---
title: Path Discrepancy in Lattice Checkerboards
url: https://www.emergentmind.com/topics/path-discrepancy
type: topic
---

# Path Discrepancy in Lattice Checkerboards

Path discrepancy, in geometric discrepancy theory, denotes the discrepancy induced on a measurable curve by a checkerboard-type coloring of the plane: it is the absolute value of the signed length of the curve lying in positively colored tiles minus the length lying in negatively colored tiles. In the general lattice-checkerboard setting, the plane is tiled by lattice translates of an arbitrary bounded, Lebesgue measurable fundamental domain \(Q\), not necessarily squares, and the principal result is a square-root law: every two-coloring of such a tiling forces arbitrarily long line segments whose discrepancy is at least a constant multiple of the square root of their length [1601.02143].

## 1. General lattice-checkerboard formulation

Let \(T \subseteq \mathbb{R}^2\) be a lattice, and let \(Q\) be a bounded, Lebesgue measurable fundamental domain of \(T\). The plane is partitioned as
\[
\mathbb{R}^2 = \bigcup_{t \in T} (t + Q).
\]
A checkerboard coloring is specified by assigning to each tile \(t+Q\) a color \(z_t \in \{-1,+1\}\). The associated coloring function is
\[
f(x) = \sum_{t \in T} z_t \chi_Q(x-t),
\]
where \(\chi_Q\) is the indicator function of \(Q\).

For any measurable curve \(I\), and especially for a line segment, the discrepancy is
\[
D(I)=\left|\int_I f(x)\,ds\right|.
\]
Equivalently, this is the signed length of \(I\) passing through positive tiles minus that through negative tiles. The formulation generalizes the square checkerboard by allowing \(Q\) to be any bounded measurable tile for the lattice rather than a square [1601.02143].

This setup preserves the lattice structure while discarding the geometric specialness of the square. The resulting notion of path discrepancy is therefore not tied to axis-aligned unit cells: the same formalism applies whenever a bounded fundamental domain tiles the plane by lattice translations.

## 2. Main lower bounds for line segments

The central estimate is first stated for finite superpositions of translated tiles with arbitrary complex weights. If \(G \subseteq T\) is finite and
\[
f(x)=\sum_{g \in G} z_g \chi_Q(x-g),
\]
then there exists a straight line \(S\) such that
\[
\left|\int_S f\right| \geq C \cdot (\#G)^{-1/2}\left(\sum_{g\in G}|z_g|^2\right)^{1/2},
\]
where \(C\) depends only on the lattice \(T\) and the shape \(Q\) [1601.02143].

In the two-coloring case \(z_g=\pm 1\), this yields the checkerboard corollary:
\[
\exists\ \text{arbitrarily long line segments } I:\quad \left|\int_I f\right| \geq C\sqrt{\mathcal{L}(I)},
\]
where \(\mathcal{L}(I)\) is the length of \(I\), and \(C\) depends only on \(Q\) and \(T\).

The theorem asserts a lower bound that is insensitive to the specific geometry of the fundamental domain, provided only that it is a bounded measurable lattice tile. In particular, the square-root growth of discrepancy is not a peculiarity of square cells. It is a robust feature of lattice checkerboards generated from arbitrary bounded fundamental domains.

## 3. Quantitative form of the square-root law

A more explicit asymptotic statement is obtained by restricting attention to colorings supported inside large squares of side \(R\). In that setting, the maximum path discrepancy for line segments crossing the square is at least \(\Omega(\sqrt{R})\) [1601.02143].

This is the quantitative form of the square-root law. It implies that as one examines larger and larger scales, there are always line segments whose imbalance between positive and negative portions grows at least like the square root of the segment length. The lower bound is therefore scale-sensitive but sublinear: it rules out uniformly small discrepancy on all long segments, while not asserting linear growth.

The dependence of the constant only on the lattice and the tile shape is essential. It shows that the phenomenon is geometric and structural rather than dependent on a particular coloring pattern. The coloring may be arbitrary, but the ambient tiling imposes a nontrivial lower bound on the discrepancy of some long line segment.

## 4. Fourier analysis, tiling, and spectral structure

Earlier square-checkerboard arguments relied strongly on the explicit zeros and decay of the Fourier transform of the square indicator function. For general \(Q\), that structure is unavailable, so the proof uses a different mechanism based on tiling and spectral properties [1601.02143].

The Fourier transform of the coloring function factors as
\[
\widehat{f}(\xi)=\widehat{\chi_Q}(\xi)\phi(\xi), \qquad
\phi(\xi)=\sum_{g\in G} z_g e^{-2\pi i g\cdot \xi}.
\]
The crucial substitute for square-specific Fourier decay is the spectral identity associated with lattice tilings. If \(Q\) tiles \(\mathbb{R}^2\) by translations of \(T\), then the exponentials \(\{e^{2\pi i t^* \cdot x}\}_{t^*\in T^*}\), where \(T^*\) is the dual lattice, form an orthonormal basis for \(L^2(Q)\), and
\[
\sum_{t^* \in T^*} |\widehat{\chi_Q}(\xi-t^*)|^2=1
\qquad \forall \xi \in \mathbb{R}^2.
\]

This identity gives global \(L^2\)-control over \(\widehat{\chi_Q}\) without requiring pointwise information such as zero sets or decay rates. Parseval’s identity and the tiling spectral identity are then used to relate line integrals of \(f\) to \(L^2\)-mass in frequency space. The proof therefore shifts from explicit shape-dependent Fourier analysis to a structural use of the fact that \(Q\) is a bounded measurable fundamental domain of a lattice.

## 5. Scope, significance, and limits of the result

The significance of the general theorem is that it removes any restriction that the fundamental domain be square. The method applies to any bounded measurable tile for the lattice, allowing shapes that are drastically different from squares, including hexagons and L-shapes [1601.02143].

This broadens checkerboard discrepancy from a special geometric model to a lattice-tiling phenomenon. The result shows that the square-root lower bound is stable under replacement of the square by an arbitrary bounded fundamental domain, provided the lattice tiling structure is retained.

A common misconception is to treat the square-root law as a universal property of all planar colorings. The result does not say that. It is formulated for checkerboard-type colorings arising from lattice tilings. The data also record that for certain pathological, non-checkerboard colorings, the lower bound can be avoided. Conversely, within the generalized checkerboard class, the square-root lower bound is described as tight up to constants, with randomized colorings achieving at most \(O(\sqrt{\ell \log \ell})\) for line segments of length \(\ell\) [1601.02143].

The theorem is therefore both robust and specific: robust with respect to the shape of the tile, specific with respect to the class of colorings under consideration.

## 6. Terminological variants in other research areas

The expression “path discrepancy” is not uniform across the literature. In multiparameter persistent homology, it denotes a generalization of matching distance in which one replaces straight-line slices in parameter space by monotone paths; the resulting path distance is
\[
d^P(\mathcal{M},\mathcal{N})=\sup_\pi d_\pi(\mathcal{M},\mathcal{N}),
\]
with the supremum taken over admissible monotone paths \(\pi\) [2507.23762].

In deployed path-aware networking, “path discrepancy” refers instead to asymmetric path availability between endpoints: the set of paths from \(A\) to \(B\) may differ from the set from \(B\) to \(A\), as observed in SCIONLab [2509.04695]. In another geometric usage, “discrepancy” between parametrized planar curves is defined as the minimum deformation energy of an admissible curve in \(SE(2)\) mapping one curve to the other [1305.3572].

These usages are conceptually distinct. The geometric discrepancy-theoretic notion studied for lattice checkerboards concerns signed length imbalance along curves in a two-colored tiling of the plane. The shared terminology reflects the general idea of measuring divergence along a path, but the underlying objects, metrics, and proof techniques are entirely different across these settings.

Source: https://www.emergentmind.com/topics/path-discrepancy