Physical distancing, one of the most widely implemented non-pharmaceutical interventions during the COVID-19 pandemic, requires maintaining a minimum separation between individuals. Meta-analyses (e.g., [1][2]) consistently show reduced transmission risk at separations of at least 1 - 2 metres. While these findings rest on epidemiological and statistical modelling, the mathematical feasibility of such constraints depends on a deeper structural assumption: that points representing individuals in physical space can be separated by non-overlapping metric neighbourhoods of controlled size.

This assumption is both topological and metric. In the Euclidean metric on ℝ³, any two distinct points admit disjoint open balls. That is, ℝ³ is Hausdorff and metrizable. The Hausdorff property ensures qualitative separability, while the metric provides quantitative scale. Without both, neighbourhoods could intersect unavoidably or collapse under identification, rendering fixed-distance policies inconsistent.

This article develops a rigorous topological and metric account of physical distancing. We formalize constraints in terms of uniform metric separation, show that Hausdorffness is necessary but insufficient, and illustrate pathologies in quotient spaces common in epidemic modelling.

Let individuals be points in a metric space ( X , d ) . A minimum-distance policy with threshold δ > 0 requires d ( x , y ) ≥ δ for all distinct x , y .

Definition3.1(UniformSeparation)

A metric space is uniformly separated (with parameter δ > 0) if d ( x , y ) ≥ δ for all distinct x , y . This property mathematically encodes the core requirement of physical-distancing policies [4].

Theorem3.1

Let ( X , d ) be a metric space. A minimum-distance policy with threshold δ > 0 is equivalent to the existence of disjoint open balls B ( x , δ 2 ) and B ( y , δ 2 ) for all distinct x , y with d ( x , y ) ≥ δ . This condition implies Hausdorffness but is strictly stronger.

Proof

⇒ If d ( x , y ) ≥ δ , then B ( x , δ 2 ) ∩ B ( y , δ 2 ) = ∅ (triangle inequality).

⇐ If such disjoint balls exist whenever d ( x , y ) ≥ δ , then the centers are separated by at least δ. Since metric balls are open, this gives Hausdorffness. However, there exist Hausdorff spaces (e.g., certain non-metrizable ones or metric spaces without a positive infimum on distances) that lack uniform separation for any δ > 0. Hence the implication is strict. ■

Here is a standard illustration of disjoint personal zones in the Euclidean plane. (See Figure 1Figure 1)

Here is a standard illustration of disjoint personal zones in the Euclidean plane:

Figure 1. Disjoint open balls of radius δ 2 around distinct points x and y separated by at least δ enforcing a minimum-distance policy (e.g., 2 meters) in Euclidean space.

Epidemiological Interpretation

In spatial SIR/SEIR models, transmission is often governed by a kernel β ( d ( x , y ) ) that decays with distance (e.g., exponential or threshold-based), becoming near-zero beyond δ . This presupposes metric balls of controlled radius exist and do not collapse. However, real models tolerate probabilistic overlaps for crowding and mobility.

Here is an example of such a distance-dependent transmission kernel. (SeeFigure 2Figure 2)

Figure 2. Distance-dependent transmission probability β ( d ) decreasing to near-zero beyond threshold δ , as commonly used in spatial epidemiological models.

Many epidemiological models use coarse-grained or aggregated spaces via quotient topologies.

Example 4.1 Grid Aggregation

Divide ℝ 2 into square cells and identify all points within each cell. The resulting quotient space may remain Hausdorff, but distances inside a cell become undefined, destroying uniform separation locally.

Example 4.2 Mean-Field Identification

In mean-field approximations, individuals in the same compartment are identified, collapsing metric structure and preventing minimum-distance enforcement.

Example 4.3 Particle Systems

Figure 3. Grid-based coarse-graining: Points within each cell are identified in the quotient space, making local uniform separation ill-defined (relevant to lattice and metapopulation models).

Configuration spaces of identical particles often quotient by permutations, where uniform separation fails even if the space is Hausdorff.

Here is a visual representation of grid aggregation leading to identification. (See Figure 3Figure 3)

Corollary 4.4

Distance-based physical-distancing models require not only Hausdorffness but a metric that survives aggregation and identification procedures [5].

Distance-dependent kernels β ( d ( x , y ) ) demand a true metric on the agent space. Aggregation must preserve metric separation; otherwise, distancing cannot be meaningfully represented. Future models should consider these structural requirements, especially in coarse-grained or manifold-based simulations [6].

Physical distancing is a mathematical constraint on the modelling space. Hausdorffness provides qualitative separability, while a compatible metric with uniform separation ensures quantitative enforcement. Both are essential for well-posed distance-based policies in spatial epidemiology. By clarifying these assumptions, we offer a more rigorous foundation for epidemic modelling, highlighting how implicit topological and metric structures underpin real-world public-health interventions.