Reduced Symmetry Tight-Binding Model

• Reduced Symmetry Tight-Binding Models are frameworks that capture effective electronic behavior by considering bond networks, orbital content, and deliberate symmetry reductions beyond the parent crystal.
• They employ a shell-resolved approach using bond equivalence classes to determine the operative symmetry of finite-range hopping Hamiltonians.
• This methodology enables the modeling of phenomena such as Rashba spin splitting, non-Hermitian transport, and low-energy moiré Hamiltonians with significant implications for spectral and topological properties.

A reduced symmetry tight-binding model is a tight-binding Hamiltonian whose operative symmetry is not fixed solely by an ideal parent crystal description, but by the retained bond network, orbital content, basis choice, and any explicit symmetry-breaking terms. The literature represented here shows that the actual symmetry of a finite-range hopping model can be lower than the maximal symmetry of the underlying point configuration, can differ from the parent space group, and can also be reduced deliberately to generate Rashba spin splitting, nonsymmorphic spectral symmetries, non-Hermitian transport regimes, or compact low-energy moiré Hamiltonians (Dagnino et al., 2024). In the specific case of “Tight-binding Rashba model and statistical field theory,” the model is presented as a pedagogical review that explains conventions for the tight-binding description of electronic states on the hexagonal Bravais lattice in two dimensions, derives Rashba spin splitting from elementary symmetry conditions, constructs a minimal tight-binding model displaying Rashba spin splitting near the center of the Brillouin zone, derives symmetry conditions for a two-particle interaction in a second-quantized framework, and summarizes temperature Green functions (Schober et al., 2015).

1. Parent symmetry, eigensymmetry, and actual tight-binding symmetry

A central distinction in contemporary tight-binding theory is the distinction between the parent space group $G$, the eigensymmetry group $E=\mathrm{Eig}_{\mathcal W}(G)$, and the symmetry $H$ of the shell-truncated hopping Hamiltonian. The parent structure fixes the available sites, but it does not, in general, determine the symmetry of the tight-binding model. Instead, the symmetry depends on the hopping range, or equivalently on which shell or set of shells is retained (Dagnino et al., 2024).

For shell-resolved $s$-wave hopping, the Hamiltonian can be written as

$$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$

where $\langle i,j\rangle_n$ denotes pairs of sites separated by the $n$-th neighbor distance. The actual symmetry group of this Hamiltonian is always a subgroup of the eigensymmetry $E$, but it need not equal the parent group $G$, and often exceeds it. This formulation directly overturns the widespread assumption that finite-range tight-binding symmetry is simply the parent-crystal symmetry (Dagnino et al., 2024).

The same literature also makes clear that reduced symmetry can arise in a different sense: the crystal can preserve its global symmetry while the little group at generic $\mathbf{k}$ is reduced relative to high-symmetry points. In rhombohedral topological insulators, the $E=\mathrm{Eig}_{\mathcal W}(G)$0 symmetry is preserved along $E=\mathrm{Eig}_{\mathcal W}(G)$1 but not along $E=\mathrm{Eig}_{\mathcal W}(G)$2 and $E=\mathrm{Eig}_{\mathcal W}(G)$3; as a result, $E=\mathrm{Eig}_{\mathcal W}(G)$4-dominated and $E=\mathrm{Eig}_{\mathcal W}(G)$5-dominated sectors can hybridize away from $E=\mathrm{Eig}_{\mathcal W}(G)$6 (Acosta et al., 2018). This suggests that “reduced symmetry” in tight-binding theory includes both shell-induced symmetry loss in real space and momentum-dependent loss of symmetry protection in reciprocal space.

2. Bond equivalence classes and shell-resolved symmetry reduction

The most systematic group-theoretic description is formulated in terms of bond equivalence classes. For a fixed shell $E=\mathrm{Eig}_{\mathcal W}(G)$7, the set of bonds is

$E=\mathrm{Eig}_{\mathcal W}(G)$8

and the parent space group acts on $E=\mathrm{Eig}_{\mathcal W}(G)$9. Two bonds are equivalent if they are related by a symmetry of $H$0, up to exchange of endpoints. The orbit of a bond under $H$1 is a bond equivalence class, and the full set of classes is $H$2 (Dagnino et al., 2024).

The operative symmetry of the shell-$H$3 Hamiltonian is then the stabilizer of this bond coloring inside the eigensymmetry group,

$H$4

If all bonds in the shell lie in a single equivalence class, the model may inherit the full eigensymmetry. If the shell splits into multiple inequivalent classes, those classes carry independent hoppings and reduce the symmetry to the subgroup that preserves the coloring (Dagnino et al., 2024).

This shell-resolved logic is illustrated by the decorated kagome and rutile examples. In the decorated kagome case, nearest-neighbor bonds form one class and the nearest-neighbor Hamiltonian regains the full symmetry of the undecorated kagome lattice, even though the parent crystal symmetry is lower. At the third shell, three bond classes appear and the symmetry becomes intermediate: larger than the parent, smaller than the full kagome eigensymmetry. In the rutile example, several short-range shells form a single bond equivalence class and exhibit enhanced symmetry, whereas in-plane diagonal $H$5-type bonds split into two inequivalent classes, producing two independent hoppings and lowering the symmetry (Dagnino et al., 2024).

The same framework yields an exact index rule in the case $H$6 and $H$7: $H$8 The paper states that this covers roughly $H$9 of the cases studied up to $s$0, and in all of them the formula works. When $s$1 or $s$2, no universal simple index formula exists, because different eigensymmetry bond classes can split differently and broken translations can be compensated on particular bond orbits by surviving point-group operations (Dagnino et al., 2024).

3. Canonical mechanisms for reducing symmetry in tight-binding Hamiltonians

Reduced symmetry can be introduced deliberately at the Hamiltonian level. One route is spin-orbit coupling constrained by lattice symmetry. The supplementary review on the tight-binding Rashba model derives Rashba spin splitting from elementary symmetry conditions on the hexagonal Bravais lattice in two dimensions and constructs a minimal tight-binding model with Rashba spin splitting near the center of the Brillouin zone; it also derives symmetry conditions for a two-particle interaction in a second-quantized framework (Schober et al., 2015).

Another route is algebraic rather than geometric. In non-trivial square-root constructions, the child Hamiltonian is engineered so that $s$3 reproduces a simpler parent model, but the square-root operation typically reduces crystal symmetry by enlarging the unit cell or changing the hopping pattern while creating new spectral symmetries, especially a chiral symmetry around $s$4. In the one-dimensional bow-tie chain, the child has four sites per unit cell rather than two, reduced translational symmetry, and a nonsymmorphic chiral symmetry implemented by a fractional lattice translation (Arkinstall et al., 2017).

Non-Hermitian models furnish a third mechanism. The non-Hermitian extension of the Rice–Mele chain includes staggered gain and loss $s$5 and complex same-sublattice hopping $s$6. In this model, Hermiticity is broken, $s$7 and $s$8 are broken separately, and $s$9 symmetry is retained for the extended periodic system. The complex next-nearest-neighbor hopping generates a drift velocity

$$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$0

and this converts broken-$$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$1 behavior from absolute to convective when $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$2 (Longhi, 2013).

A fourth mechanism is explicit time-reversal reduction in spinful topological models. The honeycomb model obtained by superimposing Kane–Mele and Haldane terms breaks ordinary time-reversal symmetry through a staggered flux while optionally also breaking exact spin conservation through Rashba coupling. The resulting model exhibits QSHI, QAHI, and QASHI phases; the reduction of symmetry permits spin-resolved band inversions that are not constrained by $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$3 (Saha et al., 2021).

Even when no spatial symmetry is intentionally broken, a minimal nearest-neighbor lattice description can be too symmetric relative to the exact microscopic problem. In the tight-binding formulation of the Kronig–Penney model, the exact band requires an effective $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$4 correction to reproduce electron–hole asymmetry away from the extreme tight-binding limit. The paper argues that this need not imply direct next-nearest-neighbor tunneling; it can arise from the nonlinear structure of the exact continuum dispersion itself (Marsiglio et al., 2017).

4. Symmetry-adapted bases, Wannier reduction, and exact versus approximate symmetry sets

Reduced symmetry is also a matter of basis construction. In a translationally symmetric $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$5 rectangular lattice with periodic boundary conditions, the Hamiltonian commutes with the discrete translation matrices $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$6 and $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$7. Standard eigensolvers recover arbitrary orthonormal vectors inside degenerate eigenspaces, but a symmetry-adapted basis is obtained by simultaneous diagonalization of $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$8, $$H=-\sum_{\langle i,j\rangle_n} t_{i,j}\, c_i^\dagger c_j+\text{h.c.},$$9, and $\langle i,j\rangle_n$0, yielding momentum-labeled eigenvectors and the dispersion

$\langle i,j\rangle_n$1

(0910.5434). This is not symmetry reduction in the sense of lowering the Hamiltonian symmetry; it is a reduction of the eigenproblem into symmetry sectors.

In moiré materials, the more consequential issue is which symmetry set is treated as exact. For twisted bilayer graphene near the magic angle, one construction works directly with the exact microscopic symmetries $\langle i,j\rangle_n$2, $\langle i,j\rangle_n$3, lattice translations, and time reversal, and does not assume an exact valley $\langle i,j\rangle_n$4 symmetry. The resulting four narrow bands admit symmetry-adapted maximally localized Wannier states centered on the dual honeycomb lattice rather than the triangular moiré sites, because the representation content at $\langle i,j\rangle_n$5 and $\langle i,j\rangle_n$6 cannot be matched by four orbitals all centered on the triangular sites (Kang et al., 2018).

A related ab-initio-informed workflow constructs full-energy atomistic Slater–Koster models for twisted bilayer graphene and then reduces them to four-band and twelve-band low-energy Hamiltonians using maximum localization with crystal- and time-reversal-symmetry constraints. In that work, the preserved symmetry is the microscopic $\langle i,j\rangle_n$7+TR symmetry of the chosen commensurate structure rather than exact valley $\langle i,j\rangle_n$8 or full $\langle i,j\rangle_n$9. The same study also reports that a non-symmetric twelve-band model can yield more compact Wannier orbitals and better flat-band accuracy inside a frozen window, at the price of losing exact crystal symmetry and robustness outside the target window (Davydov et al., 2020).

For AB-stacked MoTe$n$0/WSe$n$1, the low-energy two-band-per-valley model is constructed on a triangular moiré lattice with two Wannier orbitals at the same MM-centered site but with different $n$2 angular momenta. The topological transition is controlled by a band inversion at $n$3, where the inter-orbital coupling vanishes by symmetry and the criticality is set by $n$4. This suggests that a reduced-symmetry variant should keep the same two-orbital triangular basis while relaxing the $n$5 and $n$6 constraints on onsite and hopping matrices (Luo et al., 2022).

5. Spectral, topological, and dynamical consequences

The principal consequence of reduced symmetry is that degeneracies, band connectivity, topology, and response functions are governed by the actual symmetry of the Hamiltonian, not by the parent crystal label alone. The shell-classification study states this explicitly: assigning too much symmetry can produce false degeneracies and incorrect topological constraints, while assigning too little can miss protected features (Dagnino et al., 2024).

Whole-zone models for rhombohedral topological insulators make this point in a different language. A $n$7 model in the $n$8 sector captures the topological band inversion near $n$9, but an $E$0 model is required over the whole Brillouin zone because reduced little-group symmetry away from $E$1 allows $E$2 and $E$3 sectors to mix. This mixing reshapes the valence band, creates bulk states near the Dirac-point energy, and produces surface-projected bulk states that coexist with the topological Dirac cone (Acosta et al., 2018).

Graphene provides a complementary symmetry baseline. In the ideal lattice, $E$4 and $E$5 bands are separated by reflection symmetry in the graphene plane, and high-symmetry points carry little groups $E$6, $E$7, and $E$8. The symmetry analysis implies that breaking mirror symmetry permits $E$9-$G$0 mixing, lowering rotational symmetry splits $G$1-type degeneracies, and breaking sublattice equivalence lifts the $G$2-point doublet of the low-energy $G$3 sector (Kogan et al., 2020).

The one-dimensional catalog of two-band models shows how symmetry reduction can be classified directly at the Bloch-Hamiltonian level. In the generic form

$G$4

imposing chiral, time-reversal, or charge-conjugation symmetry removes specific Pauli components and constrains the remaining functions to be even, odd, or half-lattice harmonics. Symmorphic chiral classes admit $G$5 winding numbers, class D admits a $G$6 invariant, and nonsymmorphic classes without nonsymmorphic time reversal admit a crossing-parity $G$7 index. By contrast, nonsymmorphic time-reversal symmetry forces a metallic state with one Kramers-like degeneracy point in the Brillouin zone (McCann, 2023).

Loss of translational symmetry can also localize states without disorder. In the scaled tight-binding crystal, local symmetry dynamics destroys global translation symmetry and generates a deterministic, scale-structured chain. The resulting spectrum shows branches, minigap hierarchies, and resonator-localized eigenstates at weak coupling, followed by progressive delocalization as the hopping increases (Schmelcher, 2023). This suggests that reduced symmetry can reorganize spectral structure even when hopping amplitudes remain uniform.

6. Scope, limitations, and recurrent misconceptions

A recurrent misconception is that a shell-truncated tight-binding model automatically inherits the parent-crystal symmetry. The shell-resolved classification rejects this directly: the symmetry must be computed from the bond equivalence classes of the selected shell, and a substantial fraction of systems still have not reduced to the parent symmetry by the 20th neighbor (Dagnino et al., 2024).

A second misconception is that an additional $G$8 term in a fitted band necessarily represents literal next-nearest-neighbor tunneling. The Kronig–Penney analysis argues against this: higher-order corrections of order $G$9 can arise from the nonlinear structure of the exact continuum eigenvalue problem even in a double-well system, where no next-nearest neighbor exists (Marsiglio et al., 2017).

A third misconception is that symmetry-faithful low-energy models must preserve every approximate symmetry used in continuum discussions. In twisted bilayer graphene, one line of work constructs symmetric Wannier states precisely by keeping only the exact microscopic symmetries and not imposing exact valley $\mathbf{k}$0 (Kang et al., 2018). Another line constructs four-band and twelve-band low-energy models that preserve the microscopic commensurate $\mathbf{k}$1+TR symmetry but not exact valley symmetry, and it also shows that relaxing crystal symmetry can improve localization in practice (Davydov et al., 2020).

The present scope is also delimited by explicit restrictions in the cited literature. The shell-classification program is restricted to $\mathbf{k}$2-wave hopping on a single Wyckoff position, without orbital anisotropy, spin-orbit coupling, or magnetic symmetry, and extension to multiple Wyckoff positions, SOC, magnetic space groups, and lower-dimensional crystallographic groups is identified as an open problem (Dagnino et al., 2024). A different route, inspired by machine-learning techniques, fits tight-binding matrix elements directly to ab-initio bands while using as few orbitals and hopping terms as possible; this suggests a path toward automated reduced-symmetry model construction for interfaces, grain boundaries, and related large-scale materials calculations (Dick et al., 6 Aug 2025).

Within this landscape, a reduced symmetry tight-binding model is best understood not as a single canonical Hamiltonian, but as a class of constructions in which the physically relevant symmetry is determined at the level of the actual finite-range Hamiltonian, chosen basis, or intended low-energy reduction. The common thread is that symmetry is an emergent property of the model itself, not merely an inherited label from the parent structure.