---
title: Real-Space Second Chern Number
url: https://www.emergentmind.com/topics/real-space-second-chern-number
type: topic
---

# Real-Space Second Chern Number

Searching arXiv for recent papers on real-space second Chern number and closely related formulations.
The real-space second Chern number is a higher-dimensional topological invariant formulated directly in terms of position-space operators and the occupied-state projector, rather than exclusively through Berry curvature on a translationally invariant Brillouin torus. In four-dimensional gapped systems, it is the natural counterpart of the first Chern number in two dimensions, and it quantizes the topology underlying the 4D quantum Hall response. Recent work has established exact real-space formulas, finite-system estimators, and kernel-polynomial implementations for disordered 4D Chern insulators, while related constructions have shown that a second Chern number can also emerge in mixed momentum–real-space parameter spaces and in hyperbolic lattices with an effective 4D Brillouin zone [2502.14299], [2507.18919], [1611.01998], [2302.13217].

## 1. Definition and topological setting

In momentum space, the second Chern number for occupied bands over a four-dimensional Brillouin zone is written as
$$
C_2 = \frac{1}{32\pi^2} \int_{\mathrm{BZ}} d^4k \,\epsilon^{ijkl}\, \mathrm{Tr}\!\left(F_{ij}(k)\,F_{kl}(k)\right),
$$
which is the standard 4D Berry-curvature expression in translationally invariant systems [2502.14299]. In wedge-product form, one also encounters
$$
C_2 = \frac{1}{8\pi^2}\int_{\mathrm{BZ}_{4\mathrm{D}}} \mathrm{d}^4k\,\mathrm{tr}\bigl(\Omega\wedge\Omega\bigr),
$$
with $\Omega=dA-iA\wedge A$ and $A_\mu^{mn}(k)=i\langle u_m(k)|\partial_{k_\mu}u_n(k)\rangle$ for the occupied-band subbundle [2302.13217]. These normalizations are paper-specific conventions, but both encode the same integer invariant.

The real-space program replaces momentum derivatives by position-space derivations and expresses the invariant using the spectral projector $P$ onto occupied states. The exact marker form derived for translation-invariant systems is
$$
C_2 = \frac{1}{2!}\left(\frac{2\pi}{i}\right)^2 \epsilon^{\mu\nu\rho\sigma} \, \mathrm{tr}\!\left[P\,[X_\mu,P]\,[X_\nu,P]\,[X_\rho,P]\,[X_\sigma,P]\right]_{ii},
$$
where the trace is taken over internal degrees of freedom at a fixed site $i$ [2502.14299]. For translationally invariant systems, the expression is independent of $i$; for disordered systems, the appropriate replacement is a trace per unit volume or a windowed finite-size estimator [2502.14299].

A closely related projected-position form used for large-scale kernel-polynomial computations is
$$
C_2 = - 2 \pi^2 \epsilon^{j_1 j_2 j_3 j_4} \mathrm{Tr}\!\left[ P X_{j_1} P X_{j_2} P X_{j_3} P X_{j_4} P \right],
$$
evaluated as a bulk trace per volume in open systems [2507.18919]. The same work notes that the commutator form
$$
C_2 = - 2 \pi^2 \epsilon^{abcd} \mathrm{Tr}\!\left[ P [X_a, P] [X_b, P] [X_c, P] [X_d, P] \right]
$$
is equivalent for gapped systems with a well-defined trace per volume, although the calculations are performed with the projected-position product form [2507.18919].

This framework places the real-space second Chern number within noncommutative geometry. Defining $\delta_\mu(A)=i[X_\mu,A]$, one has
$$
C_2 = \frac{1}{2!}\left(\frac{i}{2\pi}\right)^2 \epsilon^{\mu\nu\rho\sigma}\,\tau\!\left(P\,\delta_\mu P\,\delta_\nu P\,\delta_\rho P\,\delta_\sigma P\right),
$$
which matches the marker form through $[X_\mu,P]=-i\,\delta_\mu(P)$ [2502.14299].

## 2. Exact real-space formulas and finite-system estimators

An important development is the extension of Kitaev’s real-space formulation of the first Chern number to the second Chern number [2502.14299]. In addition to the site-local marker, that work derives a projector-partition form using half-space lattice projectors $\Pi_\mu$:
$$
C_2 = \frac{1}{2!}\left(\frac{2\pi}{i}\right)^2 \,\mathrm{Tr}\!\left(\epsilon^{\mu\nu\rho\sigma}\,P\,[\Pi_\mu,P]\,[\Pi_\nu,P]\,[\Pi_\rho,P]\,[\Pi_\sigma,P]\right)
$$
and equivalently
$$
C_2 = \frac{1}{2!}\left(\frac{2\pi}{i}\right)^2 \,\mathrm{Tr}\!\left(\epsilon^{\mu\nu\rho\sigma}\,P\,\Pi_\mu\,P\,\Pi_\nu\,P\,\Pi_\rho\,P\,\Pi_\sigma\,P\right).
$$
This is presented as the recommended finite-system estimator in practice [2502.14299].

The same paper also proves a five-region deformation-invariant partition formula. Partitioning the 4D lattice into five disjoint regions $A=1,\dots,5$ with projectors $\Pi_A$, one obtains
$$
C_2 = \frac{4!}{2!}\left(\frac{2\pi}{i}\right)^2 \,\mathrm{Tr}\!\left(\epsilon^{ABCDE}\,P\,\Pi_A\,P\,\Pi_B\,P\,\Pi_C\,P\,\Pi_D\,P\,\Pi_E\right).
$$
This form is described as strictly topological in the infinite system, although on finite systems with truncation it can be slightly less accurate than the half-space projector formula [2502.14299].

For disordered systems, quantization requires a spectral gap at the Fermi energy $E_F$, or more generally a mobility gap, together with locality of the Fermi projector $P$ and decay of the commutators $[X_\mu,P]$ [2502.14299]. Under these conditions, the real-space formulas yield an integer stable under weak disorder and smooth perturbations [2502.14299]. A practical estimator uses a window projector $\Pi^{(L)}$ selecting a hypercube of linear size $L$:
$$
C_2(L) = \frac{1}{2!}\left(\frac{2\pi}{i}\right)^2 \frac{1}{L^4} \,\mathrm{Tr}\!\left(\Pi^{(L)}\,\epsilon^{\mu\nu\rho\sigma} P\,\Pi_\mu\,P\,\Pi_\nu\,P\,\Pi_\rho\,P\,\Pi_\sigma\,P\right),
$$
with $L\to\infty$ [2502.14299]. In numerics, choosing $L\approx N/2$ for a system of linear size $2N$ is reported to effectively cancel boundary terms [2502.14299].

A recurrent point in this literature is the necessity of bulk windowing. Taking the full finite-system trace can produce cancellations between artificial boundaries and drive the estimator to zero, whereas intermediate windows suppress boundary artifacts [2502.14299]. This suggests that the real-space second Chern number is best regarded not as a naive finite-volume trace, but as a bulk trace-per-volume quantity regularized to exclude nonuniversal edge contributions.

## 3. Computational frameworks

Two complementary computational strategies appear in the cited work. The first is a direct finite-size evaluation based on the half-space projector formula or the equivalent commutator marker [2502.14299]. The second is a scalable kernel polynomial method (KPM) implementation that avoids diagonalization and allows system sizes far beyond exact projector construction [2507.18919].

For the projector-based approach, the central object is the Fermi projector $P$. One route is spectral flattening: define $Q=\mathrm{sgn}(H)$ and $P=(1-Q)/2$ when the spectrum is gapped around $E_F$ [2502.14299]. The same work also describes KPM/Chebyshev approximations to the Heaviside step function and rational approximations via shifted inverses $(H-z_j)^{-1}$ [2502.14299]. The paper emphasizes that for disordered models it is often more efficient to access $P$ implicitly as a linear map on vectors, avoiding storage of dense matrices [2502.14299].

The KPM study makes this implicit strategy explicit. After rescaling the Hamiltonian to $\hat H$ with spectrum in $[-1,1]$,
$$
\hat H = (2/\Delta E) [ H - (E^+ + E^-)/2 ],
$$
the projector is approximated by a truncated Chebyshev series with Jackson kernel,
$$
P(\hat H) \approx P_M(\hat H) = \sum_{\chi=0}^M g_\chi \mu_\chi T_\chi(\hat H),
$$
with moments
$$
\mu_0(\mu) = 1 - (\arccos \mu)/\pi,\qquad
\mu_\chi(\mu) = - 2 \sin[\chi \arccos \mu]/(\chi \pi),\quad \chi \ge 1,
$$
and Jackson coefficients
$$
g_\chi = \frac{ (M - \chi + 1) \cos(\pi \chi/(M+1)) + \sin(\pi \chi/(M+1)) \cot(\pi/(M+1)) }{M+1}.
$$
The recursive action on a vector uses
$$
|v_0\rangle = |v\rangle,\qquad
|v_1\rangle = \hat H |v_0\rangle,\qquad
|v_{\chi+1}\rangle = 2 \hat H |v_\chi\rangle - |v_{\chi-1}\rangle
$$
[2507.18919].

The trace per volume is then estimated stochastically on a bulk subregion $S$:
$$
\mathrm{Tr}_S( \cdots ) \approx \frac{1}{R |S|} \sum_{i=1}^R \langle r_i| \cdots |r_i\rangle,
$$
where $|r_i\rangle$ are random phase vectors supported in $S$ [2507.18919]. For the reported 4D calculations, the paper uses $M=256$ and $R=5$ [2507.18919]. Each $C_2$ evaluation proceeds by repeated applications of $P_M(\hat H)$ and coordinate operators $X_j$, summed over the $4!=24$ nonzero Levi-Civita terms [2507.18919].

The two approaches differ mainly in scale and implementation details. The exact projector formulas establish the invariant and its finite-size regularizations [2502.14299]. The KPM formulation supplies the numerical route for large sparse Hamiltonians, including systems with $30^4$ sites, and extends the same logic to an exploratory six-dimensional calculation of the third Chern number [2507.18919].

## 4. Wilson–Dirac benchmark and disorder

The canonical test bed for real-space evaluations of $C_2$ is the four-dimensional Wilson–Dirac model. One formulation used in the projector-based study is
$$
H = -\frac{i t}{2} \sum_{\mu=1}^{4} \gamma^\mu \left(\delta_\mu - \delta_\mu^{-1}\right)
    + m\,\gamma^{5}
    + \frac{b}{2} \sum_{\mu=1}^{4} \gamma^{5}\left(\delta_\mu + \delta_\mu^{-1} - 2\right),
$$
with periodic boundary conditions on sites $-N \le i_\mu \le N-1$ and truncation window $\Pi^{(L)}$ selecting $-L \le i_\mu \le L-1$ [2502.14299]. The paper reports results for $(N,L)=(2,1)$, $(3,1)$, and $(4,2)$ [2502.14299]. In the clean phase diagram, the real-space estimator reproduces quantized plateaus as $m$ varies; for example, at $m=3$ with $t=b=1$, the exact value is $C_2=3$, while the finite-size estimate gives $C_2\approx 2.35$ for $(N,L)=(2,1)$ and approaches the integer value as the system size increases [2502.14299].

The KPM study uses the $2n$-dimensional Wilson–Dirac family
$$
H_{2n}(k) = \sum_{\mu=0}^{2n} \psi^\dagger(k) d^\mu(k) \Gamma^{2n}_\mu \psi(k),
$$
with
$$
d^0(k) = m + t \sum_{j=1}^{2n} \cos k_j,\qquad
d^j(k) = \lambda \sin k_j,
$$
on a hypercubic lattice of linear size $L$, volume $V=L^{2n}$, fixing $t=\lambda=1$ and $E_F=0$ in all numerics [2507.18919]. For the 4D case, the explicit gamma matrices are
$$
\Gamma^4 = (\sigma_x \otimes \sigma_0, \sigma_y \otimes \sigma_0, \sigma_z \otimes \sigma_x, \sigma_z \otimes \sigma_y, \sigma_z \otimes \sigma_z)
$$
[2507.18919]. The paper computes $C_2$ versus mass for $L=4,12,20$ and reports rapid convergence to the momentum-space phase diagram. At $m=-1$, where one expects $C_2=-3$, the deviation $|C_2+3|$ decays exponentially with $L$ [2507.18919].

Both studies treat disorder directly in real space. In the KPM work, Anderson-type on-site disorder is added as
$$
\Delta H = \sum_i W_i (\sigma_0 \otimes \sigma_0) c_i^\dagger c_i,
$$
with independent $W_i$ uniformly distributed in $[-W,W]$ [2507.18919]. The disorder-averaged phase diagram $C_2(m,W)$ agrees well with boundaries predicted by the self-consistent Born approximation at weak-to-moderate disorder [2507.18919]. The SCBA self-energy is written as
$$
\Sigma = (W^2 / 3) (a / 2\pi)^4 \int_{\mathrm{FBZ}} d^4k \left[ 1 / (E_F - H_4(k) - \Sigma) \right],
$$
and the practical comparison uses the lowest-order Born approximation with $\Sigma=0$ on the right-hand side to determine where the renormalized spectrum closes its gap [2507.18919].

The projector-based study distinguishes two disorder mechanisms in the 4D Wilson–Dirac model: random hopping and random mass [2502.14299]. With $t=b=1$, $m=3$, and averages over 10 disorder realizations on the $(N,L)=(2,1)$ system, random hopping drives a direct transition from the $C_2=3$ phase to a trivial phase as the half-filled gap closes and reopens, while random mass produces a broad nearly gapless regime in which $C_2$ is ill-defined rather than a trivial gapped phase [2502.14299]. These results place the real-space second Chern number in the same operational category as lower-dimensional Chern markers: it remains informative in the presence of disorder provided that the occupied projector remains local and the relevant gap structure persists.

## 5. Mixed-space and non-Euclidean realizations

Although the phrase “real-space second Chern number” often refers to projector formulas in physical space, related work shows that the same invariant can arise when one of the four coordinates is a real-space geometric parameter. In a magnetized double-gyroid photonic crystal, a helical modulation
$$
f(x,y,z) > f_0 + \Delta f \cos(\pi z/a + w \theta),\qquad
\theta=\arctan(y/x),\quad w\in\mathbb{Z},
$$
produces localized one-way modes along the helix axis [1611.01998]. Far from the helix core, the Bloch Hamiltonian depends smoothly on $(k_x,k_y,k_z,\theta)$, so the eigenstates are periodic in all four variables and define a 4D torus $T^4$ [1611.01998]. The second Chern number is then defined on this mixed momentum–real-space parameter space as
$$
C_2 = \frac{1}{4\pi^2} \int d^3k\, d\theta \,\mathrm{Tr}\!\left[ F_{xy} F_{z\theta} + F_{yz} F_{x\theta} + F_{zx} F_{y\theta} \right],
$$
using the paper’s convention and normalization [1611.01998].

In that construction, time-reversal symmetry must be broken for $C_2\neq 0$ because $F_{ij}$ with $i,j\in\{x,y,z\}$ is odd under time reversal while $F_{i\theta}$ is even [1611.01998]. The simplest helical modulation yields $C_2=w$ for $w=+1,+2,+3$, and the observed number and chirality of one-way fiber modes satisfy $N_{\mathrm{chiral}}=C_2$ [1611.01998]. This is not a projector-based real-space marker, but it is a real-space manifestation of the second Chern number in which the extra coordinate is the physical helix angle rather than a synthetic momentum dimension [1611.01998].

A different route appears in hyperbolic band topology. In the $\{8,8\}$ hyperbolic tiling, four independent translation directions generated by the Fuchsian group $\Gamma_{\{8,8\}}$ are equipped with $U(1)$ twist phases, producing a 4D Brillouin zone $T^4$ with angles $k=(k_1,k_2,k_3,k_4)$ [2302.13217]. The corresponding 4D Bloch Hamiltonian is
$$
H(\mathbf{k}) = \mathbf{d}(\mathbf{k})\cdot\boldsymbol{\Gamma} + i\,a\,\Gamma_1\Gamma_4,
$$
with
$$
\mathbf{d}(\mathbf{k}) = \{\,t_1\sin k_1,\;t_2\sin k_2,\;t_3\sin k_3,\;t_4\sin k_4,\; m + \sum_{j=1}^4 J_j \cos k_j \,\},
$$
and the second Chern number again given by
$$
C_2 = \frac{1}{8\pi^2}\int_{\mathrm{BZ}_{4\mathrm{D}}} \mathrm{d}^4k\,\mathrm{tr}\bigl(\Omega\wedge\Omega\bigr)
$$
[2302.13217].

For $m=0.7$, $a=0.2$, the central gap has $C_2=32$ while the first Chern numbers vanish; for $m=0.7$, $a=3.2$, the central 4D gap closes and the side gaps have $C_2=0$ but nontrivial first Chern numbers on certain 2D slices, such as $C_1(k_1,k_4)=\pm 1$ near $\varepsilon\approx\pm 3.2$ [2302.13217]. The authors do not implement a projector-based real-space marker; instead they construct abelian periodic-boundary-condition clusters whose discrete translation operators can be simultaneously diagonalized, enabling a discrete momentum-space evaluation of $C_2$ even in finite hyperbolic lattices [2302.13217]. This suggests a broader taxonomy: projector markers, mixed-space invariants, and twist-space invariants are distinct constructions that all realize second-Chern topology outside the standard 4D crystal setting.

## 6. Physical interpretation, responses, and limitations

The second Chern number is the topological invariant of the 4D quantum Hall effect and controls a quantized nonlinear response coefficient in four dimensions [2507.18919]. Dimensional reduction connects 4D Chern insulators to 3D time-reversal invariant topological insulators, where pumping along the extra dimension yields the magnetoelectric response $\theta$, and $C_2$ controls the topological term proportional to $\theta E\!\cdot\!B$ in the reduced description [2502.14299]. In this sense, the real-space second Chern number is not merely a numerical substitute for Berry-curvature integration; it is the disorder-compatible form of the bulk invariant governing higher-dimensional response.

The photonic mixed-space realization gives an explicit transport interpretation. A nonzero $C_2$ produces one-way fiber modes localized on the helix axis, with immunity to Rayleigh and Mie scattering, strong inhibition of stimulated Brillouin scattering and backward Raman scattering, elimination of Fresnel reflection, and single-polarization one-way propagation [1611.01998]. In that system, the coupling between momentum-space Berry curvature and the $\theta$-winding of the Dirac mass yields chiral one-dimensional defect transport [1611.01998].

The hyperbolic circuit realization provides a different boundary signature. Midgap boundary states appear only in bandgaps with nonzero Chern invariants; for the central gap with $C_2=32$ at $m=0.7$, $a=0.2$, a significant boundary-node impedance peak occurs in the frequency range $[0.58, 0.607]$ MHz, and the spatial impedance map at $0.598$ MHz is localized around partially open boundary units [2302.13217]. Because the mismatch between the 4D momentum torus and 2D real-space hyperbolic geometry complicates fixed-direction chiral transport, the observed correspondence is formulated in terms of bulk–boundary spectral signatures rather than a direct transport coefficient [2302.13217].

Several limitations are common across formulations. Exact quantization in real space requires a spectral gap or mobility gap and locality of $P$; near gap closures, finite-size estimators deviate from integers and become sensitive to boundary conditions [2502.14299]. In KPM, kernel broadening is of order $\Delta E/M$, so narrow gaps require larger $M$; otherwise the projector is smeared and $C_2$ can lose quantization [2507.18919]. Higher dimensions amplify the computational burden: in 6D, the third-Chern calculation involves $6!=720$ nonzero Levi-Civita terms and much larger Hilbert spaces, so the reported results remain qualitative rather than fully quantized [2507.18919].

A common misconception is that any “real-space” manifestation of $C_2$ must be a local projector marker. The literature considered here shows that this is too narrow. The exact real-space formulas of the Kitaev type provide one rigorous route [2502.14299], but mixed momentum–real-space constructions in photonics [1611.01998] and twist-generated 4D parameter spaces in hyperbolic lattices [2302.13217] also realize second-Chern topology in systems lacking an ordinary 4D crystal momentum. A plausible implication is that “real-space second Chern number” now names a family of related higher-dimensional invariants whose common feature is independence from a conventional 4D translationally invariant Bloch setting.

## 7. Outlook and generalization

The recent literature establishes a computational and conceptual progression. First, exact real-space formulas now place the second Chern number on the same footing as established lower-dimensional Chern markers, with explicit equivalence to the momentum-space invariant and practical finite-system estimators [2502.14299]. Second, large-scale KPM calculations demonstrate that these formulas are usable at the scale of four-dimensional lattices with $30^4$ sites and can capture disorder-driven topology in agreement with self-consistent Born approximation predictions [2507.18919]. Third, the invariant has been realized in nonstandard geometries and parameter spaces, including mixed-space photonic fibers and hyperbolic circuit networks [1611.01998], [2302.13217].

The generalization to higher Chern classes is already explicit. The KPM work extends the projected-position formula to the third Chern number in six dimensions,
$$
C_3 = - \frac{4 \pi^3 i}{3} \epsilon^{j_1 j_2 j_3 j_4 j_5 j_6} \mathrm{Tr}\!\left[ P X_{j_1} P X_{j_2} \cdots P X_{j_5} P X_{j_6} P \right],
$$
and the exact real-space framework states the broader index-theoretic structure for $2n$ dimensions [2507.18919], [2502.14299]. Although full quantization in 6D remains limited by finite-size effects, the reported qualitative agreement suggests that higher-dimensional real-space topological invariants are computationally accessible [2507.18919].

Open directions are identified most explicitly in the hyperbolic setting: developing a robust real-space local marker for $C_2$ on hyperbolic lattices, establishing chiral response measurements linked to $C_2$ in hyperbolic geometries, scaling to larger clusters while mitigating losses and parasitics, exploring disorder and interactions, and extending to other tessellations and symmetries [2302.13217]. In Euclidean lattices, the main frontier is algorithmic: combining KPM with tensor-network techniques to push accessible Hilbert spaces further, especially for higher Chern numbers [2507.18919]. Taken together, these results indicate that the second Chern number has moved from a largely momentum-space invariant of ideal 4D band theory to a real-space computable quantity relevant to disorder, finite samples, curved lattices, and synthetic higher-dimensional platforms.

Source: https://www.emergentmind.com/topics/real-space-second-chern-number