Introduction
In the field of computational geometry, "visibility" determines which areas of a polygon are illuminated by a light source. For a simple polygon—one without holes or self-intersections—the region visible directly from a point source $s$ (denoted $V_0$) is always simply connected. Previous research extended this to one diffuse reflection ($V_1$), where light scatters in all directions upon hitting a surface, finding that the visible region remains a single solid piece without holes.
However, a long-standing question remained: does this "no-holes" property persist as the number of reflections increases? While intuition suggests that the scattering nature of diffuse reflections would eventually fill any gaps, this paper proves that topological "holes" can indeed appear after just two bounces.
The Blocker Principle and Construction
The authors' proof relies on a "blocker" point $p$, which lies on every shortest path between two specific regions. By strategically placing such blockers within a complex $n$-sided polygon, light can be "shadowed" from reaching certain areas even after multiple reflections.
To demonstrate this, the paper provides a specific geometric construction. The design involves a series of pockets and edges (labeled $a$ through $w$) that guide light through specific paths while excluding others.
Proving the "Blind Spot"
The core of the paper is the mathematical demonstration that a specific triangular region, $tqr$, remains dark after two reflections. The proof follows a step-by-step exclusion:
- Direct Visibility: The triangle $tqr$ is positioned such that it cannot be seen directly from the source $s$.
- First Reflection: No point visible from the first reflection has a line of sight into the interior of the hole.
- Second Reflection: While the boundary of the triangle $tqr$ becomes illuminated after the second reflection, its interior remains entirely dark.
This leads to Theorem 2.2, which formally states that it is possible to construct a simple polygon $P$ where the region visible after at most two diffuse reflections ($V_2(s)$) is multiply connected.
Implications and Future Directions
This finding highlights a significant topological difference between one and two reflections. While diffuse reflection is more "efficient" at illuminating space than specular (mirror-like) reflection, it is not immune to creating isolated "islands" of darkness.
In specular reflection, the number of holes can grow as $\Theta(n^2)$. The authors conjecture that for diffuse reflections, the number of holes is likely much smaller—perhaps linear ($O(n)$)—though proving this upper bound remains an open problem. Additionally, while the existence of these holes is now established, developing efficient algorithms to compute these complex visibility regions $V_k(s)$ is a key area for future research.
Summary updated