Sandwich Conjecture Overview

• Sandwich Conjecture is a concept representing problems where an object is constrained between two comparators, with applications in graph theory, geometry, relativity, and Lie algebras.
• In random graph theory, the Kim–Vu conjecture demonstrates a coupling between d-regular graphs and Erdős–Rényi graphs, facilitating the transfer of monotone properties.
• Other formulations include geometric ham-sandwich cuts, Wheeler’s thin sandwich problem in relativity, and structural conditions in Lie theory that enforce intermediate constraints.

Searching arXiv for papers on “sandwich conjecture” and adjacent uses of the term. “Sandwich Conjecture” is not a single universally fixed statement. In the cited literature, the expression names several distinct programs in which an object is constrained between two comparators, or in which a “sandwich” condition governs existence, uniqueness, or structure. The most prominent current use is the Kim–Vu conjecture on coupling random $d$-regular graphs between two Erdős–Rényi graphs (Behague et al., 23 Oct 2025). Other established uses include Wheeler’s thin sandwich conjecture in general relativity (Avalos et al., 2017), Lie-theoretic sandwich conditions on nilpotent radicals and sandwich elements (Cushman, 2017), and computational formulations of ham-sandwich-type cutting problems (Chiu et al., 2020). This suggests that the phrase functions less as a single theorem-schema than as a recurrent structural motif.

1. Terminological scope and common pattern

A basic graph-theoretic formulation of “sandwich” appears in the graph sandwich problem. For a graph property $\Pi$, the corresponding sandwich problem takes as input a pair $(G_1,G_2)$ on the same vertex set with $G_1$ a subgraph of $G_2$, and asks for a graph $G$ such that $G_1 \subseteq G \subseteq G_2$ and $G\in\Pi$, or a proof that no such graph exists (Alvarado et al., 2017). For a finite set $F$ of graphs, the $F$-FREE SANDWICH PROBLEM specializes this to forbidden induced subgraphs: one asks whether the set $\Pi$0 of all $\Pi$1-free graphs $\Pi$2 with $\Pi$3 is empty (Alvarado et al., 2017).

A geometric analogue occurs in ham-sandwich theory. The classical Ham-Sandwich theorem states that for any $\Pi$4 measurable sets in $\Pi$5, there is an oriented hyperplane that simultaneously bisects them, while the $\Pi$6-Ham-Sandwich theorem replaces bisection by prescribed fractions or counts (Chiu et al., 2020). In this setting, the “sandwich” terminology no longer means graph containment; it refers to simultaneous cutting of several sets by a single hyperplane.

In gravitation, Wheeler’s thin sandwich conjecture asks whether, given nearby slices of geometry, one can uniquely recover lapse and shift from the Einstein constraint equations (Avalos et al., 2017). In Lie theory, a sandwich algebra is a complex Lie algebra whose nilpotent radical $\Pi$7 satisfies

$\Pi$8

while a sandwich element is an element $\Pi$9 such that $(G_1,G_2)$0 and $(G_1,G_2)$1 for all $(G_1,G_2)$2 (Cushman, 2017, Mattarei, 2021). The unifying feature is the imposition of a strong intermediate constraint between ambient structures.

2. The Kim–Vu sandwich conjecture in random graph theory

In random graph theory, the sandwich conjecture usually means the Kim–Vu conjecture. If $(G_1,G_2)$3 denotes the uniform random $(G_1,G_2)$4-regular graph on $(G_1,G_2)$5 and $(G_1,G_2)$6 denotes the binomial random graph, the conjecture states that when

$(G_1,G_2)$7

there should exist

$(G_1,G_2)$8

and a coupling $(G_1,G_2)$9 such that

$G_1$0

with

$G_1$1

(Behague et al., 23 Oct 2025). The motivation is transfer of monotone properties from $G_1$2, whose edges are independent, to the random regular graph (Behague et al., 23 Oct 2025).

A major intermediate step proved the conjecture when

$G_1$3

and extended the result to sufficiently near-regular degree sequences (Gao et al., 2020). That work introduced a two-round coupling and a theorem on edge probabilities in random near-regular factors of pseudorandom graphs, with the top-side coupling driven by refined switching arguments (Gao et al., 2020).

A later paper proves the conjecture in full: for each $G_1$4 there is $G_1$5 such that for all $G_1$6 one can couple

$G_1$7

so that

$G_1$8

(Behague et al., 23 Oct 2025). The proof analyzes a natural edge-by-edge coupling process introduced earlier by Gao, Isaev, and McKay and shows that in a suitable random graph $G_1$9, all edges are contained in about the same number of $G_2$0-regular subgraphs. This “edge fairness” drives both the lower and upper sandwiches (Behague et al., 23 Oct 2025).

A related coupling framework links $G_2$1 to $G_2$2 through the loopless configuration model $G_2$3 and unions or superpositions of random perfect matchings (Gao et al., 24 Oct 2025). That work verifies the Kim–Vu conjecture for all large degrees $G_2$4, proves a weakened version for $G_2$5, and shows that unions of random perfect matchings provide a natural additive intermediate model for sandwiching arguments (Gao et al., 24 Oct 2025).

3. Graph sandwich problems and forbidden induced subgraphs

The paper “Sandwiches Missing Two Ingredients of Order Four” studies the $G_2$6-FREE SANDWICH PROBLEM when the forbidden family consists of two non-isomorphic graphs of order four (Alvarado et al., 2017). It is explicitly motivated by the graph sandwich problem for trivially perfect graphs, which are exactly the $G_2$7-free graphs (Alvarado et al., 2017).

A central structural observation is that, up to complementation, there are only $G_2$8 relevant unordered pairs of non-isomorphic graphs of order four (Alvarado et al., 2017). The paper exploits complement symmetry and basic reductions such as the following: if all forbidden graphs are connected, one may assume the solution preserves the connected-component structure of $G_2$9; if the forbidden family has no universal vertex, a universal vertex in $G$0 can be deleted without changing feasibility; and if each forbidden graph has a unique $G$1-free supergraph on the same vertex set, then the sandwich problem is easy (Alvarado et al., 2017).

The resulting classification is near-complete. The paper proves polynomial-time solvability for many pairs, including the central case $G$2, and NP-completeness for several others, including $G$3, $G$4, and $G$5 (Alvarado et al., 2017). For the positive results, it repeatedly invokes classical structure theorems: Ramsey-theoretic triviality for $G$6, the fact that a $G$7-free graph of order at least $G$8 is disconnected or co-disconnected, Olariu’s characterization of paw-free connected graphs as either triangle-free or $G$9-free, and pseudo-split and claw-related decompositions (Alvarado et al., 2017).

The hardness proofs use reductions from 3-colorability, the chain graph sandwich problem, and a customized bipartite-sandwich problem (Alvarado et al., 2017). One representative example is the reduction for $G_1 \subseteq G \subseteq G_2$0: a connected graph containing a triangle is $G_1 \subseteq G \subseteq G_2$1-free exactly when it is a complete multipartite graph with at most three partite sets, so feasibility of the corresponding sandwich instance is equivalent to 3-colorability (Alvarado et al., 2017). Another technically distinctive tractable case is $G_1 \subseteq G \subseteq G_2$2, where the proof identifies a diamond-like configuration around an edge $G_1 \subseteq G \subseteq G_2$3, partitions the common neighborhood $G_1 \subseteq G \subseteq G_2$4 into sets $G_1 \subseteq G \subseteq G_2$5 and $G_1 \subseteq G \subseteq G_2$6, and encodes the resulting constraints by 2SAT clauses (Alvarado et al., 2017).

4. Ham-sandwich generalizations, uniqueness, and complexity

The $G_1 \subseteq G \subseteq G_2$7-Ham-Sandwich theorem is a discrete biased-cut version of the classical ham-sandwich theorem. For finite, well-separated point sets $G_1 \subseteq G \subseteq G_2$8, an oriented hyperplane $G_1 \subseteq G \subseteq G_2$9 is an $G\in\Pi$0-cut if it contains one point from each color and satisfies

$G\in\Pi$1

The theorem states that if an $G\in\Pi$2-cut exists, it is unique, and if the input has weak general position, then such a cut exists for every choice of $G\in\Pi$3 (Chiu et al., 2020).

The associated search problem, Alpha-HS, takes as input $G\in\Pi$4 finite point sets $G\in\Pi$5 and a vector $G\in\Pi$6, and asks for either a valid cut or a violation certificate for weak general position or well-separatedness (Chiu et al., 2020). A main complexity result places this problem in

$G\in\Pi$7

via a promise-preserving reduction to UniqueEOPL (Chiu et al., 2020). The paper contrasts this with the ordinary discrete Ham-Sandwich problem, which is PPA-complete, and explicitly notes that well-separation significantly lowers the complexity of the generalized sandwich problem (Chiu et al., 2020).

A later paper gives two new proofs of the $G\in\Pi$8-Ham-Sandwich theorem and pushes the theory beyond Euclidean hyperplane arrangements (Borzechowski et al., 11 Feb 2026). The first proof is completely combinatorial and constructs a Unique Sink Orientation on the grid

$G\in\Pi$9

where the outmap bijection for grid USOs yields existence and uniqueness of every $F$0-cut (Borzechowski et al., 11 Feb 2026). The second proof uses point-hyperplane duality and the Poincaré–Miranda theorem, and generalizes the statement to colored generalized arrangements and further to oriented matroids via rainbow arrangements (Borzechowski et al., 11 Feb 2026). The same paper proves that the realizability problem for rainbow arrangements is $F$1-complete and therefore also implies that realizability of grid Unique Sink Orientations is $F$2-complete (Borzechowski et al., 11 Feb 2026).

5. Thin sandwich and sandwich-wave problems in gravitation

Wheeler’s thin sandwich conjecture concerns the recovery of lapse and shift from nearby spatial geometries. In the formulation studied in (Avalos et al., 2017), the free initial data are

$F$3

where $F$4 is a Riemannian metric on an $F$5-dimensional manifold $F$6, $F$7 is a symmetric $F$8-tensor, $F$9 is the energy density, and $F$0 is the momentum density (Avalos et al., 2017). Writing

$F$1

one obtains

$F$2

and, from the Hamiltonian constraint under the assumption $F$3,

$F$4

Substituting this into the momentum constraint yields the reduced thin sandwich equation

$F$5

for the shift $F$6 (Avalos et al., 2017).

The paper proves a local well-posedness theorem in arbitrary dimension $F$7: if $F$8 is definite on $F$9, $\Pi$00 on $\Pi$01, and the equation $\Pi$02 has only the trivial solution $\Pi$03, then there exist neighborhoods in the relevant Sobolev spaces and a unique smooth map $\Pi$04 such that $\Pi$05 (Avalos et al., 2017). It further proves that every smooth compact $\Pi$06-dimensional manifold with $\Pi$07 admits smooth data with $\Pi$08 such that, in a neighborhood of those data, the thin sandwich problem is well-posed (Avalos et al., 2017). The result is explicitly local rather than global.

A different gravitational use of “sandwich” concerns sandwich-wave spacetimes. In a pp-wave background with a sandwich profile,

$\Pi$09

with a finite pulse region $\Pi$10, one paper studies scattering of test gravitational waves by solving the linearized Einstein equations on the background (Tang, 2023). The outgoing perturbation is reconstructed from a Debye potential, and the paper defines amplification by

$\Pi$11

It concludes that in some cases the energy of the outgoing test gravitational wave is amplified as well, with the strongest effect occurring near caustic or focusing hypersurfaces where the amplification can diverge formally (Tang, 2023). This is a different “sandwich” program from the thin sandwich conjecture, but it retains the same geometric language of a finite intermediate region.

6. Lie-theoretic sandwich algebras and sandwich elements

In Lie theory, the sandwich condition is structural rather than probabilistic. A complex Lie algebra is called a sandwich algebra if it has a nilpotent radical $\Pi$12 which is a sandwich,

$\Pi$13

the quotient by $\Pi$14 is semisimple, and the adjoint actions of a Cartan subalgebra form a maximal family of commuting semisimple endomorphisms of $\Pi$15 (Cushman, 2017). A very special sandwich algebra is a special sandwich algebra $\Pi$16 that is a subalgebra of a semisimple Lie algebra of rank one higher than the rank of $\Pi$17, with both $\Pi$18 and the ambient semisimple algebra simple (Cushman, 2017).

The paper “Very special sandwich algebras” classifies all such algebras (Cushman, 2017). It shows that every sandwich algebra decomposes as a direct sum of simple sandwich algebras, analyzes the nilpotent radical via a decomposition

$\Pi$19

and proves that nonabelian sandwich radicals contain Heisenberg subalgebras determined by symplectic root-space pieces (Cushman, 2017). In the “very special” setting, the nilradical is obtained from roots

$\Pi$20

and the paper gives a root-theoretic criterion ensuring that

$\Pi$21

It also exhibits an $\Pi$22 counterexample showing that Dynkin-diagram deletion alone does not force the sandwich condition (Cushman, 2017).

A related use of sandwich language appears in thin Lie algebras. If

$\Pi$23

is thin, then $\Pi$24, $\Pi$25 generates $\Pi$26, and the covering property

$\Pi$27

holds (Mattarei, 2021). The two-dimensional homogeneous components are called diamonds. The main theorem states that if the second diamond is $\Pi$28 with $\Pi$29, then there exists $\Pi$30 such that

$\Pi$31

(Mattarei, 2021). The paper defines a sandwich element $\Pi$32 by the conditions

$\Pi$33

and notes that in odd characteristic the second identity follows automatically from $\Pi$34 (Mattarei, 2021). The result places thin Lie algebras with late second diamonds in direct contact with the broader sandwich-element tradition associated with non-classical modular Lie theory (Mattarei, 2021).

7. Related sandwich theorems and adjacent terminology

The wider “sandwich” vocabulary in the cited literature also includes several theorems rather than conjectures. In convexity theory, a central sandwich theorem states that for functions $\Pi$35 on an interval $\Pi$36, the existence of an affine $\Pi$37 with

$\Pi$38

is equivalent to Jensen-type cross inequalities involving $\Pi$39 and $\Pi$40; an analogous result holds for set-valued functions $\Pi$41 with an affine set-valued $\Pi$42 satisfying

$\Pi$43

(Mitroi-Symeonidis, 2015). This line of work presents sandwich conditions as convexity tests and interpolation principles.

In algebraic geometry, “sandwich theorems” for Shioda–Inose structures describe K3 surfaces $\Pi$44 lying in a two-way Kummer sandwich

$\Pi$45

with both arrows degree-$\Pi$46 rational maps (Schuett, 2011). One paper gives a geometric construction of three infinite series of such K3 surfaces, while recalling Ma’s theorem that any Shioda–Inose structure admits a sandwich (Schuett, 2011).

In topological combinatorics and fair division, “Thieves can make sandwiches” proves a common generalization of the Ham Sandwich theorem and Necklace Splitting (Blagojević et al., 2017). Its main results show the existence of fair distributions of $\Pi$47 measures in $\Pi$48 among $\Pi$49 thieves using roughly $\Pi$50 convex pieces, with proofs based on a geometric realization of a topological join of partition spaces and computation of a Fadell–Husseini index (Blagojević et al., 2017).

Taken together, these usages show that “sandwich” terminology consistently signals an intermediate object, an interpolation, or a two-sided constraint. The exact content, however, varies sharply by field: monotone couplings in probabilistic combinatorics, feasible completions in algorithmic graph theory, unique biased cuts in discrete geometry, well-posedness in general relativity, and nilpotent or adjoint constraints in Lie theory.