Geometric Density of States (GDOS)

• GDOS is a framework that defines the density of quantum states by incorporating geometric features such as curvature and group velocity, crucial for band structure analysis.
• It is applied across systems—from 2D electronic bands to disordered and confined environments—using both analytical formulas and numerical simulations to characterize localization and state distribution.
• GDOS bridges algebraic approaches with experimental observables, enhancing insights into electronic, photonic, and many-body phenomena through its geometric state evaluation.

The geometric density of states (GDOS) refers to a suite of concepts in quantum condensed matter physics and many-body theory that emphasize the geometric structure—either in configuration, momentum, or energy space—underlying the density of available quantum states per energy. The term encompasses distinct but deeply related constructions for single- and many-body systems, including the directionally and curvature-resolved density of states on band-structure contours, the geometric mean (“typical”) density in disordered or localized phases, and the high-energy asymptotics in confined (e.g., billiard or box) geometries. GDOS provides crucial diagnostic power in questions ranging from localization transitions to the angular structure of STM responses, as well as revealing deep algebraic and geometric connections in density of states (DOS) calculations.

1. Geometric Density of States in Band Structure and Real-Space Responses

For electronic band structures in two dimensions, GDOS is defined as the directional and curvature-resolved density of states on a constant-energy contour (CEC) in momentum space. Given a dispersion relation $E(\mathbf{k})$, a CEC at energy $\varepsilon$ is $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$. At each $\mathbf{k}$ on the CEC, one defines the group velocity $\mathbf{v} = \nabla_{\mathbf{k}} E(\mathbf{k})$ and the local curvature $\kappa(\varepsilon, \theta)$ (where $\theta$ labels the propagation direction). The GDOS per unit energy and unit propagation angle is then

$$\rho(\varepsilon, \theta) = \frac{1}{4\pi^2 v(\mathbf{k}) \kappa(\varepsilon, \theta)},$$

where $v = |\mathbf{v}(\mathbf{k})|$ and $\kappa$ is the curvature at the point on the CEC pointed to by $\varepsilon$0. This quantifies, geometrically, how a small patch on the CEC contributes to the total DOS, with contributions enhanced where either the group velocity or curvature is small (Zhang et al., 2023).

GDOS in this context governs the amplitude of the real-space retarded Green's function at large distances: the stationary-phase contributions arise from momenta on the CEC where the group velocity is parallel to the observation direction. In the large-$\varepsilon$1 limit,

$\varepsilon$2

where $\varepsilon$3 is the GDOS in the contributing direction, and $\varepsilon$4 is the Bloch spinor (Zhang et al., 2023).

2. Geometric Mean Density of States in Disordered and Localized Systems

In disordered or Anderson-localized quantum systems, especially in numerical investigations of topological Anderson insulators, the GDOS refers to the geometric (i.e., “typical”) mean of the local density of states (LDOS) across all sites for a given energy: $\varepsilon$5 where $\varepsilon$6 is the LDOS at site $\varepsilon$7 and energy $\varepsilon$8, and $\varepsilon$9 is the total number of sites. The arithmetic (algebraic) mean $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$0 always remains finite if any nonzero local density exists, but only the geometric mean goes to zero when the overwhelming majority of sites yield exponentially suppressed $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$1, as in the localized phase (Zhang et al., 2013).

The ratio $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$2 thus functions as an “order parameter” for localization: $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$3 when states are extended, $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$4 in the thermodynamic limit for localized states. In topological Anderson insulators, phases with quantized conductance but distinct bulk properties (true band gap vs. mobility gap) are distinguished by the behavior of $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$5 and $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$6 (Zhang et al., 2013).

3. Geometric Asymptotics of Density of States in Confined Systems

For non-interacting particles (either single or many-body) in confined domains—rigid boxes, domains of arbitrary shape—the density of states exhibits a universal geometric scaling in the high-energy limit. For a single particle of mass $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$7 in a $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$8-dimensional box of volume $$C(\varepsilon) = \{ \mathbf{k} \,|\, E(\mathbf{k}) = \varepsilon \}$$9, the DOS is

$\mathbf{k}$0

with $\mathbf{k}$1 a dimension-dependent constant. This leading behavior depends only on the total volume and dimension, not on the aspect ratios or shape, expressing a geometric Weyl law (Mulhall et al., 2014).

For $\mathbf{k}$2 ideal bosons, the many-particle DOS is an $\mathbf{k}$3-fold convolution of the single-particle DOS, yielding

$\mathbf{k}$4

This encapsulates the scaling of state number with energy purely as a geometric consequence of dimensionality and volume.

4. GDOS as a Geometric Framework in Many-Body Quantum Systems

The mean DOS in many-body quantum systems of indistinguishable non-interacting particles admits a geometric decomposition organized by the group structure of particle permutations. Each term in the expansion corresponds to clusters (cycles in the permutation group) propagating on configuration submanifolds (the fixed-point sets of permutation operators), leading to terms scaling as $\mathbf{k}$5 for $\mathbf{k}$6 clusters. The asymptotic DOS at large energy is dominated by the term with all particles unclustered, reproducing the Thomas–Fermi Weyl law, while lower-$\mathbf{k}$7 corrections encode boundary and symmetry constraints (Hummel et al., 2012).

This formalism naturally incorporates quantum statistics: the signs arising from symmetric/antisymmetric projection cause nearly perfect cancellation of subleading terms for fermions below the ground state, leading to the emergence of the Fermi ground state via geometric cancellation. Analytical continuation and expansion around cluster zones thus yield results consistent with Bethe’s formula for excited states and embody the geometric essence of the mean DOS structure.

5. Algebraic-Geometric Approaches to GDOS in Tight-Binding Lattices

For tight-binding systems, particularly in two-dimensional lattices, the DOS can be reconstructed algebraically by linking the Brillouin-zone integrals to the periods of algebraic curves defined by the energy dispersion. The key is recognizing the moments of the DOS as Mellin transforms of the density,

$\mathbf{k}$8

which can be computed as combinatorial residues on the complexified algebraic curve set by the dispersion. The DOS is then recovered by an inverse Mellin transform, often expressible as a rapidly convergent sum over residues or even, in special lattices, in terms of special functions such as hypergeometric functions or complete elliptic integrals (Ray et al., 2012).

For instance, in the square lattice, the closed-form is

$\mathbf{k}$9

highlighting a van Hove singularity at $\mathbf{v} = \nabla_{\mathbf{k}} E(\mathbf{k})$0. Such algebraic-geometric frameworks allow direct linkage between lattice symmetry, spectral curve genus, and DOS singularities.

6. Practical Computation and Experimental Relevance

Calculation of GDOS in practical contexts depends on the physical regime:

• For curvature-resolved GDOS in band structures, computation involves extracting the CEC, parameterizing in angle or arc length, calculating group velocity and curvature, and assembling $\mathbf{v} = \nabla_{\mathbf{k}} E(\mathbf{k})$1 at fine angular intervals. This information, when incorporated into the Green's function analysis, directly predicts the amplitude modulation of local probes such as STM (Zhang et al., 2023).
• For geometric mean DOS in disordered systems, exact diagonalization is performed for each disorder realization to assemble histograms of LDOS at fixed energy and then average in geometric or arithmetic means for phase characterization (Zhang et al., 2013).
• For confined or many-body DOS, analytic expressions using the geometric (Weyl) formulae or geometric expansion via permutation cluster analysis yield the leading asymptotics (Mulhall et al., 2014, Hummel et al., 2012).

A salient experimental application is the protocol for extracting not only the spin texture but also the curvature of the Fermi contour by studying the angular dependence of STM amplitudes in the presence of local impurities. The GDOS enters directly into the amplitude of local DOS modulations, allowing phase and curvature information to be inferred from amplitude-only STM data, bypassing Fourier analysis and illustrating the deep geometric encoding in local response functions (Zhang et al., 2023).

7. Broader Theoretical and Experimental Implications

The concept of GDOS is universal across periodic wave systems: it governs not only electronic band structures but also photonic, phononic, and magnonic media, wherever constant-energy or isofrequency contours arise and semiclassical transport is controlled by local geometric properties of those contours. The direction-resolved and curvature-weighted state counting modulates the amplitude of Green's functions and physical responses in all such systems, providing a geometric amplitude complementary to geometric phase phenomena (e.g., Berry curvature). Its role in quantifying localization, characterizing mobility vs. true band gaps, and dictating the local measurable response underpins its foundational character in modern quantum physics (Zhang et al., 2023, Zhang et al., 2013).