Finite-range Lattice Momentum Operators for Quantum Field Theory

Abstract: We propose a Z-transform framework for the analysis and synthesis of finite range lattice momentum operators in quantum field theory. In this formulation, translation-invariant lattice operators are represented as functions of the complex variable $z$ in the unit circle, allowing their spectral properties to be analyzed using tools from digital signal processing and rational approximation theory. Within this framework, the fermion doubling problem is reinterpreted as the appearance of unwanted zeros of the discrete momentum operator on the unit circle --- an aliasing phenomenon in the sense of the Nyquist sampling theorem --- and the conditions for ghost suppression are expressed as precise constraints on the zero structure of the operator's transfer function. It is proven that no rational function can satisfy all required conditions simultaneously, motivating the finite impulse response approach developed here. This reframing naturally suggests a class of finite-range momentum operators, constructed by solving a least-squares approximation problem in the frequency domain. The resulting finite impulse response (FIR) operator approximates the continuum derivative across the full Brillouin zone, with ghost suppression achieved through the accuracy of the spectral approximation rather than through the addition of a symmetry-breaking Wilson term or the infinite-range nonlocal SLAC derivative. Numerical investigation confirms that near $θ= π$ only plane waves propagate coherently, and these exhibit group velocities far exceeding the speed of light, further distinguishing them from physical low-energy excitations. No ghost wave packet solutions exist near $θ= π$.

• The paper reformulates lattice derivative design as a Z-transform approximation problem, proving that no rational finite-range operator can simultaneously achieve exact continuum behavior, anti-Hermiticity, and a single Brillouin-zone zero.
• The authors construct antisymmetric finite-impulse-response operators by least-squares fitting to the ideal differentiator, with a 30-tap design reaching 0.03–0.06% group-velocity error in free 1+1-dimensional simulations.
• The results show that finite-range operators can confine residual doubler modes to narrow, dispersive spectral regions without Wilson terms, although gauge-field interactions, higher dimensions, and interacting-theory stability remain untested.

Overview

This paper by Olivier and Barnard reformulates the construction of translation-invariant lattice momentum operators as a frequency-domain approximation problem, using the bilateral Z-transform as its primary mathematical framework (2608.17327). Within this framework, a lattice derivative becomes a transfer function on the unit circle, fermion doubling is identified with an unwanted zero of that transfer function at the Brillouin-zone boundary — an aliasing phenomenon in the Nyquist sense — and locality becomes the statement that the impulse response has finite support. The authors prove (in an appendix) that no rational function can satisfy all desired conditions simultaneously, then construct finite impulse response (FIR) operators via least-squares approximation to the ideal differentiator $H(e^{i\theta}) = i\theta$. The resulting free-fermion propagators achieve sub-0.1% group-velocity accuracy while suppressing coherent ghost propagation without a Wilson term.

The Z-transform reformulation

The core observation is that on a periodic spatial lattice with spacing $a$, the unitary translation operator $U(a)$ maps to multiplication by $z = e^{ipa/\hbar}$, so momenta in the first Brillouin zone correspond to points on the unit circle. This is not specific to discretization: it follows from Wigner's theorem and the unitary representation of spacetime symmetries. The naive central-difference operator then has transfer function proportional to $(z^2 - 1)/(2z)$, whose numerator factors reveal two zeros on the unit circle — one physical at $z=1$ and one spurious at $z=-1$ corresponding to the fermion doubler. The doubling problem is thus recast as the design problem for a rational differentiator whose transfer function has exactly one unit-circle zero. Anti-Hermiticity of the derivative requires $D(e^{i\theta}) = i f(\theta)$ with $f$ real, odd, and nonzero away from $\theta = 0$, guaranteeing Hermiticity of $a$0 and hence unitary evolution.

This framing yields a precise impossibility statement. Requiring simultaneously (i) a unique unit-circle zero at $a$1, (ii) anti-Hermiticity, (iii) rationality (finite range), and (iv) the correct continuum limit admits no solution. The appendix proves this via a Möbius transformation mapping the unit circle to the real axis, reducing the problem to finding a real odd rational $a$2 with no real-axis poles; since $a$3 must have positive odd degree, it necessarily diverges at infinity. This is consistent with the Nielsen–Ninomiya obstruction, though the paper is explicit that it does not claim to evade the theorem — it asks instead whether the unavoidable boundary zero can be confined to a spectral window too narrow to support coherent ghost propagation.

The FIR construction

Accepting approximation rather than exactness, the paper restricts to antisymmetric FIR kernels satisfying $a$4, $a$5, giving

$a$6

Three properties hold structurally for any such kernel: a purely imaginary transfer function (anti-Hermiticity), odd symmetry, and an exact zero at $a$7. Ghost suppression therefore cannot come from removing the boundary zero — $a$8 forces $a$9 exactly by symmetry — but from its spectral width, which shrinks as $U(a)$0. The coefficients are obtained by fitting $U(a)$1 to $U(a)$2 over $U(a)$3 via pseudoinverse solution of a linear least-squares system, excluding $U(a)$4 to obtain smooth roll-off. Unlike Wilson's approach, no symmetry is broken and no additional parameter is introduced; unlike SLAC, the coupling range is strictly $U(a)$5 sites.

A practical point worth noting: implementing the full complex Wilson operator directly in the Dirac equation introduces a non-Hermitian contribution violating norm conservation, so the derivative and mass-correction parts must be separated. The FIR operator slots into the off-diagonal Dirac coupling directly, with no such complication.

Numerical validation

Free-electron propagation in $U(a)$6 dimensions ($U(a)$7, $U(a)$8, RK4 time stepping, Gaussian packets initialized as eigenstates of the operator under test) yields the following comparison of measured group velocities against continuum predictions:

$U(a)$9	Operator	Error vs continuum	Momentum drift
$z = e^{ipa/\hbar}$0	Central difference	9.15%	$z = e^{ipa/\hbar}$1%
$z = e^{ipa/\hbar}$2	Wilson	34.25%	$z = e^{ipa/\hbar}$3%
$z = e^{ipa/\hbar}$4	FIR $z = e^{ipa/\hbar}$5	0.06%	$z = e^{ipa/\hbar}$6%
$z = e^{ipa/\hbar}$7	Central difference	77.98%	$z = e^{ipa/\hbar}$8%
$z = e^{ipa/\hbar}$9	Wilson	5.77%	$(z^2 - 1)/(2z)$0%
$(z^2 - 1)/(2z)$1	FIR $(z^2 - 1)/(2z)$2	0.03%	$(z^2 - 1)/(2z)$3%

These are strong results: at high momentum the central difference fails almost completely (the wave packet stalls near $(z^2 - 1)/(2z)$4 where its group velocity vanishes), whereas even a 30-tap FIR operator reproduces the continuum velocity to within 0.06%. Momentum and norm are conserved to machine precision throughout, confirming errors reside in dispersion rather than amplitude dynamics.

The ghost-sector analysis is the conceptual heart of the paper. Numerical packet experiments show that near $(z^2 - 1)/(2z)$5 only plane waves propagate coherently — these exhibit group velocities far exceeding light speed, further distinguishing them from physical excitations — and no localized ghost wave-packet solutions exist. A finite-bandwidth packet centered in this region disperses because its components acquire substantially different group velocities. Even a 10-tap operator (21-site stencil) tracks the continuum dispersion well below roughly $(z^2 - 1)/(2z)$6, with coherent propagation failing beyond. The paper frames this as a design alternative to zero removal: restrict the spectral region over which the residual zero can support particle-like propagation, though it concedes this is not established as a general criterion.

Relation to existing formulations

Compared with overlap fermions, which realize chiral symmetry exactly through the Ginsparg–Wilson relation but require matrix sign-function evaluations per application and yield exponentially (not finitely) local kernels, the FIR operator is applied as a single precomputed convolution. Compared with SLAC, which reproduces the continuum dispersion exactly but couples every site to every other site with a $(z^2 - 1)/(2z)$7 tail, the FIR operator needs gauge transport over only a bounded number of links. The paper also notes that SLAC's discontinuity at $(z^2 - 1)/(2z)$8, identified in the literature as problematic in interacting gauge theories, is avoided by the FIR smooth roll-off. The proposed gauge-covariant extension replaces each displacement by ordered products of gauge links along straight paths; in higher dimensions path dependence introduces additional structure requiring explicit choices or path averaging.

Limitations and open questions

The paper is candid that its central conjecture — that strict finite coupling range avoids the difficulties SLAC encounters in gauge theories — is not established here. All results concern free propagation in $(z^2 - 1)/(2z)$9 dimensions. The gauge-covariant FIR operator is written down formally but remains untested, as does behavior with dynamical gauge fields, higher dimensions, and comparison against established formulations. Two further issues are flagged: the small low-energy spectral gap near $z=1$0 for moderate $z=1$1 requires study in interacting settings, and the anomalous superluminal group velocities of residual ghost plane waves suggest bounded group velocity could serve as an additional synthesis constraint — an idea noted but not pursued. The impossibility proof also relies on a private communication for part of the argumentation, and the ghost-suppression mechanism rests on the assumption that physically realizable wave packets have momentum spread large compared to $z=1$2.

Conclusion

The paper contributes a mathematically clean identification of lattice momentum operators with digital differentiator design, an explicit proof of the rational-operator no-go result within that language, and a systematic least-squares synthesis procedure producing finite-range fermion derivatives with excellent spectral fidelity. The numerical evidence — 0.03–0.06% group-velocity error versus 6–78% for conventional discretizations, with machine-precision conservation laws and symmetry-confined ghosts — supports the claim that high spectral accuracy and coherent-ghost suppression coexist under strict finite-range constraints in the free theory. Whether these properties survive gauge interactions is the decisive open question on which the practical value of the approach hinges.