and the outline is correct. The paper is clearly written and the moving-oscillator example is a useful illustration. The main caveat is that the central table is not independently reproducible from the paper alone, because the dicyclic character tables from which it is computed are not printed.

A particle at rest in a cubic lattice sees the full rotational symmetry of the cube, but a particle with nonzero momentum sees only the smaller set of rotations that leave its momentum vector fixed — the little group of that momentum. This paper computes, for every momentum direction a cubic lattice allows, precisely how a continuum state of total angular momentum $j$ and projection $m_j$ along the momentum decomposes into the irreducible representations of that little group. The result is a lookup table that runs in both directions: someone who measures a lattice state in a given representation at a given momentum can read off which continuum $m_j$ values it can descend from, and which lattice levels must merge into degenerate multiplets as the continuum limit is approached. The authors demonstrate the inverse direction on a moving isotropic harmonic oscillator on a small lattice, and are explicit that beyond the lowest few levels the inversion is ambiguous without prior knowledge of the continuum spectrum.

What carries the argument. The engine is the method of little groups applied to a semidirect product. Because translations form a normal subgroup, representations of the full space group are induced from representations of the translation subgroup labelled by momentum $k$; the residual symmetry is the little group of $k$, and the induced representation is labelled by a star of $k$ plus an irreducible representation of that little group. Characters are written down for the discrete group (Eq. 1) and for the subduced continuum representations (Eq. 2), with the rotation character read off from Table II, and the multiplicity of each lattice representation inside a given continuum representation follows from the standard character-orthogonality projection (Eq. 3). The practical consequence is that the whole reduction collapses to a statement about $O^D(2)$ and its finite subgroups, which is why the tables depend on the little group and not on the lattice size.

Core claim. The central claim is that subducing the irreducible representations of the double cover of $T^3$ ⋊ $O(3)$ (with parity) onto the cubic space group $T^3_{lat}$ ⋊ $O_h^D$ yields a reduction in which a continuum state of momentum $k$ and angular-momentum projection $m_j$ maps onto lattice states carrying the same momentum, with the rotational content given by reducing $O^D(2)$ to the little group of $k$. The authors tabulate that reduction for all the little groups that occur on a cubic lattice — $Dic_4$, $Dic_3$, $Dic_2$, $C_4$, $C_2$ — for $m_j = 0, 1/2, 1, 3/2, 2, 5/2, 3, 7/2$ (Table III), so that, for example, $k=(1,0,0)$ with $m_j=2$ becomes $B_1$ plus $B_2$. The inverse reading is the part the authors care about: lattice levels sitting in different irreducible representations at the same momentum must become degenerate in the continuum limit, and that pattern of degeneracies is what recovers the continuum angular-momentum multiplets. The moving harmonic oscillator is worked through as a test case, where the ground state $|000⟩$ is identified with $m_j = 0^+$ and higher states are matched to continuum multiplets as far as the small lattice allows.

The premise it rests on. The tables are claimed to be independent of lattice size, on the strength of a stated but not exhibited calculation showing that on a finite torus the character projection reduces to the projection of the continuous rotation group onto the little group of the momentum.

A lattice state measured at nonzero momentum in a given irreducible representation can be assigned to specific continuum $m_j$ values by inverting Table III, extending spin identification beyond the rest frame.

Levels that sit in different lattice irreducible representations at the same momentum must become degenerate in the continuum limit; observing that degeneracy is a check on both the continuum extrapolation and the spin assignment.

At nonzero momentum only the projection $m_j$ survives as a good quantum number, so a spin-$j$ multiplet is spread across several lattice representations and cannot be identified from one representation alone.

The moving oscillator provides a worked template: energies computed on the lattice, grouped by representation content, map onto continuum multiplets with definite $j$ and parity, subject to the incompleteness of a small lattice.

Because the reduction depends only on the little group, momenta along symmetry axes such as $(1,0,0)$, $(1,1,0)$ and $(1,1,1)$ each come with their own ready-made conversion table.

Combining several momentum directions should sharpen the inverse map considerably: a state of spin $j$ appears in different little groups at $(1,0,0)$, $(1,1,0)$ and $(1,1,1)$, and demanding consistency across those stars over-determines $j$ in a way a single direction cannot.

If the size-independence claim holds as stated, the tables carry over unchanged to elongated or anisotropic boxes whenever the little group of the momentum is unchanged, which is the regime moving-frame spectroscopy actually uses.

The same little-group bookkeeping should apply to matrix elements of currents, not just to the spectrum: an operator that transforms in a known lattice representation can be decomposed into continuum multipole components using the same tables, which is the stated motivation for the extension.

The paper cites one of its group-theory inputs as a forthcoming publication, so the dicyclic character tables it relies on are not independently verifiable from the reference list alone.

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