PolyGrammar: Polymer & Combinatorics Framework

• PolyGrammar is a formal grammatical framework that applies context-sensitive grammars to model digital polymers and analyze combinatorial polynomials.
• It employs symbolic hypergraph representations and a deterministic SMILES-to-PolyGrammar translation algorithm to ensure chemical and structural validity.
• The framework supports extensive enumerative applications, facilitating rapid virtual screening and e-positivity proofs in algebraic combinatorics.

PolyGrammar refers to a genus of formal grammatical frameworks—most notably, a class of parametric, context-sensitive grammars—originally designed for the systematic digital representation and generative modeling of polymers and, separately, for the structural analysis of certain combinatorial polynomials. Principal applications have arisen in both chemical informatics, where PolyGrammar defines producible polymer topologies, and algebraic combinatorics, where it underpins $e$-positivity results for polynomial families via symbolic grammar transformation. The term thus encapsulates two primary research lines: the context-sensitive grammatical representation for digital polymers (Guo et al., 2021), and the context-free grammatical calculus for positivity and generating function analysis in combinatorics (Chen et al., 2021).

1. Formal Framework of Chemical PolyGrammar

The chemical PolyGrammar for linear polyurethane chains is structured as a tuple $G = (V, \Sigma, P, S)$, where:

• Nonterminals $V = \{X, h, s\}$ serve as placeholders for chain initiation ($X$), left end-chain growth ($h$), and right end-chain growth ($s$).
• Terminals $\Sigma = \{H(x), S(x)\ |\ x \in \mathbb{N}_0\}$ represent hard-segment and soft-segment hyperedges with integer-parameterized lengths.
• The start symbol $S$ is $X$.
• Production rules $P$ are context-sensitive and parametric; rules are triggered only in specific left/right contexts and under logical constraints on segment lengths.

The 14 principal production rules for the backbone growth of polyurethanes, parametrized by the remaining segment length $x$, are categorized as follows:

Rule #	Rewrite Context	Condition	Result (substitution)
P₁	None⟨X⟩None	–	h H(L) h
P₂	None⟨X⟩None	–	S S(L) s
P₃–P₅	None⟨h⟩H(x)	$x\geq1$, $x<1$	h H(x–1), S S(x–1), ∅
...	...	...	...

Analogous productions operate for $s$ and at the right chain-end, with "null" rules ($x < 1$) terminating the segment growth (Guo et al., 2021).

2. Symbolic Hypergraph Representation

PolyGrammmar encodes a polymer as a hypergraph $\mathcal{H} = (V, E)$ where each hyperedge corresponds to a molecular fragment (e.g., a diisocyanate or polyol). Polymeric connectivity is visualized using the line graph $L(\mathcal{H})$, mapping segment adjacencies along the chain. In symbolic PolyGrammar strings, each $H$ and $S$ symbol correlates with a vertex in $L(\mathcal{H})$, directly preserving the backbone sequence. This construction enables grammatically generated polymers to be translated unambiguously into well-defined molecular graphs (Guo et al., 2021).

3. SMILES-to-PolyGrammar Translation Algorithm

Given a polymer expressed as a SMILES string, the PolyGrammar framework provides a deterministic algorithm for recovering its grammatical derivation:

• Fragmentation splits the SMILES at each urethane linkage ("–NHCOO–") resulting in atomic fragments and bond strings.
• Substring matching against reference SMILES patterns for isocyanates (H) and polyols (S) labels each fragment.
• An adjacency matrix encodes the connectivity, from which the line graph is constructed.
• Breadth-first traversal (BFS) of the line graph produces an ordered edge list, which sequentially maps to context-sensitive PolyGrammar rules via neighbor-aware lookups.

This invertibility guarantees that all molecular structures representable in SMILES can, in principle, be explicitly mapped to PolyGrammar derivations, as evidenced by 100% coverage in a test set of >600 polyurethanes (Guo et al., 2021).

4. Generative Derivation Process and Validity

PolyGrammar derivations proceed from the start symbol by successively applying context-sensitive production rules, decrementing segment counters as the polymer backbone is extended. Growth marker nonterminals ($h$ and $s$) achieve termination via "null" rules when the necessary subchain length is consumed. Each complete derivation results in a sequence of $H(x)$ and $S(x)$ terminals which, upon removing zero-parameters, yield the symbolic backbone string.

The framework ensures validity as each rule encodes only chemically permissible local reactions. At no point can atomic valency or linkage constraints be violated; all connectivity is preserved by construction. Extension to branched, block, alternating, or homopolymers, as well as global stoichiometric constraints, is achieved by introducing bracketed parallel growth and auxiliary state-propagating nonterminals (Guo et al., 2021).

5. Combinatorial PolyGrammar and $e$-Positivity

A distinct application of "PolyGrammar" arises in the context of context-free grammars for the analysis of trivariate second-order Eulerian polynomials $C_n(x, y, z)$. The transformed grammar, via variable changes inspired by Dumont, leads to:

• Nonnegative-parameter transformation: $u = x + y + z$, $v = xy + yz + zx$, $w = xyz$.
• Production rules: $u \to 3w$, $v \to 2uw$, $w \to vw$, establishing a grammatical calculus for the computation of generating functions and $e$-positivity (Chen et al., 2021).

The grammar directly yields a combinatorial labeling of 0–1–2–3 increasing plane trees, a formal-calculus proof of the generating function, and an $e$-positive expansion for $C_n(x, y, z)$. Each monomial coefficient counts trees with prescribed numbers of leaves and vertices of each degree. Generalizations encompass polynomials defined on $k$-Stirling permutations, yielding $e$-positive expansions enumerating bounded-degree increasing plane trees (Chen et al., 2021).

6. Extensions and Computational Application

PolyGrammar's architecture accommodates several extensions:

• Branched polyurethanes use bracketed duplications and specialized rules to permit parallel growth.
• Global constraint propagation employs an auxiliary nonterminal $M(l, r, t, d)$ for tracking chain-wide state, enforcing restrictions such as chain length or segment ratios.
• Copolymers and homopolymers are supported via partitioned parameter controls and secondary context-free grammars for functional group decorations.

Enumerative capacity is substantial; for chains with $L = 10$ (chain length $21$), over $2 \times 10^6$ distinct molecular structures are generated for one-component systems. Computational efficiency is demonstrated by the generation of a length-20 chain in $\sim$4 ms, and SMILES-to-PolyGrammar translation in $\sim$11 ms on a single CPU core (Guo et al., 2021). These characteristics support rapid virtual screening, machine-learning-driven property optimization, and knowledge-based synthesis planning.

7. Significance and Prospects

PolyGrammar delivers the first explicit, invertible, and fully valid generative-model representation for large polymers, bridging formal language theory and chemistry with guarantees of chemical correctness and representational completeness (Guo et al., 2021). Its symbolic hypergraph approach and grammar-driven enumerability render it a foundational system for digital polymer informatics, extensible in principle to a broad spectrum of organic and inorganic backbone topologies. In combinatorics, the grammatical calculus line embodies a unifying formalism for positivity proofs and enumerative interpretations in polynomial families (Chen et al., 2021).

A plausible implication is that continued development of PolyGrammar-based frameworks may underpin uniform proofs of algebraic properties (symmetry, unimodality) in combinatorial theory and inspire algorithmic tools for exhaustive and explainable molecular design in chemical research.