Toroidal Graphs

• Toroidal graphs are finite graphs embedded on a torus with non-crossing edges, characterized by a genus-1 Euler characteristic and distinct face-vertex configurations.
• They are classified into various types including toroidal maps, Cayley graph models, and semi-equivelar maps, each defined by specific combinatorial and symmetry constraints.
• Research on toroidal graphs reveals unique spectral, metric, and energy-minimization properties, with applications in graph drawing, topological analysis, and discrete geometry.

A toroidal graph is a finite graph that admits an embedding on the orientable surface of genus one (the standard torus) such that its edges do not cross. This concept subsumes graphs with rich structural, topological, and combinatorial properties, including numerous subclasses arising from topological graph theory, Cayley graph constructions, algebraic models, and discrete geometry. Toroidal graphs and their associated maps serve as central test cases in studies of surface-embedded graphs, spectral graph theory, metric properties, topological embeddability, and extremal combinatorics.

1. Embeddings, Topological Properties, and Basic Definitions

A toroidal embedding requires a crossing-free realization of the graph on the torus, equivalently, a graph of genus at most one. The standard construction is via coordinate identifications on the rectangle $[0,a] \times [0,b]$, with $(0,y) \sim (a,y)$ and $(x,0)\sim (x,b)$, producing a flat torus with rectangular or parallelogram fundamental domain (Biedl, 2022). The genus-1 Euler characteristic ($n - m + f = 0$ for $n,m,f$ vertices, edges, and faces, respectively) restricts possible face--vertex configurations.

Embedded graphs on the torus bifurcate into several classes:

• Toroidal maps: Embeddings where each face is a topological disk; equivalently, polyhedral 2-cell embeddings such that any two faces intersect in at most one vertex or edge (Maity et al., 2013).
• Essentially $k$-connected: The universal cover of the embedding is $k$-connected.
• Cayley graph models: Symmetric constructions such as honeycomb toroidal graphs via dihedral group Cayley graphs (Alspach, 2020) and involutory Cayley graphs for ring-based studies (Keshavarzi et al., 2 Aug 2025).
• Visibility representations: Every loopless toroidal graph admits a visibility representation on a rectangular (not merely parallelogram) torus (Biedl, 2022).

2. Classification and Enumeration of Toroidal Graphs

Enumeration of toroidal maps entails the classification of semi-equivelar maps—toroidal embeddings where every vertex has the same cyclic face-sequence. Eleven distinct semi-equivelar types exist on the torus. Eight non-classical families are fully classified via $T(r,s,k)$-representations, which encode the map as a grid with horizontal cycles, layer count $s$, and shift/gluing parameter $k$. The classification criteria for valid $T(r,s,k)$ depend on parity and divisibility constraints on $r,s,k$ and yield isomorphism via cycle-type data and symmetries such as $k \mapsto r-k$ (Maity et al., 2013).

Tables of non-isomorphic representatives for small $n$ (number of vertices) are constructed algorithmically, for instance (abbreviated):

Type	$T(r, s, k)$ criteria	Sample counts
$\{3^3,4^2\}$	$s > 2$ even, $r s> 10$, bounds on $k$	$i(10)=1, i(12)=3$
$\{3,6,3,6\}$	$2 | r$, $2rs\ge 21$, $k$ in even-offset ranges	$i(21)=1, i(24)=2$
$\{4,8^2\}$	$4|r$, $r s \ge 20$, $k \equiv 6 \pmod 4$	$i(20)=1, i(24)=2$

These representations underpin the combinatorial diversity and symmetry mechanisms in toroidal maps.

3. Structural and Spectral Properties

Toroidal graphs display unique spectral, metric, and decompositional characteristics:

• Vertex arboricity: Forbidding small cycles in toroidal graphs improves the vertex arboricity, with the exact result $a(G) \leq 2$ for toroidal graphs without $k$-cycles for $k \in \{3,4,5\}$; the result fails for $k \geq 6$ due to the complete graph $K_5$ (Choi et al., 2013).
• (d,h)-decomposability: For certain forbidden cycle conditions (e.g., having neither $i$-cycles nor $j$-cycles for specified $(i,j)$ pairs), every toroidal graph is $(2,1)$-decomposable (i.e., admits a subgraph of max degree 1 whose deletion leaves a 2-degenerate graph). A single unified discharging argument underpins the proof for a broad superclass of such graphs (Wang et al., 26 Feb 2025).
• Spectral quantities: Average effective resistance $R_{\rm avg}(T^d_n)$ on the $d$-dimensional toroidal grid is computable via Laplacian eigenvalues, yielding $R_{\rm avg}(T^d_n) = \Theta(1/d)$ for $d \geq 3$—mirroring random walk hitting time and electrical network properties (Rossi et al., 2017).

4. Symmetric Constructions: Cayley and Highly-Regular Toroidal Graphs

Special families of toroidal graphs admit highly symmetric characterizations and tight algebraic embedding:

• Honeycomb toroidal graphs are trivalent Cayley graphs on the generalized dihedral group $D_n$, yielding hexagonal-face equivelar maps on the torus. Structural invariants, such as even pancyclicity, girth, and automorphism group (up to $D_n$ or larger), are established (Alspach, 2020).
• Involutory Cayley graphs $\mathcal{G}(R)$ for finite commutative rings $R$ classify all cases when such a graph is toroidal: precisely when $R$ is one of a prescribed list of local rings of even characteristic or specified direct products of local rings. Each case is governed by adjacency via involutions and genus computations via explicit formulas (Keshavarzi et al., 2 Aug 2025).
• Generalized Heawood graphs $H(n_1,\ldots,n_{d+1})$, constructed via permutahedral tilings and quotient lattices, yield toroidal distance-regular cubic graphs dual to higher-dimensional triangulations of the torus. These admit explicit $f$-vector formulas and automorphism groups as semidirect products of translations and coordinate permutations (Ceballos et al., 2023).

5. Geometric Embedding, Graph Drawing, and Schnyder Woods

Toroidal graphs are central objects in discrete geometry and graph drawing:

• Schnyder woods: The extension of Schnyder woods to toroidal maps allows characterization of essentially 3-connected toroidal graphs via a local orientation-coloring property and global cycle-intersection restrictions. These woods facilitate periodic geodesic embeddings of the universal cover on orthogonal surfaces in $\mathbb{R}^3$ and yield straight-line representations of any essentially 3-connected toroidal map in a polynomial-sized grid. The resulting embeddings project reliably onto positive planes, preserving convexity and planarity of faces (Gonçalves et al., 2012).
• Visibility representations: Every toroidal graph admits a visibility representation on a rectangular flat torus—a result confirmed by reduction to drawing on cylinders and reinsertion of non-contractible edges via reserved columns and bipolar orientations. This construction generalizes planar visibility methods to the genus-1 setting, with direct extension to Klein bottle graphs (Biedl, 2022).
• Energy-minimizing embeddings: Weighted toroidal graphs can be embedded on optimal Euclidean tori by minimizing discrete Dirichlet energy. The unique minimizers correspond precisely to weighted Delaunay decompositions where edge weights match canonical cotangent-type quantities derived from power diagrams. The Maxwell–Cremona correspondence links harmonic map periodicity with Delaunay cellulation (Lam, 2022).

6. Topological Constraints, Knots, and Links in Toroidal Embeddings

Toroidal embeddings of abstractly planar graphs exhibit critical topological restrictions:

• Any nontrivial embedding of a planar graph on the torus necessarily includes either a knotted cycle or a nonsplit link—otherwise, the toroidal embedding is ambient-isotopic to a planar embedding. This eliminates the possibility of ravels (entangled, knot/link-free spatial graphs) on the torus, forcing entanglement-free architectures back to planar realizations (Barthel et al., 2015).

7. Open Problems and Research Directions

Key unresolved questions and further directions include:

• Algorithmic rationalization: Linear-time construction of visibility representations on rectangular tori remains open, as does efficient enumeration of semi-equivelar toroidal maps for large $n$ (Biedl, 2022, Maity et al., 2013).
• Extremal and minimal triangulations: Characterizing minimal vertex count for higher-dimensional triangulations dual to generalized Heawood graphs is conjectured but unproven for $d \geq 4$ (Ceballos et al., 2023).
• Lattice and combinatorial structure: The distributive lattice structure of Schnyder woods in the planar case seeks extension or analogs on the torus (Gonçalves et al., 2012).
• Automorphism group characterization: For generalized Heawood graphs, full classification of cases with extra symmetries beyond translations and coordinate permutations is open (Ceballos et al., 2023).
• Hamiltonicity, laceability, and diameter formulas: Honeycomb toroidal graphs provoke questions about explicit path and cycle decompositions and closed-form metric quantities (Alspach, 2020).
• Delaunay correspondence and reciprocal structures: The link between convex geometry, energy minimization, and combinatorial cellulations prompts study of higher-genus and nonorientable surfaces (Lam, 2022).
• Topological graph theory: Minimal obstructions to vertex arboricity 2 and the exact forbidden-cycle thresholds for higher-genus surface embeddings remain to be classified (Choi et al., 2013).

Toroidal graphs, with their connections to topology, combinatorics, and geometry, remain an active and multifaceted research topic with deep applications in surface graph theory, discrete optimization, spectral analysis, and mathematical chemistry.