An Efficient Triangulation of $\mathbb{R}P^5$

This paper presents a highly symmetric 6-dimensional polytope whose antipodal quotient yields a 24-vertex triangulation of $\mathbb{R}P^5$, conjectured to be minimal, while also providing improved 45- and 49-vertex triangulations for $\mathbb{R}P^6$.

The Gist

Imagine you are trying to wrap a complex, multi-dimensional gift (a mathematical shape called a Real Projective Space) using the fewest possible pieces of wrapping paper. In the world of mathematics, these "pieces" are triangles, and the way you arrange them is called a triangulation.

The fewer pieces (or vertices) you use, the more efficient your wrapping job is. Mathematicians have been trying to find the absolute minimum number of vertices needed to wrap these shapes for decades.

Here is the story of what this paper achieves, explained simply:

1. The Problem: Wrapping a "Twisted" Gift

Think of a sphere (like a basketball). It's easy to wrap; you just need a few triangles. But a Real Projective Space (denoted as $\mathbb{R}P^d$) is a weird, twisted version of a sphere. If you walk far enough in one direction on this shape, you end up back where you started, but "flipped" upside down.

Because of this twist, it's much harder to wrap. For a long time, mathematicians knew how to wrap these shapes for small dimensions (like 2D or 3D), but when they tried to wrap the 5-dimensional version ($\mathbb{R}P^5$), they were stuck. They knew a solution existed with 24 pieces, but it was a messy, computer-generated list with no clear pattern or beauty. It was like finding a solution to a puzzle that worked, but looked like a random pile of jigsaw pieces.

2. The Breakthrough: A Beautiful, Symmetrical Box

The authors of this paper (Guyer, Steinerberger, and Yang) didn't just find another way to wrap the 5D gift. They found a perfectly symmetrical way.

They built a 6-dimensional box (a polytope) that is perfectly balanced. Imagine a cube, but in 6 dimensions, where every point on one side has a matching point directly opposite it (like the North and South poles).

• The Magic Trick: They took the surface of this 6D box and glued every point to its opposite partner.
• The Result: This "gluing" process magically transformed the surface of the box into a perfect, efficient wrapping of the 5-dimensional projective space.

Why is this special?

• Symmetry: Their 6D box is incredibly symmetrical. It has 192 different ways you can rotate or flip it and have it look exactly the same. The previous "messy" solution only had 12 symmetries. It's the difference between a snowflake and a pile of gravel.
• Efficiency: They used 24 vertices (points). They believe this is the absolute minimum possible number. If you tried to use 23, the math says it's impossible.

3. The "Lego" Analogy

Imagine you are building a structure out of Lego bricks.

• Old Method: You had a pile of 24 bricks, but they were all different shapes and colors, and you had to glue them together in a very specific, confusing order to make the shape.
• New Method: The authors found a set of 24 bricks that are all identical twins. They fit together in a perfect, repeating pattern. Not only does it build the shape, but the whole structure is so balanced that if you spin it, it looks the same from many angles.

4. Bonus: Wrapping a 6-Dimensional Gift

The paper didn't stop there. Using the same clever tricks, they also improved the wrapping for the 6-dimensional version ($\mathbb{R}P^6$).

• The previous best record required 53 pieces.
• They found a way to do it with 45 pieces (and even a theoretical version with 49).
• This is a huge jump in efficiency, breaking a record that had stood for a while.

5. How Did They Do It? (The "AI" Part)

This is the most modern part of the story. The authors didn't just sit at a blackboard and guess.

1. The Search: They used a powerful AI tool (Google DeepMind's AlphaEvolve) to search through millions of random arrangements of points in 6D space.
2. The "Sparsity" Trick: The AI found a messy arrangement that worked well. The authors then realized that this messy arrangement was actually a "rotated" version of a very clean, simple pattern (one with lots of zeros in the numbers).
3. The Discovery: By "un-rotating" the AI's messy solution, they uncovered the beautiful, symmetrical 6D box described above.

Summary

In short, this paper is about finding the most efficient and beautiful way to wrap a complex, twisted mathematical shape.

• Before: We had a messy, ugly solution for the 5D shape.
• Now: We have a beautiful, highly symmetrical solution that uses the minimum number of pieces possible.
• Bonus: We also improved the record for the 6D shape.

It's a victory for both human intuition (recognizing the pattern in the AI's mess) and computational power, proving that sometimes the most efficient solutions are also the most beautiful.

Technical Summary

Here is a detailed technical summary of the paper "An Efficient Triangulation of $\mathbb{RP}^5$" by Dan Guyer, Stefan Steinerberger, and Yirong Yang.

1. Problem Statement

The paper addresses the fundamental problem in piecewise-linear topology of finding vertex-minimal triangulations of real projective spaces, denoted as $\mathbb{RP}^d$.

• Context: While minimal triangulations are known for spheres ($S^d$) and low-dimensional projective spaces ($\mathbb{RP}^1$ through $\mathbb{RP}^4$), the bounds for $\mathbb{RP}^d$ for $d \geq 5$ remain widely separated.
• Lower Bounds: The best known lower bound is derived from the Arnoux–Marin result (1991), which states that any triangulation of $\mathbb{RP}^d$ ($d \geq 3$) requires at least $\frac{(d+2)(d+1)}{2} + 1$ vertices. For $\mathbb{RP}^5$, this lower bound is 22 vertices.
• Upper Bounds (Prior State):
  • For $\mathbb{RP}^5$, the best known construction (Lutz, 2006) used 24 vertices, but it was a computational output lacking a clear geometric interpretation (specifically, it was not known to arise from an antipodal quotient of a polytope).
  • For $\mathbb{RP}^6$, the best known construction (Venturello–Zheng, 2021) used 53 vertices.
• Goal: To construct explicit, geometrically interpretable triangulations of $\mathbb{RP}^5$ and $\mathbb{RP}^6$ that minimize the number of vertices, ideally proving minimality or approaching it closely.

2. Methodology

The authors employ a hybrid approach combining geometric optimization, combinatorial topology, and computer-assisted search.

A. Theoretical Framework

The core strategy relies on constructing a centrally symmetric simplicial polytope $P$ in dimension $d+1$.

• If $P$ is a centrally symmetric simplicial $(d+1)$-polytope with vertex set $V$, and its boundary $\partial P$ satisfies a specific "disjoint star" condition (Condition 1), then the antipodal quotient $\partial P / \tau$ (identifying $v$ with $-v$) forms a triangulation of $\mathbb{RP}^d$.
• Condition (1): For every vertex $v$, the closed star of $v$ and the closed star of its antipode $\tau(v)$ must be disjoint. In graph terms, no vertex can be adjacent to both $v$ and $-v$ in the 1-skeleton.

B. Optimization and Discovery of $P_{6,48}$

To find the polytope for $\mathbb{RP}^5$ (requiring a 6-polytope with 48 vertices), the authors formulated an optimization problem:

1. Objective: Arrange 48 points on the 5-sphere ($S^5$) such that the minimal inner product between any two points connected by an edge in the convex hull is maximized (ideally $>0$). This ensures the "disjoint star" condition.
2. Black-Box Optimization: They utilized Google DeepMind's AlphaEvolve to search the space of 240 variables (48 points $\times$ 5 coordinates).
3. Sparsity Trick: The initial optimization yielded unstructured points. To recover a geometric structure, the authors applied an $L_1$-sparsity minimization technique. They sought an orthogonal matrix $Q$ that minimizes $\sum \|Qx_i\|_{\ell_1}$. This promoted solutions with many zero coordinates, revealing a highly symmetric structure.
4. Verification: The resulting candidate set was verified using SageMath to confirm it forms a simplicial polytope satisfying Condition (1).

C. Extension to $\mathbb{RP}^6$

Using the same computational approach, the authors searched for a 7-polytope to triangulate $\mathbb{RP}^6$, successfully finding a configuration with 90 vertices (yielding a 45-vertex triangulation).

3. Key Contributions and Results

A. Main Result: A 24-Vertex Triangulation of $\mathbb{RP}^5$

The authors present a new triangulation of $\mathbb{RP}^5$ with 24 vertices.

• Construction: It is the antipodal quotient of the boundary of a centrally symmetric simplicial 6-polytope, denoted $P_{6,48}$, which has 48 vertices.
• Explicit Coordinates: The vertices of $P_{6,48}$ are explicitly defined using parameters $\alpha=3/7, \beta=4/7, \gamma=5/7$. The coordinates are simple rational numbers with a rich combinatorial structure.
• Symmetry: The polytope has an automorphism group of order 192. This is significantly higher than the 24-vertex triangulation by Lutz (which has an automorphism group of order 12).
• Structure: The polytope can be viewed as a system of overlapping combinatorial 3-cubes. The 1-skeleton is a 23-regular graph on 48 vertices.
• Comparison: The resulting triangulation has an $f$-vector of $(24, 273, 1174, 2277, 2028, 676)$. It is non-isomorphic to Lutz's triangulation.

B. Improved Triangulations for $\mathbb{RP}^6$

The authors provide two new constructions for $\mathbb{RP}^6$:

1. Conceptual Example: A triangulation with 49 vertices derived directly from $P_{6,48}$ via a cone construction.
2. Optimized Example: A triangulation with 45 vertices derived from a new 7-polytope $P_{7,90}$ (90 vertices).

• Significance: The 45-vertex construction improves upon the previous best known result of 53 vertices (Venturello–Zheng).

C. Conjectures

• Vertex Minimality: The authors conjecture that their 24-vertex triangulation of $\mathbb{RP}^5$ is vertex-minimal.
• Reasoning:
  1. Neither their method nor the BISTELLAR program (the standard computational tool for minimal triangulations) could find a solution with fewer vertices.
  2. The 1-skeleton of their triangulation is "tight" regarding the disjoint star condition (every vertex is adjacent to exactly one of any antipodal pair).
  3. High symmetry (48 vertices is highly composite) likely allows for structures impossible with 44 or 46 vertices.

4. Significance

• Geometric Interpretation: Unlike previous computational results (like Lutz's), this work provides a geometric realization of a minimal triangulation via a convex polytope. This bridges the gap between combinatorial existence and geometric construction.
• Symmetry: The discovery of a polytope with an automorphism group of order 192 offers new insights into the role of symmetry in minimizing triangulations.
• Methodological Innovation: The combination of black-box optimization (AlphaEvolve) with $L_1$-sparsity tricks to recover structured mathematical objects is a novel approach for discovering combinatorial manifolds.
• Bounds: The work significantly narrows the gap between the lower bound (29 vertices for $\mathbb{RP}^6$) and the upper bound (45 vertices), challenging the community to find even more efficient constructions.

Summary Table of Results

Dimension	Space	Previous Best (Vertices)	New Result (Vertices)	Method
5	$\mathbb{RP}^5$	24 (Lutz, non-geometric)	24 (Geometric, $P_{6,48}$)	Polytope Quotient
6	$\mathbb{RP}^6$	53 (Venturello–Zheng)	45 ($P_{7,90}$)	Polytope Quotient
6	$\mathbb{RP}^6$	53	49 (Conceptual)	Cone over $P_{6,48}$

The paper concludes that while the 45-vertex bound for $\mathbb{RP}^6$ is likely not minimal (given the lower bound of 29), the 24-vertex bound for $\mathbb{RP}^5$ is a strong candidate for the true minimum, supported by the high symmetry and tightness of the construction.