Narayana Polynomials¶

Package the Narayana distribution of a Catalan family into a polynomial whose coefficient of each power counts objects with a specified statistic, retaining the Catalan total at one while exposing symmetry, unimodality, real-rootedness, and specializations.

Version: v2 · 2026-09-06 · History
Domain-specific #: 2346
Origin domain: mathematics
Subdomain: enumerative combinatorics
Aliases: Narayana polynomial, Type-A Narayana polynomial

Core Idea¶

The Narayana polynomials are finite generating polynomials whose coefficients are the type-A Narayana numbers. For a positive integer n and 1 <= k <= n, the Narayana number is N(n,k) = (1/n) binom(n,k) binom(n,k-1). One common convention is N_n(t) = sum_{k=1}^n N(n,k)t^k, with N_0(t)=1; another shifts the exponent and uses t^(k-1). The two conventions carry the same coefficient triangle but differ by a factor of t, so the exponent convention must be declared before identities, reciprocal polynomials, or constant terms are compared.

The polynomial is a statistic-refined enumerator. In the Dyck-path interpretation, N(n,k) counts Dyck paths of semilength n having exactly k peaks; the same numbers refine other Catalan families, including noncrossing partitions by number of blocks and suitable pattern-avoiding permutations by descents.^[1] The coefficient index therefore records a meaningful combinatorial statistic, not merely a sequence position. Summing the coefficients, equivalently evaluating the unshifted polynomial at t=1, returns the Catalan number. The polynomial preserves the distribution that the Catalan total alone erases.

The row symmetry N(n,k)=N(n,n+1-k) makes the coefficient sequence palindromic after accounting for the selected exponent convention. Positive coefficients, real nonpositive zeros, log-concavity, and unimodality tie the enumerative interpretation to polynomial geometry. Wang and Yeh place Narayana polynomials among classical applications of interlacing methods for real-rooted polynomial sequences.^[2] Agapito treats them as distinguished symmetric real-rooted examples with nonnegative gamma-vectors.^[3] Those results are properties of the established polynomial family, not part of its bare definition; a draft must not define Narayana polynomial as any palindromic or real-rooted Catalan refinement.

Specializations must be convention-aware. Under the unshifted convention used here, N_n(1) is the nth Catalan number. Evaluation at other values weights every Catalan object by its statistic rather than merely counting it; for example, t=2 produces the large Schröder-number specialization under the stated indexing convention. The ordinary bivariate generating function packages all n and the statistic variable together and satisfies an algebraic square-root form. Recurrences, reciprocal forms, Legendre-polynomial relations, and generalized or type-B Narayana polynomials are downstream structures, not interchangeable names.

The candidate survives collision review because Narayana Number is a coefficient family, Catalan Number is the unrefined total, and a generic generating polynomial does not select this triangle or statistic. The abstraction has a stable map: Catalan object plus marked statistic becomes a monomial; aggregation over all objects becomes a polynomial; coefficient extraction recovers refined counts; specialization collapses the refinement. That map supports recognition across Dyck paths, noncrossing partitions, and pattern-avoiding permutations while preserving the type-A identity.

Structural Signature¶

The size index. A nonnegative integer n fixes one Catalan family row.
The refinement index. An integer k records peaks, blocks, descents, or another equidistributed Narayana statistic.
The Narayana coefficient. N(n,k)=(1/n)binom(n,k)binom(n,k-1) for 1 <= k <= n.
The declared exponent convention. Powers are either t^k or t^(k-1) and may not be silently mixed.
The finite coefficient sum. The row is packaged as a polynomial rather than as an unstructured list.
The statistic interpretation. Coefficient extraction counts objects of size n with exactly the selected statistic value.
The Catalan collapse. Evaluation at t=1 forgets the statistic and returns the full Catalan count.
The reciprocal symmetry. Complementary statistic values have equal coefficients under the type-A row symmetry.
The polynomial properties. Real-rootedness, log-concavity, unimodality, and gamma-positivity constrain the coefficient distribution.
The family-level relations. Recurrences and generating functions connect adjacent size indices without redefining the coefficients.

What It Is Not¶

Not a Narayana number. A number is one coefficient; the polynomial packages an entire refined row.
Not the Catalan generating function. Catalan totals omit the within-row statistic distribution.
Not every Catalan refinement. A different statistic need not have the Narayana distribution.
Not a q-Narayana polynomial by default. Additional deformation parameters define distinct generalized families.
Not a type-B Narayana polynomial. Coxeter-type variants use different coefficients and combinatorial objects.
Not defined by real-rootedness. Many real-rooted polynomials are unrelated; the coefficient triangle is constitutive.
Not convention-free. Shifting exponents changes constant terms, reciprocal formulas, and displayed degree.

Scope of Application¶

Narayana polynomials are literal when a type-A Narayana row is encoded as a statistic-generating polynomial with an explicit size, statistic, and exponent convention.

Dyck paths. Counting semilength-n paths by number of peaks.
Noncrossing partitions. Counting partitions by number of blocks.
Pattern avoidance. Recording descent or related distributions on Catalan permutation classes.
Plane trees. Transporting the coefficient distribution through a stated Catalan bijection.
Polynomial analysis. Studying zeros, interlacing, log-concavity, and gamma expansions.
Generating functions. Combining the size and statistic variables in algebraic series.
Specialization identities. Recovering Catalan, Schröder, or weighted totals under verified conventions.
Bijection design. Proving equidistribution by mapping one statistic-refined Catalan family to another.

Clarity¶

State whether n begins at zero, the allowed range of k, whether N(0,0) is separately defined, and whether the exponent is k or k-1. Give at least one coefficient formula and one combinatorial interpretation. A table should be checked from the definition: for the unshifted convention, N_1(t)=t, N_2(t)=t+t^2, N_3(t)=t+3t^2+t^3, and N_4(t)=t+6t^2+6t^3+t^4. Then N_4(1)=14, the fourth Catalan number. If a reciprocal polynomial is introduced, specify its degree and normalization. Identify whether a claimed property applies to type A, type B, generalized, q-, or modified Narayana polynomials. Distinguish coefficient symmetry from a combinatorial involution proving it.

Manages Complexity¶

A Catalan number compresses a large family to one total. The Narayana polynomial restores one dimension of structure without listing every object: each coefficient is a fiber size of the statistic map, evaluation weights those fibers, differentiation yields statistic moments after normalization, and roots constrain coefficient shape. The representation enables algebraic proofs of distributional facts and bijective transfers between Catalan families. Complexity returns through convention drift and proliferating generalizations. The same label is used for shifted, reciprocal, type-B, and q-analog forms, so formulas can appear contradictory when they merely use different normalizations. Reference-grade work pins the convention before using identities and treats a special value as a consequence, not the definition.

Abstract Reasoning¶

Choose a Catalan family of size n and a statistic with Narayana distribution.
Fix the coefficient convention for N(n,k) and the valid range of k.
Choose the exponent convention and record any shift or reciprocal normalization.
Map each object with statistic k to the monomial carrying the corresponding power.
Sum monomials and group equal powers to obtain the coefficient polynomial.
Extract coefficients to recover refined counts and evaluate at one to recover the total.
Use coefficient symmetry or a bijection to relate complementary statistic values.
Apply recurrence, generating-function, or interlacing tools under the same normalization.
Check small rows directly to catch index shifts before asserting a general identity.
Label every type, q-, generalized, or modified extension rather than importing its properties silently.

Knowledge Transfer¶

The abstraction transfers a general enumerative pattern: replace a total count by a polynomial whose exponents mark a statistic and whose coefficients count its fibers. This move turns combinatorial refinement into algebra. It supports moments, symmetry tests, specializations, real-rootedness arguments, and bijections. The Narayana family provides a disciplined exemplar because several Catalan models yield the same row. Transfer requires proving equidistribution; sharing a Catalan total does not guarantee sharing the Narayana polynomial.

Examples¶

Canonical¶

Dyck paths of semilength four are counted by the Catalan number 14. Classifying them by peaks yields Narayana counts 1,6,6,1 for one through four peaks. Under the unshifted convention the polynomial is N_4(t)=t+6t^2+6t^3+t^4. The coefficient of t^2 says that six paths have two peaks, while evaluation at one returns all fourteen paths. Reversing peak count across the row exposes palindromicity. The example demonstrates refinement, not merely a formula table.^[1]

Mapped back: Dyck paths of fixed semilength → peak-count statistic → Narayana coefficient fibers → polynomial t+6t^2+6t^3+t^4 → Catalan total at one.

Applied / In Practice¶

Suppose a bijection maps noncrossing partitions of [n] with k blocks to a Catalan path statistic known to be counted by N(n,k). Summing t^k over partitions produces the same Narayana polynomial. Polynomial real-rootedness then implies log-concavity of the block-count sequence, but only after the coefficient identification is proved. Wang and Yeh's interlacing framework establishes real-zero behavior for classical Narayana sequences and illustrates how an enumerative identity unlocks analytic consequences.^[2]

Mapped back: noncrossing partitions + block statistic → coefficient identification → Narayana polynomial → real-rootedness theorem → log-concave refined counts.

Structural Tensions¶

Total count vs. refined distribution. Catalan evaluation is simple but discards statistic shape. Diagnostic: Which coefficient information is lost at t=1?
Equivalent conventions vs. formula mismatch. Exponent shifts preserve coefficients but change display. Diagnostic: Is the polynomial divisible by t, and why?
Algebraic property vs. combinatorial explanation. Real roots imply shape without giving a bijection. Diagnostic: Is symmetry proved analytically or by an object-level involution?
Classical family vs. generalizations. Type-B and q-forms are useful but distinct. Diagnostic: Do the coefficients still equal the type-A binomial product?
Coefficient formula vs. statistic identity. A triangle may be computed without saying what it counts. Diagnostic: Which statistic fiber does N(n,k) represent?
Reciprocity vs. normalization. Reciprocal polynomials depend on degree convention. Diagnostic: Has the exponent range been fixed before reversal?
Special values vs. defining structure. Evaluations yield famous sequences. Diagnostic: Does the claimed specialization follow from the exact convention used?

Structural–Framed Character¶

The structure is a size-indexed coefficient triangle, statistic-indexed monomials, Catalan collapse at one, reciprocal symmetry, and family-level recurrence or generating function. The frame is the chosen Catalan objects, statistic interpretation, exponent shift, variable name, and normalization. Different frames can instantiate the same polynomial through equidistribution; different normalizations can present it differently without changing the coefficient data.

Structural Core vs. Domain Accent¶

The transferable core is finite object family + integer-valued statistic → coefficient-generating polynomial preserving the statistic distribution. The domain accent is the Narayana binomial coefficient, Catalan families, peaks or blocks, palindromic row, real nonpositive zeros, and named specializations. Remove the accent and Representation remains; retain it and Narayana Polynomials form a distinct enumerative family.

Representation is the strict parent by specialization. The polynomial represents a statistic-refined Catalan family by encoding fiber counts as coefficients and statistic values as exponents. Representation is broader and does not require Narayana numbers, Catalan objects, or polynomial operations.

The prospective workspace queue contains one strict upward edge to prime:representation. No live DAG mutation is authorized.

Relationships to Other Abstractions¶

Current abstraction Narayana Polynomials Domain-specific

Parents (1) — more general patterns this builds on

Narayana Polynomials is a kind of Representation Prime

Representation is the strict parent by specialization.

Hierarchy path (1) — routes to 1 parentless root

Narayana Polynomials → Representation → Abstraction

Neighborhood in Abstraction Space¶

Narayana Polynomials sits in a sparse region of the domain-specific corpus (97^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With¶

Narayana Number. One coefficient in the triangle.
Catalan Number. Sum of a Narayana row, with the statistic forgotten.
Catalan Generating Function. Series across size indices rather than a fixed-row statistic polynomial.
Eulerian Polynomial. Descent enumerator on all permutations with a different coefficient triangle.
Type-B Narayana Polynomial. Coxeter-type analogue with different coefficients.
q-Narayana Polynomial. Parameter-deformed family whose definition must be separately given.
Reciprocal Polynomial. Operation applicable to many polynomials, not the Narayana identity itself.

References¶

[1] T. Kyle Petersen, Inquiry-Based Enumerative Combinatorics (2013), chapter on Catalan and Narayana numbers, https://math.depaul.edu/tpeter21/302Book.pdf. registry ↩a ↩b

[2] Y. Wang and Y.-N. Yeh, A Unified Approach to Polynomial Sequences with Only Real Zeros, Advances in Applied Mathematics 38, no. 4 (2007): 542–560, https://doi.org/10.1016/j.aam.2006.02.003. registry ↩a ↩b

[3] José Agapito, On Symmetric Polynomials with Only Real Zeros and Nonnegative Gamma-Vectors, Linear Algebra and its Applications 451 (2014): 260–289, https://doi.org/10.1016/j.laa.2014.03.018. registry ↩