A quadratic form generalization of rational dinv
Abstract: We introduce a quadratic form on the space of functions on the gap poset of the numerical semigroup . We prove combinatorially that when evaluated on the indicator function of an upward closed subset , this quadratic form precisely recovers the Gorsky--Mazin statistic of , viewed as a Young subdiagram of . Furthermore, we prove Theorem~1.2 that when evaluated on a pair of subdiagrams of , the symmetric bilinear form associated with is equal to a novel cross- statistic, which is nonnegative. Combining these, we prove the inequality [ Q(\mathbf{n})\geq \dfrac{1}{|G|}\,|\mathbf{n}|_\infty2] if is a real-valued decreasing function on , showing an effective positive definiteness of on the corresponding cone. Theorem~1.2, the main engine of the paper, was autoformalized in Lean/Mathlib by AxiomProver.
Paper Prompts
Sign up for free to create and run prompts on this paper.
Top Community Prompts
Explain it Like I'm 14
Overview
This paper is about finding a clean, number-based way to measure how “balanced” certain paths in a grid are. These paths are called rational Dyck paths, and they live inside a rectangle of size (with and being positive whole numbers that share no common factors). There is a famous statistic on these paths called dinv (short for “diagonal inversions”). The authors build a new formula—called a quadratic form —that, when you plug in a path, exactly reproduces its dinv. They then extend this idea to pairs of paths and prove that is “positive” in a strong, useful way.
What the paper is trying to figure out
In simple terms, the paper asks:
- Can we write the dinv statistic as a simple, general-purpose formula that looks like “sum of (weights × values × values)”?
- Can this be extended to pairs of paths to define a fair, symmetric “cross-dinv” that measures how two paths interact?
- Is always nonnegative (and even bounded below in a useful way) when applied to nicely ordered data? If so, can that help prove that some big infinite sum they care about actually converges?
How the authors approach the problem
Think of an rectangle filled with small boxes. Focus on the boxes strictly below the diagonal from the top-left to the bottom-right. Those boxes form a “Young diagram,” and a rational Dyck path is like a staircase path staying above (or hugging) that diagram. The paper uses three ideas:
- The grid of boxes as numbers
- Each box gets a number using (you don’t need the details; just think: each box has a label).
- The collection of labels that don’t come from (with nonnegative ) is called the “gap set.” These “gaps” correspond exactly to the boxes below the diagonal.
- Dinv as “hook tilt”
- For each box in a path’s diagram, you draw a small “hook”: how far you can go left (the “arm”) and down (the “leg”) before leaving the shape.
- The dinv of a path counts how many boxes have a hook whose “tilt” (a fraction comparing leg and arm) straddles the slope . If the tilt is too flat or too steep, that box doesn’t count.
- A new quadratic form
- Instead of counting hook tilts directly (which is fiddly), the authors define a formula that sums up pairwise interactions between boxes. It adds or subtracts based on simple rules about how far apart the box labels are.
- When the input is the “indicator” of a path (a 1 for boxes in the path, and 0 otherwise), turns out to equal the path’s dinv exactly.
- They also define a related “bilinear form” so that . This lets them talk about interactions between two different paths (cross-dinv).
A friendly analogy: if a path is a shape built from boxes, is like an “energy” that sums the pushes and pulls between boxes according to easy rules. Amazingly, that energy equals dinv.
Main results and why they matter
Here are the core findings, explained simply:
- Recovering dinv from
- If you plug in the 0/1 vector of a path into , you get exactly . So is a clean, formula-based version of dinv. This gives a new, algebraic way to understand and compute dinv.
- A new cross-dinv for pairs of paths
- The authors define a symmetric “cross-dinv,” , that measures how two paths interact. They prove that the bilinear form satisfies . This extends dinv from one path to two paths in a natural, balanced way.
- Positivity and a concrete lower bound
- On a natural cone of “decreasing” or “nested” data (think: numbers that don’t increase as you move in certain directions in the grid), they show is always nonnegative.
- Even better, they prove a strong bound:
- In words: is at least a constant times the square of the largest entry of , where is the number of boxes under the diagonal.
- This is like saying: the “energy” is not just nonnegative—it has a firm floor depending on how big your entries get.
- Convergence of an important infinite sum
- In related work, the authors need to sum an infinite series built from . Because is positive and has the lower bound above, that huge series behaves nicely and converges. This is crucial for making sense of a new generating function that ties into deep identities similar to the Rogers–Ramanujan identities.
- Computer verification
- The key theorem connecting and cross-dinv was fully formalized and checked by a theorem-proving system (Lean/Mathlib), via an AI tool called AxiomProver. That adds confidence that the combinatorial heart of the paper is correct.
Why is this important? Dinv, rational Dyck paths, and -Catalan numbers show up in surprising places—like knot theory and geometry (for example, spaces connected to the singular curve ). Turning dinv into a clean algebraic object helps unify ideas, extend them to higher “ranks” (multiple nested paths), and supports new identities that relate complicated sums to elegant infinite products.
What this could lead to
- A “high-rank” dinv: For several nested paths (think: stacking paths inside each other), naturally defines a higher-level dinv that sums all pairwise cross-dinvs. This opens doors to studying families of paths at once.
- New sum=product identities: With ensuring convergence, the authors conjecture and start proving fresh identities similar to the classic Rogers–Ramanujan identities—beautiful equalities connecting infinite sums and infinite products.
- Bridges across fields: These results tie together combinatorics (counting paths), algebra (quadratic forms), geometry (spaces of points on curves), and knot theory. A more algebraic dinv may make those bridges stronger and easier to cross.
Key takeaways
- The paper builds a quadratic form that exactly matches the classic dinv statistic for rational Dyck paths.
- It defines and connects a new cross-dinv for pairs of paths with a bilinear form .
- It proves strong positivity for , including an explicit inequality, which ensures useful infinite sums converge.
- It lays groundwork for high-rank versions (multiple nested paths) and new deep identities.
- Parts of the proof were verified by a computer proof assistant, boosting reliability.
Knowledge Gaps
Knowledge gaps, limitations, and open questions
Below is a consolidated list of unresolved issues and concrete directions suggested (explicitly or implicitly) by the paper.
- Generalization beyond two generators: Does an analogue of the quadratic form Q and the cone C_R exist for gap posets of numerical semigroups with ≥3 generators? Are there corresponding “rational dinv” and cross-dinv statistics and positivity results in that setting?
- Non-coprime parameters: The paper assumes gcd(a,b)=1. What breaks (e.g., in the gap poset structure) and which results still hold if a and b are not coprime? Is there a meaningful reformulation of Q and cross-dinv in that case?
- Integrality of cross-dinv: Since B(D,E)=(B′(D,E)+B′(E,D))/2 is an average of two integers, is dinv(D,E) always an integer? If yes, can one prove a structural parity result; if not, can one characterize when half-integers occur?
- Optimal lower bound constant: The paper proves Q(n) ≥ (1/|G|) * ||n||∞2 on C_R. What is the best constant c(a,b) such that Q(n) ≥ c(a,b) * ||n||∞2 holds for all n∈C_R? Characterize extremizers achieving equality.
- Orthogonality characterization: The bilinear form B can vanish for nonzero inputs (Remark 1.5). Precisely characterize pairs (D,E) (or general n,n′∈C_R) with B(n,n′)=0. Is there a combinatorial notion of “orthogonality” for diagrams corresponding to B=0?
- Positive combinatorial formula for general integer-valued n: The paper gives Q(1_D)=dinv(D) and Q(∑ 1_{D_i}) via cross-dinv. Is there a direct, positive, cell-wise combinatorial expression for Q(n) when n∈ℤG∩C_R with arbitrary multiplicities, not presented as a sum over nested indicators?
- Comparison to existing high-rank/nested dinv: The paper defines a “high-rank dinv” via Q for nested paths and notes it appears different from Loehr–Warrington (2008). Determine precise relationships (e.g., bijections, equidistribution, or transforms) between this new statistic and prior notions of nested dinv.
- q,t-structures for higher rank: For nested Dyck paths, is there an “area-like” companion statistic and a corresponding q,t-symmetry (analogous to rational q,t-Catalan) involving Q? Do shuffle- or nabla-type phenomena extend to the high-rank setting?
- Algorithmics for Z_{a,b}(z): The truncation bound ||n||∞ ≤ √(N|G|) ensures finiteness, but what are practical algorithms and complexity bounds to compute Z_{a,b}(z) to large orders N and for large (a,b)? Can dynamic programming or structural decompositions accelerate enumeration?
- Infinite product conjecture (HJO26+): Fully proving the conjectural infinite product for Z_{a,b}(z) for all coprime (a,b), identifying the analytic/modular structure behind the Rogers–Ramanujan-type identities, and clarifying for which (a,b) product formulas hold.
- Spectral analysis of Q outside the cone: Q is positive definite on C_R but indefinite on the full space ℝG. Determine the spectrum/signature of the associated kernel operator on ℝG, and characterize the maximal cones on which Q is positive (semi-)definite.
- Upper bounds and norm comparisons: Establish sharp upper bounds on Q(n) in terms of norms (e.g., ||n||1, ||n||_2, or mixed norms), with constants depending on (a,b), to complement the proven lower bound and refine truncation/estimation in Z{a,b}(z).
- Asymptotics and distributions: For random rational Dyck paths (or nested systems), study the typical size and limiting distribution of Q (i.e., dinv and cross-dinv), possible limit shapes, and variance asymptotics as a,b grow.
- Geometric meaning of cross-dinv: In the Quot-scheme/curve singularity context motivating Q, what is the geometric interpretation of cross-dinv? Does it correspond to intersection numbers, weight filtrations, or Ext-graded dimensions?
- Robustness to diagram classes: The results are stated for upward-closed subsets of the gap poset/Young subdiagrams in the ab-rectangle below the diagonal. Do the definitions of Q and cross-dinv extend (with positivity) to other posets or shapes (e.g., skew shapes, other rational shapes)?
- Alternative kernels and deformations: The kernel K(d) has a specific inclusion–exclusion form. Are there deformations K_t(d) (e.g., introducing parameters) that interpolate between area/dinv-like forms? Which deformations preserve positivity on C_R?
- Symmetry in (a,b): Classical Catalan statistics exhibit q,t-symmetry. Beyond Q(1_D)=dinv(D), are there symmetries or dualities in (a,b) reflected in B(D,E) or in Z_{a,b}(z)? Does cross-dinv enjoy a natural involutive symmetry exchanging a and b?
- Tight combinatorial descriptions without projections: The proofs use projections/antiprojections to rows; can cross-dinv be characterized purely in terms of local hooks/arm–leg data without boundary reference sets U_D, yielding more direct counting formulas?
- Formalization scope and methods: Only Theorem 1.2 was fully autoformalized in Lean/Mathlib. Can the entire argument (including Theorem 1.1 and Theorem 1.3) be formalized? What abstractions are needed in Mathlib to support such combinatorial–algebraic proofs more generally?
- Extending convergence results: The paper establishes formal convergence of Z_{a,b}(z). What can be said about analytic continuation, modularity, or asymptotics of coefficients as functions of (a,b)? Under what conditions does Z_{a,b}(z) correspond to known q-series with established analytic properties?
Practical Applications
Immediate Applications
Below are concrete ways the paper’s findings can be put to use now, with sectors, potential tools/products/workflows, and key assumptions/dependencies.
- Fast, reliable computation of rational dinv and cross-dinv via a quadratic form
- Sectors: academia (algebraic combinatorics; algebraic geometry), software (computer algebra systems)
- What to deploy: Implement the quadratic form Q and its associated symmetric bilinear form B to compute Gorsky–Mazin dinv(D) and the new cross-dinv(D,E) for subdiagrams D,E⊆G. Use the nested-path decomposition (Lemma 5.1) to evaluate high-rank dinv for n-tuples of nested Dyck paths by Q(1_{D1}+…+1_{Dn}). Package as a SageMath/Julia/Python module with Lean-backed unit tests.
- Assumptions/dependencies: Requires coprime integers a<b and Dyck paths as upward-closed subsets in the gap poset G; correctness depends on the theorems proved in the paper; performance depends on enumerating cells in G (size |G|=(a−1)(b−1)/2).
- Efficient coefficient extraction of Z_{a,b}(q) using truncation bounds
- Sectors: academia (enumerative geometry, q-series), software (HPC enumeration)
- What to deploy: Use the proven inequality Q(n) ≥ (1/|G|)‖n‖∞2 to truncate the infinite sum for Z{a,b}(q) to vectors n with ‖n‖∞ ≤ √(N|G|) to compute all coefficients up to qN. Implement enumeration over the cone Cℝ (monotone functions on G) and accumulate the contributions z{Q(n)}(1+O_n(z)).
- Assumptions/dependencies: Definitions of Z_{a,b} and O_n(·) follow forthcoming work (HJO26+); correctness here assumes the positivity results in the paper and trusted implementation of the cone constraints; computational feasibility depends on |G| and N.
- Formal verification workflow for combinatorial statements using AxiomProver/Lean
- Sectors: academia (research reproducibility), software (formal methods), policy (publishing/peer review)
- What to deploy: Adopt the demonstrated pipeline—natural-language theorem → AxiomProver → Lean (4.28.0) proof scripts—to verify core combinatorial lemmas and identities (e.g., Theorem 1.2). Integrate AXLE to verify the proofs and enforce definition integrity.
- Assumptions/dependencies: Requires availability of Lean 4.28.0/Mathlib and AxiomProver; scope currently demonstrated on the main bilinear theorem; generalization to other areas requires further engineering and curation.
- Educational visualization and curriculum modules on Dyck paths and posets
- Sectors: education (undergraduate/graduate mathematics), outreach
- What to deploy: Interactive notebooks/web apps that illustrate the bijections and “blue/red arrow” decompositions underlying dinv and cross-dinv; exercises where students compute Q(1_D) and see it match dinv(D) on examples.
- Assumptions/dependencies: Requires only basic combinatorics tooling (e.g., Jupyter, Sage, Mathematica, or JavaScript/d3); depends on clear rendering of the gap diagram G and the order relation.
- Support for small-parameter torus-knot invariant computations
- Sectors: mathematical physics (knot homology), academia
- What to deploy: Use fast Q-based dinv evaluation to accelerate the combinatorial parts of computations related to HOMFLY-PT homology of (a,b)-torus knots (where dyck-path statistics enter), enabling larger-a,b tabulations than previously practical.
- Assumptions/dependencies: Effectiveness strongest for torus knots and rational Catalan contexts; requires integration with existing homology computation pipelines; limited to modest sizes by combinatorial explosion in |G|.
Long-Term Applications
These directions require further research, scaling, or development before broad deployment.
- High-rank dinv for nested Dyck paths and symmetric-function computations
- Sectors: academia (algebraic combinatorics, representation theory)
- What could emerge: A general high-rank dinv statistic for chains of Dyck paths via Q(n_D)=∑_{i,j} dinv(D_i,D_j); potential new formulas for rational (q,t)-Catalan objects and links to Macdonald polynomial phenomena; software for nested path enumeration and statistics.
- Assumptions/dependencies: Requires theoretical comparisons to existing nested/high-rank dinv notions (e.g., Loehr–Warrington) and proofs of desired properties (e.g., symmetry, positivity, equidistributions).
- Proof and exploitation of the conjectured product formulas for Z_{a,b}(z)
- Sectors: academia (number theory, combinatorics), physics (statistical mechanics, CFT)
- What could emerge: New bi-infinite Rogers–Ramanujan-type identities; analytic tools for partition-theoretic and q-series problems; better asymptotics and modular-type phenomena; improved algorithms for evaluating Z_{a,b} via infinite-product structures.
- Assumptions/dependencies: Requires full proof of the product conjectures beyond initial cases; analytic control over convergence/continuation; validated computational pipelines for large parameters.
- Poset-based convex regularizers for machine learning and optimization
- Sectors: software (optimization libraries), AI/ML (structured prediction)
- What could emerge: Use Q as a positive semidefinite energy on the monotone cone C_ℝ to regularize learning over order-constrained features (e.g., monotone risk scores or demand curves on partial orders); incorporate into convex solvers as a structure-aware penalty.
- Assumptions/dependencies: Generalization of the quadratic-form/positivity framework from the gap poset G to task-specific posets/DAGs; efficient evaluation/gradients; empirical validation against standard regularizers.
- Scalable, end-to-end formal verification in mathematical publishing and education
- Sectors: software, academia, policy
- What could emerge: Journal submission workflows that accept natural-language manuscripts and require formally-verified cores for main theorems; classroom tools that auto-generate and check formal versions of homework and exam problems.
- Assumptions/dependencies: Advances in autoformalization coverage, usability, and robustness; community standards for what must be verified; sustained maintenance of theorem-proving ecosystems (Lean/Mathlib or equivalents).
- Generalized quadratic forms on other numerical semigroups/posets for counting and structure discovery
- Sectors: academia (combinatorics, algebra), software (computational algebra)
- What could emerge: Families of kernel-defined quadratic forms with cross-statistics on broader semigroups and posets yielding positive definiteness on natural cones; new counting formulas for modules over singular rings and related moduli; reusable libraries for poset-based enumeration.
- Assumptions/dependencies: Theoretical development of kernels K and cones with provable positivity; identification of geometric/combinatorial settings where such forms control convergence or encode statistics; scalable implementations.
Notes on feasibility across all items:
- Mathematical assumptions: coprime a<b; functions n lie in the monotonicity cone C_ℝ; results are currently established for the gap poset G of ⟨a,b⟩.
- Software dependencies: Lean 4.28.0, Mathlib, AxiomProver/AXLE (as used in the appendix), and standard CAS or numerical libraries for implementations.
- Research dependencies: For Z_{a,b}(·) and related products, several claims rely on forthcoming or conjectural results (HJO26+ and product conjectures); long-term items presuppose those developments.
Glossary
- affine cell decomposition: A partition of a space into pieces isomorphic to affine spaces, often used to analyze geometry or cohomology. "the construction of an affine cell decomposition of the moduli space motivates a novel quadratic form"
- antiprojection: In this paper’s combinatorial grid, the unique “backward” mapping of a cell to a specified row that keeps the value below or equal to the source’s value. "called the projection and the antiprojection of onto the -th row."
- AxiomProver: An AI system used to produce formal Lean proofs of mathematical statements. "was autoformalized in Lean/Mathlib by AxiomProver."
- AXLE (Axiom Lean Engine): A tool for verifying Lean proof files developed by Axiom Math. "using AXLE (Axiom Lean Engine), a tool developed by Axiom Math"
- bilinear form: A function of two vector arguments that is linear in each argument separately. "A bilinear form "
- cohomology: A suite of algebraic invariants of topological or algebraic spaces capturing global structure. "to the cohomology of the compactified Jacobian"
- compactified Jacobian: A moduli space extending the Jacobian of a curve to include rank-1 torsion-free sheaves over singular curves. "the cohomology of the compactified Jacobian (or Hilbert scheme of points)"
- cone: Here, a convex subset of a vector space closed under nonnegative scaling and addition, denoted . "It is conjectured that despite the negative signs, is positive definite on the cone"
- cross-dinv: A bilinear extension of the dinv statistic to pairs of Dyck paths, defined via mixed hook slopes. "a novel cross- statistic"
- dinv: A statistic on (rational) Dyck paths counting certain cells determined by arm/leg inequalities. "the statistic"
- Dyck path: A lattice path (here, rational) staying under a diagonal in a rectangle, encoded as an upward-closed Young subdiagram. "Gorsky and Mazin defined the statistic for a Dyck path ."
- gap poset: The partially ordered set of gaps (non-semigroup elements) of a numerical semigroup, ordered by addition in the semigroup. "functions on the gap poset of the numerical semigroup ."
- groupoid volume: A weighted count of objects in a groupoid, summing inverses of automorphism group sizes. "(a.k.a. the groupoid volume)"
- Hilbert scheme of points: A moduli space parameterizing 0-dimensional subschemes (points with multiplicity) on a variety. "the cohomology of the compactified Jacobian (or Hilbert scheme of points)"
- HOMFLY-PT homology: A link homology theory categorifying the HOMFLY-PT polynomial, relating knot invariants to geometry. "to the HOMFLY-PT homology of the -torus knot"
- hook slope: Ratios derived from arm and leg lengths (small/large), used to define dinv conditions. "the small and large mixed cross hook slopes"
- indicator function: The function taking value 1 on a set and 0 elsewhere; here on subdiagrams. "the indicator function of an upward closed subset "
- indicator vector: The vector in with 1s on a subdiagram and 0s elsewhere. "let be the indicator vector of ."
- Lean/Mathlib: The Lean theorem prover and its community mathematical library used for formal verification. "was autoformalized in Lean/Mathlib by AxiomProver."
- leg length: For a cell in a Young diagram, the number of cells below it in its column within the subdiagram. "the arm and leg lengths of "
- numerical semigroup: A cofinite additive submonoid of the nonnegative integers; here generated by and . "the numerical semigroup "
- poset: A partially ordered set; here is ordered by adding elements of the numerical semigroup. "We treat as a poset with "
- positive definite: A quadratic form satisfying for all nonzero in the domain considered. "positive definite on the cone"
- positive semi-definite: A bilinear/quadratic form taking nonnegative values on all inputs. "is already positive semi-definite on $C_{\mathbb{R}$."
- projection: In this grid model, the unique “forward” mapping of a cell to a specified row preserving order via -values. "the projection and the antiprojection of onto the -th row."
- q,t-Catalan numbers: A bivariate -analogue of Catalan numbers counting Dyck paths by area and dinv. "The classical -Catalan numbers "
- quadratic form: A homogeneous polynomial of degree 2 on a vector space; here . "We introduce a quadratic form "
- Quot scheme: A moduli space parameterizing quotient sheaves of a fixed sheaf on a variety. "the geometry of the Quot scheme of points on the curve "
- rational (a,b)-Catalan numbers: Generalizations of -Catalan numbers indexed by coprime , enumerating rational Dyck paths. "the rational -Catalan numbers "
- rational Dyck paths: Lattice paths staying under the diagonal in an rectangle for coprime . "corresponding to rational Dyck paths in an rectangle."
- Rogers--Ramanujan type: Refers to identities resembling the classical Rogers–Ramanujan sum–product identities. "sumproduct identities of Rogers--Ramanujan type."
- torus knot: A knot lying on the surface of a torus, specified here by the pair . "the -torus knot"
- upward closed subset: A subset of a poset closed under moving to greater elements; here equivalent to Young subdiagrams. "it is an upward closed subset of "
- Young diagram: A left-justified arrangement of boxes representing a partition; here realized below a diagonal in a rectangle. "realizing as a Young diagram"
- Young subdiagram: A subset of a Young diagram that is itself a Young diagram (upward-closed). "viewed as a Young subdiagram of "