Tangent-Fermion Lattice Formulation

• Tangent-Fermion Lattice is defined by replacing the sine dispersion with a tangent function in the Dirac operator to achieve a single Dirac cone and eliminate fermion doubling.
• It employs a generalized eigenvalue problem with local, Hermitian operators that maintain exact chiral and time-reversal symmetries on the lattice.
• The approach has been extended to handle boundary conditions, magnetic fields, interacting one-dimensional systems, and quantum-circuit implementations, demonstrating its versatility.

Searching arXiv for the cited tangent-fermion papers to ground the article in the current literature.
{"query":"tangent fermions Dirac lattice Beenakker Pacholski Donis Vela arXiv", "max_results": 10}
{"query":"2411.11564", "max_results": 5}
A tangent-fermion lattice is a lattice formulation of the Dirac equation in which the continuum linear dispersion is replaced by a tangent dispersion, while the dynamics are recast as a generalized eigenvalue problem with local Hermitian operators. In the two-dimensional setting, this construction was introduced to place a massless Dirac or Majorana fermion on a lattice with only a single Dirac cone in the Brillouin zone, preserving chiral and time-reversal symmetries and avoiding fermion doubling [2302.12793]. Subsequent work extended the same framework to boundary-value problems for Dirac edge states [2411.11564], to the magnetic-field problem and the zeroth Landau level [2505.13658], to interacting one-dimensional chiral and helical systems [2601.09563, 2606.24713], and to quantum-circuit block encodings of the generalized Dirac operator [2606.19020].

1. Definition and formal construction

In the two-dimensional massless case, the continuum starting point is the Dirac equation
[
-i\hbar v(\sigma_x \partial_x + \sigma_y \partial_y)\psi(x,y) = E \psi(x,y),
]
with (\psi=(\psi_1,\psi_2)^T). A naive nearest-neighbor discretization replaces (\partial_x) by a symmetric finite difference and produces a sine dispersion (E(k)\propto \sin(ak)), which has zeros both at (k=0) and (k=\pi/a). In two dimensions this generates extra Dirac cones at Brillouin-zone boundaries and hence fermion doubling [2411.11564, 2302.12793].

The tangent-fermion construction replaces the sine dispersion by a tangent dispersion. In the static two-dimensional formulation, the Bloch Hamiltonian is
[
H_{\rm Stacey}
= \frac{2\hbar v}{a}\bigl[\sigma_x\tan(ak_x/2)+\sigma_y\tan(ak_y/2)\bigr],
]
with dispersion
[
E_{\rm Stacey}^2(\mathbf{k}) =
\left(\frac{2\hbar v}{a}\right)^2
\bigl[\tan^2(ak_x/2)+\tan^2(ak_y/2)\bigr].
]
Near (\mathbf{k}=0), (\tan(ak_\alpha/2)\approx ak_\alpha/2), so the continuum cone is recovered. At the Brillouin-zone boundary the tangent diverges, giving a pole rather than an additional zero [2302.12793].

The central technical step is to write the Dirac problem as a generalized eigenvalue problem
[
\mathcal H \Psi = E\,\mathcal P \Psi,
]
with
[
\mathcal P = \tfrac{1}{4}(1+\cos a\hat{k}_x)(1+\cos a\hat{k}_y),
]
[
\mathcal H =
\frac{\hbar v}{2a}\Big[
\sigma_x (1+\cos a\hat{k}_y)\sin a\hat{k}_x +
\sigma_y (1+\cos a\hat{k}_x)\sin a\hat{k}_y
\Big].
]
Both (\mathcal H) and (\mathcal P) are local, Hermitian, and sparse; (\mathcal P) is positive definite. For Bloch states this yields
[
E(\mathbf k) = \pm \frac{2\hbar v}{a}\sqrt{
\tan^2\left(\frac{a k_x}{2}\right) +
\tan^2\left(\frac{a k_y}{2}\right) }.
]
This is the characteristic tangent-fermion lattice dispersion: one Dirac cone per Brillouin zone, with the nonlinearity pushed to the zone edge as a tangent pole [2411.11564].

2. Fermion doubling, locality, and symmetry

The tangent-fermion lattice was developed against the background of the Nielsen–Ninomiya obstruction. In the standard lattice-Hamiltonian setting, a local discretization preserving the relevant symmetries cannot realize a single massless Dirac cone without doublers. Tangent fermions evade this obstruction in two linked ways: the underlying tangent derivative is nonlocal, and the practical formulation uses a generalized eigenvalue problem with two local operators rather than a single local Hamiltonian [2302.12793, 2606.19020].

In one dimension, Stacey’s derivative is
[
\partial^{\rm tangent}
= 2\,\frac{T-1}{T+1}
= 2\sum_{j=1}^\infty (-1)^j \left(T^{-j} - T^j\right),
]
with Fourier symbol (2 i \tan(k/2)). This has only one zero in the Brillouin zone and a pole at (k=\pi), so there is no second low-energy cone. Pacholski et al. showed that the same spectrum can be implemented through a local operator pencil ((\sigma_x K,B)), with
[
K = \tfrac{1}{2} i (T^{-1} - T),\qquad
B = \tfrac{1}{4}(1 + T^{-1})(1 + T),
]
so that
[
\sigma_x K \Psi = E B \Psi.
]
This factorization retains the tangent dispersion while restoring locality at the level of the generalized eigenproblem [2606.19020].

A defining feature of the two-dimensional tangent-fermion lattice is exact preservation of chiral symmetry and time-reversal symmetry. In the square-lattice formulation, ({\mathcal H,\sigma_z}=0), and under (\mathbf k\to -\mathbf k) with (\boldsymbol\sigma\to -\boldsymbol\sigma), both (\mathcal H) and (\mathcal P) remain invariant [2411.11564]. The earlier tangent-fermion analysis identifies this symmetry content as the reason the Dirac cone is topologically protected against disorder and staggered perturbations: attempts to couple the Dirac point at (\mathbf{k}=0) to Brillouin-zone-boundary states encounter the tangent divergence, so a gap cannot be opened by the same mechanism that destabilizes sine-dispersion fermions [2302.12793].

The literature distinguishes this construction sharply from Wilson and staggered formulations. Wilson fermions remove doublers by adding a momentum-dependent mass term, but this breaks chiral and time-reversal symmetries. Staggered schemes reduce doubling but retain multiple Dirac points in two dimensions and do not preserve the full symmetry structure of a single topological-insulator surface cone [2302.12793, 2411.11564].

3. Boundary conditions and single-cone Dirac edge states

A major extension of the tangent-fermion lattice is the treatment of boundary conditions for confined Dirac fermions. In the continuum, current conservation and self-adjointness do not permit simply setting (\psi=0) on the boundary. For a straight edge at (x=0), the admissible one-parameter family is
[
\psi(0,y) = (\sigma_y\cos\theta + \sigma_z\sin\theta)\,\psi(0,y).
]
Special cases include the infinite-mass boundary condition at (\theta=0) and the zigzag boundary condition at (\theta=\pi/2) [2411.11564].

On the surface of a three-dimensional topological insulator, these boundary conditions arise from a magnetic insulator. A magnetization
[
\mathbf M = \operatorname{sign}(n_x) M_0 (0,-\sin\theta,\cos\theta)
]
adds a large term (\mathbf M\cdot\boldsymbol\sigma), and in the limit (M_0\to\infty) the surface Dirac fermion acquires exactly the boundary condition above. Changing the magnetization direction sweeps out the full one-parameter family [2411.11564].

The lattice implementation proceeds by restricting the infinite-lattice operators (\mathcal H) and (\mathcal P) to a finite domain, rotating the spinor basis locally at each boundary site with a unitary (U_n) chosen so that
[
U_n (\mathbf t_n\cdot\boldsymbol\sigma) U_n^\dagger = \sigma_z,
]
and then removing the spin-down rows and columns at the boundary sites. The resulting matrices (\tilde{\mathcal H}) and (\tilde{\mathcal P}) satisfy
[
\tilde{\mathcal H}\psi = E\,\tilde{\mathcal P}\psi.
]
Because only principal submatrices are taken and the intermediate transformation is unitary, Hermiticity and positive definiteness are preserved [2411.11564].

In channel geometry, with width (W) and conserved longitudinal momentum (q=k_y), the continuum quantization condition is
[
E\cos\alpha_- = \frac{k(E)\sin\alpha_-}{\tan[k(E)W]} + q\cos\alpha_+,
]
with (k(E)=\sqrt{E^2-q^2}) and (\alpha_\pm=\tfrac12(\theta_1\pm\theta_2)). The tangent-fermion strip spectrum reproduces the expected cases: no edge states for infinite-mass boundaries, a flat zero-energy edge band for zigzag boundaries, and dispersive edge states for intermediate (\theta_{1,2}). The low-energy lattice spectra agree almost exactly with the analytic continuum results, while remaining built from a single Dirac cone in the Brillouin zone [2411.11564].

The same study also identifies limitations. For oblique boundaries, the simple sharp-boundary prescription can couple low-energy states to the tangent pole in the folded Brillouin zone and generate spurious oscillations; a finite-mass boundary layer remedies this. For perfectly zigzag boundaries, the flat edge band exhibits a doubled degeneracy, and a small perturbation of (\theta) splits it into one physical and one spurious branch [2411.11564].

4. Magnetic fields and the zeroth Landau level

The tangent-fermion lattice also has a gauge-invariant magnetic-field formulation. On a square lattice, the field is introduced through gauge-covariant translation operators
[
{\cal T}\alpha = \sum{\mathbf{n}} e^{i\phi_\alpha(\mathbf{n})}\,
|\mathbf{n}\rangle\langle \mathbf{n}+\mathbf{e}\alpha|,
]
with
[
\phi\alpha(\mathbf{n}) =
e\int_{\mathbf{n}+\mathbf{e}\alpha}^{\mathbf{n}}
A\alpha(\mathbf{r})\,dx_\alpha.
]
The lattice generalized eigenproblem uses
[
{\cal H} =
\frac{\hbar v}{8ia}\,\sigma_x\,(1+{\cal T}_y)({\cal T}_x-{\cal T}_x^\dagger)(1+{\cal T}_y^\dagger)
+
\frac{\hbar v}{8ia}\,\sigma_y\,(1+{\cal T}_x)({\cal T}_y-{\cal T}_y^\dagger)(1+{\cal T}_x^\dagger),
]
[
{\cal C} =
\tfrac{1}{8}(1+{\cal T}_x)(1+{\cal T}_y)+\tfrac{1}{8}(1+{\cal T}_y)(1+{\cal T}_x),
\qquad
{\cal P} = {\cal C}{\cal C}^\dagger.
]
This formulation preserves exact lattice chiral symmetry, ({\sigma_z,{\cal H}}=0) and ([\sigma_z,{\cal P}]=0) [2505.13658].

In the continuum, the massless Dirac equation in a perpendicular field (B\hat z) has Landau levels
[
E_n = \pm v\sqrt{2n\hbar e|B|},\qquad n=0,1,2,\ldots,
]
and the zeroth Landau level is exactly at zero energy, independent of (|B|), with definite chirality
[
\langle 0|\sigma_z|0\rangle = \mathrm{sign}(B).
]
On an infinite lattice, however, Stacey’s theorem enforces equal numbers of zero modes of opposite chirality, so the zeroth level becomes doubly degenerate and loses its continuum-style topological protection [2505.13658].

The resolution proposed in the 2025 work is boundary-condition-assisted chirality selection. On a finite domain, the boundary condition
[
\psi(x,y)=\sigma_z\psi(x,y),\qquad (x,y)\in\partial\mathcal{D},
]
removes the spin-down degree of freedom at boundary sites and changes the zero-mode count. In channel geometry at (B=0), the finite tangent-fermion lattice acquires a doubly degenerate zero-energy mode consisting of two spin-up edge states, giving
[
{\cal I}(0)=N_{\uparrow}^{(0)}-N_{\downarrow}^{(0)}=2.
]
Turning on a uniform magnetic field then produces a zero-energy sector composed of a spin-polarized edge state and a spin-polarized bulk Landau state localized around (x=q/eB) [2505.13658].

The central robustness result is that, in a non-uniform magnetic field preserving chiral symmetry, higher Landau levels broaden but the zero mode remains exactly flat. In the fully two-dimensional rectangular geometry with (B(\mathbf r)\in(B_0-\delta B,B_0+\delta B)) and (\delta B=10B_0), all (n\neq 0) Landau levels broaden into smeared steps in the integrated density of states, while the zero-mode plateau remains perfectly sharp. The zero-energy band has total degeneracy (2L/a) and consists of bulk zeroth-Landau-level states together with additional spin-polarized edge states, all of the same chirality [2505.13658].

5. One-dimensional interacting tangent fermions

The tangent-fermion idea was extended to one-dimensional helical and chiral systems by replacing the usual sine dispersion with
[
E(k) = \frac{2\hbar v_{\rm F}}{a}\,\tan\frac{ka}{2},
\qquad |k|<\pi/a,
]
or, in condensed notation,
[
E(k)=2t_0\tan\frac{k}{2}.
]
In real space this corresponds to long-range hopping with alternating sign. For the helical Luttinger liquid,
[
H_0
=\sum_{n>m} t_{nm}\Bigl( c_{n\uparrow}^\dagger c_{m\uparrow}
- c_{n\downarrow}^\dagger c_{m\downarrow} \Bigr)+\text{H.c.},
\qquad
t_{nm}=2 i t_0 (-1)^{n-m},
]
which preserves time-reversal symmetry while eliminating the doubler at the Brillouin-zone edge [2601.09563].

That work uses the tangent-fermion kinetic term together with forward-scattering and Umklapp interactions to study spontaneous time-reversal symmetry breaking in a helical Luttinger liquid. Bosonization gives the Luttinger parameter
[
K=\sqrt{\frac{1+\kappa_{\rm intra}-\kappa_{\rm inter}}
{1+\kappa_{\rm intra}+\kappa_{\rm inter}}},
]
and the two-particle backscattering term becomes relevant for (K<1/2). Density-matrix renormalization group calculations on finite tangent-fermion lattices confirm the expected transition: at half filling and with nonzero two-particle Umklapp, the propagator crosses over from Luttinger scaling to exponential decay and transverse spin correlators saturate, indicating a gapped phase with spontaneous time-reversal symmetry breaking [2601.09563].

A distinct one-dimensional application is the anomaly-free 3–4–5–0 chiral model. There the free Hamiltonian is
[
H_0 =
\sum_{n>m} t_{nm}\Bigl(
c^\dagger_{n,3} c_{m,3} + c^\dagger_{n,4} c_{m,4}
- c^\dagger_{n,5} c_{m,5} - c^\dagger_{n,0} c_{m,0}
\Bigr)+\text{H.c.},
]
again with
[
t_{nm} = 2 i t_0\,(-1)^{n-m}.
]
Because (E(k)=2t_0\tan(k/2)) is strictly monotonic on (|k|<\pi), each species has a single chiral branch and no mirror node at (k=\pi) [2606.24713].

The interacting terms include the six-fermion 3–4–5–0 gapping operators and a Hubbard-type density-density interaction that tunes the Luttinger parameter to
[
K_p = \sqrt{
\frac{2\pi t_0 - C U^{(p)}_{\rm H}}
{2\pi t_0 + C U^{(p)}_{\rm H}} },
\qquad C=5.
]
The scaling dimension of the six-fermion interaction becomes
[
D^{(p)}_{3450}=5K_p,
]
so in the symmetric case the interaction is relevant for
[
K<\frac{2}{5}.
]
The DMRG results show the opening of an excitation gap in this regime without the appearance of a degenerate ground state, which the paper identifies as the hallmark of symmetric mass generation [2606.24713].

In both one-dimensional studies, the computational practicality of tangent fermions relies on the fact that the nonlocal hopping admits an exact matrix-product-operator representation with finite bond dimension independent of system size [2601.09563, 2606.24713]. This suggests that the generalized-eigenproblem logic of tangent fermions is not limited to single-particle lattice Dirac equations.

6. Quantum algorithms, computational representation, and scope

A recent development concerns quantum-circuit decomposition of the tangent-fermion Dirac operator. Direct linear-combination-of-unitaries representations of nonlocal discretizations such as the tangent derivative require a number of terms and a subnormalization factor that grow with lattice size. The generalized-eigenproblem formulation avoids this by block-encoding each member of the operator pencil separately [2606.19020].

In one dimension, with antiperiodic translation operator (\tilde T), the local operators are
[
\tilde{K} = \frac{1}{2} i \tilde{T}^\dagger - \frac{1}{2} i \tilde{T},
\qquad
\tilde{B} = \frac{1}{2}\openone + \frac{1}{4}\tilde{T}^\dagger + \frac{1}{4}\tilde{T}.
]
These have exact LCU decompositions with
[
L(\tilde K)=2,\qquad \lambda(\tilde K)=1,
]
[
L(\tilde B)=3,\qquad \lambda(\tilde B)=1.
]
In (d) dimensions,
[
\lambda_{\rm tangent}({\cal P}) = 1,\qquad
\lambda_{\rm tangent}({\cal H}) = d,
]
and
[
L_{\rm tangent}({\cal P}) = 3^d,\qquad
L_{\rm tangent}({\cal H}) = 2 d\,3^{d-1},
]
all independent of lattice size [2606.19020].

This provides an efficient block-encoding primitive for generalized eigenvalue solvers and quantum linear-system methods applied to Dirac spectra and Green functions without fermion doubling. The paper’s characterization is that the tangent-fermion pencil has complexity “on a par with elliptic operators” in the LCU sense [2606.19020].

Across these developments, the term “tangent-fermion lattice” has a precise and consistent meaning. It does not refer to a curved-space tangent bundle; in the one-dimensional chiral-fermion literature, the phrase means a lattice whose discretized Dirac operator has a tangent dispersion instead of the standard sine dispersion [2606.24713]. In the foundational two-dimensional work, it denotes a symmetry-preserving lattice regularization of a single Dirac cone, implemented through a local generalized eigenproblem and extendable to magnetic fields, boundaries, edge modes, Landau levels, interacting one-dimensional phases, and quantum algorithms [2302.12793, 2411.11564, 2505.13658, 2606.19020].

A plausible implication is that tangent fermions occupy a distinct position among lattice-fermion regularizations: they trade the standard single-operator eigenproblem for an operator pencil, and in return realize a single Dirac cone without Wilson mass terms, preserve exact lattice chiral symmetry, and remain compatible with sparse-matrix, tensor-network, and block-encoding methods.