Gorenstein Planar Ternary Graphs

• Gorenstein planar ternary graphs are defined as simple graphs with no induced cycle of length divisible by 3, where the associated independence complex is Gorenstein.
• The topic unifies ring-theoretic properties of the edge ideal with combinatorial criteria such as the Eulerian independence complex and the well-covered (W2) condition.
• Studies reveal that for planar ternary graphs, spherical independence complexes transform into vertex decomposable flag spheres with explicit Delannoy-number based enumerative invariants.

Gorenstein planar ternary graphs are studied through the Stanley–Reisner algebra of edge ideals and the topology of independence complexes. In the recent independence-complex literature, a ternary graph is a finite simple graph with no induced cycle whose length is divisible by $3$, while a graph $G$ is Gorenstein when $R/I(G)$ is Gorenstein, equivalently when its independence complex $\Ind(G)=\Delta(G)$ is Gorenstein (Eom et al., 1 Aug 2025, Trung, 2016). The subject links exact planar criteria, triangle-free and girth-restricted graph classifications, and a planar-ternary theory in which spherical independence complexes become vertex decomposable flag spheres with explicit combinatorial and enumerative structure (Hoang et al., 2015, Bayer et al., 26 Sep 2025).

1. Foundational definitions and terminological scope

For a simple graph $G$ on vertices $\{x_1,\dots,x_n\}$, the edge ideal is

$$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$

and the independence complex $\Delta(G)=\Ind(G)$ has as faces the independent sets of $G$. Its dimension is

$\dim \Delta(G)=\alpha(G)-1,$

where $G$0 is the independence number (Hoang et al., 2015). In this framework, $G$1 is Cohen–Macaulay or Gorenstein precisely when $G$2, equivalently $G$3, has the corresponding property (Trung, 2016).

Two auxiliary notions recur throughout the literature. A graph is well-covered if every maximal independent set has the same size. The class $G$4 consists of graphs $G$5 that are well-covered and satisfy

$G$6

for every vertex $G$7 (Hoang et al., 2015). For an independent set $G$8, the localization

$G$9

satisfies

$R/I(G)$0

which is the main graph-theoretic mechanism for inductive link calculations (Trung, 2016).

The modern planar-ternary literature uses ternary to mean that $R/I(G)$1 has no induced cycle whose length is $R/I(G)$2, and $R/I(G)$3-ternary to mean that $R/I(G)$4 has no induced cycles of length $R/I(G)$5 or $R/I(G)$6 (Eom et al., 1 Aug 2025). Adjacent papers note other uses of related language: one girth-based classification treats “ternary/subcubic” only as an external interpretation, and the graphic-matroid literature remarks that ternary representability is not part of its explicit Gorenstein classification (Hoang et al., 2012, Hibi et al., 2019). As a result, the phrase Gorenstein planar ternary graphs is technically meaningful only after the operative notion of ternaryness has been fixed.

2. Planar Gorenstein graphs and the Eulerian criterion

The central planar recognition theorem is that a planar graph is Gorenstein if and only if its independence complex is Eulerian (Trung, 2016). The paper proves the more general statement that if $R/I(G)$7 is a pseudo-planar graph without isolated vertices, then

$R/I(G)$8

and every planar graph is pseudo-planar (Trung, 2016). For planar graphs, Gorensteinness is therefore converted from a ring-theoretic condition on $R/I(G)$9 into a combinatorial-topological condition on $\Ind(G)=\Delta(G)$0.

The proof uses a local-to-global strategy. If $\Ind(G)=\Delta(G)$1 is Eulerian, then it is semi-Eulerian, and this implies $\Ind(G)=\Delta(G)$2 (Trung, 2016). A key identity expresses the reduced Euler characteristic of $\Ind(G)=\Delta(G)$3 in terms of a neighborhood graph: $\Ind(G)=\Delta(G)$4 where $\Ind(G)=\Delta(G)$5 (Trung, 2016). Hence

$\Ind(G)=\Delta(G)$6

for every, equivalently for some, vertex $\Ind(G)=\Delta(G)$7 (Trung, 2016). The remaining difficulty is Cohen–Macaulayness; the proof resolves it via chain-level reductions on localizations $\Ind(G)=\Delta(G)$8, Reisner’s criterion for vanishing link homology, and Stanley’s criterion that a simplicial complex is Gorenstein iff its core is Cohen–Macaulay and Eulerian (Trung, 2016).

This characterization is especially important because general Gorensteinness of graphs can depend on the characteristic of the base field, whereas the planar theorem isolates a purely combinatorial criterion in the planar setting (Hoang et al., 2015, Trung, 2016).

3. Triangle-free, girth-restricted, and recursively generated planar families

For triangle-free graphs, the Gorenstein property becomes purely combinatorial. The basic theorem states that if $\Ind(G)=\Delta(G)$9 is a simple graph without isolated vertices, then the following are equivalent: $G$0

$G$1

$G$2

(Hoang et al., 2015). In this regime, Gorensteinness is controlled by well-coveredness together with the vertex-deletion stability encoded in $G$3. The same paper gives an edge-local reformulation: for a triangle-free graph without isolated vertices,

$G$4

where

$G$5

(Hoang et al., 2015). This criterion is what allows the equivalence with Cohen–Macaulayness of $G$6.

The planar girth-$G$7 case is completely explicit. Using Pinter’s classification of connected planar $G$8-graphs of girth $G$9, one obtains a family $\{x_1,\dots,x_n\}$0 that exhausts the connected planar Gorenstein graphs of girth $\{x_1,\dots,x_n\}$1 (Hoang et al., 2015). In a parallel formulation, a connected planar graph of girth $\{x_1,\dots,x_n\}$2 is Gorenstein if and only if it lies in the recursive family $\{x_1,\dots,x_n\}$3, equivalently if and only if it is isomorphic to $\{x_1,\dots,x_n\}$4 for some $\{x_1,\dots,x_n\}$5 (Hoang et al., 2012). The recursive construction starts from $\{x_1,\dots,x_n\}$6; given adjacent degree-$\{x_1,\dots,x_n\}$7 vertices $\{x_1,\dots,x_n\}$8, if $\{x_1,\dots,x_n\}$9 is the other neighbor of $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$0, then one adds vertices $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$1 and edges

$$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$2

to obtain the next graph in the family (Hoang et al., 2012).

These graphs are structurally rigid. For all $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$3, both $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$4 and $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$5 are well-covered and vertex decomposable, with

$$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$6

hence Cohen–Macaulay (Hoang et al., 2012). Moreover,

$$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$7

for all $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$8 (Hoang et al., 2012). The triangle-free classification also implies that for connected graphs in $$I(G)=(x_ix_j\mid x_ix_j\in E(G))\subseteq R=k[x_1,\dots,x_n],$$9, if the graph is not $\Delta(G)=\Ind(G)$0 or $\Delta(G)=\Ind(G)$1, then its girth is at most $\Delta(G)=\Ind(G)$2; thus the connected Gorenstein graphs of girth at least $\Delta(G)=\Ind(G)$3 are only $\Delta(G)=\Ind(G)$4, $\Delta(G)=\Ind(G)$5, and $\Delta(G)=\Ind(G)$6 (Hoang et al., 2015).

4. Ternary graphs and spherical independence complexes

The ternary condition imposes a strong homotopical dichotomy. Kim’s theorem, as quoted in the planar-ternary literature, states that

$\Delta(G)=\Ind(G)$7

(Bayer et al., 26 Sep 2025). Thus ternaryness is detected by the topology of all induced-subgraph independence complexes, not merely by the absence of certain cycles.

A decisive strengthening is that for a ternary graph without isolated vertices,

$\Delta(G)=\Ind(G)$8

(Bayer et al., 26 Sep 2025). In particular, when the independence complex is homotopy equivalent to a sphere of its full dimension, the complex is Gorenstein; equivalently, $\Delta(G)=\Ind(G)$9 is $G$0-well-covered (Bayer et al., 26 Sep 2025). The proof isolates two mechanisms: if $G$1 is ternary and $G$2 with $G$3, then $G$4 is Cohen–Macaulay; if $G$5 is ternary and $G$6 is Cohen–Macaulay, then $G$7 is a pseudomanifold (Bayer et al., 26 Sep 2025). Combined with top-dimensional homology, this yields the Gorenstein conclusion.

For the narrower class of $G$8-ternary graphs, the topology is numerically controlled by domination parameters. If $G$9, then

$\dim \Delta(G)=\alpha(G)-1,$0

where $\dim \Delta(G)=\alpha(G)-1,$1 is the independent domination number, $\dim \Delta(G)=\alpha(G)-1,$2 the domination number, and $\dim \Delta(G)=\alpha(G)-1,$3 the domination number of the line graph (Eom et al., 1 Aug 2025). This identity is not valid for all ternary graphs: for $\dim \Delta(G)=\alpha(G)-1,$4,

$\dim \Delta(G)=\alpha(G)-1,$5

but

$\dim \Delta(G)=\alpha(G)-1,$6

(Eom et al., 1 Aug 2025). The failure on $\dim \Delta(G)=\alpha(G)-1,$7 shows that ternaryness alone does not force the sharper domination equalities.

5. Planar ternary Gorenstein graphs as polytopal flag spheres

For planar ternary graphs, the spherical case is substantially stronger than ordinary Gorensteinness. If $\dim \Delta(G)=\alpha(G)-1,$8 is a planar ternary graph and $\dim \Delta(G)=\alpha(G)-1,$9 is a homology sphere, then

$G$00

(Bayer et al., 26 Sep 2025). Since independence complexes are flag complexes, this places planar ternary Gorenstein graphs inside a tightly constrained class of flag spheres with strong shellability-type behavior.

The mechanism is classification-theoretic. The planar-ternary paper states that Trung’s classification makes connected planar Gorenstein graphs extremely restricted: essentially they are built from a family $G$01, plus one exceptional graph $G$02 (Bayer et al., 26 Sep 2025). It then shows that each $G$03 is ternary and that $G$04 is obtained from the boundary of a crosspolytope by edge subdivisions (Bayer et al., 26 Sep 2025). Edge subdivisions preserve the relevant polytopal structure, so these complexes remain boundaries of simplicial polytopes and remain vertex decomposable. For disjoint unions, one has

$G$05

so the general planar case is obtained by joins of the connected building blocks (Bayer et al., 26 Sep 2025).

As a result, for planar ternary graphs,

$G$06

(Bayer et al., 26 Sep 2025). This is a stronger conclusion than the planar Eulerian criterion alone: it does not merely certify Gorensteinness, but identifies the independence complex within a very rigid polytopal subclass.

6. Transformations, partition refinement, and Delannoy $G$07-polynomials

The mutation structure of these planar ternary flag spheres is governed by edge subdivisions and contractions. Let $G$08 denote the Lutz–Nevo graph whose vertices are flag PL spheres and whose edges correspond to edge subdivisions and contractions. The planar-ternary paper defines the partition refinement graph $G$09 as the Hasse diagram of the refinement poset on partitions of $G$10, and proves that

$G$11

for every $G$12 (Bayer et al., 26 Sep 2025). More precisely, if $G$13 is the induced subgraph of $G$14 consisting of independence complexes of planar ternary graphs with independence number $G$15, then

$G$16

(Bayer et al., 26 Sep 2025).

This correspondence is explicit. Vertices of $G$17 are indexed by disjoint unions

$G$18

so they are naturally indexed by partitions of $G$19 (Bayer et al., 26 Sep 2025). An edge in the refinement graph corresponds to an allowed edge subdivision that merges two connected components $G$20 and $G$21 into $G$22. A crucial restriction is that subdividing an edge whose endpoints lie in the same $G$23-component leaves the planar ternary class (Bayer et al., 26 Sep 2025). Thus the admissible moves are exactly the partition-merging moves.

The same rigidity appears in the $G$24-vector. For connected planar ternary $G$25,

$G$26

where $G$27 and $G$28 is a Delannoy number (Bayer et al., 26 Sep 2025). Equivalently,

$G$29

so the $G$30-polynomial is a Delannoy polynomial (Bayer et al., 26 Sep 2025). Since the zeros of these Delannoy polynomials are negative real numbers, the $G$31-polynomial is real-rooted; for disjoint unions, multiplicativity under joins gives

$G$32

(Bayer et al., 26 Sep 2025). The paper further notes that real-rootedness implies $G$33-positivity in this setting (Bayer et al., 26 Sep 2025).

7. Adjacent notions, limitations, and common sources of confusion

Several neighboring theories are relevant but non-equivalent. First, the Gorenstein property of graphs can depend on the characteristic of the base field: one example is obtained from a graph whose independence complex comes from a triangulation of $G$34, which is Gorenstein only when $G$35 (Hoang et al., 2015). The triangle-free classification is exceptional precisely because it yields a characteristic-free graph-theoretic criterion via $G$36 and Cohen–Macaulayness of $G$37 (Hoang et al., 2015).

Second, pseudo-Gorenstein$G$38 is strictly weaker than Gorensteinness. It is defined by the conditions that the top coefficient of the $G$39-polynomial is $G$40 and $G$41, equivalently

$G$42

(Hibi et al., 9 Mar 2026). This criterion supplies a concrete extremal test for planar ternary candidates, but it does not characterize Gorensteinness because Cohen–Macaulayness is not assumed (Hibi et al., 9 Mar 2026). The same paper gives exact congruence classifications for paths and cycles, including

$G$43

and

$G$44

(Hibi et al., 9 Mar 2026).

Third, the phrase Gorenstein graphic matroids refers to a different algebraic object. The toric ring $G$45 of the base polytope of a graphic matroid is Gorenstein exactly when each $G$46-connected component is built, for $G$47, from a $G$48-cycle by repeated edge-gluing and $G$49-subdivision of weight-$G$50 edges, or, for $G$51, from $G$52 by repeated collision operations (Hibi et al., 2019). These classified graphs are planar in the relevant cases, but the paper does not formulate a ternary-specific criterion, and its Gorenstein notion is toric-matroidal rather than the edge-ideal notion used for independence complexes (Hibi et al., 2019).

Taken together, the literature provides three complementary recognition principles for the edge-ideal notion of Gorensteinness: Eulerianity of $G$53 for planar graphs, membership in $G$54 together with triangle-freeness for the $G$55-criterion, and full-dimensional spherical homotopy type for ternary graphs (Trung, 2016, Hoang et al., 2015, Bayer et al., 26 Sep 2025). In the planar ternary setting, these principles converge on a remarkably rigid class of flag spheres whose graph-theoretic, topological, and enumerative invariants are all tightly constrained.