Introduction
Macdonald polynomials P μ ( x ; q , t ) P_\mu(x;q,t) P μ ( x ; q , t ) are a family of symmetric functions introduced by Macdonald in 1988 [14 , 15 ] . Macdonald polynomials generalize many well-known bases of symmetric functions, including Schur functions and Hall–Littlewood polynomials. Moreover, due to their deep connections with various areas such as special function theory, representation theory, algebraic geometry, and quantum mechanics, they have been the subject of extensive study since their introduction. In the same work, Macdonald also defined the integral form Macdonald polynomials J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) with coefficients in Z [ q , t ] \mathbb{Z}[q,t] Z [ q , t ] :
J μ ( x ; q , t ) = P μ ( x ; q , t ) ∏ c ∈ μ ( 1 − q arm ( c ) t leg ( c ) + 1 ) . (1) J_\mu(x;q,t)=P_\mu(x;q,t)\prod_{c\in\mu}(1-q^{\operatorname{arm}(c)}t^{\operatorname{leg}(c)+1}). \tag*{(1)} J μ ( x ; q , t ) = P μ ( x ; q , t ) c ∈ μ ∏ ( 1 − q arm ( c ) t leg ( c ) + 1 ) . (1)
See § 2 for the precise definitions.
In 2005, Haglund, Haiman, and Loehr [8 ] made a major breakthrough in Macdonald theory by discovering a combinatorial formula for the modified Macdonald polynomials H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) , defined by
H ~ μ ( x ; q , t ) = t n ( μ ) J μ [ X 1 − 1 / t ; q , 1 / t ] , (2) \widetilde{H}_\mu(x;q,t)=t^{n(\mu)}J_\mu\left[\frac{X}{1-1/t};q,1/t\right], \tag*{(2)} H μ ( x ; q , t ) = t n ( μ ) J μ [ 1 − 1/ t X ; q , 1/ t ] , (2)
where X = x 1 + x 2 + ⋯ X=x_1+x_2+\cdots X = x 1 + x 2 + ⋯ , the square brackets indicate plethystic substitution , and n ( μ ) = ∑ i ≥ 1 ( i − 1 ) μ i n(\mu)=\sum_{i\geq1}(i-1)\mu_i n ( μ ) = ∑ i ≥ 1 ( i − 1 ) μ i . Haiman [11 ] proved that H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) is Schur positive, i.e., its Schur coefficients belong to Z ≥ 0 [ q , t ] \mathbb{Z}_{\geq0}[q,t] Z ≥ 0 [ q , t ] .
In [8 ] , Haglund, Haiman, and Loehr expressed H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) as a positive linear combination of LLT polynomials. Introduced by Lascoux, Leclerc, and Thibon [13 ] , the LLT polynomials LLT ν ( x ; q ) \operatorname{LLT}_\nu(x;q) LLT ν ( x ; q ) are symmetric functions that are known to be Schur positive [6 ] . In general, LLT polynomials are indexed by tuples of skew shapes. The special case in which each skew shape consists of a single cell gives rise to the so-called unicellular LLT polynomials , which can also be indexed by Dyck paths.
The unicellular LLT polynomials are closely related to the chromatic symmetric functions χ γ ( x ; q ) \chi_\gamma(x;q) χ γ ( x ; q ) introduced in [17 ] when γ \gamma γ is a Dyck path of length 2 n 2n 2 n corresponding to the incomparability graph of a natural unit interval order. More precisely, Haglund and Wilson [10 Eq. (6)]
2020 Mathematics Subject Classification . Primary: 05E05; Secondary: 05A30, 33D15.
Key words and phrases. integral form Macdonald polynomials, e e e -positivity, chromatic symmetric functions.
The second author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government RS-2025-00557835. The third author was supported by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Education (RS-2026-25469155). The fourth author was supported by funding for the academic research program of Chungbuk National University in 2026 and NRF grant RS-2024-00344076. observed that the result of Carlsson and Mellit [3 Proposition 3.5] is equivalent to
χ γ ( x ; q ) = LLT γ [ ( q − 1 ) X ; q ] ( q − 1 ) n . (3) \chi_\gamma(x;q)=\frac{\operatorname{LLT}_\gamma[(q-1)X;q]}{(q-1)^n}. \tag*{(3)} χ γ ( x ; q ) = ( q − 1 ) n LLT γ [( q − 1 ) X ; q ] . (3)
The chromatic symmetric functions χ γ ( x ; q ) \chi_\gamma(x;q) χ γ ( x ; q ) have been shown to expand positively in the elementary symmetric function basis (a property we refer to as e -positivity henceforth) when q = 1 q=1 q = 1 [12 , 5 ] , and were conjectured by Shareshian and Wachs [17 ] to have e -coefficients in $\mathbb{Z}_{\geq 0}[q]`.1
We note that (2) is equivalent to
J μ ( x ; q , t − 1 ) = t − ∑ i ≥ 1 i μ i H ~ μ [ ( t − 1 ) X ; q , t ] . (4) J_\mu(x;q,t^{-1})=t^{-\sum_{i\geq1}i\mu_i}\widetilde{H}_\mu[(t-1)X;q,t]. \tag*{(4)} J μ ( x ; q , t − 1 ) = t − ∑ i ≥ 1 i μ i H μ [( t − 1 ) X ; q , t ] . (4)
The present work is motivated by the analogy between (3) and (4) . On the right-hand sides, the unicellular LLT polynomials LLT γ ( x ; q ) \operatorname{LLT}_\gamma(x;q) LLT γ ( x ; q ) and the modified Macdonald polynomials H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) are both Schur positive. In view of this, the conjectured e -positivity of χ γ ( x ; q ) \chi_\gamma(x;q) χ γ ( x ; q ) motivates the investigation of e -positivity for suitably normalized specializations of integral form Macdonald polynomials J μ ( x ; q , t − 1 ) J_\mu(x;q,t^{-1}) J μ ( x ; q , t − 1 ) .
With regard to Schur positivity, Haglund conjectured (see [1 Eq. (1)] ) that, for every partition μ \mu μ and every positive integer k k k , the symmetric function
J μ ( x ; q , q k ) ( 1 − q ) ∣ μ ∣ \frac{J_\mu(x;q,q^k)}{(1-q)^{|\mu|}} ( 1 − q ) ∣ μ ∣ J μ ( x ; q , q k )
has Schur coefficients in Z ≥ 0 [ q ] \mathbb{Z}_{\geq0}[q] Z ≥ 0 [ q ] . Several special cases have been proved in [2 , 21 , 22 ] . Furthermore, Wilson [20 ] conjectured that, for every negative integer k k k , the expression obtained by replacing ( 1 − q ) (1-q) ( 1 − q ) with ( q − 1 ) (q-1) ( q − 1 ) and multiplying by a suitable power of q q q is also Schur positive.
Building on this conjecture and the preceding discussion of the e -positivity of J μ ( x ; q , t − 1 ) J_\mu(x;q,t^{-1}) J μ ( x ; q , t − 1 ) under an appropriate normalization, we propose the following conjecture.
Conjecture 1.1. For any partition μ \mu μ and any positive integer k k k , the following is e -positive:
q k ∑ i ≥ 1 i μ i J μ ( x ; q , q − k ) ( q − 1 ) ∣ μ ∣ . (5) q^{k\sum_{i\geq1}i\mu_i}\frac{J_\mu(x;q,q^{-k})}{(q-1)^{|\mu|}}. \tag*{(5)} q k ∑ i ≥ 1 i μ i ( q − 1 ) ∣ μ ∣ J μ ( x ; q , q − k ) . (5)
In this paper, we prove this conjecture when μ \mu μ is either a hook or a two-column partition. To do this, we first consider the e -expansion of the integral form Macdonald polynomial:
J μ ( x ; q , t ) = ∑ λ ⊢ ∣ μ ∣ c μ , λ ( q , t ) e λ . J_\mu(x;q,t)=\sum_{\lambda\vdash|\mu|}c_{\mu,\lambda}(q,t)e_\lambda. J μ ( x ; q , t ) = λ ⊢ ∣ μ ∣ ∑ c μ , λ ( q , t ) e λ .
When μ \mu μ is a hook, we express the coefficients c μ , λ ( q , t ) = c μ , λ c_{\mu,\lambda}(q,t)=c_{\mu,\lambda} c μ , λ ( q , t ) = c μ , λ as weighted sums over compositions (Theorem 4.2 ). As a consequence, we obtain a positive e -expansion of (5) (Corollary 4.6 ). When μ \mu μ has at most two columns, we compute the coefficients explicitly as products (Theorem 5.1 ), which implies the e -positivity of (5) . In addition, for an arbitrary partition μ \mu μ of n n n , we give a product formula for the coefficient c μ , ( n ) c_{\mu,(n)} c μ , ( n ) of e n e_n e n (Theorem 6.1 ).
The rest of the paper is organized as follows. In Section 2 we recall basic definitions and known results. In Section 3 we give an e -expansion formula for J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) when μ \mu μ is a one-row partition. In Section 4 we extend this result to hooks and prove Conjecture 1.1 in this case. Section 5 treats the two-column case, and Section 6 the e -coefficient. Finally, in Section 7, we explain a connection to α \alpha α -chromatic symmetric functions and propose a conjecture on the q → 1 q\to1 q → 1 limit of (5) .
Declaration on Generative AI
The ChatGPT 5.5, 5.6, and 6 Pro models and Claude Opus 4.7 were used for the following tasks:
Simplifying the authors’ original combinatorial proof of Theorem 4.2 in Section 3.1.
Finding a new recursive proof of Theorem 4.2 in Section 3.2.
Finding and proving Proposition 4.5 and Corollary 4.6.
Proving Theorem 6.1.
Proofreading the manuscript and conducting computational experiments with Sage.
All proofs obtained with the assistance of these AI tools have been independently verified and rewritten by the authors.
Preliminaries
In this section, we introduce the basic definitions and known results that will be used throughout the paper.
Basic definitions
A composition of n n n is a sequence α = ( α 1 , α 2 , … , α ℓ ) \alpha=(\alpha_1,\alpha_2,\ldots,\alpha_\ell) α = ( α 1 , α 2 , … , α ℓ ) of positive integers satisfying ∑ i α i = n \sum_i\alpha_i=n ∑ i α i = n , denoted by α ⊨ n \alpha\vDash n α ⊨ n . Each α i \alpha_i α i is called a part of α \alpha α , and the number of parts is called the length of α \alpha α , denoted by ℓ ( α ) \ell(\alpha) ℓ ( α ) . The sum of the parts of α \alpha α , denoted by ∣ α ∣ |\alpha| ∣ α ∣ , is called the size of α \alpha α . A partition is a composition such that the parts are weakly decreasing. If ∑ i λ i = n \sum_i\lambda_i=n ∑ i λ i = n , we say that λ \lambda λ is a partition of n n n and write λ ⊢ n \lambda\vdash n λ ⊢ n . For notational convenience, we also write ( 0 ) (0) ( 0 ) for the empty partition ∅ \varnothing ∅ .
If a composition α \alpha α is equal to a rearrangement of a partition λ \lambda λ , we write α ∼ λ \alpha\sim\lambda α ∼ λ . For partitions λ , μ ⊢ n \lambda,\mu\vdash n λ , μ ⊢ n , we define the dominance order by μ ≤ λ \mu\leq\lambda μ ≤ λ if
μ 1 + ⋯ + μ i ≤ λ 1 + ⋯ + λ i for all i ≥ 1. \mu_1+\cdots+\mu_i\leq\lambda_1+\cdots+\lambda_i \qquad\text{for all } i\geq1. μ 1 + ⋯ + μ i ≤ λ 1 + ⋯ + λ i for all i ≥ 1.
where both partitions are extended by zeros. We also write ( 0 ) (0) ( 0 ) for the empty composition ∅ \varnothing ∅ .
Given a partition λ = ( λ 1 , … , λ k ) \lambda=(\lambda_1,\ldots,\lambda_k) λ = ( λ 1 , … , λ k ) , its diagram is an array of unit squares, called cells , with λ i \lambda_i λ i cells in the i i i th row from the bottom, left-justified. We often identify λ \lambda λ with its diagram. A cell in row i i i and column j j j is denoted by ( i , j ) (i,j) ( i , j ) . The conjugate λ ′ \lambda^{\prime} λ ′ of λ \lambda λ is the partition whose diagram is obtained from λ \lambda λ by reflecting across the diagonal y = x y=x y = x . The reading order of λ \lambda λ is the order of its cells obtained by reading the rows from top to bottom and the cells in each row from left to right.
A partition of the form ( r , 1 s ) (r,1^s) ( r , 1 s ) with r ≥ 1 r\geq1 r ≥ 1 and s ≥ 0 s\geq0 s ≥ 0 is called a hook . Here, a b a^b a b means the sequence of b b b a a a 's. Note that hooks include one-row partitions ( n ) (n) ( n ) and one-column partitions ( 1 n ) = ( 1 , 1 n − 1 ) (1^n)=(1,1^{n-1}) ( 1 n ) = ( 1 , 1 n − 1 ) .
Given a cell c ∈ λ c\in\lambda c ∈ λ , the arm (resp. coarm ) length of c = ( i , j ) c=(i,j) c = ( i , j ) is the number of cells in row i i i to the right (resp. left) of c c c in λ \lambda λ , denoted by arm ( c ) \operatorname{arm}(c) arm ( c ) (resp. arm ′ ( c ) \operatorname{arm}^{\prime}(c) arm ′ ( c ) ). The leg (resp. coleg ) length of c = ( i , j ) c=(i,j) c = ( i , j ) is the number of cells in column j j j strictly above (resp. below) c c c in λ \lambda λ , denoted by leg ( c ) \operatorname{leg}(c) leg ( c ) (resp. leg ′ ( c ) \operatorname{leg}^{\prime}(c) leg ′ ( c ) ). See Figure 1 for an example.
Figure 1. For the cell c = ( 2 , 4 ) c=(2,4) c = ( 2 , 4 ) in λ = ( 9 , 8 , 6 , 6 , 3 ) \lambda=(9,8,6,6,3) λ = ( 9 , 8 , 6 , 6 , 3 ) , the arm, coarm, leg, and coleg lengths are arm ( c ) = 4 \operatorname{arm}(c)=4 arm ( c ) = 4 , arm ′ ( c ) = 3 \operatorname{arm}^{\prime}(c)=3 arm ′ ( c ) = 3 , leg ( c ) = 2 \operatorname{leg}(c)=2 leg ( c ) = 2 and leg ′ ( c ) = 1 \operatorname{leg}^{\prime}(c)=1 leg ′ ( c ) = 1 , respectively.
For a partition λ \lambda λ , a semistandard Young tableau of shape λ \lambda λ is a filling of λ \lambda λ with positive integers such that the entries are weakly increasing from left to right in each row and strictly increasing from bottom to top in each column. We use SSYT ( λ ) \operatorname{SSYT}(\lambda) SSYT ( λ ) to denote the set of semistandard Young tableaux of shape λ \lambda λ .
For a nonnegative integer r r r and a partition μ \mu μ , we use the q q q -Pochhammer symbols
( a ; q ) r = ∏ i = 0 r − 1 ( 1 − a q i ) , ( a ; q ) μ = ∏ i = 1 ℓ ( μ ) ( a ; q ) μ i . (a;q)_r=\prod_{i=0}^{r-1}(1-aq^i), \qquad(a;q)_\mu=\prod_{i=1}^{\ell(\mu)}(a;q)_{\mu_i}. ( a ; q ) r = i = 0 ∏ r − 1 ( 1 − a q i ) , ( a ; q ) μ = i = 1 ∏ ℓ ( μ ) ( a ; q ) μ i .
where ( a ; q ) 0 = 1 (a;q)_0=1 ( a ; q ) 0 = 1 . For integers n , k n,k n , k with 0 ≤ k ≤ n 0\leq k\leq n 0 ≤ k ≤ n , the q q q -binomial coefficient is defined by
[ n k ] q = ( q ; q ) n ( q ; q ) k ( q ; q ) n − k . \left[\begin{matrix} n \\ k \end{matrix}\right]_q
=
\frac{(q;q)_n}{(q;q)_k(q;q)_{n-k}}. [ n k ] q = ( q ; q ) k ( q ; q ) n − k ( q ; q ) n .
We also define
[ n ] q = 1 + q + ⋯ + q n − 1 , [ n ] q ! = ∏ i = 1 n [ i ] q . [n]_q = 1 + q + \cdots+ q^{n-1}, \qquad[n]_q! = \prod_{i=1}^{n} [i]_q. [ n ] q = 1 + q + ⋯ + q n − 1 , [ n ] q ! = i = 1 ∏ n [ i ] q .
Symmetric functions
Let Λ F \Lambda_F Λ F denote the F F F -algebra of symmetric functions in infinitely many variables x = ( x 1 , x 2 , … ) \boldsymbol{x}=(x_1,x_2,\ldots) x = ( x 1 , x 2 , … ) with coefficients in F = Q ( q , t ) F=\mathbb{Q}(q,t) F = Q ( q , t ) . For a partition λ = ( λ 1 , … , λ l ) \lambda=(\lambda_1,\ldots,\lambda_l) λ = ( λ 1 , … , λ l ) , the monomial symmetric function m λ m_\lambda m λ is defined by m λ = ∑ α x α m_\lambda=\sum_\alpha x^\alpha m λ = ∑ α x α , where the sum is over all sequences α = ( α 1 , α 2 , … ) \alpha=(\alpha_1,\alpha_2,\ldots) α = ( α 1 , α 2 , … ) obtained by rearranging ( λ 1 , … , λ l , 0 , 0 , … ) (\lambda_1,\ldots,\lambda_l,0,0,\ldots) ( λ 1 , … , λ l , 0 , 0 , … ) , and x α = x 1 α 1 x 2 α 2 ⋯ x^\alpha=x_1^{\alpha_1}x_2^{\alpha_2}\cdots x α = x 1 α 1 x 2 α 2 ⋯ . The elementary symmetric function e λ e_\lambda e λ is defined by e λ = ∏ i e λ i e_\lambda=\prod_i e_{\lambda_i} e λ = ∏ i e λ i , where e n = ∑ i 1 < ⋯ < i n x i 1 ⋯ x i n e_n=\sum_{i_1<\cdots<i_n}x_{i_1}\cdots x_{i_n} e n = ∑ i 1 < ⋯ < i n x i 1 ⋯ x i n . The complete homogeneous symmetric function h λ h_\lambda h λ is defined by h λ = ∏ i h λ i h_\lambda=\prod_i h_{\lambda_i} h λ = ∏ i h λ i , where h n = ∑ i 1 ≤ ⋯ ≤ i n x i 1 ⋯ x i n h_n=\sum_{i_1\leq\cdots\leq i_n}x_{i_1}\cdots x_{i_n} h n = ∑ i 1 ≤ ⋯ ≤ i n x i 1 ⋯ x i n . The power sum symmetric function p λ p_\lambda p λ is defined by p λ = ∏ i p λ i p_\lambda=\prod_i p_{\lambda_i} p λ = ∏ i p λ i , where p n = ∑ i x i n p_n=\sum_i x_i^n p n = ∑ i x i n . We also define e 0 = h 0 = p 0 = 1 e_0=h_0=p_0=1 e 0 = h 0 = p 0 = 1 . The Schur function s λ s_\lambda s λ is defined combinatorially by
s λ = ∑ T ∈ SSYT ( λ ) x T . s_\lambda=\sum_{T\in\operatorname{SSYT}(\lambda)}x^T. s λ = T ∈ SSYT ( λ ) ∑ x T .
where x T = ∏ i x i a i x^T=\prod_i x_i^{a_i} x T = ∏ i x i a i with a i a_i a i the number of occurrences of i i i in T T T .
We note that each of { m λ } \{m_\lambda\} { m λ } , { e λ } \{e_\lambda\} { e λ } , { h λ } \{h_\lambda\} { h λ } , { p λ } \{p_\lambda\} { p λ } , and { s λ } \{s_\lambda\} { s λ } forms a basis of the space of symmetric functions. With a slight abuse of notation, for any of the bases { g λ } \{g_\lambda\} { g λ } listed above, we write [ g λ ] f [g_\lambda]f [ g λ ] f for the coefficient of g λ g_\lambda g λ in the expansion of a symmetric function f ∈ Λ F f\in\Lambda_F f ∈ Λ F in that basis.
Let E = E ( t 1 , t 2 , … ) E=E(t_1,t_2,\ldots) E = E ( t 1 , t 2 , … ) be a formal Laurent series in t 1 , t 2 , … t_1,t_2,\ldots t 1 , t 2 , … with rational coefficients. For f = ∑ λ c λ p λ ∈ Λ F f=\sum_\lambda c_\lambda p_\lambda\in\Lambda_F f = ∑ λ c λ p λ ∈ Λ F , the plethystic substitution f [ E ] f[E] f [ E ] is defined by
f [ E ] : = ∑ λ c λ p λ [ E ] , f[E]:=\sum_\lambda c_\lambda p_\lambda[E], f [ E ] := λ ∑ c λ p λ [ E ] ,
where p λ [ E ] = ∏ i p λ i [ E ] p_\lambda[E]=\prod_i p_{\lambda_i}[E] p λ [ E ] = ∏ i p λ i [ E ] and p k [ E ] = E ( t 1 k , t 2 k , … ) p_k[E]=E(t_1^k,t_2^k,\ldots) p k [ E ] = E ( t 1 k , t 2 k , … ) . We refer the reader to [15 Ch. I, §8] for more details on plethysm.
Macdonald polynomials
The Macdonald polynomials P μ = P μ ( x ; q , t ) P_\mu=P_\mu(x;q,t) P μ = P μ ( x ; q , t ) are the unique symmetric functions satisfying
P μ = m μ + ∑ λ < μ u μ λ ( q , t ) m λ , ⟨ P λ , P μ ⟩ q , t = 0 if λ ≠ μ , P_\mu=m_\mu+\sum_{\lambda<\mu}u_{\mu\lambda}(q,t)m_\lambda,\qquad\langle P_\lambda,P_\mu\rangle_{q,t}=0\quad\text{if }\lambda\ne\mu, P μ = m μ + λ < μ ∑ u μ λ ( q , t ) m λ , ⟨ P λ , P μ ⟩ q , t = 0 if λ = μ ,
for suitable coefficients u μ λ ( q , t ) ∈ F u_{\mu\lambda}(q,t)\in F u μ λ ( q , t ) ∈ F ; see [14 , 15 ] . Here, ⟨ ⋅ , ⋅ ⟩ q , t \langle\cdot,\cdot\rangle_{q,t} ⟨ ⋅ , ⋅ ⟩ q , t is the inner product on Λ F \Lambda_F Λ F determined by
⟨ p λ , p μ ⟩ q , t = δ λ μ z λ ∏ i = 1 ℓ ( λ ) 1 − q λ i 1 − t λ i , z λ = ∏ r ≥ 1 r m r ( λ ) m r ( λ ) ! . (6) \langle p_\lambda,p_\mu\rangle_{q,t}=\delta_{\lambda\mu}z_\lambda\prod_{i=1}^{\ell(\lambda)}\frac{1-q^{\lambda_i}}{1-t^{\lambda_i}},\qquad z_\lambda=\prod_{r\geq1}r^{m_r(\lambda)}m_r(\lambda)!. \tag*{(6)} ⟨ p λ , p μ ⟩ q , t = δ λ μ z λ i = 1 ∏ ℓ ( λ ) 1 − t λ i 1 − q λ i , z λ = r ≥ 1 ∏ r m r ( λ ) m r ( λ )! . (6)
where m r ( λ ) m_r(\lambda) m r ( λ ) is the multiplicity of r r r in $\lambda. The integral form Macdonald polynomials J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) and the modified Macdonald polynomials H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) are defined by
J μ ( x ; q , t ) = P μ ( x ; q , t ) ∏ c ∈ μ ( 1 − q arm ( c ) t leg ( c ) + 1 ) , J_\mu(x;q,t)=P_\mu(x;q,t)\prod_{c\in\mu}(1-q^{\operatorname{arm}(c)}t^{\operatorname{leg}(c)+1}), J μ ( x ; q , t ) = P μ ( x ; q , t ) c ∈ μ ∏ ( 1 − q arm ( c ) t leg ( c ) + 1 ) ,
H ~ μ [ X ; q , t ] = t n ( μ ) J μ [ X 1 − t − 1 ; q , t − 1 ] , (7) \widetilde{H}_\mu[X;q,t]=t^{n(\mu)}J_\mu\left[\frac{X}{1-t^{-1}};q,t^{-1}\right], \tag*{(7)} H μ [ X ; q , t ] = t n ( μ ) J μ [ 1 − t − 1 X ; q , t − 1 ] , (7)
where n ( μ ) = ∑ i ( i − 1 ) μ i n(\mu)=\sum_i(i-1)\mu_i n ( μ ) = ∑ i ( i − 1 ) μ i . In [8 ] , Haglund, Haiman, and Loehr discovered a remarkable combinatorial formula for H ~ μ ( x ; q , t ) \widetilde{H}_\mu(x;q,t) H μ ( x ; q , t ) . To state their result, we need some definitions.
For a partition μ ⊢ n \mu\vdash n μ ⊢ n , a filling σ : μ → Z + \sigma:\mu\to\mathbb{Z}_{+} σ : μ → Z + is an assignment of positive integers to the cells in the diagram of μ \mu μ . A filling is standard if the entries are precisely 1 , 2 , … , ∣ μ ∣ 1,2,\ldots,|\mu| 1 , 2 , … , ∣ μ ∣ . The monomial x σ x^\sigma x σ associated with the filling σ \sigma σ is defined by
x σ = ∏ c ∈ μ x σ ( c ) . x^\sigma=\prod_{c\in\mu}x_{\sigma(c)}. x σ = c ∈ μ ∏ x σ ( c ) .
For a cell c = ( i , j ) ∈ μ c=(i,j)\in\mu c = ( i , j ) ∈ μ , let South ( i , j ) = ( i − 1 , j ) \operatorname{South}(i,j)=(i-1,j) South ( i , j ) = ( i − 1 , j ) . Given a filling σ : μ → Z + \sigma:\mu\to\mathbb{Z}^{+} σ : μ → Z + , the descent set Des ( σ ) \operatorname{Des}(\sigma) Des ( σ ) and the major index maj ( σ ) \operatorname{maj}(\sigma) maj ( σ ) are defined by
Des ( σ ) = { c ∈ μ : South ( c ) ∈ μ , σ ( c ) > σ ( South ( c ) ) } , \operatorname{Des}(\sigma)=\{c\in\mu:\operatorname{South}(c)\in\mu,\ \sigma(c)>\sigma(\operatorname{South}(c))\}, Des ( σ ) = { c ∈ μ : South ( c ) ∈ μ , σ ( c ) > σ ( South ( c ))} ,
maj ( σ ) = ∑ c ∈ Des ( σ ) ( leg ( c ) + 1 ) . \operatorname{maj}(\sigma)=\sum_{c\in\operatorname{Des}(\sigma)}(\operatorname{leg}(c)+1). maj ( σ ) = c ∈ Des ( σ ) ∑ ( leg ( c ) + 1 ) .
A triple of μ \mu μ is a set { u , v , w } \{u,v,w\} { u , v , w } of three cells with u , v ∈ μ u,v\in\mu u , v ∈ μ such that v v v is in the same row as u u u and strictly to its right, and w = South ( u ) w=\operatorname{South}(u) w = South ( u ) , which may be below the first row. Let σ \sigma σ be a filling of μ \mu μ . We orient a triple by starting at the cell whose entry is smallest, proceeding to the cell with the next-largest entry, and ending at the cell with the largest entry. If two cells have the same entry, the entry of the cell that appears earlier in the reading order of μ \mu μ is considered smaller. The triple is an inversion triple if this circular motion is counterclockwise, and a coinversion triple if it is clockwise. We denote by inv ( σ ) \operatorname{inv}(\sigma) inv ( σ ) (resp. coinv ( σ ) \operatorname{coinv}(\sigma) coinv ( σ ) ) the number of inversion triples (resp. coinversion triples) of σ \sigma σ . Here, we define σ ( w ) = ∞ \sigma(w)=\infty σ ( w ) = ∞ if w w w is below the first row. Note that
coinv ( σ ) = n ( μ ′ ) − inv ( σ ) . \operatorname{coinv}(\sigma)=n(\mu')-\operatorname{inv}(\sigma). coinv ( σ ) = n ( μ ′ ) − inv ( σ ) .
Theorem 2.1 ([8 Theorem 2.2] ). For a partition μ \mu μ , the modified Macdonald polynomial H ~ μ ( x ; q , t ) \widetilde{H}_{\mu}(x;q,t) H μ ( x ; q , t ) is given by
H ~ μ ( x ; q , t ) = ∑ σ : μ → Z + x σ q inv ( σ ) t maj ( σ ) . (8) \widetilde{H}_{\mu}(x;q,t)=\sum_{\sigma:\mu\to\mathbb{Z}^{+}}x^{\sigma}q^{\operatorname{inv}(\sigma)}t^{\operatorname{maj}(\sigma)}. \tag*{(8)} H μ ( x ; q , t ) = σ : μ → Z + ∑ x σ q inv ( σ ) t maj ( σ ) . (8)
By (7) , we have
J μ ( x ; q , t ) = t n ( μ ) H ~ μ [ X ( 1 − t ) ; q , t − 1 ] = t n ( μ ) + n H ~ μ ′ [ X ( t − 1 − 1 ) ; t − 1 , q ] . (9) J_{\mu}(x;q,t)=t^{n(\mu)}\widetilde{H}_{\mu}[X(1-t);q,t^{-1}]=t^{n(\mu)+n}\widetilde{H}_{\mu'}[X(t^{-1}-1);t^{-1},q]. \tag*{(9)} J μ ( x ; q , t ) = t n ( μ ) H μ [ X ( 1 − t ) ; q , t − 1 ] = t n ( μ ) + n H μ ′ [ X ( t − 1 − 1 ) ; t − 1 , q ] . (9)
To state a combinatorial interpretation for J μ ( x ; q , t ) J_{\mu}(x;q,t) J μ ( x ; q , t ) , we introduce the necessary definitions.
We say that a cell u ∈ μ u\in\mu u ∈ μ attacks the cells in the same row and strictly to its right, and the cells in the row below and strictly to its left. A filling σ : μ → Z + \sigma:\mu\to\mathbb{Z}^{+} σ : μ → Z + is nonattacking if there are no cells u , v ∈ μ u,v\in\mu u , v ∈ μ such that u u u attacks v v v and σ ( u ) = σ ( v ) \sigma(u)=\sigma(v) σ ( u ) = σ ( v ) .
Let A + = { 1 , 2 , … } \mathcal{A}_{+}=\{1,2,\ldots\} A + = { 1 , 2 , … } , A − = { 1 ‾ , 2 ‾ , … } \mathcal{A}_{-}=\{\overline{1},\overline{2},\ldots\} A − = { 1 , 2 , … } , and A ± = A + ∪ A − \mathcal{A}_{\pm}=\mathcal{A}_{+}\cup\mathcal{A}_{-} A ± = A + ∪ A − . We write ∣ i ∣ = i |i|=i ∣ i ∣ = i and ∣ i ‾ ∣ = i |\overline{i}|=i ∣ i ∣ = i . A superfilling is an assignment σ : μ → A ± \sigma:\mu\to\mathcal{A}_{\pm} σ : μ → A ± of elements in A ± \mathcal{A}_{\pm} A ± to the cells in μ \mu μ . We denote by ∣ σ ∣ |\sigma| ∣ σ ∣ the filling obtained from σ \sigma σ by replacing each entry i ‾ \overline{i} i with i i i . We say that a superfilling σ \sigma σ is nonattacking if ∣ σ ∣ |\sigma| ∣ σ ∣ is nonattacking. We denote by neg ( σ ) \operatorname{neg}(\sigma) neg ( σ ) the number of entries of σ \sigma σ in A − \mathcal{A}_{-} A − . Let NASF ( μ ) \operatorname{NASF}(\mu) NASF ( μ ) denote the set of nonattacking superfillings of μ \mu μ .
Given a total order ≺ \prec ≺ on A ± \mathcal{A}_{\pm} A ± , the standardization of σ \sigma σ with respect to ≺ \prec ≺ is the bijective filling σ ≺ : μ → { 1 , … , n } \sigma^{\prec}:\mu\to\{1,\ldots,n\} σ ≺ : μ → { 1 , … , n } obtained by relabeling the cells in increasing order of their entries, where ties between cells holding the same letter are broken by reading order: a letter in A + \mathcal{A}_{+} A + read earlier is regarded as smaller, and a letter in A − \mathcal{A}_{-} A − read earlier is regarded as larger.
Haglund, Haiman, and Loehr [8 Eq. (63)] found a combinatorial interpretation for J μ ( x ; q , t ) J_{\mu}(x;q,t) J μ ( x ; q , t ) . We will use the following reformulation from [7 (A.30)] .
Theorem 2.2. We have
J μ ( x ; q , t ) = ∑ σ ∈ NASF ( μ ′ ) x ∣ σ ∣ q maj ( σ ) t coinv ( σ ) ( − t ) neg ( σ ) . (10) J_{\mu}(x;q,t)=\sum_{\sigma\in\operatorname{NASF}(\mu')}x^{|\sigma|}q^{\operatorname{maj}(\sigma)}t^{\operatorname{coinv}(\sigma)}(-t)^{\operatorname{neg}(\sigma)}. \tag*{(10)} J μ ( x ; q , t ) = σ ∈ NASF ( μ ′ ) ∑ x ∣ σ ∣ q maj ( σ ) t coinv ( σ ) ( − t ) neg ( σ ) . (10) where maj ( σ ) = maj ( σ s t ) \operatorname{maj}(\sigma)=\operatorname{maj}(\sigma^{\mathrm{st}}) maj ( σ ) = maj ( σ st ) and coinv ( σ ) = coinv ( σ s t ) \operatorname{coinv}(\sigma)=\operatorname{coinv}(\sigma^{\mathrm{st}}) coinv ( σ ) = coinv ( σ st ) , with standardization taken with respect to the ordering
1 ≺ 1 ‾ ≺ 2 ≺ 2 ‾ ≺ ⋯ ≺ n ≺ n ‾ ≺ ⋯ . (11) 1\prec\overline{1}\prec2\prec\overline{2}\prec\cdots\prec n\prec\overline{n}\prec\cdots. \tag*{(11)} 1 ≺ 1 ≺ 2 ≺ 2 ≺ ⋯ ≺ n ≺ n ≺ ⋯ . (11)
One-row case
In this section, we give an explicit combinatorial formula for the e e e -coefficients of J ( n ) ( x ; q , t ) J_{(n)}(x;q,t) J ( n ) ( x ; q , t ) . We give two proofs of this formula. We first define a weight of a composition with respect to a one-row partition.
Definition 3.1. For a composition α = ( α 1 , … , α k ) \alpha=(\alpha_1,\ldots,\alpha_k) α = ( α 1 , … , α k ) of n n n , let
wt ( i , α ) = { 1 − q i − 1 t α j if i = α 1 + ⋯ + α j − 1 + 1 for some 1 ≤ j ≤ k , q i − 1 − 1 otherwise. (12) \operatorname{wt}(i,\alpha)=
\begin{cases}
1-q^{i-1}t^{\alpha_j} & \text{if }i=\alpha_1+\cdots+\alpha_{j-1}+1\text{ for some }1\leq j\leq k,\\
q^{i-1}-1 & \text{otherwise.}
\end{cases} \tag*{(12)} wt ( i , α ) = { 1 − q i − 1 t α j q i − 1 − 1 if i = α 1 + ⋯ + α j − 1 + 1 for some 1 ≤ j ≤ k , otherwise. (12)
For the one-row partition ( n ) (n) ( n ) , define
wt ( n ) ( α ) = ∏ i = 1 n wt ( i , α ) . (13) \operatorname{wt}_{(n)}(\alpha)=\prod_{i=1}^{n}\operatorname{wt}(i,\alpha). \tag*{(13)} wt ( n ) ( α ) = i = 1 ∏ n wt ( i , α ) . (13)
It is helpful to visualize the weight wt ( n ) ( α ) \operatorname{wt}_{(n)}(\alpha) wt ( n ) ( α ) as in the next example.
Example 3.2. Let n = 10 n=10 n = 10 and α = ( 2 , 1 , 3 , 4 ) \alpha=(2,1,3,4) α = ( 2 , 1 , 3 , 4 ) . Draw a row divided into blocks of sizes α 1 , α 2 , α 3 , α 4 \alpha_1,\alpha_2,\alpha_3,\alpha_4 α 1 , α 2 , α 3 , α 4 , that is, 2 , 1 , 3 , 4 2,1,3,4 2 , 1 , 3 , 4 , from left to right, as illustrated in Figure 2 . The first cells of the blocks have indices i = 1 , 3 , 4 , 7 i=1,3,4,7 i = 1 , 3 , 4 , 7 . We have
Figure 2. For α = ( 2 , 1 , 3 , 4 ) \alpha=(2,1,3,4) α = ( 2 , 1 , 3 , 4 ) , a row divided into blocks of sizes α 1 , α 2 , α 3 , α 4 \alpha_1,\alpha_2,\alpha_3,\alpha_4 α 1 , α 2 , α 3 , α 4 from left to right. A dot marks the first cell of each block, and the weight wt ( i , α ) \operatorname{wt}(i,\alpha) wt ( i , α ) is written above the i i i th cell.
wt ( 10 ) ( 2 , 1 , 3 , 4 ) = ( 1 − t 2 ) ( q − 1 ) ( 1 − q 2 t ) ( 1 − q 3 t 3 ) \operatorname{wt}_{(10)}(2,1,3,4)=(1-t^2)(q-1)(1-q^2t)(1-q^3t^3) wt ( 10 ) ( 2 , 1 , 3 , 4 ) = ( 1 − t 2 ) ( q − 1 ) ( 1 − q 2 t ) ( 1 − q 3 t 3 )
× ( q 4 − 1 ) ( q 5 − 1 ) ( 1 − q 6 t 4 ) ( q 7 − 1 ) ( q 8 − 1 ) ( q 9 − 1 ) . {}\times(q^4-1)(q^5-1)(1-q^6t^4)(q^7-1)(q^8-1)(q^9-1). × ( q 4 − 1 ) ( q 5 − 1 ) ( 1 − q 6 t 4 ) ( q 7 − 1 ) ( q 8 − 1 ) ( q 9 − 1 ) .
Theorem 3.3. For n ≥ 0 n\geq0 n ≥ 0 , we have
J ( n ) ( x ; q , t ) = ∑ λ ⊢ n ( ∑ α ⊨ λ wt ( n ) ( α ) ) e λ . J_{(n)}(x;q,t)=\sum_{\lambda\vdash n}\left(\sum_{\alpha\models\lambda}\operatorname{wt}_{(n)}(\alpha)\right)e_\lambda. J ( n ) ( x ; q , t ) = λ ⊢ n ∑ α ⊨ λ ∑ wt ( n ) ( α ) e λ .
To prove this theorem, it is convenient to use the following notation. For a composition α \alpha α , write
e α = ∏ j e α j e_\alpha=\prod_j e_{\alpha_j} e α = j ∏ e α j
and let
Y ( n ) = Y ( n ) ( x ; q , t ) = ∑ α ⊨ n wt ( n ) ( α ) e α . (14) Y_{(n)}=Y_{(n)}(x;q,t)=\sum_{\alpha\models n}\operatorname{wt}_{(n)}(\alpha)e_\alpha. \tag*{(14)} Y ( n ) = Y ( n ) ( x ; q , t ) = α ⊨ n ∑ wt ( n ) ( α ) e α . (14)
Then the identity in Theorem 3.3 is equivalent to
J ( n ) ( x ; q , t ) = Y ( n ) ( x ; q , t ) . (15) J_{(n)}(x;q,t)=Y_{(n)}(x;q,t). \tag*{(15)} J ( n ) ( x ; q , t ) = Y ( n ) ( x ; q , t ) . (15)
We provide two proofs of this identity in the next two subsections. The first proof is combinatorial and the second proof is recursive.
Combinatorial proof
In this subsection, we prove (3.4) by comparing monomial coefficients on both sides using the Haglund–Haiman–Loehr formula (2.5).
For the one-row partition μ = ( n ) \mu=(n) μ = ( n ) , we can identify each superfilling of μ ′ = ( 1 n ) \mu'=(1^n) μ ′ = ( 1 n ) with its reading word. Using this identification, we first restate (2.5) in terms of words on A ± \mathcal{A}_{\pm} A ± . We denote by W n ( A ± ) W_n(\mathcal{A}_{\pm}) W n ( A ± ) the set of words of length n n n on A ± \mathcal{A}_{\pm} A ± . We define W n ( A + ) W_n(\mathcal{A}_{+}) W n ( A + ) and W n ( A − ) W_n(\mathcal{A}_{-}) W n ( A − ) similarly. For π = π 1 ⋯ π n ∈ W n ( A ± ) \pi=\pi_1\cdots\pi_n\in W_n(\mathcal{A}_{\pm}) π = π 1 ⋯ π n ∈ W n ( A ± ) , we define ∣ π ∣ = ∣ π 1 ∣ ⋯ ∣ π n ∣ ∈ W n ( A + ) |\pi|=|\pi_1|\cdots|\pi_n|\in W_n(\mathcal{A}_{+}) ∣ π ∣ = ∣ π 1 ∣ ⋯ ∣ π n ∣ ∈ W n ( A + ) . For π = π 1 ⋯ π n ∈ W n ( A + ) \pi=\pi_1\cdots\pi_n\in W_n(\mathcal{A}_{+}) π = π 1 ⋯ π n ∈ W n ( A + ) , we write x π = x π 1 ⋯ x π n x^\pi=x_{\pi_1}\cdots x_{\pi_n} x π = x π 1 ⋯ x π n .
For π ∈ W n ( A ± ) \pi\in W_n(\mathcal{A}_{\pm}) π ∈ W n ( A ± ) , let neg ( π ) \operatorname{neg}(\pi) neg ( π ) denote the number of letters of π \pi π in A − \mathcal{A}_{-} A − . We define
Des ( π ) = { 1 ≤ i ≤ n − 1 : π i ≻ π i + 1 or π i = π i + 1 ∈ A − } , maj ( π ) = ∑ i ∈ Des ( π ) i , \operatorname{Des}(\pi)=\{1\leq i\leq n-1:\pi_i\succ\pi_{i+1}\text{ or }\pi_i=\pi_{i+1}\in\mathcal{A}_{-}\},\qquad
\operatorname{maj}(\pi)=\sum_{i\in\operatorname{Des}(\pi)}i, Des ( π ) = { 1 ≤ i ≤ n − 1 : π i ≻ π i + 1 or π i = π i + 1 ∈ A − } , maj ( π ) = i ∈ Des ( π ) ∑ i ,
where ≺ \prec ≺ is the ordering of A ± \mathcal{A}_{\pm} A ± in (2.6). For later use, we also define
Des ∗ ( π ) = { 1 ≤ i ≤ n − 1 : π i ≻ ∗ π i + 1 or π i = π i + 1 ∈ A − } , maj ∗ ( π ) = ∑ i ∈ Des ∗ ( π ) i . \operatorname{Des}^{*}(\pi)=\{1\leq i\leq n-1:\pi_i\succ^{*}\pi_{i+1}\text{ or }\pi_i=\pi_{i+1}\in\mathcal{A}_{-}\},\qquad
\operatorname{maj}^{*}(\pi)=\sum_{i\in\operatorname{Des}^{*}(\pi)}i. Des ∗ ( π ) = { 1 ≤ i ≤ n − 1 : π i ≻ ∗ π i + 1 or π i = π i + 1 ∈ A − } , maj ∗ ( π ) = i ∈ Des ∗ ( π ) ∑ i .
where ≺ ∗ \prec^{*} ≺ ∗ is the ordering of A ± \mathcal{A}_{\pm} A ± given by
1 ‾ ≺ ∗ 2 ‾ ≺ ∗ ⋯ ≺ ∗ 1 ≺ ∗ 2 ≺ ∗ ⋯ . (16) \overline{1}\prec^{*}\overline{2}\prec^{*}\cdots\prec^{*}1\prec^{*}2\prec^{*}\cdots. \tag*{(16)} 1 ≺ ∗ 2 ≺ ∗ ⋯ ≺ ∗ 1 ≺ ∗ 2 ≺ ∗ ⋯ . (16)
Now consider σ ∈ NASF ( ( 1 n ) ) \sigma\in\operatorname{NASF}((1^{n})) σ ∈ NASF (( 1 n )) and let π \pi π be the reading word of σ \sigma σ . Then maj ( σ ) = maj ( π ) \operatorname{maj}(\sigma)=\operatorname{maj}(\pi) maj ( σ ) = maj ( π ) and neg ( σ ) = neg ( π ) \operatorname{neg}(\sigma)=\operatorname{neg}(\pi) neg ( σ ) = neg ( π ) . Since the one-column partition ( 1 n ) (1^{n}) ( 1 n ) contains no attacking pairs, the map σ ↦ π \sigma\mapsto\pi σ ↦ π is a bijection from NASF ( ( 1 n ) ) \operatorname{NASF}((1^{n})) NASF (( 1 n )) to W n ( A ± ) W_{n}(\mathcal{A}_{\pm}) W n ( A ± ) . Moreover, since ( 1 n ) (1^{n}) ( 1 n ) has no triples, we always have coinv ( σ ) = 0 \operatorname{coinv}(\sigma)=0 coinv ( σ ) = 0 . Therefore, (10) gives
J ( n ) ( x ; q , t ) = ∑ π ∈ W n ( A ± ) x ∣ π ∣ q maj ( π ) ( − t ) neg ( π ) . (17) J_{(n)}(x;q,t)=\sum_{\pi\in W_{n}(\mathcal{A}_{\pm})}x^{|\pi|}q^{\operatorname{maj}(\pi)}(-t)^{\operatorname{neg}(\pi)}. \tag*{(17)} J ( n ) ( x ; q , t ) = π ∈ W n ( A ± ) ∑ x ∣ π ∣ q maj ( π ) ( − t ) neg ( π ) . (17)
Lemma 3.4. We have
Y ( n ) ( x ; q , t ) = ∑ π ∈ W n ( A ± ) x ∣ π ∣ q maj ∗ ( π ) ( − t ) neg ( π ) . Y_{(n)}(x;q,t)=\sum_{\pi\in W_{n}(\mathcal{A}_{\pm})}x^{|\pi|}q^{\operatorname{maj}^{*}(\pi)}(-t)^{\operatorname{neg}(\pi)}. Y ( n ) ( x ; q , t ) = π ∈ W n ( A ± ) ∑ x ∣ π ∣ q maj ∗ ( π ) ( − t ) neg ( π ) .
Proof. For a composition α = ( α 1 , … , α k ) ⊨ n \alpha=(\alpha_{1},\ldots,\alpha_{k})\models n α = ( α 1 , … , α k ) ⊨ n and 1 ≤ i ≤ k 1\le i\le k 1 ≤ i ≤ k , we define the i i i th block of α \alpha α to be
B i ( α ) = { α 1 + ⋯ + α i − 1 + 1 , α 1 + ⋯ + α i − 1 + 2 , … , α 1 + ⋯ + α i } . B_i(\alpha)=\{\alpha_1+\cdots+\alpha_{i-1}+1,\alpha_1+\cdots+\alpha_{i-1}+2,\ldots,\alpha_1+\cdots+\alpha_i\}. B i ( α ) = { α 1 + ⋯ + α i − 1 + 1 , α 1 + ⋯ + α i − 1 + 2 , … , α 1 + ⋯ + α i } .
By definition, we have
Y ( n ) ( x ; q , t ) = ∑ α ⊨ n wt ( n ) ( α ) e α = ∑ ( α , π ) ∈ C n wt ( n ) ( α ) x π , (18) Y_{(n)}(x;q,t)=\sum_{\alpha\models n}\operatorname{wt}_{(n)}(\alpha)e_{\alpha}=\sum_{(\alpha,\pi)\in C_n}\operatorname{wt}_{(n)}(\alpha)x^{\pi}, \tag*{(18)} Y ( n ) ( x ; q , t ) = α ⊨ n ∑ wt ( n ) ( α ) e α = ( α , π ) ∈ C n ∑ wt ( n ) ( α ) x π , (18)
where C n C_n C n is the set of pairs ( α , π ) (\alpha,\pi) ( α , π ) such that α = ( α 1 , … , α k ) ⊨ n \alpha=(\alpha_1,\ldots,\alpha_k)\models n α = ( α 1 , … , α k ) ⊨ n , π = π 1 ⋯ π n ∈ W n ( A + ) \pi=\pi_1\cdots\pi_n\in W_n(\mathcal{A}_{+}) π = π 1 ⋯ π n ∈ W n ( A + ) , and the subword of π \pi π at positions in B i ( α ) B_i(\alpha) B i ( α ) is strictly decreasing for each 1 ≤ i ≤ k 1\le i\le k 1 ≤ i ≤ k .
Now we reformulate (18) using words on A ± \mathcal{A}_{\pm} A ± . Writing B i ( α ) = { r i , r i + 1 , … , s i } B_i(\alpha)=\{r_i,r_i+1,\ldots,s_i\} B i ( α ) = { r i , r i + 1 , … , s i } , we have
wt ( n ) ( α ) = ∏ i = 1 k ( ( 1 − q r i − 1 ( − t ) α i ) ∏ j = r i + 1 s i ( q j − 1 − 1 ) ) = ∏ i = 1 k ( ∏ j = r i + 1 s i ( q j − 1 − 1 ) + q r i − 1 ( − t ) α i ∏ j = r i + 1 s i ( 1 − q j − 1 ) ) . \begin{aligned}
\operatorname{wt}_{(n)}(\alpha)=\prod_{i=1}^{k}\left((1-q^{r_i-1}(-t)^{\alpha_i})\prod_{j=r_i+1}^{s_i}(q^{j-1}-1)\right) \\
&=\prod_{i=1}^{k}\left(\prod_{j=r_i+1}^{s_i}(q^{j-1}-1)+q^{r_i-1}(-t)^{\alpha_i}\prod_{j=r_i+1}^{s_i}(1-q^{j-1})\right).
\end{aligned} wt ( n ) ( α ) = i = 1 ∏ k ( ( 1 − q r i − 1 ( − t ) α i ) j = r i + 1 ∏ s i ( q j − 1 − 1 ) ) = i = 1 ∏ k ( j = r i + 1 ∏ s i ( q j − 1 − 1 ) + q r i − 1 ( − t ) α i j = r i + 1 ∏ s i ( 1 − q j − 1 ) ) .
Choosing the first or second term in the i i i th factor can be interpreted as leaving the subword π r i ⋯ π s i \pi_{r_i}\cdots\pi_{s_i} π r i ⋯ π s i unchanged or replacing it with π s i ⋯ π r i \pi_{s_i}\cdots\pi_{r_i} π s i ⋯ π r i , respectively. Using this idea, we define D n D_n D n to be the set of pairs ( α , π ) (\alpha,\pi) ( α , π ) such that α = ( α 1 , … , α k ) ⊨ n \alpha=(\alpha_1,\ldots,\alpha_k)\models n α = ( α 1 , … , α k ) ⊨ n , π = π 1 ⋯ π n ∈ W n ( A ± ) \pi=\pi_1\cdots\pi_n\in W_n(\mathcal{A}_{\pm}) π = π 1 ⋯ π n ∈ W n ( A ± ) , and for each 1 ≤ i ≤ k 1\le i\le k 1 ≤ i ≤ k , the subword of π \pi π at positions in B i ( α ) B_i(\alpha) B i ( α ) is either a strictly decreasing word on A + \mathcal{A}_{+} A + or a strictly increasing word on A − \mathcal{A}_{-} A − with respect to the order ≺ \prec ≺ in (11) . For ( α , π ) ∈ D n (\alpha,\pi)\in D_n ( α , π ) ∈ D n , we define
wt ( α , π ) = ∏ i = 1 n wt ( i , α , π ) , \operatorname{wt}(\alpha,\pi)=\prod_{i=1}^{n}\operatorname{wt}(i,\alpha,\pi), wt ( α , π ) = i = 1 ∏ n wt ( i , α , π ) ,
where
wt ( i , α , π ) = { 1 if i = min ( B j ( α ) ) for some 1 ≤ j ≤ ℓ ( α ) and π i ∈ A + , q i − 1 ( − t ) α j if i = min ( B j ( α ) ) for some 1 ≤ j ≤ ℓ ( α ) and π i ∈ A − , q i − 1 − 1 if i ≠ min ( B j ( α ) ) for all 1 ≤ j ≤ ℓ ( α ) and π i ∈ A + , 1 − q i − 1 if i ≠ min ( B j ( α ) ) for all 1 ≤ j ≤ ℓ ( α ) and π i ∈ A − . \operatorname{wt}(i,\alpha,\pi)=
\begin{cases}
1 & \text{if }i=\min(B_j(\alpha))\text{ for some }1\le j\le\ell(\alpha)\text{ and }\pi_i\in\mathcal{A}_{+},\\
q^{i-1}(-t)^{\alpha_j} & \text{if }i=\min(B_j(\alpha))\text{ for some }1\le j\le\ell(\alpha)\text{ and }\pi_i\in\mathcal{A}_{-},\\
q^{i-1}-1 & \text{if }i\ne\min(B_j(\alpha))\text{ for all }1\le j\le\ell(\alpha)\text{ and }\pi_i\in\mathcal{A}_{+},\\
1-q^{i-1} & \text{if }i\ne\min(B_j(\alpha))\text{ for all }1\le j\le\ell(\alpha)\text{ and }\pi_i\in\mathcal{A}_{-}.
\end{cases} wt ( i , α , π ) = ⎩ ⎨ ⎧ 1 q i − 1 ( − t ) α j q i − 1 − 1 1 − q i − 1 if i = min ( B j ( α )) for some 1 ≤ j ≤ ℓ ( α ) and π i ∈ A + , if i = min ( B j ( α )) for some 1 ≤ j ≤ ℓ ( α ) and π i ∈ A − , if i = min ( B j ( α )) for all 1 ≤ j ≤ ℓ ( α ) and π i ∈ A + , if i = min ( B j ( α )) for all 1 ≤ j ≤ ℓ ( α ) and π i ∈ A − .
Then (18) can be rewritten as
Y ( n ) ( x ; q , t ) = ∑ ( α , π ) ∈ D n wt ( α , π ) x ∣ π ∣ = ∑ π ∈ W n ( A ± ) x ∣ π ∣ ∑ α : ( α , π ) ∈ D n wt ( α , π ) , Y_{(n)}(x;q,t)=\sum_{(\alpha,\pi)\in D_n}\operatorname{wt}(\alpha,\pi)x^{|\pi|}
=\sum_{\pi\in W_n(\mathcal{A}_{\pm})}x^{|\pi|}\sum_{\alpha:(\alpha,\pi)\in D_n}\operatorname{wt}(\alpha,\pi), Y ( n ) ( x ; q , t ) = ( α , π ) ∈ D n ∑ wt ( α , π ) x ∣ π ∣ = π ∈ W n ( A ± ) ∑ x ∣ π ∣ α : ( α , π ) ∈ D n ∑ wt ( α , π ) ,
where x ∣ π ∣ = x ∣ π 1 ∣ ⋯ x ∣ π n ∣ x^{|\pi|}=x_{|\pi_1|}\cdots x_{|\pi_n|} x ∣ π ∣ = x ∣ π 1 ∣ ⋯ x ∣ π n ∣ . Therefore, it suffices to show that for each π ∈ W n ( A ± ) \pi\in W_n(\mathcal{A}_{\pm}) π ∈ W n ( A ± ) ,
∑ α ⊨ n wt ( α , π ) = q maj ∗ ( π ) ( − t ) neg ( π ) , (19) \sum_{\alpha\models n}\operatorname{wt}(\alpha,\pi)=q^{\operatorname{maj}^{*}(\pi)}(-t)^{\operatorname{neg}(\pi)}, \tag*{(19)} α ⊨ n ∑ wt ( α , π ) = q maj ∗ ( π ) ( − t ) neg ( π ) , (19)
where we extend the definition of wt ( α , π ) \operatorname{wt}(\alpha,\pi) wt ( α , π ) to all compositions α ⊨ n \alpha\models n α ⊨ n by setting wt ( α , π ) = 0 \operatorname{wt}(\alpha,\pi)=0 wt ( α , π ) = 0 unless ( α , π ) ∈ D n (\alpha,\pi)\in D_n ( α , π ) ∈ D n .
Identifying a composition α = ( α 1 , … , α k ) \alpha=(\alpha_1,\ldots,\alpha_k) α = ( α 1 , … , α k ) with its blocks B 1 ( α ) , … , B k ( α ) B_1(\alpha),\ldots,B_k(\alpha) B 1 ( α ) , … , B k ( α ) , we can construct a composition of n n n by choosing whether i i i and i + 1 i+1 i + 1 belong to the same block for each 1 ≤ i ≤ n − 1 1\leq i\leq n-1 1 ≤ i ≤ n − 1 . We say that there is a cut between i i i and i + 1 i+1 i + 1 if they belong to different blocks. Note that the cuts can be chosen independently, so there are 2 n − 1 2^{n-1} 2 n − 1 compositions of n n n in total.
Now we fix π ∈ W n ( A ± ) \pi\in W_n(\mathcal{A}_{\pm}) π ∈ W n ( A ± ) . We define the weights c i ( π ) c_i(\pi) c i ( π ) and d i ( π ) d_i(\pi) d i ( π ) for cutting and not cutting, respectively, between i i i and i + 1 i+1 i + 1 according to the signs of π i \pi_i π i and π i + 1 \pi_{i+1} π i + 1 , as follows:
Table 1.
Here χ ( P ) = 1 \chi(P)=1 χ ( P ) = 1 if P P P holds and χ ( P ) = 0 \chi(P)=0 χ ( P ) = 0 otherwise. Then, by the construction and the definition of wt ( α , π ) \operatorname{wt}(\alpha,\pi) wt ( α , π ) , we have
∑ α ⊨ n wt ( α , π ) = ( − t ) neg ( π ) ∏ i = 1 n − 1 ( c i ( π ) + d i ( π ) ) . (20) \sum_{\alpha\vDash n}\operatorname{wt}(\alpha,\pi)=(-t)^{\operatorname{neg}(\pi)}\prod_{i=1}^{n-1}(c_i(\pi)+d_i(\pi)). \tag*{(20)} α ⊨ n ∑ wt ( α , π ) = ( − t ) neg ( π ) i = 1 ∏ n − 1 ( c i ( π ) + d i ( π )) . (20)
On the other hand, for each 1 ≤ i ≤ n − 1 1\leq i\leq n-1 1 ≤ i ≤ n − 1 , we have
c i ( π ) + d i ( π ) = { q i if π i , π i + 1 ∈ A + and π i > π i + 1 , 1 if π i , π i + 1 ∈ A + and π i ≤ π i + 1 , 1 if π i , π i + 1 ∈ A − and ∣ π i ∣ < ∣ π i + 1 ∣ , q i if π i , π i + 1 ∈ A − and ∣ π i ∣ ≥ ∣ π i + 1 ∣ , q i if π i ∈ A + and π i + 1 ∈ A − , 1 if π i ∈ A − and π i + 1 ∈ A + . c_i(\pi)+d_i(\pi)=
\begin{cases}
q^i & \text{if }\pi_i,\pi_{i+1}\in\mathcal{A}_+ \text{ and }\pi_i>\pi_{i+1},\\
1 & \text{if }\pi_i,\pi_{i+1}\in\mathcal{A}_+ \text{ and }\pi_i\leq\pi_{i+1},\\
1 & \text{if }\pi_i,\pi_{i+1}\in\mathcal{A}_- \text{ and }|\pi_i|<|\pi_{i+1}|,\\
q^i & \text{if }\pi_i,\pi_{i+1}\in\mathcal{A}_- \text{ and }|\pi_i|\geq|\pi_{i+1}|,\\
q^i & \text{if }\pi_i\in\mathcal{A}_+ \text{ and }\pi_{i+1}\in\mathcal{A}_-,\\
1 & \text{if }\pi_i\in\mathcal{A}_- \text{ and }\pi_{i+1}\in\mathcal{A}_+.
\end{cases} c i ( π ) + d i ( π ) = ⎩ ⎨ ⎧ q i 1 1 q i q i 1 if π i , π i + 1 ∈ A + and π i > π i + 1 , if π i , π i + 1 ∈ A + and π i ≤ π i + 1 , if π i , π i + 1 ∈ A − and ∣ π i ∣ < ∣ π i + 1 ∣ , if π i , π i + 1 ∈ A − and ∣ π i ∣ ≥ ∣ π i + 1 ∣ , if π i ∈ A + and π i + 1 ∈ A − , if π i ∈ A − and π i + 1 ∈ A + .
This implies that c i ( π ) + d i ( π ) c_i(\pi)+d_i(\pi) c i ( π ) + d i ( π ) is equal to q i q^i q i if i ∈ Des ∗ ( π ) i\in\operatorname{Des}^{*}(\pi) i ∈ Des ∗ ( π ) and 1 1 1 otherwise. Therefore, (19) follows from (20) .
For a composition β = ( β 1 , … , β d ) ⊨ n \beta=(\beta_1,\ldots,\beta_d)\vDash n β = ( β 1 , … , β d ) ⊨ n , let B ( β ) B(\beta) B ( β ) be the set of words π ∈ W n ( A ± ) \pi\in W_n(\mathcal{A}_{\pm}) π ∈ W n ( A ± ) in which i i i and i ˉ \bar{i} i ˉ occur β i \beta_i β i times in total for each 1 ≤ i ≤ d 1\leq i\leq d 1 ≤ i ≤ d . Equivalently, π ∈ B ( β ) \pi\in B(\beta) π ∈ B ( β ) if and only if x ∣ π ∣ = x β x^{|\pi|}=x^\beta x ∣ π ∣ = x β .
Lemma 3.5. For β ⊨ n \beta\vDash n β ⊨ n , we have
∑ π ∈ B ( β ) q maj ( π ) ( − t ) neg ( π ) = ∑ π ∈ B ( β ) q maj ∗ ( π ) ( − t ) neg ( π ) . \sum_{\pi\in B(\beta)}q^{\operatorname{maj}(\pi)}(-t)^{\operatorname{neg}(\pi)}
=
\sum_{\pi\in B(\beta)}q^{\operatorname{maj}^{*}(\pi)}(-t)^{\operatorname{neg}(\pi)}. π ∈ B ( β ) ∑ q maj ( π ) ( − t ) neg ( π ) = π ∈ B ( β ) ∑ q maj ∗ ( π ) ( − t ) neg ( π ) .
Proof. Let β = ( β 1 , … , β d ) ⊨ n \beta=(\beta_1,\ldots,\beta_d)\vDash n β = ( β 1 , … , β d ) ⊨ n . For a word π + \pi^+ π + on A + \mathcal{A}_+ A + and a word π − \pi^- π − on A − \mathcal{A}_- A − , a shuffle of π + \pi^+ π + and π − \pi^- π − is a word π \pi π on A ± \mathcal{A}_{\pm} A ± such that the subword of π \pi π consisting of letters in A + \mathcal{A}_+ A + (resp. A − \mathcal{A}_- A − ) is equal to π + \pi^+ π + (resp. π − \pi^- π − ). Let sh ( π + , π − ) \operatorname{sh}(\pi^+,\pi^-) sh ( π + , π − ) denote the set of such shuffles. Let V ( β ) V(\beta) V ( β ) be the set of pairs ( π + , π − ) (\pi^+,\pi^-) ( π + , π − ) of words on A + \mathcal{A}_+ A + and A − \mathcal{A}_- A − , respectively, such that, for each 1 ≤ i ≤ d 1\leq i\leq d 1 ≤ i ≤ d , the total number of occurrences of i i i in π + \pi^+ π + and i ˉ \bar{i} i ˉ in π − \pi^- π − is β i \beta_i β i . Then we have
∑ π ∈ B ( β ) q maj ( π ) ( − t ) neg ( π ) = ∑ ( π + , π − ) ∈ V ( β ) ( − t ) k ∑ π ∈ sh ( π + , π − ) q maj ( π ) , (21) \sum_{\pi\in B(\beta)}q^{\operatorname{maj}(\pi)}(-t)^{\operatorname{neg}(\pi)}
=
\sum_{(\pi^+,\pi^-)\in V(\beta)}(-t)^k
\sum_{\pi\in\operatorname{sh}(\pi^+,\pi^-)}q^{\operatorname{maj}(\pi)}, \tag*{(21)} π ∈ B ( β ) ∑ q maj ( π ) ( − t ) neg ( π ) = ( π + , π − ) ∈ V ( β ) ∑ ( − t ) k π ∈ sh ( π + , π − ) ∑ q maj ( π ) , (21)
where k k k is the total number of letters in π − \pi^- π − . Since the identity (21) with each maj \operatorname{maj} maj replaced by maj ∗ \operatorname{maj}^{*} maj ∗ also holds, it suffices to show that for ( π + , π − ) ∈ V ( β ) (\pi^+,\pi^-)\in V(\beta) ( π + , π − ) ∈ V ( β ) , we have
∑ π ∈ sh ( π + , π − ) q maj ( π ) = ∑ π ∈ sh ( π + , π − ) q maj ∗ ( π ) . (22) \sum_{\pi\in\operatorname{sh}(\pi^+,\pi^-)}q^{\operatorname{maj}(\pi)}
=
\sum_{\pi\in\operatorname{sh}(\pi^+,\pi^-)}q^{\operatorname{maj}^{*}(\pi)}. \tag*{(22)} π ∈ sh ( π + , π − ) ∑ q maj ( π ) = π ∈ sh ( π + , π − ) ∑ q maj ∗ ( π ) . (22)
Let σ + \sigma^+ σ + and σ − \sigma^- σ − (resp. τ + \tau^+ τ + and τ − \tau^- τ − ) be the words such that the concatenation σ + σ − \sigma^+\sigma^- σ + σ − (resp. τ + τ − \tau^+\tau^- τ + τ − ) is the standardization of π + π − \pi^+\pi^- π + π − with respect to the order ≺ \prec ≺ in (11) (resp. ≺ ∗ \prec^{*} ≺ ∗ in (16) ). Then by the definitions of maj and maj* and by the well-known fact [19 Ch. 3, Exercise 162(b)] , we have
∑ π ∈ sh ( π + , π − ) q maj ( π ) = ∑ σ ∈ sh ( σ + , σ − ) q maj ( σ ) = q maj ( σ + ) + maj ( σ − ) [ n k ] q \sum_{\pi\in\operatorname{sh}(\pi^+,\pi^-)}q^{\operatorname{maj}(\pi)}
=
\sum_{\sigma\in\operatorname{sh}(\sigma^+,\sigma^-)}q^{\operatorname{maj}(\sigma)}
=
q^{\operatorname{maj}(\sigma^+)+\operatorname{maj}(\sigma^-)}
\begin{bmatrix} n \\ k \end{bmatrix}_q π ∈ sh ( π + , π − ) ∑ q maj ( π ) = σ ∈ sh ( σ + , σ − ) ∑ q maj ( σ ) = q maj ( σ + ) + maj ( σ − ) [ n k ] q
∑ π ∈ sh ( π + , π − ) q maj ∗ ( π ) = ∑ τ ∈ sh ( τ + , τ − ) q maj ( τ ) = q maj ( τ + ) + maj ( τ − ) [ n k ] q . \sum_{\pi\in\operatorname{sh}(\pi^+,\pi^-)}q^{\operatorname{maj}^*(\pi)}
=
\sum_{\tau\in\operatorname{sh}(\tau^+,\tau^-)}q^{\operatorname{maj}(\tau)}
=
q^{\operatorname{maj}(\tau^+)+\operatorname{maj}(\tau^-)}
\begin{bmatrix} n \\ k \end{bmatrix}_q. π ∈ sh ( π + , π − ) ∑ q maj ∗ ( π ) = τ ∈ sh ( τ + , τ − ) ∑ q maj ( τ ) = q maj ( τ + ) + maj ( τ − ) [ n k ] q .
Since the orderings (2.6) and (3.5) agree on A + \mathcal{A}_+ A + and on A − \mathcal{A}_- A − , we have maj ( σ + ) = maj ( τ + ) \operatorname{maj}(\sigma^+)=\operatorname{maj}(\tau^+) maj ( σ + ) = maj ( τ + ) and maj ( σ − ) = maj ( τ − ) \operatorname{maj}(\sigma^-)=\operatorname{maj}(\tau^-) maj ( σ − ) = maj ( τ − ) . Therefore, we obtain (3.11), which completes the proof. □
Now we prove (3.4).
First proof of (3.4). For every β ⊨ n \beta\models n β ⊨ n , (3.6) and Lemmas 3.4 and 3.5 give
[ x β ] J ( n ) = ∑ π ∈ B ( β ) q maj ( π ) ( − t ) neg ( π ) = ∑ π ∈ B ( β ) q maj ∗ ( π ) ( − t ) neg ( π ) = [ x β ] Y ( n ) . [x^\beta]J_{(n)}=\sum_{\pi\in B(\beta)}q^{\operatorname{maj}(\pi)}(-t)^{\operatorname{neg}(\pi)}
=\sum_{\pi\in B(\beta)}q^{\operatorname{maj}^*(\pi)}(-t)^{\operatorname{neg}(\pi)}
=[x^\beta]Y_{(n)}. [ x β ] J ( n ) = π ∈ B ( β ) ∑ q maj ( π ) ( − t ) neg ( π ) = π ∈ B ( β ) ∑ q maj ∗ ( π ) ( − t ) neg ( π ) = [ x β ] Y ( n ) .
Since both sides of (3.4) are symmetric functions, the identity follows. □
Recursive proof
In this subsection, we give a recursive proof of (3.4). We first recall the generating function of J ( n ) ( x ; q , t ) J_{(n)}(x;q,t) J ( n ) ( x ; q , t ) .
Lemma 3.6. We have
∑ n ≥ 0 J ( n ) ( x ; q , t ) ( q ; q ) n u n = ∏ i ≥ 1 A ( x i u ) , A ( v ) : = ∑ m ≥ 0 ( t ; q ) m ( q ; q ) m v m . \sum_{n\geq0}\frac{J_{(n)}(x;q,t)}{(q;q)_n}u^n
=
\prod_{i\geq1}A(x_i u),\qquad
A(v):=\sum_{m\geq0}\frac{(t;q)_m}{(q;q)_m}v^m. n ≥ 0 ∑ ( q ; q ) n J ( n ) ( x ; q , t ) u n = i ≥ 1 ∏ A ( x i u ) , A ( v ) := m ≥ 0 ∑ ( q ; q ) m ( t ; q ) m v m .
Proof. Let g n ( x ; q , t ) g_n(x;q,t) g n ( x ; q , t ) be the symmetric functions given by [15 Ch. VI, (2.8), p. 311]
∑ n ≥ 0 g n ( x ; q , t ) u n = ∏ i ≥ 1 ( t x i u ; q ) ∞ ( x i u ; q ) ∞ . \sum_{n\geq0}g_n(x;q,t)u^n
=
\prod_{i\geq1}\frac{(t x_i u;q)_\infty}{(x_i u;q)_\infty}. n ≥ 0 ∑ g n ( x ; q , t ) u n = i ≥ 1 ∏ ( x i u ; q ) ∞ ( t x i u ; q ) ∞ .
By [15 Ch. VI, (4.9), p. 323] and [15 Ch. VI, (8.1) and (8.3), p. 352] , we have
J ( n ) ( x ; q , t ) = ( t ; q ) n P ( n ) ( x ; q , t ) = ( q ; q ) n g n ( x ; q , t ) . J_{(n)}(x;q,t)=(t;q)_nP_{(n)}(x;q,t)=(q;q)_n g_n(x;q,t). J ( n ) ( x ; q , t ) = ( t ; q ) n P ( n ) ( x ; q , t ) = ( q ; q ) n g n ( x ; q , t ) .
The q q q -binomial identity [15 Ch. VI, §2, Example 1, p. 314] gives
A ( v ) = ∑ m ≥ 0 ( t ; q ) m ( q ; q ) m v m = ( t v ; q ) ∞ ( v ; q ) ∞ . A(v)=\sum_{m\geq0}\frac{(t;q)_m}{(q;q)_m}v^m=\frac{(tv;q)_\infty}{(v;q)_\infty}. A ( v ) = m ≥ 0 ∑ ( q ; q ) m ( t ; q ) m v m = ( v ; q ) ∞ ( t v ; q ) ∞ .
Combining these identities, we obtain
∑ n ≥ 0 J ( n ) ( x ; q , t ) ( q ; q ) n u n = ∑ n ≥ 0 g n ( x ; q , t ) u n = ∏ i ≥ 1 A ( x i u ) , \sum_{n\geq0}\frac{J_{(n)}(x;q,t)}{(q;q)_n}u^n
=
\sum_{n\geq0}g_n(x;q,t)u^n
=
\prod_{i\geq1}A(x_i u), n ≥ 0 ∑ ( q ; q ) n J ( n ) ( x ; q , t ) u n = n ≥ 0 ∑ g n ( x ; q , t ) u n = i ≥ 1 ∏ A ( x i u ) ,
as desired. □
Second proof of (3.4). We first derive a recurrence for Y ( n ) Y_{(n)} Y ( n ) by considering the last part of each composition of n ≥ 1 n\geq1 n ≥ 1 . For 1 ≤ k ≤ n 1\leq k\leq n 1 ≤ k ≤ n , write α = ( β , k ) \alpha=(\beta,k) α = ( β , k ) with β ⊨ n − k \beta\models n-k β ⊨ n − k . By (3.1) and (3.2), we have
wt ( n ) ( α ) = wt ( n − k ) ( β ) ( 1 − q n − k t k ) ∏ j = n − k + 1 n − 1 ( q j − 1 ) = ( − 1 ) k − 1 ( 1 − q n − k t k ) ( q ; q ) n − 1 ( q ; q ) n − k wt ( n − k ) ( β ) . \begin{aligned}
\operatorname{wt}_{(n)}(\alpha)
=
\operatorname{wt}_{(n-k)}(\beta)(1-q^{n-k}t^k)
\prod_{j=n-k+1}^{n-1}(q^j-1) \\
&=(-1)^{k-1}(1-q^{n-k}t^k)\frac{(q;q)_{n-1}}{(q;q)_{n-k}}\operatorname{wt}_{(n-k)}(\beta).
\end{aligned} wt ( n ) ( α ) = wt ( n − k ) ( β ) ( 1 − q n − k t k ) j = n − k + 1 ∏ n − 1 ( q j − 1 ) = ( − 1 ) k − 1 ( 1 − q n − k t k ) ( q ; q ) n − k ( q ; q ) n − 1 wt ( n − k ) ( β ) .
Summing over β \beta β and k k k , using e α = e β e k e_\alpha=e_\beta e_k e α = e β e k , and dividing by ( q ; q ) n − 1 (q;q)_{n-1} ( q ; q ) n − 1 , we obtain
Y ( n ) ( q ; q ) n − 1 = ∑ k = 1 n ( − 1 ) k − 1 ( 1 − q n − k t k ) e k Y ( n − k ) ( q ; q ) n − k . (23) \frac{Y_{(n)}}{(q;q)_{n-1}}
=
\sum_{k=1}^{n}(-1)^{k-1}(1-q^{n-k}t^k)e_k\frac{Y_{(n-k)}}{(q;q)_{n-k}}. \tag*{(23)} ( q ; q ) n − 1 Y ( n ) = k = 1 ∑ n ( − 1 ) k − 1 ( 1 − q n − k t k ) e k ( q ; q ) n − k Y ( n − k ) . (23)
On the other hand, for the function A ( v ) A(v) A ( v ) in Lemma 3.6, we have
( 1 − v ) A ( v ) = ( 1 − t v ) A ( q v ) . (1-v)A(v)=(1-tv)A(qv). ( 1 − v ) A ( v ) = ( 1 − t v ) A ( q v ) .
Multiplying these identities over i ≥ 1 i \ge1 i ≥ 1 with v = x i u v=x_i u v = x i u and applying Lemma 3.6 gives
( ∏ i ≥ 1 ( 1 − x i u ) ) ∑ n ≥ 0 J ( n ) ( q ; q ) n u n = ( ∏ i ≥ 1 ( 1 − t x i u ) ) ∑ n ≥ 0 J ( n ) ( q ; q ) n ( q u ) n . \left(\prod_{i\geq1}(1-x_i u)\right)\sum_{n\geq0}\frac{J^{(n)}}{(q;q)_n}u^n=\left(\prod_{i\geq1}(1-tx_i u)\right)\sum_{n\geq0}\frac{J^{(n)}}{(q;q)_n}(qu)^n. ( i ≥ 1 ∏ ( 1 − x i u ) ) n ≥ 0 ∑ ( q ; q ) n J ( n ) u n = ( i ≥ 1 ∏ ( 1 − t x i u ) ) n ≥ 0 ∑ ( q ; q ) n J ( n ) ( q u ) n .
Using ∏ i ≥ 1 ( 1 − x i u ) = ∑ k ≥ 0 ( − 1 ) k e k u k \prod_{i\geq1}(1-x_i u)=\sum_{k\geq0}(-1)^k e_k u^k ∏ i ≥ 1 ( 1 − x i u ) = ∑ k ≥ 0 ( − 1 ) k e k u k and comparing coefficients of u n u^n u n , we obtain, for n ≥ 1 n\geq1 n ≥ 1 ,
J ( n ) ( q ; q ) n − 1 = ∑ k = 1 n ( − 1 ) k − 1 ( 1 − q n − k t k ) e k J ( n − k ) ( q ; q ) n − k . \frac{J^{(n)}}{(q;q)_{n-1}}=\sum_{k=1}^{n}(-1)^{k-1}(1-q^{n-k}t^k)e_k\frac{J^{(n-k)}}{(q;q)_{n-k}}. ( q ; q ) n − 1 J ( n ) = k = 1 ∑ n ( − 1 ) k − 1 ( 1 − q n − k t k ) e k ( q ; q ) n − k J ( n − k ) .
This is the same recurrence as (3.12). Since J ( 0 ) = Y ( 0 ) = 1 J_{(0)}=Y_{(0)}=1 J ( 0 ) = Y ( 0 ) = 1 , we obtain J ( n ) = Y ( n ) J_{(n)}=Y_{(n)} J ( n ) = Y ( n ) for all n ≥ 0 n\geq0 n ≥ 0 by induction. □
Hook case
In this section, we give an e e e -expansion of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) when μ \mu μ is a hook using the one-row case in Section 3. As a consequence, we prove Conjecture 1.1 in this case.
Main formula
We first extend the definition of the weight wt ( n ) ( α ) \operatorname{wt}_{(n)}(\alpha) wt ( n ) ( α ) given in the previous section. Recall that χ ( P ) = 1 \chi(P)=1 χ ( P ) = 1 if P P P holds and χ ( P ) = 0 \chi(P)=0 χ ( P ) = 0 otherwise.
Definition 4.1. For a hook μ = ( r , 1 s ) \mu=(r,1^s) μ = ( r , 1 s ) with r ≥ 1 r\geq1 r ≥ 1 , s ≥ 0 s\geq0 s ≥ 0 , and r + s = n r+s=n r + s = n , and a composition α = ( α 1 , … , α k ) \alpha=(\alpha_1,\ldots,\alpha_k) α = ( α 1 , … , α k ) of n n n , define
wt ( r , 1 s ) ( α ) = χ ( α k > s ) ( t ; t ) s ∏ i = 1 r wt ( i , α ) . \operatorname{wt}_{(r,1^s)}(\alpha)=\chi(\alpha_k>s)(t;t)_s\prod_{i=1}^{r}\operatorname{wt}(i,\alpha). wt ( r , 1 s ) ( α ) = χ ( α k > s ) ( t ; t ) s i = 1 ∏ r wt ( i , α ) .
The e e e -coefficients of J ( r , 1 s ) ( x ; q , t ) J_{(r,1^s)}(x;q,t) J ( r , 1 s ) ( x ; q , t ) have a combinatorial description in terms of these weights, as follows.
Theorem 4.2. Let n , r ≥ 1 n,r\geq1 n , r ≥ 1 and s ≥ 0 s\geq0 s ≥ 0 be integers with r + s = n r+s=n r + s = n . Then we have
J ( r , 1 s ) ( x ; q , t ) = ∑ λ ⊢ n ( ∑ α ∼ λ wt ( r , 1 s ) ( α ) ) e λ . J_{(r,1^s)}(x;q,t)=\sum_{\lambda\vdash n}\left(\sum_{\alpha\sim\lambda}\operatorname{wt}_{(r,1^s)}(\alpha)\right)e_\lambda. J ( r , 1 s ) ( x ; q , t ) = λ ⊢ n ∑ ( α ∼ λ ∑ wt ( r , 1 s ) ( α ) ) e λ . where the second sum is over all compositions α \alpha α with α ∼ λ \alpha\sim\lambda α ∼ λ .
This theorem will be proved in the next subsection. In what follows, we provide an equivalent description of (4.2) using ribbon tilings.
A ribbon is a connected skew shape containing no 2 × 2 2\times2 2 × 2 block. The content of a cell ( i , j ) (i,j) ( i , j ) is j − i j-i j − i , and the head of a ribbon is its cell of smallest content.
A ribbon tiling of a partition ν \nu ν is a tiling of its diagram by ribbons. Its type is the partition formed by the ribbon sizes. Write RT ( ν ) \operatorname{RT}(\nu) RT ( ν ) for the set of ribbon tilings of ν \nu ν , and RT ( ν , λ ) \operatorname{RT}(\nu,\lambda) RT ( ν , λ ) for those of type λ \lambda λ . A ribbon tiling is special if each ribbon contains a cell in the first column of ν \nu ν . Denote the corresponding sets by SRT ( ν ) \operatorname{SRT}(\nu) SRT ( ν ) and SRT ( ν , λ ) \operatorname{SRT}(\nu,\lambda) SRT ( ν , λ ) .
For T ∈ RT ( ν ) T\in\operatorname{RT}(\nu) T ∈ RT ( ν ) , define
wt ( T ) = ∏ c ∈ ν wt ( c , T , ν ) , wt ( c , T , ν ) = { 1 − q leg ( c ) t ∣ θ ∣ , if c is the head of a ribbon θ ∈ T , q leg ( c ) − t arm ′ ( c ) , otherwise. \operatorname{wt}(T)=\prod_{c\in\nu}\operatorname{wt}(c,T,\nu),\qquad
\operatorname{wt}(c,T,\nu)=
\begin{cases}
1-q^{\operatorname{leg}(c)}t^{|\theta|}, & \text{if $c$ is the head of a ribbon $\theta\in T$,}\\
q^{\operatorname{leg}(c)}-t^{\operatorname{arm}'(c)}, & \text{otherwise.}
\end{cases} wt ( T ) = c ∈ ν ∏ wt ( c , T , ν ) , wt ( c , T , ν ) = { 1 − q leg ( c ) t ∣ θ ∣ , q leg ( c ) − t arm ′ ( c ) , if c is the head of a ribbon θ ∈ T , otherwise.
Here the leg and coarm lengths are measured in ν \nu ν .
Theorem 4.3. Let μ = ( r , 1 s ) \mu=(r,1^s) μ = ( r , 1 s ) , where r ≥ 1 r\geq1 r ≥ 1 , s ≥ 0 s\geq0 s ≥ 0 , and r + s = n r+s=n r + s = n . Then
J μ ( x ; q , t ) = ∑ λ ⊢ n ( ∑ T ∈ SRT ( μ ′ , λ ) wt ( T ) ) e λ . J_\mu(x;q,t)=\sum_{\lambda\vdash n}\left(\sum_{T\in\operatorname{SRT}(\mu',\lambda)}\operatorname{wt}(T)\right)e_\lambda. J μ ( x ; q , t ) = λ ⊢ n ∑ T ∈ SRT ( μ ′ , λ ) ∑ wt ( T ) e λ .
This follows directly from Theorem 4.2 by interpreting each composition with nonzero weight as a special ribbon tiling of μ ′ \mu' μ ′ with the same weight.
Proof of thm:main
Let
Y ( r , 1 s ) = Y ( r , 1 s ) ( x ; q , t ) = ∑ α ⊨ n wt ( r , 1 s ) ( α ) e α . (24) Y_{(r,1^{s})}=Y_{(r,1^{s})}(x;q,t)=\sum_{\alpha\models n}\operatorname{wt}_{(r,1^{s})}(\alpha)e_{\alpha}. \tag*{(24)} Y ( r , 1 s ) = Y ( r , 1 s ) ( x ; q , t ) = α ⊨ n ∑ wt ( r , 1 s ) ( α ) e α . (24)
Then Theorem 4.2 is equivalent to
J ( r , 1 s ) ( x ; q , t ) = Y ( r , 1 s ) ( x ; q , t ) . (25) J_{(r,1^{s})}(x;q,t)=Y_{(r,1^{s})}(x;q,t). \tag*{(25)} J ( r , 1 s ) ( x ; q , t ) = Y ( r , 1 s ) ( x ; q , t ) . (25)
It is known [15 p. 363] that for r , s ≥ 1 r,s\geq1 r , s ≥ 1 ,
( 1 − q r t s ) J ( r ) J ( 1 s ) = ( 1 − t s ) J ( r + 1 , 1 s − 1 ) + ( 1 − q r ) J ( r , 1 s ) . (26) (1-q^{r}t^{s})J_{(r)}J_{(1^{s})}=(1-t^{s})J_{(r+1,1^{s-1})}+(1-q^{r})J_{(r,1^{s})}. \tag*{(26)} ( 1 − q r t s ) J ( r ) J ( 1 s ) = ( 1 − t s ) J ( r + 1 , 1 s − 1 ) + ( 1 − q r ) J ( r , 1 s ) . (26)
We will show that Y ( r , 1 s ) ( x ; q , t ) Y_{(r,1^{s})}(x;q,t) Y ( r , 1 s ) ( x ; q , t ) satisfies the same recurrence relation and initial conditions. Note that if ( r , 1 s ) = ( 1 , 1 n − 1 ) = ( 1 n ) (r,1^{s})=(1,1^{n-1})=(1^{n}) ( r , 1 s ) = ( 1 , 1 n − 1 ) = ( 1 n ) , then by definition we have wt ( 1 , 1 n − 1 ) ( α ) = 0 \operatorname{wt}_{(1,1^{n-1})}(\alpha)=0 wt ( 1 , 1 n − 1 ) ( α ) = 0 unless α = ( n ) \alpha=(n) α = ( n ) . Thus,
Y ( 1 n ) = ( t ; t ) n e n . (27) Y_{(1^{n})}=(t;t)_{n}e_{n}. \tag*{(27)} Y ( 1 n ) = ( t ; t ) n e n . (27)
Lemma 4.4. For integers r , s ≥ 1 r,s\geq1 r , s ≥ 1 , we have
( 1 − q r t s ) Y ( r ) Y ( 1 s ) = ( 1 − t s ) Y ( r + 1 , 1 s − 1 ) + ( 1 − q r ) Y ( r , 1 s ) . (28) (1-q^{r}t^{s})Y_{(r)}Y_{(1^{s})}=(1-t^{s})Y_{(r+1,1^{s-1})}+(1-q^{r})Y_{(r,1^{s})}. \tag*{(28)} ( 1 − q r t s ) Y ( r ) Y ( 1 s ) = ( 1 − t s ) Y ( r + 1 , 1 s − 1 ) + ( 1 − q r ) Y ( r , 1 s ) . (28)
Proof. Let n = r + s n=r+s n = r + s . By (4.4), the right-hand side of (4.8) is equal to
∑ α ⊨ n ( ( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) + ( 1 − q r ) wt ( r , 1 s ) ( α ) ) e α . (29) \sum_{\alpha\models n}\left((1-t^{s})\operatorname{wt}_{(r+1,1^{s-1})}(\alpha)+(1-q^{r})\operatorname{wt}_{(r,1^{s})}(\alpha)\right)e_{\alpha}. \tag*{(29)} α ⊨ n ∑ ( ( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) + ( 1 − q r ) wt ( r , 1 s ) ( α ) ) e α . (29)
Fix α = ( α 1 , … , α k ) ⊨ n \alpha=(\alpha_{1},\ldots,\alpha_{k})\vDash n α = ( α 1 , … , α k ) ⊨ n . If α k ≤ s − 1 \alpha_{k}\leq s-1 α k ≤ s − 1 , then both weights in (4.9) vanish. If α k ≥ s + 1 \alpha_{k}\geq s+1 α k ≥ s + 1 , then r > α 1 + ⋯ + α k − 1 r>\alpha_{1}+\cdots+\alpha_{k-1} r > α 1 + ⋯ + α k − 1 , so wt ( r + 1 , α ) = q r − 1 \operatorname{wt}_{(r+1,\alpha)}=q^{r}-1 wt ( r + 1 , α ) = q r − 1 . Hence, the summand in (4.9) also vanishes:
( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) + ( 1 − q r ) wt ( r , 1 s ) ( α ) = ( t ; t ) s ( ( q r − 1 ) + ( 1 − q r ) ) ∏ i = 1 r wt ( i , α ) = 0. (1-t^{s})\operatorname{wt}_{(r+1,1^{s-1})}(\alpha)+(1-q^{r})\operatorname{wt}_{(r,1^{s})}(\alpha)=(t;t)_{s}\left((q^{r}-1)+(1-q^{r})\right)\prod_{i=1}^{r}\operatorname{wt}(i,\alpha)=0. ( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) + ( 1 − q r ) wt ( r , 1 s ) ( α ) = ( t ; t ) s ( ( q r − 1 ) + ( 1 − q r ) ) i = 1 ∏ r wt ( i , α ) = 0.
Therefore, we can restrict the sum in (4.9) to the compositions with α k = s \alpha_{k}=s α k = s .
Suppose α = ( β , s ) \alpha=(\beta,s) α = ( β , s ) with β ⊨ r \beta\vDash r β ⊨ r . Then wt ( r , 1 s ) ( α ) = 0 \operatorname{wt}_{(r,1^{s})}(\alpha)=0 wt ( r , 1 s ) ( α ) = 0 and
( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) = ( 1 − q r t s ) ( t ; t ) s wt ( r ) ( β ) . (1-t^{s})\operatorname{wt}_{(r+1,1^{s-1})}(\alpha)=(1-q^{r}t^{s})(t;t)_{s}\operatorname{wt}_{(r)}(\beta). ( 1 − t s ) wt ( r + 1 , 1 s − 1 ) ( α ) = ( 1 − q r t s ) ( t ; t ) s wt ( r ) ( β ) .
This, together with e α = e β e s e_{\alpha}=e_{\beta}e_{s} e α = e β e s and (4.7), implies that (4.9) is equal to
( 1 − q r t s ) ( t ; t ) s e s ∑ β ⊨ r wt ( r ) ( β ) e β = ( 1 − q r t s ) Y ( 1 s ) Y ( r ) . (1-q^{r}t^{s})(t;t)_{s}e_{s}\sum_{\beta\vDash r}\operatorname{wt}_{(r)}(\beta)e_{\beta}=(1-q^{r}t^{s})Y_{(1^{s})}Y_{(r)}. ( 1 − q r t s ) ( t ; t ) s e s β ⊨ r ∑ wt ( r ) ( β ) e β = ( 1 − q r t s ) Y ( 1 s ) Y ( r ) .
This is the left-hand side of (4.8), completing the proof.
Now we are ready to prove Theorem 4.2.
Proof of Theorem 4.2. By Theorem 3.3, we have
J ( k ) ( x ; q , t ) = Y ( k ) ( x ; q , t ) for all k ≥ 1. J_{(k)}(x;q,t)=Y_{(k)}(x;q,t)\qquad\text{for all }k\geq1. J ( k ) ( x ; q , t ) = Y ( k ) ( x ; q , t ) for all k ≥ 1.
By [15 p. 339] and (4.7), we have
J ( 1 k ) ( x ; q , t ) = ( t ; t ) k e k = Y ( 1 k ) ( x ; q , t ) for all k ≥ 1. J_{(1^{k})}(x;q,t)=(t;t)_{k}e_{k}=Y_{(1^{k})}(x;q,t)\qquad\text{for all }k\geq1. J ( 1 k ) ( x ; q , t ) = ( t ; t ) k e k = Y ( 1 k ) ( x ; q , t ) for all k ≥ 1.
Thus, using (4.6), (4.8), and the above results, we obtain (4.5) by induction on s s s .
Proof of conj:epositivity for hooks
In this subsection we prove the e e e -positivity in Conjecture 1.1 when μ \mu μ is a hook.
Let μ = ( r , 1 s ) \mu=(r,1^{s}) μ = ( r , 1 s ) with n = r + s n=r+s n = r + s , and fix an integer k ≥ 1 k\geq1 k ≥ 1 . For a composition α = ( α 1 , … , α ℓ ) ⊨ n \alpha=(\alpha_1,\ldots,\alpha_\ell)\vDash n α = ( α 1 , … , α ℓ ) ⊨ n and 1 ≤ i ≤ r 1\leq i\leq r 1 ≤ i ≤ r , define
wt k ( i , α ) = { [ k α j ] q − [ i − 1 ] q if i = 1 + α 1 + ⋯ + α j − 1 for some 1 ≤ j ≤ ℓ , [ i − 1 ] q otherwise . \operatorname{wt}_k(i,\alpha)=
\begin{cases}
[k\alpha_j]_q-[i-1]_q & \text{if }i=1+\alpha_1+\cdots+\alpha_{j-1}\text{ for some }1\leq j\leq\ell,\\
[i-1]_q & \text{otherwise}.
\end{cases} wt k ( i , α ) = { [ k α j ] q − [ i − 1 ] q [ i − 1 ] q if i = 1 + α 1 + ⋯ + α j − 1 for some 1 ≤ j ≤ ℓ , otherwise .
Then Theorem 4.2 with t = q − k t=q^{-k} t = q − k gives
q k ( n + ( s + 1 2 ) ) J ( r , 1 s ) ( x ; q , q − k ) ( q − 1 ) n = ( ∏ j = 1 s [ k j ] q ) ∑ α ⊨ n ℓ ( α ) > s ( ∏ i = 1 r wt k ( i , α ) ) e α . (30) q^{k\left(n+\binom{s+1}{2}\right)}
\frac{J_{(r,1^s)}(x;q,q^{-k})}{(q-1)^n}
=
\left(\prod_{j=1}^{s}[kj]_q\right)
\sum_{\substack{\alpha\vDash n\\ \ell(\alpha)>s}}
\left(\prod_{i=1}^{r}\operatorname{wt}_k(i,\alpha)\right)e_\alpha. \tag*{(30)} q k ( n + ( 2 s + 1 ) ) ( q − 1 ) n J ( r , 1 s ) ( x ; q , q − k ) = ( j = 1 ∏ s [ k j ] q ) α ⊨ n ℓ ( α ) > s ∑ ( i = 1 ∏ r wt k ( i , α ) ) e α . (30)
If k ≥ r − 1 k\geq r-1 k ≥ r − 1 , then
[ k α j ] q − [ i − 1 ] q = q i − 1 [ k α j − i + 1 ] q ∈ Z ≥ 0 [ q ] . [k\alpha_j]_q-[i-1]_q=q^{i-1}[k\alpha_j-i+1]_q\in\mathbb{Z}_{\geq0}[q]. [ k α j ] q − [ i − 1 ] q = q i − 1 [ k α j − i + 1 ] q ∈ Z ≥ 0 [ q ] .
Thus, Conjecture 1.1 holds in this case. For an arbitrary k ≥ 1 k\geq1 k ≥ 1 , we first establish the following one-row formula, which is different from Theorem 3.3 for t = q − k t=q^{-k} t = q − k . Throughout this section, the notation ∑ c 0 + ⋯ + c k − 1 = n \sum_{c_0+\cdots+c_{k-1}=n} ∑ c 0 + ⋯ + c k − 1 = n means the sum over all nonnegative integers c 0 , … , c k − 1 c_0,\ldots,c_{k-1} c 0 , … , c k − 1 satisfying c 0 + ⋯ + c k − 1 = n c_0+\cdots+c_{k-1}=n c 0 + ⋯ + c k − 1 = n .
Proposition 4.5. Let n , k ≥ 1 n,k\geq1 n , k ≥ 1 be integers. Then
q k n J ( n ) ( x ; q , q − k ) ( q − 1 ) n = [ n ] q ! ∑ c 0 + ⋯ + c k − 1 = n q ∑ i = 0 k − 1 i c i ∏ i = 0 k − 1 e c i . (31) q^{kn}\frac{J_{(n)}(x;q,q^{-k})}{(q-1)^n}
=
[n]_q!\sum_{c_0+\cdots+c_{k-1}=n}q^{\sum_{i=0}^{k-1}ic_i}\prod_{i=0}^{k-1}e_{c_i}. \tag*{(31)} q k n ( q − 1 ) n J ( n ) ( x ; q , q − k ) = [ n ] q ! c 0 + ⋯ + c k − 1 = n ∑ q ∑ i = 0 k − 1 i c i i = 0 ∏ k − 1 e c i . (31)
In particular, the left-hand side has e e e -coefficients in Z ≥ 0 [ q ] \mathbb{Z}_{\geq0}[q] Z ≥ 0 [ q ] .
Proof. Fix k ≥ 1 k\geq1 k ≥ 1 , and let
H n = q k n J ( n ) ( x ; q , q − k ) ( q − 1 ) n [ n ] q ! , F n = ∑ c 0 + ⋯ + c k − 1 = n q ∑ i = 0 k − 1 i c i ∏ i = 0 k − 1 e c i . (32) H_n=\frac{q^{kn}J_{(n)}(x;q,q^{-k})}{(q-1)^n[n]_q!},\qquad
F_n=\sum_{c_0+\cdots+c_{k-1}=n}q^{\sum_{i=0}^{k-1}ic_i}\prod_{i=0}^{k-1}e_{c_i}. \tag*{(32)} H n = ( q − 1 ) n [ n ] q ! q k n J ( n ) ( x ; q , q − k ) , F n = c 0 + ⋯ + c k − 1 = n ∑ q ∑ i = 0 k − 1 i c i i = 0 ∏ k − 1 e c i . (32)
Then we must show that H n = F n H_n=F_n H n = F n .
We first derive a recurrence for H n H_n H n . By (30) with r = n r=n r = n and s = 0 s=0 s = 0 , we have
[ n ] q ! H n = ∑ α ⊨ n ( ∏ i = 1 n wt k ( i , α ) ) e α . (33) [n]_q!H_n=\sum_{\alpha\vDash n}\left(\prod_{i=1}^{n}\operatorname{wt}_k(i,\alpha)\right)e_\alpha. \tag*{(33)} [ n ] q ! H n = α ⊨ n ∑ ( i = 1 ∏ n wt k ( i , α ) ) e α . (33)
If α = ( β , a ) ⊨ n \alpha=(\beta,a)\vDash n α = ( β , a ) ⊨ n , then the last block of α \alpha α contributes the weight
e a ∏ i = n − a + 1 n wt k ( i , α ) = e a ( [ k a ] q − [ n − a ] q ) [ n − 1 ] q ! [ n − a ] q ! . e_a\prod_{i=n-a+1}^{n}\operatorname{wt}_k(i,\alpha)
=e_a([ka]_q-[n-a]_q)\frac{[n-1]_q!}{[n-a]_q!}. e a i = n − a + 1 ∏ n wt k ( i , α ) = e a ([ k a ] q − [ n − a ] q ) [ n − a ] q ! [ n − 1 ] q ! .
Hence, dividing (33) by [ n − 1 ] q ! [n-1]_q! [ n − 1 ] q ! gives
[ n ] q H n = ∑ a = 1 n ( [ k a ] q − [ n − a ] q ) e a H n − a ( n ≥ 1 ) . (34) [n]_qH_n=\sum_{a=1}^{n}([ka]_q-[n-a]_q)e_aH_{n-a}\qquad(n\geq1). \tag*{(34)} [ n ] q H n = a = 1 ∑ n ([ k a ] q − [ n − a ] q ) e a H n − a ( n ≥ 1 ) . (34)
To derive the recurrence for F n F_n F n , we claim that
∑ a = 0 n q k a e a F n − a = ∑ d 0 + ⋯ + d k = n q ∑ i = 0 k i d i ∏ i = 0 k e d i = ∑ a = 0 n q n − a e a F n − a . (35) \sum_{a=0}^{n}q^{ka}e_aF_{n-a}
=
\sum_{d_0+\cdots+d_k=n}q^{\sum_{i=0}^{k}id_i}\prod_{i=0}^{k}e_{d_i}
=
\sum_{a=0}^{n}q^{n-a}e_aF_{n-a}. \tag*{(35)} a = 0 ∑ n q k a e a F n − a = d 0 + ⋯ + d k = n ∑ q ∑ i = 0 k i d i i = 0 ∏ k e d i = a = 0 ∑ n q n − a e a F n − a . (35)
The first equality is obtained by setting d k = a d_k=a d k = a and d i = c i d_i=c_i d i = c i for $0\leq i\leq k-1. The second is obtained by setting d 0 = a d_0=a d 0 = a and d i = c i − 1 d_i=c_{i-1} d i = c i − 1 for 1 ≤ i ≤ k 1\leq i\leq k 1 ≤ i ≤ k , since
∑ i = 1 k i d i = ∑ i = 0 k − 1 ( i + 1 ) c i = n − a + ∑ i = 0 k − 1 i c i . \sum_{i=1}^{k}id_i=\sum_{i=0}^{k-1}(i+1)c_i=n-a+\sum_{i=0}^{k-1}ic_i. i = 1 ∑ k i d i = i = 0 ∑ k − 1 ( i + 1 ) c i = n − a + i = 0 ∑ k − 1 i c i .
Solving for F n F_n F n in (4.15) yields
( q n − 1 ) F n = ∑ a = 1 n ( q k a − q n − a ) e a F n − a . (q^n-1)F_n=\sum_{a=1}^{n}(q^{ka}-q^{n-a})e_aF_{n-a}. ( q n − 1 ) F n = a = 1 ∑ n ( q k a − q n − a ) e a F n − a .
Dividing by q − 1 q-1 q − 1 shows that F n F_n F n satisfies the same recurrence (4.14). Since F 0 = H 0 = 1 F_0=H_0=1 F 0 = H 0 = 1 , induction gives F n = H n F_n=H_n F n = H n for all n ≥ 0 n\geq0 n ≥ 0 . □
We now use the one-row formula to obtain a positive formula for every hook.
Corollary 4.6. Let r , k ≥ 1 r,k\geq1 r , k ≥ 1 and s ≥ 0 s\geq0 s ≥ 0 be integers with n = r + s n=r+s n = r + s . Then
q k ( n + ( s + 1 2 ) ) J ( r , 1 s ) ( x ; q , q − k ) ( q − 1 ) n = [ r − 1 ] q ! ∏ j = 1 s [ k j ] q ∑ d 0 + ⋯ + d k = n d k ≤ s < d 0 q ∑ i = 0 k i d i [ k ( d 0 − d k ) ] q ∏ i = 0 k e d i . (36) \frac{q^{k\left(n+\binom{s+1}{2}\right)}J_{(r,1^s)}(x;q,q^{-k})}{(q-1)^n}
=[r-1]_q!\prod_{j=1}^{s}[kj]_q
\sum_{\substack{d_0+\cdots+d_k=n\\d_k\leq s<d_0}}
q^{\sum_{i=0}^{k}id_i}[k(d_0-d_k)]_q\prod_{i=0}^{k}e_{d_i}. \tag*{(36)} ( q − 1 ) n q k ( n + ( 2 s + 1 ) ) J ( r , 1 s ) ( x ; q , q − k ) = [ r − 1 ] q ! j = 1 ∏ s [ k j ] q d 0 + ⋯ + d k = n d k ≤ s < d 0 ∑ q ∑ i = 0 k i d i [ k ( d 0 − d k ) ] q i = 0 ∏ k e d i . (36) In particular, the expression in (1.5) has e e e -coefficients in Z ≥ 0 [ q ] \mathbb{Z}_{\geq0}[q] Z ≥ 0 [ q ] for every hook partition.
Proof. Fix k ≥ 1 k\geq1 k ≥ 1 . Let
B r , s = q k ( n + ( s + 1 2 ) ) J ( r , 1 s ) ( x ; q , q − k ) ( q − 1 ) n , C s = ∏ j = 1 s [ k j ] q . B_{r,s}=\frac{q^{k\left(n+\binom{s+1}{2}\right)}J_{(r,1^s)}(x;q,q^{-k})}{(q-1)^n},
\qquad
C_s=\prod_{j=1}^{s}[kj]_q. B r , s = ( q − 1 ) n q k ( n + ( 2 s + 1 ) ) J ( r , 1 s ) ( x ; q , q − k ) , C s = j = 1 ∏ s [ k j ] q .
By the same argument as in the proof of Proposition 4.5, separating the last part a > s a>s a > s of the compositions α = ( β , a ) ⊨ n \alpha=(\beta,a)\vDash n α = ( β , a ) ⊨ n in (4.10) gives
B r , s = C s ∑ a = s + 1 n e a ( [ k a ] q − [ n − a ] q ) [ r − 1 ] q ! [ n − a ] q ! ∑ β ⊨ n − a ( ∏ i = 1 n − a wt k ( i , β ) ) e β = C s [ r − 1 ] q ! ∑ a = s + 1 n ( [ k a ] q − [ n − a ] q ) e a B n − a , 0 [ n − a ] q ! . \begin{aligned}
B_{r,s}
=C_s\sum_{a=s+1}^{n}e_a([ka]_q-[n-a]_q)\frac{[r-1]_q!}{[n-a]_q!}
\sum_{\beta\vDash n-a}\left(\prod_{i=1}^{n-a}\operatorname{wt}_k(i,\beta)\right)e_\beta\\
&=C_s[r-1]_q!\sum_{a=s+1}^{n}([ka]_q-[n-a]_q)e_a\frac{B_{n-a,0}}{[n-a]_q!}.
\end{aligned} B r , s = C s a = s + 1 ∑ n e a ([ k a ] q − [ n − a ] q ) [ n − a ] q ! [ r − 1 ] q ! β ⊨ n − a ∑ ( i = 1 ∏ n − a wt k ( i , β ) ) e β = C s [ r − 1 ] q ! a = s + 1 ∑ n ([ k a ] q − [ n − a ] q ) e a [ n − a ] q ! B n − a , 0 .
Applying Proposition 4.5 to B n − a , 0 B_{n-a,0} B n − a , 0 and using [ k a ] q − [ n − a ] q = ( q k a − q n − a ) / ( q − 1 ) [ka]_q-[n-a]_q=(q^{ka}-q^{n-a})/(q-1) [ k a ] q − [ n − a ] q = ( q k a − q n − a ) / ( q − 1 ) , we obtain
B r , s C s [ r − 1 ] q ! = ∑ a = s + 1 n q k a − q n − a q − 1 e a ∑ c 0 + ⋯ + c k − 1 = n − a q ∑ i = 0 k − 1 i c i ∏ i = 0 k − 1 e c i . (37) \frac{B_{r,s}}{C_s[r-1]_q!}
=\sum_{a=s+1}^{n}\frac{q^{ka}-q^{n-a}}{q-1}e_a
\sum_{\substack{c_0+\cdots+c_{k-1}=n-a}}
q^{\sum_{i=0}^{k-1}ic_i}\prod_{i=0}^{k-1}e_{c_i}. \tag*{(37)} C s [ r − 1 ] q ! B r , s = a = s + 1 ∑ n q − 1 q k a − q n − a e a c 0 + ⋯ + c k − 1 = n − a ∑ q ∑ i = 0 k − 1 i c i i = 0 ∏ k − 1 e c i . (37)
Consider a tuple ( c 0 , … , c k − 1 ) (c_0,\ldots,c_{k-1}) ( c 0 , … , c k − 1 ) of nonnegative integers with c 0 + ⋯ + c k − 1 = n − a c_0+\cdots+c_{k-1}=n-a c 0 + ⋯ + c k − 1 = n − a . If we let ( d 0 , … , d k ) = ( c 0 , … , c k − 1 , a ) (d_0,\ldots,d_k)=(c_0,\ldots,c_{k-1},a) ( d 0 , … , d k ) = ( c 0 , … , c k − 1 , a ) , then d 0 + ⋯ + d k = n d_0+\cdots+d_k=n d 0 + ⋯ + d k = n , d k = a > s d_k=a>s d k = a > s , and
∑ i = 0 k i d i = k a + ∑ i = 0 k − 1 i c i . \sum_{i=0}^{k}id_i=ka+\sum_{i=0}^{k-1}ic_i. i = 0 ∑ k i d i = k a + i = 0 ∑ k − 1 i c i .
If we let ( d 0 , … , d k ) = ( a , c 0 , … , c k − 1 ) (d_0,\ldots,d_k)=(a,c_0,\ldots,c_{k-1}) ( d 0 , … , d k ) = ( a , c 0 , … , c k − 1 ) , then d 0 + ⋯ + d k = n d_0+\cdots+d_k=n d 0 + ⋯ + d k = n , d 0 = a > s d_0=a>s d 0 = a > s , and
∑ i = 0 k i d i = ∑ i = 0 k − 1 ( i + 1 ) c i = ( n − a ) + ∑ i = 0 k − 1 i c i . \sum_{i=0}^{k}id_i=\sum_{i=0}^{k-1}(i+1)c_i=(n-a)+\sum_{i=0}^{k-1}ic_i. i = 0 ∑ k i d i = i = 0 ∑ k − 1 ( i + 1 ) c i = ( n − a ) + i = 0 ∑ k − 1 i c i .
Thus, we can rewrite (4.17) as
B r , s C s [ r − 1 ] q ! = ∑ d 0 + ⋯ + d k = n d k > s q ∑ i = 0 k i d i q − 1 ∏ i = 0 k e d i − ∑ d 0 + ⋯ + d k = n d 0 > s q ∑ i = 0 k i d i q − 1 ∏ i = 0 k e d i . (38) \frac{B_{r,s}}{C_s[r-1]_q!}
=\sum_{\substack{d_0+\cdots+d_k=n\\d_k>s}}
\frac{q^{\sum_{i=0}^{k}id_i}}{q-1}\prod_{i=0}^{k}e_{d_i}
-
\sum_{\substack{d_0+\cdots+d_k=n\\d_0>s}}
\frac{q^{\sum_{i=0}^{k}id_i}}{q-1}\prod_{i=0}^{k}e_{d_i}. \tag*{(38)} C s [ r − 1 ] q ! B r , s = d 0 + ⋯ + d k = n d k > s ∑ q − 1 q ∑ i = 0 k i d i i = 0 ∏ k e d i − d 0 + ⋯ + d k = n d 0 > s ∑ q − 1 q ∑ i = 0 k i d i i = 0 ∏ k e d i . (38)
Interchanging d 0 d_0 d 0 and d k d_k d k gives
∑ d 0 + ⋯ + d k = n d k > s q ∑ i = 0 k i d i ∏ i = 0 k e d i = ∑ d 0 + ⋯ + d k = n d 0 > s q k d 0 + ∑ i = 0 k − 1 i d i ∏ i = 0 k e d i . \sum_{\substack{d_0+\cdots+d_k=n\\d_k>s}}
q^{\sum_{i=0}^{k}id_i}\prod_{i=0}^{k}e_{d_i}
=
\sum_{\substack{d_0+\cdots+d_k=n\\d_0>s}}
q^{kd_0+\sum_{i=0}^{k-1}id_i}\prod_{i=0}^{k}e_{d_i}. d 0 + ⋯ + d k = n d k > s ∑ q ∑ i = 0 k i d i i = 0 ∏ k e d i = d 0 + ⋯ + d k = n d 0 > s ∑ q k d 0 + ∑ i = 0 k − 1 i d i i = 0 ∏ k e d i .
Thus (38) is equivalent to
B r , s C s [ r − 1 ] q ! = ∑ d 0 + ⋯ + d k = n d 0 > s q k d 0 − q k d k q − 1 q ∑ i = 0 k − 1 i d i ∏ i = 0 k e d i . \frac{B_{r,s}}{C_s[r-1]_q!}
=
\sum_{\substack{d_0+\cdots+d_k=n\\d_0>s}}
\frac{q^{kd_0}-q^{kd_k}}{q-1}q^{\sum_{i=0}^{k-1}id_i}\prod_{i=0}^{k}e_{d_i}. C s [ r − 1 ] q ! B r , s = d 0 + ⋯ + d k = n d 0 > s ∑ q − 1 q k d 0 − q k d k q ∑ i = 0 k − 1 i d i i = 0 ∏ k e d i .
Since the summand is antisymmetric in d 0 d_0 d 0 and d k d_k d k , the sum over nonnegative integers d i d_i d i satisfying both d 0 > s d_0>s d 0 > s and d k > s d_k>s d k > s vanishes. Thus, we obtain
B r , s C s [ r − 1 ] q ! = ∑ d 0 + ⋯ + d k = n d k ≤ s < d 0 q k d 0 − q k d k q − 1 q ∑ i = 0 k − 1 i d i ∏ i = 0 k e d i = ∑ d 0 + ⋯ + d k = n d k ≤ s < d 0 q k d k [ k ( d 0 − d k ) ] q q ∑ i = 0 k − 1 i d i ∏ i = 0 k e d i , \begin{aligned}
\frac{B_{r,s}}{C_s[r-1]_q!}
=
\sum_{\substack{d_0+\cdots+d_k=n\\d_k\le s<d_0}}
\frac{q^{kd_0}-q^{kd_k}}{q-1}q^{\sum_{i=0}^{k-1}id_i}\prod_{i=0}^{k}e_{d_i} \\
&=
\sum_{\substack{d_0+\cdots+d_k=n\\d_k\le s<d_0}}
q^{kd_k}[k(d_0-d_k)]_q q^{\sum_{i=0}^{k-1}id_i}\prod_{i=0}^{k}e_{d_i},
\end{aligned} C s [ r − 1 ] q ! B r , s = d 0 + ⋯ + d k = n d k ≤ s < d 0 ∑ q − 1 q k d 0 − q k d k q ∑ i = 0 k − 1 i d i i = 0 ∏ k e d i = d 0 + ⋯ + d k = n d k ≤ s < d 0 ∑ q k d k [ k ( d 0 − d k ) ] q q ∑ i = 0 k − 1 i d i i = 0 ∏ k e d i ,
which is the desired formula. □
Two-column case
In this section, we provide a formula for the e e e -coefficients of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) and prove Conjecture 1.1 when μ \mu μ consists of two columns. In this case, each nonzero e e e -coefficient of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) admits a factorization into ∣ μ ∣ |\mu| ∣ μ ∣ factors.
Theorem 5.1. Let 0 ≤ ℓ ≤ ⌊ n / 2 ⌋ 0\le\ell\le\lfloor n/2\rfloor 0 ≤ ℓ ≤ ⌊ n /2 ⌋ and μ = ( 2 ℓ , 1 n − 2 ℓ ) \mu=(2^\ell,1^{n-2\ell}) μ = ( 2 ℓ , 1 n − 2 ℓ ) . Then we have
J μ ( x ; q , t ) = ∑ p = 0 ℓ ( q ℓ − p ( 1 − t n − 2 p ) ( t ; t ) n − ℓ − p − 1 ( q − 1 ; t ) ℓ − p ( t ; t ) ℓ ( t ; t ) ℓ − p ( q t ; t ) n − ℓ ( q t ; t ) n − ℓ − p ) e ( n − p , p ) , (39) J_\mu(x;q,t)=\sum_{p=0}^{\ell}\left(q^{\ell-p}(1-t^{n-2p})(t;t)_{n-\ell-p-1}(q^{-1};t)_{\ell-p}\frac{(t;t)_\ell}{(t;t)_{\ell-p}}\frac{(qt;t)_{n-\ell}}{(qt;t)_{n-\ell-p}}\right)e_{(n-p,p)}, \tag*{(39)} J μ ( x ; q , t ) = p = 0 ∑ ℓ ( q ℓ − p ( 1 − t n − 2 p ) ( t ; t ) n − ℓ − p − 1 ( q − 1 ; t ) ℓ − p ( t ; t ) ℓ − p ( t ; t ) ℓ ( q t ; t ) n − ℓ − p ( q t ; t ) n − ℓ ) e ( n − p , p ) , (39) where we use the convention that ( 1 − t n − 2 p ) ( t ; t ) n − ℓ − p − 1 = ( t ; t ) n − 2 p (1-t^{n-2p})(t;t)_{n-\ell-p-1}=(t;t)_{n-2p} ( 1 − t n − 2 p ) ( t ; t ) n − ℓ − p − 1 = ( t ; t ) n − 2 p when p = ℓ p=\ell p = ℓ .
Equivalently, we have
J μ ( x ; q , t ) = ∑ p = 0 ℓ ( ∏ c ∈ μ wt μ , p ( c ) ) e ( n − p , p ) . J_\mu(x;q,t)=\sum_{p=0}^{\ell}\left(\prod_{c\in\mu}\operatorname{wt}_{\mu,p}(c)\right)e_{(n-p,p)}. J μ ( x ; q , t ) = p = 0 ∑ ℓ ( c ∈ μ ∏ wt μ , p ( c ) ) e ( n − p , p ) . where the weight wt μ , p ( c ) \operatorname{wt}_{\mu,p}(c) wt μ , p ( c ) of a cell c = ( i , j ) ∈ μ c=(i,j)\in\mu c = ( i , j ) ∈ μ is defined by
wt μ , p ( c ) = { 1 − q arm ( c ) t leg ( c ) + 1 , if i ≤ p , 1 − t n − 2 p , if c = ( p + 1 , 1 ) , q j − 1 − t i − p − 1 , if i > p and c ≠ ( p + 1 , 1 ) , \operatorname{wt}_{\mu,p}(c)=
\begin{cases}
1-q^{\operatorname{arm}(c)}t^{\operatorname{leg}(c)+1}, & \text{if }i\le p,\\
1-t^{n-2p}, & \text{if }c=(p+1,1),\\
q^{j-1}-t^{i-p-1}, & \text{if }i>p\text{ and }c\ne(p+1,1),
\end{cases} wt μ , p ( c ) = ⎩ ⎨ ⎧ 1 − q arm ( c ) t leg ( c ) + 1 , 1 − t n − 2 p , q j − 1 − t i − p − 1 , if i ≤ p , if c = ( p + 1 , 1 ) , if i > p and c = ( p + 1 , 1 ) , with arm and leg lengths measured in μ \mu μ . See Figure 3 .
To prove Theorem 5.1, we use the following lemma.
Figure 3. The weight wt μ , p ( c ) \operatorname{wt}_{\mu,p}(c) wt μ , p ( c ) of each cell c ∈ μ c\in\mu c ∈ μ for μ = ( 2 4 , 1 2 ) \mu=(2^4,1^2) μ = ( 2 4 , 1 2 ) and p = 2 p=2 p = 2 .
Lemma 5.2. For a ≥ b a \ge b a ≥ b with n = a + b n=a+b n = a + b , we have
e a ⋅ e b = ∑ j = 0 b ( q ; t ) a − j ( q ; t ) b − j ( 1 − q t n − 2 j ) ( t ; t ) a − j ( t ; t ) b − j ( t ; t ) j ( q ; t ) n − j + 1 J ( 2 j , 1 n − 2 j ) ( x ; q , t ) . (40) e_a\cdot e_b=\sum_{j=0}^{b}\frac{(q;t)_{a-j}(q;t)_{b-j}(1-qt^{n-2j})}{(t;t)_{a-j}(t;t)_{b-j}(t;t)_j(q;t)_{n-j+1}}J_{(2^j,1^{n-2j})}(x;q,t). \tag*{(40)} e a ⋅ e b = j = 0 ∑ b ( t ; t ) a − j ( t ; t ) b − j ( t ; t ) j ( q ; t ) n − j + 1 ( q ; t ) a − j ( q ; t ) b − j ( 1 − q t n − 2 j ) J ( 2 j , 1 n − 2 j ) ( x ; q , t ) . (40)
Proof. Note that P ( 1 n ) ( x ; q , t ) = e n ( x ) P_{(1^n)}(x;q,t)=e_n(x) P ( 1 n ) ( x ; q , t ) = e n ( x ) . Then the lemma follows directly from the Pieri rule for Macdonald polynomials in [15 Ch. VI, (6.24) (iv)] by taking μ \mu μ to be a one-column partition. □
Proof of Theorem 5.1. We use Lemma 5.2 to verify the e e e -coefficients of J ( 2 ℓ , 1 n − 2 ℓ ) ( x ; q , t ) J_{(2^\ell,1^{n-2\ell})}(x;q,t) J ( 2 ℓ , 1 n − 2 ℓ ) ( x ; q , t ) . Note that the identities (40) with ( a , b ) = ( n − ℓ , ℓ ) (a,b)=(n-\ell,\ell) ( a , b ) = ( n − ℓ , ℓ ) for 0 ≤ ℓ ≤ ⌊ n / 2 ⌋ 0\le\ell\le\lfloor n/2\rfloor 0 ≤ ℓ ≤ ⌊ n /2 ⌋ determine the polynomials J ( 2 ℓ , 1 n − 2 ℓ ) ( x ; q , t ) J_{(2^\ell,1^{n-2\ell})}(x;q,t) J ( 2 ℓ , 1 n − 2 ℓ ) ( x ; q , t ) uniquely. Indeed, the coefficient of J ( 2 ℓ , 1 n − 2 ℓ ) J_{(2^\ell,1^{n-2\ell})} J ( 2 ℓ , 1 n − 2 ℓ ) in (40) when j = ℓ j=\ell j = ℓ is
( q ; t ) n − 2 ℓ ( 1 − q t n − 2 ℓ ) ( t ; t ) n − 2 ℓ ( t ; t ) ℓ ( q ; t ) n − ℓ + 1 ≠ 0 , \frac{(q;t)_{n-2\ell}(1-qt^{n-2\ell})}{(t;t)_{n-2\ell}(t;t)_\ell(q;t)_{n-\ell+1}}\ne0, ( t ; t ) n − 2 ℓ ( t ; t ) ℓ ( q ; t ) n − ℓ + 1 ( q ; t ) n − 2 ℓ ( 1 − q t n − 2 ℓ ) = 0 ,
so (40) with ( a , b ) = ( n − ℓ , ℓ ) (a,b)=(n-\ell,\ell) ( a , b ) = ( n − ℓ , ℓ ) expresses J ( 2 ℓ , 1 n − 2 ℓ ) J_{(2^\ell,1^{n-2\ell})} J ( 2 ℓ , 1 n − 2 ℓ ) in terms of e n − ℓ e ℓ e_{n-\ell}e_\ell e n − ℓ e ℓ and the J ( 2 j , 1 n − 2 j ) J_{(2^j,1^{n-2j})} J ( 2 j , 1 n − 2 j ) with j < ℓ j<\ell j < ℓ , and induction on ℓ \ell ℓ proves the claim. Therefore, it suffices to show that the right-hand sides of (39) satisfy (40) with ( a , b ) = ( n − ℓ , ℓ ) (a,b)=(n-\ell,\ell) ( a , b ) = ( n − ℓ , ℓ ) for every 0 ≤ ℓ ≤ ⌊ n / 2 ⌋ 0\le\ell\le\lfloor n/2\rfloor 0 ≤ ℓ ≤ ⌊ n /2 ⌋ , that is,
e ( n − ℓ , ℓ ) ( x ) = ∑ p = 0 ℓ S n , ℓ , p e ( n − p , p ) ( x ) , (41) e_{(n-\ell,\ell)}(x)=\sum_{p=0}^{\ell}S_{n,\ell,p}e_{(n-p,p)}(x), \tag*{(41)} e ( n − ℓ , ℓ ) ( x ) = p = 0 ∑ ℓ S n , ℓ , p e ( n − p , p ) ( x ) , (41)
where
S n , ℓ , p = ∑ j = p ℓ q j − p ( 1 − t n − 2 p ) ( 1 − q t n − 2 j ) ( 1 − q ) ( q − 1 ; t ) j − p ( t ; t ) n − j − p − 1 ( q ; t ) n − ℓ − j ( q ; t ) ℓ − j ( t ; t ) j − p ( q t ; t ) n − j − p ( t ; t ) n − ℓ − j ( t ; t ) ℓ − j . S_{n,\ell,p}=\sum_{j=p}^{\ell}q^{j-p}\frac{(1-t^{n-2p})(1-qt^{n-2j})}{(1-q)}
\frac{(q^{-1};t)_{j-p}(t;t)_{n-j-p-1}(q;t)_{n-\ell-j}(q;t)_{\ell-j}}
{(t;t)_{j-p}(qt;t)_{n-j-p}(t;t)_{n-\ell-j}(t;t)_{\ell-j}}. S n , ℓ , p = j = p ∑ ℓ q j − p ( 1 − q ) ( 1 − t n − 2 p ) ( 1 − q t n − 2 j ) ( t ; t ) j − p ( q t ; t ) n − j − p ( t ; t ) n − ℓ − j ( t ; t ) ℓ − j ( q − 1 ; t ) j − p ( t ; t ) n − j − p − 1 ( q ; t ) n − ℓ − j ( q ; t ) ℓ − j .
Since { e λ } \{e_\lambda\} { e λ } forms a basis, (41) is equivalent to S n , ℓ , p = χ ( p = ℓ ) S_{n,\ell,p}=\chi(p=\ell) S n , ℓ , p = χ ( p = ℓ ) . The case p = ℓ p=\ell p = ℓ is immediate, so it remains to show that S n , ℓ , p = 0 S_{n,\ell,p}=0 S n , ℓ , p = 0 for 0 ≤ p < ℓ ≤ ⌊ n / 2 ⌋ 0\le p<\ell\le\lfloor n/2\rfloor 0 ≤ p < ℓ ≤ ⌊ n /2 ⌋ .
Reversing the summation by replacing j j j with ℓ − j \ell-j ℓ − j and using the identities
( a ; t ) r + j = ( a ; t ) r ( a t r ; t ) j , ( a ; t ) r − j = ( a ; t ) r ( − a ) j t ( j 2 ) − r j ( a − 1 t 1 − r ; t ) j , (a;t)_{r+j}=(a;t)_r(at^r;t)_j,\qquad
(a;t)_{r-j}=\frac{(a;t)_r}{(-a)^jt^{\binom{j}{2}-rj}(a^{-1}t^{1-r};t)_j}, ( a ; t ) r + j = ( a ; t ) r ( a t r ; t ) j , ( a ; t ) r − j = ( − a ) j t ( 2 j ) − r j ( a − 1 t 1 − r ; t ) j ( a ; t ) r ,
where the second identity holds for 0 ≤ j ≤ r 0\le j\le r 0 ≤ j ≤ r , we obtain
S n , ℓ , p = q ℓ − p ( 1 − t n − 2 p ) ( q − 1 ; t ) ℓ − p ( t ; t ) n − ℓ − p − 1 ( q ; t ) n − 2 ℓ ( 1 − q ) ( t ; t ) ℓ − p ( q t ; t ) n − ℓ − p ( t ; t ) n − 2 ℓ ∑ j = 0 ℓ − p t j ( q , t p − ℓ , t n − ℓ − p , q t n − 2 ℓ ; t ) j ( 1 − q t n − 2 ℓ + 2 j ) ( t , q t p − ℓ + 1 , q t n − ℓ − p + 1 , t n − 2 ℓ + 1 ; t ) j . S_{n,\ell,p}=
\frac{q^{\ell-p}(1-t^{n-2p})(q^{-1};t)_{\ell-p}(t;t)_{n-\ell-p-1}(q;t)_{n-2\ell}}
{(1-q)(t;t)_{\ell-p}(qt;t)_{n-\ell-p}(t;t)_{n-2\ell}}
\sum_{j=0}^{\ell-p}
t^j\frac{(q,t^{p-\ell},t^{n-\ell-p},qt^{n-2\ell};t)_j(1-qt^{n-2\ell+2j})}
{(t,qt^{p-\ell+1},qt^{n-\ell-p+1},t^{n-2\ell+1};t)_j}. S n , ℓ , p = ( 1 − q ) ( t ; t ) ℓ − p ( q t ; t ) n − ℓ − p ( t ; t ) n − 2 ℓ q ℓ − p ( 1 − t n − 2 p ) ( q − 1 ; t ) ℓ − p ( t ; t ) n − ℓ − p − 1 ( q ; t ) n − 2 ℓ j = 0 ∑ ℓ − p t j ( t , q t p − ℓ + 1 , q t n − ℓ − p + 1 , t n − 2 ℓ + 1 ; t ) j ( q , t p − ℓ , t n − ℓ − p , q t n − 2 ℓ ; t ) j ( 1 − q t n − 2 ℓ + 2 j ) .
We apply the terminating very-well-poised 6 ϕ 5 {}_6\phi_5 6 ϕ 5 summation formula [4 (II.21)] with base t t t and parameters
a = q t n − 2 ℓ , b = t n − ℓ − p , c = q , a=qt^{n-2\ell},\qquad b=t^{n-\ell-p},\qquad c=q, a = q t n − 2 ℓ , b = t n − ℓ − p , c = q ,
and terminating parameter t p − ℓ t^{p-\ell} t p − ℓ . Since a t ℓ − p + 1 / ( b c ) = t at^{\ell-p+1}/(bc)=t a t ℓ − p + 1 / ( b c ) = t , the sum above equals
( 1 − a ) ( a t , a t / ( b c ) ; t ) ℓ − p ( a t / b , a t / c ; t ) ℓ − p = ( 1 − q t n − 2 ℓ ) ( q t n − 2 ℓ + 1 , t p − ℓ + 1 ; t ) ℓ − p ( q t p − ℓ + 1 , t n − 2 ℓ + 1 ; t ) ℓ − p = 0. (1-a)\frac{(at,at/(bc);t)_{\ell-p}}{(at/b,at/c;t)_{\ell-p}}
=(1-qt^{n-2\ell})\frac{(qt^{n-2\ell+1},t^{p-\ell+1};t)_{\ell-p}}
{(qt^{p-\ell+1},t^{n-2\ell+1};t)_{\ell-p}}=0. ( 1 − a ) ( a t / b , a t / c ; t ) ℓ − p ( a t , a t / ( b c ) ; t ) ℓ − p = ( 1 − q t n − 2 ℓ ) ( q t p − ℓ + 1 , t n − 2 ℓ + 1 ; t ) ℓ − p ( q t n − 2 ℓ + 1 , t p − ℓ + 1 ; t ) ℓ − p = 0.
Indeed, since ℓ − p ≥ 1 \ell-p\ge1 ℓ − p ≥ 1 , the factor ( t p − ℓ + 1 ; t ) ℓ − p (t^{p-\ell+1};t)_{\ell-p} ( t p − ℓ + 1 ; t ) ℓ − p contains 1 − t 0 = 0 1-t^0=0 1 − t 0 = 0 . Thus S n , ℓ , p = 0 S_{n,\ell,p}=0 S n , ℓ , p = 0 , completing the proof. □
Using Theorem 5.1, we can prove Conjecture 1.1 when μ \mu μ has at most two columns.
Corollary 5.3. For every partition μ \mu μ with at most two columns and every positive integer k k k , the expression in (5) has e e e -coefficients in Z ≥ 0 [ q ] \mathbb{Z}_{\ge0}[q] Z ≥ 0 [ q ] .
Proof. For μ = ( 2 ℓ , 1 n − 2 ℓ ) \mu=(2^\ell,1^{n-2\ell}) μ = ( 2 ℓ , 1 n − 2 ℓ ) , consider the coefficients
A j = [ e ( n − j , j ) ] ( q k ∑ i t i J μ ( x ; q , q − k ) ( q − 1 ) n ) , 0 ≤ j ≤ ℓ . A_j=[e_{(n-j,j)}]\left(\frac{q^k\sum_i t_i\frac{J_\mu(x;q,q^{-k})}{(q-1)^n}}{}\right),\qquad0\le j\le\ell. A j = [ e ( n − j , j ) ] q k ∑ i t i ( q − 1 ) n J μ ( x ; q , q − k ) , 0 ≤ j ≤ ℓ .
Substituting t = q − k t=q^{-k} t = q − k into Theorem 5.1 and using 1 − q − a = q − a ( q − 1 ) [ a ] q 1-q^{-a}=q^{-a}(q-1)[a]_q 1 − q − a = q − a ( q − 1 ) [ a ] q , we obtain, for 0 ≤ j < ℓ 0\le j<\ell 0 ≤ j < ℓ ,
A j = q j [ k ( n − 2 j ) ] q ∏ i = 0 ℓ − j − 1 [ k i + 1 ] q ∏ i = 1 n − ℓ − j − 1 [ k i ] q ∏ i = ℓ − j + 1 ℓ [ k i ] q ∏ i = n − ℓ − j + 1 n − ℓ [ k i − 1 ] q . A_j=q^j[k(n-2j)]_q\prod_{i=0}^{\ell-j-1}[ki+1]_q\prod_{i=1}^{n-\ell-j-1}[ki]_q\prod_{i=\ell-j+1}^{\ell}[ki]_q\prod_{i=n-\ell-j+1}^{n-\ell}[ki-1]_q. A j = q j [ k ( n − 2 j ) ] q i = 0 ∏ ℓ − j − 1 [ k i + 1 ] q i = 1 ∏ n − ℓ − j − 1 [ k i ] q i = ℓ − j + 1 ∏ ℓ [ k i ] q i = n − ℓ − j + 1 ∏ n − ℓ [ k i − 1 ] q .
For j = ℓ j=\ell j = ℓ , the convention in Theorem 5.1 gives
A ℓ = q ℓ ∏ i = 1 n − 2 ℓ [ k i ] q ∏ i = 1 ℓ [ k i ] q ∏ i = n − 2 ℓ + 1 n − ℓ [ k i − 1 ] q . A_\ell=q^\ell\prod_{i=1}^{n-2\ell}[ki]_q\prod_{i=1}^{\ell}[ki]_q\prod_{i=n-2\ell+1}^{n-\ell}[ki-1]_q. A ℓ = q ℓ i = 1 ∏ n − 2 ℓ [ k i ] q i = 1 ∏ ℓ [ k i ] q i = n − 2 ℓ + 1 ∏ n − ℓ [ k i − 1 ] q .
Thus A j ∈ Z ≥ 0 [ q ] A_j\in\mathbb{Z}_{\geq0}[q] A j ∈ Z ≥ 0 [ q ] for every j j j , proving the corollary. □
The e n e_n e n -coefficient
Combining the relation between P μ ( x ; q , t ) P_\mu(x;q,t) P μ ( x ; q , t ) and J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) with the triangularity of P μ ( x ; q , t ) P_\mu(x;q,t) P μ ( x ; q , t ) in the monomial symmetric function basis, we obtain
J μ ( x ; q , t ) = ∑ λ ≥ μ ′ c μ , λ e λ , J_\mu(x;q,t)=\sum_{\lambda\geq\mu'}c_{\mu,\lambda}e_\lambda, J μ ( x ; q , t ) = λ ≥ μ ′ ∑ c μ , λ e λ ,
where
c μ , μ ′ = ∏ c ∈ μ ( 1 − q arm ( c ) t leg ( c ) + 1 ) = ∏ c ∈ μ ′ ( 1 − q leg ( c ) t arm ( c ) + 1 ) . (42) c_{\mu,\mu'}=\prod_{c\in\mu}(1-q^{\operatorname{arm}(c)}t^{\operatorname{leg}(c)+1})=\prod_{c\in\mu'}(1-q^{\operatorname{leg}(c)}t^{\operatorname{arm}(c)+1}). \tag*{(42)} c μ , μ ′ = c ∈ μ ∏ ( 1 − q arm ( c ) t leg ( c ) + 1 ) = c ∈ μ ′ ∏ ( 1 − q leg ( c ) t arm ( c ) + 1 ) . (42)
In particular,
J ( 1 n ) ( x ; q , t ) = ( t ; t ) n e n . J_{(1^n)}(x;q,t)=(t;t)_n e_n. J ( 1 n ) ( x ; q , t ) = ( t ; t ) n e n .
More generally, we compute the e n e_n e n -coefficient of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) for any μ ⊢ n \mu\vdash n μ ⊢ n .
Theorem 6.1. For n ≥ 1 n\geq1 n ≥ 1 and μ ⊢ n \mu\vdash n μ ⊢ n , we have
[ e n ] J μ ( x ; q , t ) = ( 1 − t n ) ( q − 1 ) ⋯ ( q μ 1 − 1 − 1 ) ∏ c ∉ Row 1 ( μ ) ( q arm ′ ( c ) − t leg ′ ( c ) ) [e_n]J_\mu(x;q,t)=(1-t^n)(q-1)\cdots(q^{\mu_1-1}-1)\prod_{c\notin\operatorname{Row}_1(\mu)}(q^{\operatorname{arm}'(c)}-t^{\operatorname{leg}'(c)}) [ e n ] J μ ( x ; q , t ) = ( 1 − t n ) ( q − 1 ) ⋯ ( q μ 1 − 1 − 1 ) c ∈ / Row 1 ( μ ) ∏ ( q arm ′ ( c ) − t leg ′ ( c ) ) = ( 1 − t n ) ( q − 1 ) ⋯ ( q μ 1 − 1 − 1 ) ∏ c ∉ Col 1 ( μ ′ ) ( q leg ′ ( c ) − t arm ′ ( c ) ) , =(1-t^n)(q-1)\cdots(q^{\mu_1-1}-1)\prod_{c\notin\operatorname{Col}_1(\mu')}(q^{\operatorname{leg}'(c)}-t^{\operatorname{arm}'(c)}), = ( 1 − t n ) ( q − 1 ) ⋯ ( q μ 1 − 1 − 1 ) c ∈ / Col 1 ( μ ′ ) ∏ ( q leg ′ ( c ) − t arm ′ ( c ) ) , where Row 1 ( μ ) \operatorname{Row}_1(\mu) Row 1 ( μ ) (resp. Col 1 ( μ ) \operatorname{Col}_1(\mu) Col 1 ( μ ) ) is the set of cells in the first row (resp. column) of μ \mu μ . Equivalently,
[ e n ] J μ ( x ; q , t ) = ( 1 − t n ) ∏ c ∈ μ ′ ∖ { ( 1 , 1 ) } ( q leg ′ ( c ) − t arm ′ ( c ) ) . (43) [e_n]J_\mu(x;q,t)=(1-t^n)\prod_{c\in\mu'\setminus\{(1,1)\}}(q^{\operatorname{leg}'(c)}-t^{\operatorname{arm}'(c)}). \tag*{(43)} [ e n ] J μ ( x ; q , t ) = ( 1 − t n ) c ∈ μ ′ ∖ {( 1 , 1 )} ∏ ( q leg ′ ( c ) − t arm ′ ( c ) ) . (43)
To prove Theorem 6.1, we need the following lemma.
Lemma 6.2. Let n ≥ 1 n\geq1 n ≥ 1 . If
f = ∑ λ ⊢ n a λ e λ , f=\sum_{\lambda\vdash n}a_\lambda e_\lambda, f = λ ⊢ n ∑ a λ e λ , then
[ e n ] f = ( − 1 ) n − 1 ( 1 − t n ) lim u → 1 f [ 1 − u 1 − t ] / ( 1 − u ) . (44) [e_n]f=(-1)^{n-1}(1-t^n)\lim_{u\to1}f\left[\frac{1-u}{1-t}\right]/(1-u). \tag*{(44)} [ e n ] f = ( − 1 ) n − 1 ( 1 − t n ) u → 1 lim f [ 1 − t 1 − u ] / ( 1 − u ) . (44)
Proof. By [15 Ch. I, (2.14′)] , we have
e m [ 1 − u 1 − t ] = ∑ λ ⊢ m ( − 1 ) m − ℓ ( λ ) z λ − 1 p λ [ 1 − u 1 − t ] , (45) e_m\left[\frac{1-u}{1-t}\right]=\sum_{\lambda\vdash m}(-1)^{m-\ell(\lambda)}z_\lambda^{-1}p_\lambda\left[\frac{1-u}{1-t}\right], \tag*{(45)} e m [ 1 − t 1 − u ] = λ ⊢ m ∑ ( − 1 ) m − ℓ ( λ ) z λ − 1 p λ [ 1 − t 1 − u ] , (45)
where z λ z_\lambda z λ is given in (2.1). Since
p k [ 1 − u 1 − t ] = 1 − u k 1 − t k = k ( 1 − u ) 1 − t k + O ( ( 1 − u ) 2 ) , p_k\left[\frac{1-u}{1-t}\right]=\frac{1-u^k}{1-t^k}=\frac{k(1-u)}{1-t^k}+O((1-u)^2), p k [ 1 − t 1 − u ] = 1 − t k 1 − u k = 1 − t k k ( 1 − u ) + O (( 1 − u ) 2 ) ,
we see that only the term λ = ( m ) \lambda=(m) λ = ( m ) of the right-hand side of (6.4) contributes to first order in 1 − u 1-u 1 − u . Therefore,
f [ 1 − u 1 − t ] = a ( n ) e n [ 1 − u 1 − t ] + ∑ λ < ( n ) a λ e λ [ 1 − u 1 − t ] = a ( n ) ( − 1 ) n − 1 1 − u 1 − t n + O ( ( 1 − u ) 2 ) . \begin{aligned}
f\left[\frac{1-u}{1-t}\right]=a_{(n)}e_n\left[\frac{1-u}{1-t}\right]+\sum_{\lambda<(n)}a_\lambda e_\lambda\left[\frac{1-u}{1-t}\right] \\
&=a_{(n)}(-1)^{n-1}\frac{1-u}{1-t^n}+O((1-u)^2).
\end{aligned} f [ 1 − t 1 − u ] = a ( n ) e n [ 1 − t 1 − u ] + λ < ( n ) ∑ a λ e λ [ 1 − t 1 − u ] = a ( n ) ( − 1 ) n − 1 1 − t n 1 − u + O (( 1 − u ) 2 ) .
Then the result follows by taking the limit as u u u approaches 1 and solving for the coefficient a ( n ) a_{(n)} a ( n ) . □
Proof of Theorem 6.1. By [15 Ch. VI, (8.8)] , we have
J μ [ 1 − u 1 − t ; q , t ] = ∏ c ∈ μ ( t leg ′ ( c ) − q arm ′ ( c ) u ) . J_\mu\left[\frac{1-u}{1-t};q,t\right]=\prod_{c\in\mu}\left(t^{\operatorname{leg}'(c)}-q^{\operatorname{arm}'(c)}u\right). J μ [ 1 − t 1 − u ; q , t ] = c ∈ μ ∏ ( t leg ′ ( c ) − q arm ′ ( c ) u ) .
Since the cell ( 1 , 1 ) (1,1) ( 1 , 1 ) corresponds to the factor ( 1 − u ) (1-u) ( 1 − u ) , we have
lim u → 1 J μ [ 1 − u 1 − t ] / ( 1 − u ) = ∏ c ∈ μ ∖ { ( 1 , 1 ) } ( t leg ′ ( c ) − q arm ′ ( c ) ) . \lim_{u\to1}J_\mu\left[\frac{1-u}{1-t}\right]/(1-u)=\prod_{c\in\mu\setminus\{(1,1)\}}\left(t^{\operatorname{leg}'(c)}-q^{\operatorname{arm}'(c)}\right). u → 1 lim J μ [ 1 − t 1 − u ] / ( 1 − u ) = c ∈ μ ∖ {( 1 , 1 )} ∏ ( t leg ′ ( c ) − q arm ′ ( c ) ) .
By applying Lemma 6.2, we obtain
[ e n ] J μ ( x ; q , t ) = ( − 1 ) n − 1 ( 1 − t n ) ∏ c ∈ μ ∖ { ( 1 , 1 ) } ( t leg ′ ( c ) − q arm ′ ( c ) ) = ( 1 − t n ) ∏ c ∈ μ ∖ { ( 1 , 1 ) } ( q arm ′ ( c ) − t leg ′ ( c ) ) . \begin{aligned}
[e_n]J_\mu(x;q,t)=(-1)^{n-1}(1-t^n)\prod_{c\in\mu\setminus\{(1,1)\}}\left(t^{\operatorname{leg}'(c)}-q^{\operatorname{arm}'(c)}\right) \\
&=(1-t^n)\prod_{c\in\mu\setminus\{(1,1)\}}\left(q^{\operatorname{arm}'(c)}-t^{\operatorname{leg}'(c)}\right).
\end{aligned} [ e n ] J μ ( x ; q , t ) = ( − 1 ) n − 1 ( 1 − t n ) c ∈ μ ∖ {( 1 , 1 )} ∏ ( t leg ′ ( c ) − q arm ′ ( c ) ) = ( 1 − t n ) c ∈ μ ∖ {( 1 , 1 )} ∏ ( q arm ′ ( c ) − t leg ′ ( c ) ) .
Since all cells in the first row have coleg length 0, we obtain the first identity. The remaining expressions follow directly from analyzing the diagram of μ \mu μ . □
Final remarks
In this section, we explain the analogy between the normalization of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) in Conjecture 1.1 and the α \alpha α -chromatic symmetric functions introduced by Haglund, Oh, and Yoo [9 ] . We then propose a conjecture on an expansion of the normalized J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) in terms of chromatic symmetric functions when q = 1 q=1 q = 1 .
For a graph G G G with vertex set V = { v 1 , … , v n } V=\{v_1,\ldots,v_n\} V = { v 1 , … , v n } , the chromatic symmetric function χ G ( x ) \chi_G(x) χ G ( x ) is defined by
χ G ( x ) = ∑ c : V → Z + x c ( v 1 ) ⋯ x c ( v n ) , \chi_G(x)=\sum_{c:V\to\mathbb{Z}_{+}}x_{c(v_1)}\cdots x_{c(v_n)}, χ G ( x ) = c : V → Z + ∑ x c ( v 1 ) ⋯ x c ( v n ) ,
where the sum is over all proper colorings c c c of G G G . If G G G is the incomparability graph of a natural unit interval order, then it can be identified with a Dyck path γ \gamma γ of length 2 n 2n 2 n . From now on, we consider only such graphs and write χ γ ( x ) \chi_\gamma(x) χ γ ( x ) in place of χ G ( x ) \chi_G(x) χ G ( x ) . See [18 , 9 ] for more details.
By (1.4), the normalization of J μ ( x ; q , t ) J_\mu(x;q,t) J μ ( x ; q , t ) in Conjecture 1.1 satisfies
q k ∑ i ≥ 1 i μ i J μ ( x ; q , q − k ) ( q − 1 ) n = H ~ μ [ ( q k − 1 ) X ; q , q k ] ( q − 1 ) n . q^k\sum_{i\ge1}i\mu_i\frac{J_\mu(x;q,q^{-k})}{(q-1)^n}=\frac{\widetilde{H}_\mu[(q^k-1)X;q,q^k]}{(q-1)^n}. q k i ≥ 1 ∑ i μ i ( q − 1 ) n J μ ( x ; q , q − k ) = ( q − 1 ) n H μ [( q k − 1 ) X ; q , q k ] .
Applying the same plethystic substitution to the unicellular LLT polynomial LLT γ ( x ; q ) \operatorname{LLT}_\gamma(x;q) LLT γ ( x ; q ) and dividing by ( q − 1 ) n (q-1)^n ( q − 1 ) n gives the α \alpha α -chromatic symmetric function at α = k \alpha=k α = k :
χ γ ( k ) ( x ; q ) = LLT γ [ ( q k − 1 ) X ; q ] ( q − 1 ) n . \chi_\gamma^{(k)}(x;q)=\frac{\operatorname{LLT}_\gamma[(q^k-1)X;q]}{(q-1)^n}. χ γ ( k ) ( x ; q ) = ( q − 1 ) n LLT γ [( q k − 1 ) X ; q ] .
When q = 1 q=1 q = 1 , the α \alpha α -chromatic symmetric functions are polynomials in α \alpha α whose coefficients are symmetric functions. Hence, they can be expanded in terms of the basis { ( α r ) } r ≥ 0 \left\{\binom{\alpha}{r}\right\}_{r\ge0} { ( r α ) } r ≥ 0 of the polynomials in α \alpha α . Haglund, Oh, and Yoo [9 Theorem 4.4] showed that each coefficient of ( α r ) \binom{\alpha}{r} ( r α ) has a positive expansion in terms of chromatic symmetric functions:
χ γ ( α ) ( x ; 1 ) = ∑ r = 1 n ( α r ) r ! ∑ S ∈ S ( n , r ) χ β ( γ , S ) ( x ) . (46) \chi_{\gamma}^{(\alpha)}(x;1)=\sum_{r=1}^{n}\binom{\alpha}{r}r!\sum_{S\in\mathcal{S}(n,r)}\chi_{\beta(\gamma,S)}(x). \tag*{(46)} χ γ ( α ) ( x ; 1 ) = r = 1 ∑ n ( r α ) r ! S ∈ S ( n , r ) ∑ χ β ( γ , S ) ( x ) . (46)
Here S ( n , r ) \mathcal{S}(n,r) S ( n , r ) denotes the set of all set partitions of { 1 , 2 , … , n } \{1,2,\ldots,n\} { 1 , 2 , … , n } into r r r blocks, and β ( γ , S ) \beta(\gamma,S) β ( γ , S ) is a certain Dyck path determined by γ \gamma γ and S S S . By Hikita’s theorem [12 ] , this implies that χ γ ( α ) ( x ; 1 ) \chi_{\gamma}^{(\alpha)}(x;1) χ γ ( α ) ( x ; 1 ) is e e e -positive for every positive integer α \alpha α .
Now we define
J ^ μ ( α ) ( x ) = lim q → 1 J μ ( x ; q , q − α ) ( q − 1 ) ∣ μ ∣ . \widehat{J}_{\mu}^{(\alpha)}(x)=\lim_{q\to1}\frac{J_{\mu}(x;q,q^{-\alpha})}{(q-1)^{|\mu|}}. J μ ( α ) ( x ) = q → 1 lim ( q − 1 ) ∣ μ ∣ J μ ( x ; q , q − α ) .
Using the Haglund–Haiman–Loehr formula [8 Proposition 8.1] , one can see that this is also a polynomial in α \alpha α . Inspired by (7.1), we propose the following question.
Question 7.1. Let n , r n,r n , r be positive integers, μ ⊢ n \mu\vdash n μ ⊢ n and 0 ≤ r ≤ n 0\le r\le n 0 ≤ r ≤ n . Denote by O P ( n , r ) \mathcal{OP}(n,r) O P ( n , r ) the set of ordered set partitions of { 1 , 2 , … , n } \{1,2,\ldots,n\} { 1 , 2 , … , n } into r r r blocks. For each S ∈ O P ( n , r ) S\in\mathcal{OP}(n,r) S ∈ O P ( n , r ) , can we find a Dyck path π ( μ , S ) \pi(\mu,S) π ( μ , S ) such that
J ^ μ ( α ) ( x ) = ∑ S ∈ O P ( n , r ) χ π ( μ , S ) ( x ) ? (47) \widehat{J}_{\mu}^{(\alpha)}(x)=\sum_{S\in\mathcal{OP}(n,r)}\chi_{\pi(\mu,S)}(x)\ ? \tag*{(47)} J μ ( α ) ( x ) = S ∈ O P ( n , r ) ∑ χ π ( μ , S ) ( x ) ? (47)
Again by the result of Hikita [12 ] , an affirmative answer to this question would imply the e e e -positivity of J ^ μ ( α ) ( x ) \widehat{J}_{\mu}^{(\alpha)}(x) J μ ( α ) ( x ) for every positive integer α \alpha α .
We note that the chromatic symmetric functions { χ γ ( x ) } γ ∈ D \{\chi_{\gamma}(x)\}_{\gamma\in\mathcal{D}} { χ γ ( x ) } γ ∈ D , where D \mathcal{D} D is the set of Dyck paths, do not form a basis of the space of symmetric functions. Thus, an expansion of the form (7.2) need not be unique. Note also that if π ( μ , S ) \pi(\mu,S) π ( μ , S ) depends only on μ \mu μ and the underlying set partition of S S S , then one can rewrite (7.2) as
J ^ μ ( α ) ( x ) = ∑ r = 1 n ( α r ) r ! ∑ S ∈ S ( n , r ) χ π ( μ , S ) ( x ) , \widehat{J}_{\mu}^{(\alpha)}(x)=\sum_{r=1}^{n}\binom{\alpha}{r}r!\sum_{S\in\mathcal{S}(n,r)}\chi_{\pi(\mu,S)}(x), J μ ( α ) ( x ) = r = 1 ∑ n ( r α ) r ! S ∈ S ( n , r ) ∑ χ π ( μ , S ) ( x ) ,
which has the same form as (7.1). However, such a rewriting is not possible in general. For example, if μ = ( 2 , 2 , 1 ) \mu=(2,2,1) μ = ( 2 , 2 , 1 ) , then the coefficient of ( α 4 ) \binom{\alpha}{4} ( 4 α ) has no expansion in terms of chromatic symmetric functions in which every coefficient is a nonnegative integer multiple of 4 ! 4! 4 ! .
Department of Mathematics, University of Pennsylvania, Philadelphia, PA 19104, USA Email address: jhaglund@math.upenn.edu
Department of Mathematics, Sungkyunkwan University, Suwon, Gyeonggi-do 16419, South Korea Email address: jangsookim@skku.edu
Department of Mathematics, Sungkyunkwan University, Suwon, Gyeonggi-do 16419, South Korea Email address: jaeseongoh@skku.edu
Department of Mathematics, Chungbuk National University, Cheongju 28644, South Korea Email address: meesueyoo@chungbuk.ac.kr