Hostname: page-component-7d8f8d645b-h5t26 Total loading time: 0 Render date: 2023-05-28T14:37:21.864Z Has data issue: false Feature Flags: { "useRatesEcommerce": true } hasContentIssue false

THE STABLE LIMIT DAHA AND THE DOUBLE DYCK PATH ALGEBRA

Published online by Cambridge University Press:  26 September 2022

Bogdan Ion*
Affiliation:
Department of Mathematics, University of Pittsburgh, Pittsburgh, PA 15260
Dongyu Wu
Affiliation:
Department of Mathematics, University of Pittsburgh, Pittsburgh, PA 15260; Beijing Institute of Mathematical Sciences And Applications, Beijing 101408 (dow16@pitt.edu, wudongyu@bimsa.cn)
Rights & Permissions[Opens in a new window]

Abstract

We study the compatibility of the action of the DAHA of type GL with two inverse systems of polynomial rings obtained from the standard Laurent polynomial representations. In both cases, the crucial analysis is that of the compatibility of the action of the Cherednik operators. Each case leads to a representation of a limit structure (the +/– stable limit DAHA) on a space of almost symmetric polynomials in infinitely many variables (the standard representation). As an application, we show that the defining representation of the double Dyck path algebra arises from the standard representation of the +stable limit DAHA.

Type
Research Article
Creative Commons
Creative Common License - CCCreative Common License - BYCreative Common License - NCCreative Common License - SA
This is an Open Access article, distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike licence (http://creativecommons.org/licenses/by-nc-sa/4.0), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the same Creative Commons licence is used to distribute the re-used or adapted article and the original article is properly cited. The written permission of Cambridge University Press must be obtained prior to any commercial use.
Copyright
© The Author(s), 2022. Published by Cambridge University Press

1 Introduction

There are subtle phenomena in Macdonald theory (the positivity of the composition Kostka functions [Reference Knop21, Conjecture 11.2]) that arise in the so-called stable limit, that is, they describe properties of objects that are associated to certain limits of polynomial functions when the number of variables approaches infinity. The object that provides the appropriate finite rank algebraic context in this case is the double affine Hecke algebra (DAHA) of type $\mathrm {GL}_k$. However, the only stable limit structure that was investigated more systematically is the stable limit of the spherical DAHA of type $\mathrm {GL}_k$, which is intimately related to several interconnected geometric contexts: the spherical subalgebra of the Hall algebra of the category of coherent sheaves on an elliptic curve [Reference Burban and Schiffmann4, Reference Schiffmann and Vasserot25], the convolution algebra in the equivariant K-theory of the Hilbert scheme of $\mathbb {A}^2$ [Reference Schiffmann and Vasserot26], the Feigin–Odesskii [Reference Feĭgin and Odesskiĭ9] shuffle algebra [Reference Feigin and Tsymbaliuk10, Reference Schiffmann and Vasserot26, Reference Negut22].

In the case of the spherical DAHA, the stable limit structure arises as an inverse limit of the corresponding finite rank structure. The DAHA $\mathscr {H}_k$ of type $\mathrm {GL}_k$ do not form an inverse system and a conceivable residual limit structure will not capture some relations that hold in particular finite ranks. The basic idea towards understanding the residual limit structure is to undertake a study of the limiting behaviour of individual elements (e.g., for a specific set of generators) in an inverse system of representations and understand the algebraic structure defined by the limit operators.

We pursue here this idea, as follows. Let $\mathscr {P}_k=\mathbb Q(\textbf {t},\textbf {q})[x_1^{\pm 1},\dots ,x_k^{\pm 1}]$ be the standard Laurent polynomial representation of $\mathscr {H}_k$; $\mathscr {P}_k$ is a faithful representation of $\mathscr {H}_k$. Let

$$ \begin{align*} \mathscr{P}_k^+=\mathbb Q(\textbf{t},\textbf{q})[x_1,\dots,x_k]\quad \text{and}\quad \mathscr{P}_k^-=\mathbb Q(\textbf{t},\textbf{q})[x_1^{-1},\dots,x_k^{-1}], \end{align*} $$

each being a faithful representation of a corresponding subalgebra of $\mathscr {H}_k$ denoted by $\mathscr {H}_k^+$ and, respectively, $\mathscr {H}_k^-$. Both $(\mathscr {P}_k^+)_{k\geq 2}$ and $(\mathscr {P}_k^-)_{k\geq 2}$ form graded inverse systems, with structure maps that specialize the last variable to $0$. The corresponding graded inverse limits are denoted by $\mathscr {P}_{\infty }^{\pm }$; they are sometimes referred to as the rings of formal polynomials in the variables $x_i^{\pm }$, $i\geq 1$. For us, an important role is played by the graded subrings $\mathscr {P}_{\mathrm {as}}^{\pm }\subset \mathscr {P}_{\infty }^{\pm }$ consisting of elements fixed by all simple transpositions of the variables, with the possible exception of finitely many. Following Knop [Reference Knop21], we refer to $\mathscr {P}_{\mathrm {as}}^{\pm }$ as the almost symmetric modules and to its elements as almost symmetric functions.

The action of the generators of $\mathscr {H}^{\pm }_k$ that act as Demazure–Lusztig operators or multiplications operators is immediately seen to be compatible with the inverse system; each operator will consequently induce a limit operator acting on the corresponding graded inverse limit $\mathscr {P}_{\infty }^{\pm }$. Therefore, in both cases, the crucial analysis is that of the action of the Cherednik operators and of their compatibility with the inverse system.

The compatibility of the action of the Cherednik operators in $\mathscr {H}_k^-$ with the inverse system $\mathscr {P}_k^-$ was established in [Reference Knop21]; the corresponding operators act on $\mathscr {P}_{\infty }^-$. Knop has also pointed out that $\mathscr {P}_{\mathrm {as}}^-\subset \mathscr {P}_{\infty }^-$ is stable under the action of the limit Cherednik operators and provides the natural context for the theory of nonsymmetric Macdonald functions (the stable limits of nonsymmetric Macdonald polynomials).

On the other hand, the analysis of the $\mathscr {H}_k^+$ stable limit action requires a more sophisticated approach. The action of the Cherednik operators in $\mathscr {H}_k^+$ is no longer compatible with the inverse system $\mathscr {P}_k^+$ and requires a careful investigation of their action in order to understand the obstructions to such a compatibility and what particular features might have a chance to manifest in the stable limit. We briefly discuss the main new developments here, postponing a more technical discussion for the later sections.

The failure of the compatibility of the action of the Cherednik operators in $\mathscr {H}_k^+$ with the inverse system $\mathscr {P}_k^+$ can be witnessed at the level of their spectrum and allows for a precise identification of the obstruction. As it turns out, the Cherednik operator $Y_i$ is compatible with the inverse system $x_i\mathscr {P}_k^+$, inducing an action of the limit operator on $x_i\mathscr {P}_{\infty }^+$. While this might be satisfactory from the point of view of a single Cherednik operator, it does not lead to a nonzero common domain for all Cherednik operators $Y_i$, $i\geq 1$. It is therefore necessary to adjust the action of $Y_i$ on a complement of $x_i\mathscr {P}_k^+$. We address this first difficulty uniformly, by introducing a new finite rank algebraic structure $\widetilde {\mathscr {H}}_k^+$, which we call the deformed DAHA, and its standard representation on $\mathscr {P}_k^+$. The actions of the Cherednik operator $Y_i$ and the deformed Cherednik operator $\widetilde {Y}_i$ coincide on $x_i\mathscr {P}_k^+$.

Nevertheless, $\widetilde {\mathscr {H}}_k^+$ is a more complicated structure, and, most notably, the deformed Cherednik operators no longer commute. Their action is still not compatible with the inverse system, but they satisfy instead a more subtle compatibility. To express this compatibility, we formulate a concept of limit that takes into consideration not only the inverse system but also the $\textbf {t}$-adic topology. This allows us to define limit operators $\widetilde {Y}_i$ acting, not on $\mathscr {P}_{\infty }^+$ but only on $\mathscr {P}_{\mathrm {as}}^+$, attesting again to the canonical nature of the almost symmetric module. On $x_i\mathscr {P}_{\mathrm {as}}^+$, the action of the limit operator $\widetilde {Y}_i$ coincides with the action of the limit Cherednik operator $Y_i$. This analysis occupies most of §6, culminating with the proof of Proposition 6.29.

We relate the action of the limit operators on $\mathscr {P}_{\mathrm {as}}^+$ to the following algebraic structure. Let $\mathscr {H}^+$ be the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by the elements $T_i$,$X_i$, and $Y_i$, $i\geq 1$, satisfying the following relations:

(1.1a)$$ \begin{align} \begin{gathered} T_{i}T_{j}=T_{j}T_{i}, \quad |i-j|>1,\\ T_{i}T_{i+1}T_{i}=T_{i+1}T_{i}T_{i+1}, \quad i\geq 1, \end{gathered} \end{align} $$
(1.1b)$$ \begin{align} (T_{i}-1)(T_{i}+\textbf{t})=0, \quad i\geq 1, \end{align} $$
(1.1c)$$ \begin{align} \begin{gathered} \textbf{t} T_i^{-1} X_i T_i^{-1}=X_{i+1}, \quad i\geq 1\\ T_{i}X_{j}=X_{j}T_{i}, \quad j\neq i,i+1,\\ X_i X_j=X_j X_i,\quad i,j\geq 1, \end{gathered} \end{align} $$
(1.1d)$$ \begin{align} \begin{gathered} \textbf{t}^{-1} T_i Y_i T_i=Y_{i+1}, \quad i\geq 1\\ T_{i}Y_{j}=Y_{j}T_{i}, \quad j\neq i,i+1,\\ Y_i Y_j=Y_j Y_i, \quad i,j\geq 1, \end{gathered} \end{align} $$
(1.1e)$$ \begin{align} Y_1 T_1 X_1=X_2 Y_1T_1. \end{align} $$

We call $\mathscr {H}^+$ the $^+$stable limit DAHA. There is a corresponding $^-$stable limit DAHA $\mathscr {H}^-$ and a canonical $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra anti-isomorphism between $\mathscr {H}^+$ and $\mathscr {H}^-$. As a result, we use the terminology stable limit DAHA to refer to either of them, depending on the context. Both algebras are closely related to the inductive limit of the group algebras of the braid groups $\mathcal {B}_k$ of k distinct points on the punctured torus. We refer to §4.2 for the precise relationship.

The action of the limit operators on $\mathscr {P}_{\mathrm {as}}^-$ can be easily seen to define a representation of $\mathscr {H}^-$. Our first main result is the following (Theorem 6.34).

Theorem A. The action of the limit operators on $\mathscr {P}_{\mathrm {as}}^+$ defines a representation of $\mathscr {H}^+$.

We call the representation of $\mathscr {H}^{\pm }$ on $\mathscr {P}_{\mathrm {as}}^{\pm }$ the standard representation of the stable limit DAHA. The main difficulty in establishing this result is the proof of the commutativity of the limit $\widetilde {Y}_i$ operators. On this account, the commutativity of the Cherednik operators that was lost by deforming their action on $\mathscr {P}_k^+$ is restored in the limit. The action of the operators $T_i$, $X_i$ and $Y_i\in \mathscr {H}^+$ on $\mathscr {P}_{\mathrm {as}}^+$ is denoted by $\mathscr {T}_i$, $\mathscr {X}_i$ and $\mathscr {Y}_i$, respectively.

One aspect that underscores the special nature of the positivity phenomenon in the Macdonald theory is the fact that the work around the former Macdonald positivity conjecture (first proved by Haiman [Reference Haiman17] in the equivalent formulation of the $n!$ conjecture) and the closely related work on the ring of diagonal harmonics [Reference Haiman18] relies on the geometry of Hilbert schemes of points in $\mathbb {A}^2$, and so far, a workable explanation based on DAHA technology—which was successful in settling all the other Macdonald conjectures—is missing. Another apparent dissimilarity is the fact that the representation theory relevant for the $n!$ conjecture is that of the symmetric group $S_n$, while all the other Macdonald conjectures are addressing phenomena rooted in the representation theory of reductive groups.

Following the work in [Reference Haiman18], a precise conjecture (the so-called Shuffle conjecture) was formulated in [Reference Haglund, Haiman, Loehr, Remmel and Ulyanov14] (see also [Reference Haglund13]) for the monomial expansion of the bigraded Frobenius characteristic of the ring of double harmonics. This conjecture and a number of further refinements and generalizations were recently proved [Reference Carlsson and Mellit6, Reference Mellit24]. In the process, Carlsson and Mellit revealed a new algebraic structure (the double Dyck path algebra $\mathbb {A}_{\textbf {q},\textbf {t}}$) together with a specific representation $V_{\bullet }$, which, aside from governing the combinatorics relevant for the shuffle conjecture, exhibits striking structural similarities with the family of DAHA of type $\mathrm {GL}_k$, $k\geq 2$. In [Reference Carlsson, Gorsky and Mellit5], a connection was established between $\mathbb {A}_{\textbf {q},\textbf {t}}$ and the geometry of Hilbert schemes; it shows that the representation $V_{\bullet }$ is closely connected with a geometric action in the equivariant K-theory of certain smooth strata in the parabolic flag Hilbert schemes of points in $\mathbb {A}^2$. This action is akin to the action of stable limit spherical DAHA on the equivariant K-theory of the Hilbert schemes of points in $\mathbb {A}^2$ [Reference Schiffmann and Vasserot25]. We refer to Definition 7.3, Proposition 7.5 and Theorem 7.7 for the definition of $\mathbb {A}_{\textbf {t},\textbf {q}}$, $V_{\bullet }$ and other structural details.

The structural similarities between $\mathbb {A}_{\textbf {q},\textbf {t}}$ and the family of DAHA of type $\mathrm {GL}_k$ are therefore important for connecting the geometry of Hilbert schemes with the DAHA. The algebra $\mathbb {A}_{\textbf {q},\textbf {t}}$ is a quiver algebra and its representation $V_{\bullet }=(V_k)_{k\geq 0}$ is a quiver representation; the subalgebra of $\mathbb {A}_{\textbf {q},\textbf {t}}$ generated by certain elements that correspond to loops that act at node k resembles quite closely the DAHA of type $\mathrm {GL}_k$. In fact, the generators in such a subalgebra satisfy all the expected DAHA relations except for one (see discussion in [Reference Carlsson and Mellit6, pg. 693–694]). However, it was expected that an explicit relationship with the DAHA exists. The main difficulty in establishing such a connection rests in explaining the relationship between the Cherednik operators and the operators $z_i$ in $\mathbb {A}_{\textbf {q},\textbf {t}}$.

We establish a direct connection between the standard representation of the $^+$stable limit DAHA and the double Dyck path algebra. To facilitate the comparison and not deviate from the traditional notation for parameters in double affine Hecke algebras, we will swap the role of the parameters $\textbf {q}$ and $\textbf {t}$ in the original definition of the double Dyck path algebra. Therefore, the connection we establish is between the $^+$stable limit DAHA and the algebra $\mathbb {A}_{\textbf {t},\textbf {q}}$. Specifically, we show that the standard representation of $\mathscr {H}^+$ can be used to construct a quiver representation of $\mathbb {A}_{\textbf {t},\textbf {q}}$. The standard representation $\mathscr {P}_{\mathrm {as}}^+$ has a natural filtration $\mathscr {P}(k)^+$, ${k\geq 0}$; the subspace $\mathscr {P}(k)^+$ consists of the elements of $\mathscr {P}_{\mathrm {as}}^+$ that are symmetric in the variables $x_i$, $i>k$. We denote by $\mathscr {P}_{\bullet }$ the complex of vector spaces $(\mathscr {P}(k)^+)_{k\geq 0}$. The vector spaces $\mathscr {P}(k)^+$ form an direct system and $\mathscr {P}_{\mathrm {as}}^+$ is their direct limit. Our second main result is the following (Theorem 7.13).

Theorem B. There is a $\mathbb {A}_{\textbf {t},\textbf {q}}$ action on $\mathscr {P}_{\bullet }$ such that all generators of $\mathbb {A}_{\textbf {t},\textbf {q}}$ corresponding to loops in Definition 7.3 (i.e., $T_i$, $y_i$, $z_i$) act as $\mathscr {T}_i$, $\mathscr {X}_i$, $\mathscr {Y}_i$.

In fact, this representation turns out to be precisely the defining representation of $\mathbb {A}_{\textbf {t},\textbf {q}}$ (Theorem 7.14)

Theorem C. The representations of $\mathbb {A}_{\textbf {t},\textbf {q}}$ on $\mathscr {P}_{\bullet }$ and $V_{\bullet }$ are isomorphic.

Since the notation in [Reference Carlsson and Mellit6] for the elements of the double Dyck path algebra that are relevant for its comparison with the action of $\mathscr {H}^+$ does not match the more traditional notation used in the DAHA literature, we include below, for the reader’s convenience, the conversion table between notation used here and the notation used in [Reference Carlsson and Mellit6] and [Reference Mellit24]:

It is important to note that all the generators of $\mathbb {A}_{\textbf {t},\textbf {q}}$ corresponding to loops act globally in $\mathscr {P}_{\bullet }$, by which we mean that their action on each $\mathscr {P}(k)^+$ is the restriction of their action on $\mathscr {P}_{\mathrm {as}}^+$. The vector spaces $V_k$, $k\geq 0$, also form a direct system, but the action of the elements $z_i\in \mathbb {A}_{\textbf {t},\textbf {q}}$ act only locally, by which we mean that the action is not compatible with the structure maps of the direct system. This obscures the fact that the representation $V_{\bullet }$ arises from a more fundamental representation on the direct limit of $V_k$, $k\geq 0$. In some sense, the isomorphism between $\mathscr {P}_{\bullet }$ and $V_{\bullet }$ straightens out this apparent incompatibility and explains how the limit Cherednik operators lead to the operators $z_i$ in $\mathbb {A}_{\textbf {t},\textbf {q}}$.

In [Reference Carlsson and Mellit6, pg. 694] (see also [Reference Carlsson, Gorsky and Mellit5, Remark 2.6]), there is one phrase commenting on a possible algebra isomorphism between $e_k\mathbb {A}_{\textbf {t},\textbf {q}} e_k$ (the subalgebra of $\mathbb {A}_{\textbf {t},\textbf {q}}$ generated by loops based at node k) and a ‘partially symmetrized’ version of the stable limit spherical DAHA, still to be defined for $k>0$. Our results, Theorem B and Theorem C, establish a different relationship between different objects, namely between the standard representations of $\mathscr {H}^+$ and $\mathbb {A}_{\textbf {t},\textbf {q}}$.

Our construction of the standard representations of the stable limit DAHA opens a number of immediate questions on the spectral theory of the limit Cherednik operators as well as on the structure of these representations. A second set of questions is related the possible applications of the stable limit DAHA—through the path opened by the work of Carlsson and Mellit—to the combinatorics and geometry of Hilbert schemes. We hope to pursue these questions in future publications.

2 Notation

2.1

We denote by $\mathbf {X}$ an infinite alphabet $x_1,x_2,\dots $ and by $\operatorname {\mathrm {Sym}}[\mathbf {X}]$ the ring of symmetric functions in $\mathbf {X}$. The field or ring of coefficients $\mathscr {K}\supseteq \mathbb Q$ will depend on the context. For any $k\geq 1$, we denote by $\overline {\mathbf {X}}_k$ the finite alphabet $x_1,x_2,\dots , x_k$ and by ${\mathbf {X}}_k$ the infinite alphabet $x_{k+1},x_{k+2},\dots $. $\operatorname {\mathrm {Sym}}[{\mathbf {X}}_k]$ will denote the ring of symmetric functions in ${\mathbf {X}}_k$. Furthermore, for any $1\leq k\leq m$, we denote by $\overline {\mathbf {X}}_{[k,m]}$ the finite alphabet $x_k,\dots , x_m$. As usual, we denote by $h_n[\mathbf {X}]$ (or $h_n[\overline {\mathbf {X}}_k]$ or $h_n[\mathbf {X}_k]$ or $h_n[\overline {\mathbf {X}}_{[k,m]}]$) the n-th complete symmetric functions (or polynomials) in the indicated alphabet, by $p_n[\mathbf {X}]$ (or $p_n[\overline {\mathbf {X}}_k]$ or $p_n[\mathbf {X}_k]$ or $p_n[\overline {\mathbf {X}}_{[k,m]}]$) the n-th power sum symmetric functions (or polynomials), and by $e_n[\mathbf {X}]$ (or $e_n[\overline {\mathbf {X}}_k]$ or $e_n[\mathbf {X}_k]$ or $e_n[\overline {\mathbf {X}}_{[k,m]}]$) the n-th elementary symmetric functions (or polynomials). The symmetric function $p_1[\mathbf {X}]=h_1[\mathbf {X}]=e_1[\mathbf {X}]$ is also denoted by $\mathbf {X}=x_1+x_2+\cdots $. For a partition $\lambda $, $m_{\lambda }[\mathbf {X}]$ (or $m_{\lambda }[\overline {\mathbf {X}}_k]$ or $m_{\lambda }[\mathbf {X}_k]$) denotes the monomial symmetric function (or polynomial) in the indicated alphabet.

2.2

Any action of the monoid $(\mathbb Z_{>0},\cdot )$ on the ring $\mathscr {K}$ extends to a canonical action by $\mathbb Q$-algebra morphisms on $\operatorname {\mathrm {Sym}}[\mathbf {X}]$. The morphism corresponding to the action of $n\in \mathbb Q_{>0}$ is denoted by $\mathfrak {p}_n$ and is defined by

$$ \begin{align*} \mathfrak{p}_n\cdot p_k[\mathbf{X}]=p_{nk}[\mathbf{X}], \quad k\geq 1. \end{align*} $$

In our context, $\mathscr {K}$ will be a ring of polynomials or a field of fractions generated by some finite set of parameters, the action of $(\mathbb Z_{>0},\cdot )$ on $\mathscr {K}$ is $\mathbb Q$-linear, and $\mathfrak {p}_n$ acts on parameters by raising them to the n-th power. For example, if $\mathscr {K}=\mathbb Q(\textbf {t},\textbf {q})$, then $\mathfrak {p}_n\cdot \textbf {t}=\textbf {t}^n, ~\mathfrak {p}_n\cdot \textbf {q}=\textbf {q}^n$.

Let R be a ring with an action of $(\mathbb Z_{>0},\cdot )$ by ring morphisms. Any ring morphism $\varphi : \operatorname {\mathrm {Sym}}[\mathbf {X}]\to R$ that is compatible with the action of $(\mathbb Z_{>0},\cdot )$ is uniquely determined by the image of $p_1[\mathbf {X}]=\mathbf {X}$. The image of $F[\mathbf {X}]\in \operatorname {\mathrm {Sym}}[\mathbf {X}]$ through $\varphi $ is usually denoted by $F[\varphi (\mathbf {X})]$ and called the plethystic evaluation (or substitution) of F at $\varphi (\mathbf {X})$.

The plethystic exponential $\textrm {Exp}[\mathbf {X}]$ is defined as

$$ \begin{align*} \textrm{Exp}[\mathbf{X}]=\sum_{n=0}^{\infty}h_n[\mathbf{X}]=\exp\left(\sum_{n=1}^{\infty} \frac{p_n[\mathbf{X}]}{n}\right). \end{align*} $$

2.3

We will use some symmetric polynomials that are related to the complete homogeneous symmetric functions via plethystic substitution. More precisely, let $h_n[(1-\textbf {t})\overline {\mathbf {X}}_k]$ be the symmetric polynomial obtained from the symmetric function $h_n[(1-\textbf {t})\mathbf {X}]$ by specializing to $0$ the elements of the alphabet ${\mathbf {X}}_k$. The corresponding notation applies to $h_n[(\textbf {t}-1)\mathbf {X}]$ and other plethystic substitutions.

For example, from the Newton identities for the complete homogeneous symmetric functions, we can easily see that (for $x=x_1$)

(2.1)$$ \begin{align} h_n[(1-\textbf{t})x]=(1-\textbf{t})x^n, \quad h_n[(\textbf{t}-1)x]=\textbf{t}^{n-1}(\textbf{t}-1)x^n, n\geq 1, \end{align} $$

and $h_0[(1-\textbf {t})x]=h_0[(\textbf {t}-1)x]=1$. The classical convolution formula $\displaystyle h_n[\mathbf {X}]=\sum _{i+j=n} h_i[\mathbf {X}] h_j[\mathbf {X}]$ has the following counterparts

(2.2)$$ \begin{align} h_n[(1-\textbf{t})\overline{\mathbf{X}}_k]=\sum_{i+j=n} h_i[(1-\textbf{t})\overline{\mathbf{X}}_{k-1}]h_j[(1-\textbf{t})x_k] \end{align} $$

and

(2.3)$$ \begin{align} h_n[(1-\textbf{t})\overline{\mathbf{X}}_{k-1}]=\sum_{i+j=n} h_i[(1-\textbf{t})\overline{\mathbf{X}}_{k}]h_j[(\textbf{t}-1)x_k]. \end{align} $$

We will make use of such formulas applied to similar situations—for example, with $\overline {\mathbf {X}}_{[k+1,m]}$, $m>k$, replacing $\overline {\mathbf {X}}_{k}$ in the formulas above.

2.4

For any $k\geq 1$, let $\mathscr {P}_k=\mathbb Q(\textbf {t},\textbf {q})[x_1^{\pm 1},\dots ,x_k^{\pm 1}]$ be the ring of Laurent polynomials in the variables $x_1,\dots , x_k$. The symmetric group $S_k$ acts on $\mathscr {P}_k$ by permuting the variables. We denote by $s_i$ the simple transposition that interchanges $x_i$ and $x_{i+1}$ and is fixing all the other variables. The polynomial subrings

$$ \begin{align*} \mathscr{P}_k^+=\mathbb Q(\textbf{t},\textbf{q})[x_1,\dots,x_k]\quad \text{and}\quad \mathscr{P}_k^-=\mathbb Q (\textbf{t}, \textbf{q})[x_1^{-1},\dots,x_k^{-1}] \end{align*} $$

are stable under the action of $S_k$.

2.5

Let

$$ \begin{align*} \pi_k:\mathscr{P}^+_k\rightarrow \mathscr{P}^+_{k-1},\quad \text{and } \quad \iota_k:\mathscr{P}_{k-1}^{+}\rightarrow \mathscr{P}_{k}^{+} \end{align*} $$

be the evaluation morphism that maps $x_{k}$ to $0$ and, respectively, the canonical inclusion. The definitions and facts that we discuss below have straightforward analogues associated to the sequence of rings $\mathscr {P}_k^-$, $k\geq 1$. To avoid introducing additional notation, we will use $\pi _k$ and $\iota _k$ for the corresponding morphisms between the rings $\mathscr {P}_k^-$ and $\mathscr {P}_{k-1}^-$ (to be clear, in this case, the evaluation morphism $\pi _k$ maps $x_k^{-1}$ to $0$), with the hope that the necessary distinction will be clear from the context (in fact, the objects associated with $\mathscr {P}_k^-$ will appear only in §5). We adopt the same convention for all the other morphisms considered in this subsection: $\iota _{n,k}$, $\Pi _k$, $I_k$, $J_k$, J.

The rings $\mathscr {P}_k^{\pm }$, $k\geq 1$ form an graded inverse system. We will use the notation $\mathscr {P}_{\infty }^{\pm }$ for the graded inverse limit ring $\displaystyle \lim _{\longleftarrow }\mathscr {P}_k^{\pm }$. The graded inverse limit rings are sometimes referred to in the literature as the rings of formal polynomials in the variables $x_i^{\pm }$, $i\geq 1$. We denote by $\Pi _k: \displaystyle \lim _{\longleftarrow }\mathscr {P}_k^{\pm }\to \mathscr {P}_k^{\pm }$ the canonical morphism.

The inductive limit $\displaystyle \lim _{{\longrightarrow }} \mathscr {P}_k^+$ is canonically isomorphic with the ring $\mathbb {Q}(\textbf {t},\textbf {q})[x_1,x_2,\dots ]$ of polynomials in infinitely many variables (and similarly for $\displaystyle \lim _{{\longrightarrow }} \mathscr {P}_k^-$). As with the projective limit, the rings $\displaystyle \lim _{\substack {\longrightarrow \\ k\geq n}} \mathscr {P}_k^{\pm }$ and $\displaystyle \lim _{{\longrightarrow }} \mathscr {P}_k^{\pm }$ are canonically isomorphic. We denote by $I_k: \displaystyle \mathscr {P}_k^{\pm }\to \lim _{{\longrightarrow }} \mathscr {P}_k^{\pm }$ the canonical morphism.

The following diagram is commutative:

where each horizontal map represents the identity map. For a fixed $n\geq 1$, denote by $\iota _{n,k}: \mathscr {P}_n^{\pm }\to \mathscr {P}_k^{\pm }$, $n\leq k$, the canonical inclusion. The sequence of maps $\iota _{n,k}: \mathscr {P}_n^{\pm }\to \mathscr {P}_k^{\pm }$, $k\geq n$ is compatible with the structure maps $\pi _k$. Therefore, they induce a morphism

$$ \begin{align*}\mathscr{P}_n^{\pm}\to \lim_{\substack{\longrightarrow \\ k\geq n}} \mathscr{P}_k^{\pm}\cong \displaystyle \lim_{{\longrightarrow }} \mathscr{P}_k^{\pm}. \end{align*} $$

Furthermore, these maps are compatible with the structure maps $\iota _n$ and therefore induce a morphism

$$ \begin{align*}J: \lim_{{\longrightarrow}} \mathscr{P}_k^{\pm} \to \lim_{\longleftarrow}\mathscr{P}_k^{\pm}. \end{align*} $$

By construction, $\Pi _kJI_k: \mathscr {P}_k^{\pm }\to \mathscr {P}_k^{\pm }$ is the identity function. We denote $J_k=JI_k: \mathscr {P}_k^{\pm }\to \mathscr {P}_{\infty }^{\pm }$.

2.6

For any $n\geq 1$, a sequence of operators $A_k:\mathscr {P}_k^{\pm }\to \mathscr {P}_k^{\pm }$, $k\geq n$ compatible with the inverse system induces a (limit) operator $A: \mathscr {P}_{\infty }^{\pm }\to \mathscr {P}_{\infty }^{\pm }$. We have $A_k=\Pi _k A J_k$. For example, the sequence $A_k=s_n$, $k\geq n$, given by the action of the simple transposition $s_n$, induces a limit operator $s_n$ acting on $\mathscr {P}_{\infty }^{\pm }$. In turn, this leads to an action of the infinite symmetric group $S(\infty )$ (the inductive limit of $S_k$, $k\geq 1$) on $P_{\infty }^{\pm }$.

For any $k\geq 0$, denote

$$ \begin{align*} \mathscr{P}(k)^{\pm}=\{F\in \mathscr{P}_{\infty}^{-}\ |\ s_i F=F,~ \text{for all } i>k \}. \end{align*} $$

From the definition, it is clear that ${\mathscr {P}}(k)^{\pm }\subset {\mathscr {P}}(k+1)^{\pm }$. Also, ${\mathscr {P}}(0)^{+}$ is the ring of symmetric functions $\operatorname {\mathrm {Sym}}[\mathbf {X}]$, and, more generally, for any $k\leq 1$, the multiplication map

$$ \begin{align*} \mathscr{P}_k^{+}\otimes \operatorname{\mathrm{Sym}}[\mathbf{X}_k] \cong\mathscr{P}(k)^+ \end{align*} $$

is an algebra isomorphism.

2.7

An important role will be played by the graded subrings $\mathscr {P}_{\mathrm {as}}^{\pm }\subset \mathscr {P}_{\infty }^{\pm }$, defined as the inductive limit of the spaces $\mathscr {P}(k)^{\pm }$:

$$ \begin{align*} {\mathscr{P}}_{\text{as}}^{\pm}=\bigcup_{k\geq 0}{\mathscr{P}}(k)^{\pm}. \end{align*} $$

More concretely, an element of $\mathscr {P}_{\mathrm {as}}^{\pm }$ must be fixed by all simple transpositions with the possible exception of finitely many. Following Knop [Reference Knop21], we refer to $\mathscr {P}_{\mathrm {as}}^{\pm }$ as the almost symmetric modules.

3 Preliminaries

3.1 Affine Hecke Algebras

Definition 3.1. The affine Hecke algebra $\mathscr {A}_k$ of type $GL_k$ is the $\mathbb {Q}(\textbf {t})$-algebra generated by

$$ \begin{align*} T_1,\dots,T_{k-1},X_1^{\pm1},\dots,X_k^{\pm1} \end{align*} $$

satisfying the following relations:

(3.1a)$$ \begin{align} \begin{gathered} T_{i}T_{j}=T_{j}T_{i}, \quad |i-j|>1,\\ T_{i}T_{i+1}T_{i}=T_{i+1}T_{i}T_{i+1}, \quad 1\leq i\leq k-2, \end{gathered} \end{align} $$

(3.1b)$$ \begin{align} (T_{i}-1)(T_{i}+\textbf{t})=0, \quad 1\leq i\leq k-1, \end{align} $$
(3.1c)$$ \begin{align} \begin{gathered} \textbf{t} T_i^{-1} X_i T_i^{-1}=X_{i+1}, \quad 1\leq i\leq k-1\\ T_{i}X_{j}=X_{j}T_{i},\quad j\neq i,i+1,\\ X_i X_j=X_j X_i \quad 1\leq i,j\leq k. \end{gathered} \end{align} $$

The definition above is based on the Bernstein presentation of the affine Hecke algebra $\mathscr {A}_k$. The following presentation of $\mathscr {A}_k$ will also be used.

Proposition 3.2. The affine Hecke algebra $\mathscr {A}_k$ is generated by

$$ \begin{align*} \widetilde{T}_0,T_1,\dots,T_{k-1},\widetilde{\omega}_k^{\pm 1} \end{align*} $$

satisfying the relations:

(3.2a)$$ \begin{align} \begin{gathered} T_{i}T_{j}=T_{j}T_{i}, \quad |i-j|>1,\\ T_{i}T_{i+1}T_{i}=T_{i+1}T_{i}T_{i+1}, \quad 1\leq i\leq k-2,\\ (T_{i}-1)(T_{i}+\textbf{t})=0, \quad 1\leq i\leq k-1, \end{gathered} \end{align} $$
(3.2b)$$ \begin{align} \begin{gathered} T_{i} \widetilde{T}_0=\widetilde{T}_0 T_{i}, \quad 2\leq i\leq k-1,\\ \widetilde{T}_0 T_{1}\widetilde{T}_0=T_{1}\widetilde{T}_0 T_{1}, \\ (\widetilde{T}_0-1)(\widetilde{T}_0+\textbf{t})=0, \end{gathered} \end{align} $$
(3.2c)$$ \begin{align} \widetilde{\omega}_k T_i \widetilde{\omega}_k^{-1}=T_{i-1} \textrm{ for }1\leq i\leq k-1,\quad \widetilde{\omega}_k \widetilde{T}_0 \widetilde{\omega}_k^{-1}=T_{k-1}. \end{align} $$

Remark 3.3. $\widetilde {T}_0,\widetilde {\omega }_k$ and $X_1,\dots ,X_k$ are related by

$$ \begin{align*}\widetilde{\omega}_k=\textbf{t}^{1-k} T_{k-1}\dots T_{1}X_{1}^{-1}= X_k^{-1}T_{k-1}^{-1}\dots T_{1}^{-1}, \end{align*} $$
$$ \begin{align*} \widetilde{T}_0=\widetilde{\omega}_k^{-1}T_{k-1}\widetilde{\omega}_k=\widetilde{\omega}_k T_1\widetilde{\omega}_k^{-1} =\textbf{t}^{k-1}X_1X_k^{-1}T_1^{-1}\dots T_{k-1}^{-1}\dots T_1^{-1}. \end{align*} $$

Notation 3.4. We denote by $\mathscr {A}_k^+$ the subalgebra of $\mathscr {A}_k$ generated by $T_i$, $i\leq k-1$, and $X_i$, $i\leq k$, or equivalently, by $T_i$, $i\leq k-1$, and $\widetilde {\omega }_k^{-1}$. Similarly, we denote by $\mathscr {A}_k^-$ the subalgebra of $\mathscr {A}_k$ generated by $T_i$, $i\leq k-1$, and $X^{-1}_i$, $i\leq k$, or equivalently, by $T_i$, $i\leq k-1$, and $\widetilde {\omega }_k$.

3.2 Double Affine Hecke Algebras

Definition 3.5. The double affine Hecke algebra (DAHA) $\mathscr {H}_k$, $k\geq 2$, of type $GL_{k}$ is the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by $T_0,T_1,\dots ,T_{k-1}$, $X_1^{\pm 1},\dots ,X_k^{\pm 1}$, and $\omega _k^{\pm 1}$ satisfying equations (3.1a), (3.1b) and (3.1c) and the following relations:

(3.3a)$$ \begin{align} \begin{gathered} T_{i} T_0=T_0 T_{i}, \quad 2\leq i\leq k-1,\\ T_0 T_{1}T_0=T_{1}T_0 T_{1}, \end{gathered} \end{align} $$

(3.3b)$$ \begin{align} (T_0-1)(T_0+\textbf{t})=0, \end{align} $$
(3.3c)$$ \begin{align} \begin{gathered} \omega_k T_i \omega_k^{-1}=T_{i-1}, \quad 2\leq i\leq k-1,\\ \omega_k T_1 \omega_k^{-1}=T_0, \quad \omega_k T_0 \omega_k^{-1}=T_{k-1}, \end{gathered} \end{align} $$
(3.3d)$$ \begin{align} \begin{gathered} \omega_k X_{i+1}\omega_k^{-1}=X_i, \quad 1\leq i\leq k-1,\quad \omega_k X_1 \omega_k^{-1}=\textbf{q}^{-1}X_k. \end{gathered} \end{align} $$

The double affine Hecke algebras were first introduced, in greater generality, by Cherednik [Reference Cherednik7]. Under one of the possible sets of conventions, the definition of the DAHA of type $GL_k$ can be found, for example, in [Reference Cherednik8, §3.7]. The presentation in Definition 3.5 is consistent with the conventions in [Reference Ion and Sahi19]. We will also make use of the following equivalent presentation of $\mathscr {H}_k$.

Proposition 3.6. The algebra $\mathscr {H}_k$, $k\geq 2$, is the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by the elements $T_1,\dots ,T_{k-1}$, $X_1^{\pm 1},\dots ,X_k^{\pm 1}$, $Y_1^{\pm 1},\dots ,Y_k^{\pm 1}$ satisfying equations (3.1a), (3.1b) and (3.1c) and the following relations:

(3.4a)$$ \begin{align} \begin{gathered} \textbf{t}^{-1} T_i Y_i T_i=Y_{i+1}, \quad 1\leq i\leq k-1\\ T_{i}Y_{j}=Y_{j}T_{i}, \quad j\neq i,i+1,\\ Y_i Y_j=Y_j Y_i \quad 1\leq i,j\leq k, \end{gathered} \end{align} $$
(3.4b)$$ \begin{align} Y_1 T_1 X_1=X_2 Y_1T_1, \end{align} $$
(3.4c)$$ \begin{align} Y_1 X_1\dots X_k= \textbf{q} X_1\dots X_k Y_1. \end{align} $$

Remark 3.7. $Y_1,\dots ,Y_k$ and $\omega _k$ are related by

$$ \begin{align*} Y_i = \textbf{t}^{k+1-i}T_{i-1} \dots T_{1}\omega_k^{-1} T_{k-1}^{-1}\dots T_{i}^{-1}. \end{align*} $$

The presentation in Proposition 3.6 can be derived from Definition 3.5. See also [Reference Schiffmann and Vasserot25, §2.1]; the generators $X_i,Y_i$ in [Reference Schiffmann and Vasserot25] correspond to $X_i^{-1}, Y_i^{-1}$ in our notation, and the relation [Reference Schiffmann and Vasserot25, (2.7)] is equivalent to equation (3.4b) modulo the relations (3.1c) and (3.4a).

Notation 3.8. We denote by $\mathscr {H}_k^+$ the subalgebra of $\mathscr {H}_k$ generated by $T_i$, $i\leq k-1$, and $X_i$, $Y_i$, $i\leq k$. Similarly, we denote by $\mathscr {H}_k^-$ the subalgebra of $\mathscr {H}_k$ generated by $T_i$,$i\leq k-1$, and $X^{-1}_i$, $Y^{-1}_i$, $i\leq k$.

3.3

The representation below is called the standard representation of $\mathscr {H}_k$ [Reference Cherednik7, Theorem 2.3]

Proposition 3.9. The following formulas define a faithful representation of $\mathscr {H}_k$ on $\mathscr {P}_k$:

(3.5)$$ \begin{align} \begin{split} T_i f(x_1,\dots,x_k) &= s_i f(x_1,\dots,x_k) +(1-\textbf{t})x_i\frac{1-s_i}{x_i-x_{i+1}}f(x_1,\dots,x_k), \\ \widetilde{\omega}_k f(x_1,\dots,x_k) &= \textbf{t}^{1-k}T_{k-1}\dots T_{1}x_1^{-1} f(x_1,\dots,x_k),\\ \omega_k f(x_1,\dots,x_k) &= f(\textbf{q}^{-1} x_k,x_1,\dots,x_{k-1}). \end{split} \end{align} $$

Furthermore, for any $1\leq i\leq k$, the elements $X_i$ act as left multiplication by $x_i$.

By restriction, we obtain the corresponding standard representations of $\mathscr {H}^+_k$ and $\mathscr {H}^-_k$.

Corollary 3.10. The subspace $\mathscr {P}_k^{\pm }$ is stable under the action of $\mathscr {H}_k^{\pm }$. The corresponding representation of $\mathscr {H}_k^{\pm }$ on $\mathscr {P}_k^{\pm }$ is faithful.

All the standard representations are induced representations from a one-dimensional representation of the corresponding affine Hecke subalgebra.

Convention 3.11. There will be elements denoted by the same symbol that belong to several (often infinitely many) algebras. The notation does not keep track of this information if it is implicit from the context. When necessary, we will add the superscript $(k)$ (e.g., $Y_i^{(k)}\in \mathscr {H}_k$) to make such information explicit.

4 The stable limit DAHA

4.1

We introduce a pair of closely related algebras with infinitely many generators, which we call stable limit DAHAs.

Definition 4.1. Let $\mathscr {H}^+$ be the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by the elements $T_i$,$X_i$, and $Y_i$, $i\geq 1$, satisfying the following relations:

(4.1a)$$ \begin{align} \begin{gathered} T_{i}T_{j}=T_{j}T_{i}, \quad |i-j|>1,\\ T_{i}T_{i+1}T_{i}=T_{i+1}T_{i}T_{i+1}, \quad i\geq 1, \end{gathered} \end{align} $$
(4.1b)$$ \begin{align} (T_{i}-1)(T_{i}+\textbf{t})=0, \quad i\geq 1, \end{align} $$
(4.1c)$$ \begin{align} \begin{gathered} \textbf{t} T_i^{-1} X_i T_i^{-1}=X_{i+1}, \quad i\geq 1\\ T_{i}X_{j}=X_{j}T_{i}, \quad j\neq i,i+1,\\ X_i X_j=X_j X_i,\quad i,j\geq 1, \end{gathered} \end{align} $$
(4.1d)$$ \begin{align} \begin{gathered} \textbf{t}^{-1} T_i Y_i T_i=Y_{i+1}, \quad i\geq 1\\ T_{i}Y_{j}=Y_{j}T_{i}, \quad j\neq i,i+1,\\ Y_i Y_j=Y_j Y_i, \quad i,j\geq 1, \end{gathered} \end{align} $$
(4.1e)$$ \begin{align} Y_1 T_1 X_1=X_2 Y_1T_1. \end{align} $$

We will call $\mathscr {H}^+$ the $^+$stable limit DAHA.

Note that, as opposed to the corresponding elements of $\mathscr {H}_k$, the elements $X_i$, $Y_i$ are not invertible in $\mathscr {H}^+$. Similarly, we define the $^-$stable limit DAHA.

Definition 4.2. Let $\mathscr {H}^-$ be the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by the elements $T_i$,$X^{-1}_i$, and $Y^{-1}_i$, $i\geq 1$, satisfying the following relations:

(4.2a)$$ \begin{align} \begin{gathered} T_{i}T_{j}=T_{j}T_{i}, \quad |i-j|>1,\\ T_{i}T_{i+1}T_{i}=T_{i+1}T_{i}T_{i+1}, \quad i\geq 1, \end{gathered} \end{align} $$
(4.2b)$$ \begin{align} (T_{i}-1)(T_{i}+\textbf{t})=0, \quad i\geq 1, \end{align} $$

(4.2c)$$ \begin{align} \begin{gathered} \textbf{t}^{-1} T_i X_i^{-1} T_i=X_{i+1}^{-1}, \quad i\geq 1\\ T_{i}X^{-1}_{j}=X^{-1}_{j}T_{i}, \quad j\neq i,i+1,,\\ X^{-1}_i X^{-1}_j=X^{-1}_j X^{-1}_i,\quad i,j\geq 1, \end{gathered} \end{align} $$
(4.2d)$$ \begin{align} \begin{gathered} \textbf{t} T_i^{-1} Y^{-1}_i T^{-1}_i=Y^{-1}_{i+1}, \quad i\geq 1\\ T_{i}Y^{-1}_{j}=Y^{-1}_{j}T_{i}, \quad j\neq i,i+1,\\ Y^{-1}_i Y^{-1}_j=Y^{-1}_j Y^{-1}_i, \quad i,j\geq 1, \end{gathered} \end{align} $$
(4.2e)$$ \begin{align} X^{-1}_1 T^{-1}_1 Y^{-1}_1=T^{-1}_1 Y^{-1}_1 X^{-1}_2. \end{align} $$

We will call $\mathscr {H}^-$ the $^-$stable limit DAHA.

The map

$$ \begin{align*} \mathfrak{e}: \mathscr{H}^+\to \mathscr{H}^- \end{align*} $$

that sends $T_i$, $X_i$, $Y_i$ to $T_i$, $Y^{-1}_i$, $X^{-1}_i$, respectively, extends to an anti-isomorphism of $\mathbb {Q}(\textbf {t},\textbf {q})$-algebras. For this reason, we regard $\mathscr {H}^+$ and $\mathscr {H}^-$ as capturing the same structure, and we will use the terminology stable limit DAHA to refer to either of them, depending on the context.

The terminology is justified by the fact that these structures arise from analyzing the stabilization phenomena for the standard representations of the algebras $\mathscr {H}^{\pm }_k$ as k approaches infinity. We will construct natural representations of both $\mathscr {H}^{\pm }$ in this fashion. To point more directly to his connection consider the following.

Definition 4.3. For any $k\geq 2$, denote by $\mathscr {H}^+(k)$ the subalgebra of $\mathscr {H}^+$ generated by $T_i$, $i\leq k-1$, and $X_i$, $Y_i$, $1\leq i\leq k$; $\mathscr {H}^-(k)$ denotes the subalgebra of $\mathscr {H}^-$ generated by $T_i$, $i\leq k-1$, and $X^{-1}_i$, $Y^{-1}_i$, $1\leq i\leq k$.

Remark 4.4. We have a canonical surjective $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra morphisms $\mathscr {H}^{\pm }(k)\to \mathscr {H}^{\pm }_k$.

4.2

The stable limit DAHA is also closely related to the inductive limit of the braid groups $\mathcal {B}_k$ of k distinct points on the punctured torus. Indeed, we have the following presentation of $\mathcal {B}_k$ [Reference Bellingeri1, Theorem 1.1] (see also [Reference Mellit24, Theorem 5.1]).

Theorem 4.5. For $k\geq 2$, the group $\mathcal {B}_k$ is generated by the elements $\sigma _i$, $1\leq i\leq k-1$, and $X_i$, $Y_i$, $1\leq i\leq k$ satisfying the following relations:

(4.3a)$$ \begin{align} \begin{gathered} \sigma_{i}\sigma_{j}=\sigma_{j}\sigma_{i}, \quad |i-j|>1,\\ \sigma_{i}\sigma_{i+1}\sigma_{i}=\sigma_{i+1}\sigma_{i}\sigma_{i+1}, \quad1\leq i\leq k-2, \end{gathered} \end{align} $$
(4.3b)$$ \begin{align} \begin{gathered} \sigma_i^{-1} X_i \sigma_i^{-1}=X_{i+1}, \quad 1\leq i\leq k-1\\ \sigma_{i}X_{j}=X_{j}\sigma_{i}, \quad j\neq i,i+1,\\ X_i X_j=X_j X_i,\quad 1\leq i,j\leq k, \end{gathered} \end{align} $$

(4.3c)$$ \begin{align} \begin{gathered} \sigma_i Y_i \sigma_i=Y_{i+1}, \quad1\leq i\leq k-1\\ \sigma_{i}Y_{j}=Y_{j}\sigma_{i}, \quad j\neq i,i+1,\\ Y_i Y_j=Y_j Y_i, \quad 1\leq i,j\leq k, \end{gathered} \end{align} $$
(4.3d)$$ \begin{align} Y_1 \sigma_1 X_1=X_2 Y_1\sigma_1. \end{align} $$

The notation here and the one in [Reference Bellingeri1] are related as follows: $X_1=b_1$, $Y_1=a_1^{-1}$.

Denote by $\mathcal {B}_k^+$ the submonoid of $\mathcal {B}_k$ generated by $\sigma _i,\sigma _i^{-1}$, $1\leq i\leq k$, and $X_i$, $Y_i$, $1\leq i\leq k$. The canonical inclusion maps between the monoids $\mathcal {B}^+_k$ consitute a direct system. The direct limit monoid is denoted by $\mathcal {B}^+_{\infty }$. A comparison between the relations in Theorem 4.5 and those in Definition 4.1 (with $\sigma _i$ mapping to $\textbf {t}^{-\frac {1}{2}}T_i$, $i\geq 1$) shows that $\mathscr {H}^+$ is the quotient of the monoid ring of $\mathcal {B}^+_{\infty }$ by the ideal generated by the quadratic relations (4.1b). A similar relationship connects $\mathscr {H}^-$ and the monoid $\mathcal {B}^-_{\infty }$, the inductive limit of the submonoids $\mathcal {B}_k^-$ of $\mathcal {B}_k$ generated by $\sigma _i,\sigma _i^{-1}$, $1\leq i\leq k$, and $X_i^{-1}$, $Y_i^{-1}$, $1\leq i\leq k$.

4.3

Let

$$ \begin{align*} \iota: \mathscr{H}^+\to \mathscr{H}^+ \end{align*} $$

be the $\mathbb {Q}$-algebra automorphism that sends $T_i$ to $T_i^{-1}$, swaps $X_i$ and $Y_i$ and inverts the parameters $\textbf {t}$ and $\textbf {q}$. It is easy to check that all relations are preserved and that $\iota $ is an involution.

Finally, let us define two $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra endomorphisms

$$ \begin{align*} \mathfrak{a},\mathfrak{b}: \mathscr{H}^+\to \mathscr{H}^+ \end{align*} $$

as follows. For all $i\geq 1$,

$$ \begin{align*} \mathfrak{a}(T_i)&=T_i, & \mathfrak{a}(X_i)&=X_i, &\mathfrak{a}(Y_i)&=\textbf{t}^{1-i}(T_{i-1}\cdots T_1)(T_1\cdots T_{i-1})X_iY_i, \\ \mathfrak{b}(T_i)&=T_i, & \mathfrak{b}(Y_i)&=Y_i, &\mathfrak{b}(X_i)&=\textbf{t}^{i-1}(T^{-1}_{i-1}\cdots T^{-1}_1)(T^{-1}_1\cdots T^{-1}_{i-1})Y_iX_i. \end{align*} $$

The fact that $\mathfrak {a}$ preserves the defining relations can be directly checked. The only relations that require a small computation are the commutativity relations between the $\mathfrak {a}(Y_i)$. All the commutativity relations follow, in fact, from the commutativity relation between $\mathfrak {a}(Y_1)$ and $\mathfrak {a}(Y_2)$. This can be verified as follows:

$$ \begin{align*} T_1X_1 Y_1T_1X_1 Y_1 &= T_1X_1 X_2Y_1T_1 Y_1 & & \text{by (4.1e)}\\ &= X_1X_2 T_1Y_1T_1 Y_1 & & \text{by (4.1c)} \\ &= X_1 X_2 Y_1 T_1Y_1T_1 & & \text{by (4.1d)} \\ &= X_1 Y_1T_1X_1 Y_1 T_1. & & \text{by (4.1e)} \end{align*} $$

The fact that $\mathfrak {a}$ is a morphism implies that $\mathfrak {b}$ is a morphism since

$$ \begin{align*}\mathfrak{b}=\iota \mathfrak{a}\iota. \end{align*} $$

Remark that the quadratic relation (4.1b) was not used in the verification of the other relations. Therefore, $\mathfrak {a}$, $\mathfrak {b}$ and $\iota $ are not only endomorphisms of $\mathscr {H}^+$ but also endomorphisms of the underlying monoid.

Proposition 4.6. The endomorphism $\mathfrak {a}$ and $\mathfrak {b}$ of $\mathscr {H}^+$ generate a free monoid.

Proof. The double affine Hecke algebra $\mathscr {H}_k$ has endomorphisms defined by the same formulas as $\mathfrak {a}$, $\mathfrak {b}$, with the obvious constraint on the label i. Indeed, with this definition, $\mathfrak {a}$ and $\mathfrak {b}$ correspond, respectively, to $\rho _2$ and $\rho _1$ in [Reference Schiffmann and Vasserot25, §2.1] or to $\mathfrak {a}$ and $\mathfrak {b}^{-1}$ in [Reference Ion and Sahi19, Chapter 6]. Furthermore, the corresponding $\mathfrak {a}$, $\mathfrak {b}$ are in fact automorphisms of $\mathscr {H}_k$ and generate a copy of the braid group on three strands (see, e.g., [Reference Ion and Sahi19, Theorem 6.4]). They stabilize the subalgebra $\mathscr {H}^+_k$, and the restrictions on $\mathfrak {a}$, $\mathfrak {b}$ to $\mathscr {H}^+_k$ generate the free monoid on two generators. Since, as pointed out in Remark 4.4, $\mathscr {H}^+_k$ is a quotient of the subalgebra $\mathscr {H}^+(k)\subseteq \mathscr {H}^+$ and $\mathscr {H}^+(k)$ is stable under the action of $\mathfrak {a}$ and $\mathfrak {b}$, we obtain that the endomorphisms $\mathfrak {a}$, $\mathfrak {b}$ of $\mathscr {H}^+$ generate a free monoid.

Remark 4.7. We can regard the braid group on three strands $B_3$ as generated by $\mathfrak {a}, \mathfrak {b}^{-1}$ satisfying the braid relation

$$ \begin{align*}\mathfrak{a}\mathfrak{b}^{-1}\mathfrak{a}=\mathfrak{b}^{-1}\mathfrak{a}\mathfrak{b}^{-1}. \end{align*} $$

There is a surjective group morphism $B_3\to \operatorname {\mathrm {SL}}(2,\mathbb Z)$ which maps $\mathfrak {a}$ to $A= \begin {bmatrix}1&1\\ 0&1 \end {bmatrix}$ and $\mathfrak {b}$ to $ B=\begin {bmatrix}1&0\\ 1&1 \end {bmatrix}$. The image of the free monoid $B_3^+$ generated by $\mathfrak {a}$ and $\mathfrak {b}$ maps bijectively to the free monoid generated by A and B which consists of the set $\operatorname {\mathrm {SL}}(2,\mathbb Z)^+$ of matrices with nonnegative integer entries.

Similar facts, pertaining to the algebra $\mathscr {H}^-$, can be recorded with the help of the antimorphism $\mathfrak {e}$.

4.4

It is important to remark that the algebras $\mathscr {H}_k^+$, $k\geq 2$, do not form an inverse system in any natural way and therefore $\mathscr {H}^+$ cannot be directly thought of as an inverse limit of $\mathscr {H}_k^+$, $k\geq 2$. However, we will show that both $\mathscr {H}^{\pm }$ acquire natural representations that are constructed by considering inverse systems built from the standard representations of $\mathscr {H}_k^{\pm }$, $k\geq 2$.

5 The standard representation of the ${}^{-}$stable limit DAHA

5.1

The results in the remainder of this section are due to Knop [Reference Knop21]. We briefly recall here the emerging structure. We refer to §2.5 for the relevant notation. The map $\pi _k:\mathscr {P}^-_k\rightarrow \mathscr {P}^-_{k-1}$ is partially compatible with the actions of $\mathscr {A}_k^-$ and $\mathscr {A}_{k-1}^-$. More precisely,

(5.1)$$ \begin{align} \begin{split} \pi_k T_i &= T_i \pi_k,\quad 1\leq i\leq k-2,\\ \pi_k \widetilde{\omega}_k &= 0, \\ \pi_k T_{k-1}\widetilde{\omega}_k & = \widetilde{\omega}_{k-1}\pi_k,\\ \pi_k X_i^{-1} &= X_{i}^{-1}\pi_k,\quad 1\leq i\leq k-1,\\ \pi_k X_k^{-1} &= 0. \end{split} \end{align} $$

We refer to [Reference Knop21, Theorem 9.1] for the details.

5.2

The compatibility with the actions of $\mathscr {H}_k^{-}$ and $\mathscr {H}_{k-1}^{-}$ can also be investigated, but the verification is more delicate. More precisely, we have the following result [Reference Knop21, Proposition 9.11].

Proposition 5.1. For any $1\leq i\leq k-1$, we have

(5.2)$$ \begin{align} \pi_k Y_i= Y_{i}\pi_k. \end{align} $$

Furthermore, the operator $Y^{-1}_i$ stabilizes both $\mathscr {P}_k^{-}$ and $\mathscr {P}_{k-1}^-$ and

(5.3)$$ \begin{align} \pi_k Y^{-1}_i= Y^{-1}_{i}\pi_k. \end{align} $$

5.3

For any $n\geq 1$, the sequence of operators $(A_k)_{k\geq 1}$ defined by

$$ \begin{align*} A_k:=Y_n^{(k)}, \quad k\geq n, \end{align*} $$

induces the limit operator $\displaystyle \mathscr {Y}_n:\lim _{\substack {\longleftarrow \\ k\geq n}} \mathscr {P}_k^-\to \lim _{\substack {\longleftarrow \\ k\geq n}} \mathscr {P}_k^-$. Since $\displaystyle \lim _{\substack {\longleftarrow \\ k\geq n}} \mathscr {P}_k$ and $\mathscr {P}^-_{\infty }$ are canonically isomorphic, we obtain an operator $\displaystyle \mathscr {Y}_n:\mathscr {P}^-\to \mathscr {P}^-$. Similarly, we obtain the limit operators $\mathscr {T}_i, \mathscr {X}^{-1}_i$, and $\mathscr {Y}^{-1}_i$, $i\geq 1$. It is important to remark that the operators $\mathscr {Y}^{-1}_i$ are invertible (with inverse $\mathscr {Y}_i$). The following result immediately follows.

Theorem 5.2. The limit operators $\mathscr {T}_i, \mathscr {X}^{-1}_i$, and $\mathscr {Y}^{-1}_i$, $i\geq 1$, define an $\mathscr {H}^{-}$ action on $\mathscr {P}^-_{\infty }$.

Proof. All relations are satisfied because they are satisfied by the corresponding operators acting on each $\mathscr {P}_k^{-}$.

5.4

As it was pointed out in [Reference Knop21, §10] the almost symmetric module $\mathscr {P}_{\mathrm {as}}^-$, defined in §2.7, is a $\mathscr {H}^-$-stable subspace of $\mathscr {P}_{\infty }^-$ and is more canonical from a certain point of view. Specifically, each $\mathscr {P}_k^-$ is a parabolic module for the affine Hecke algebra $\mathscr {A}_k^-$ and has a standard basis (in the sense of Kazhdan-Lusztig theory) indexed by compositions with at most k parts. The sequences consisting of the standard basis elements indexed by the same composition (in all $\mathscr {P}_k^-$, $k\geq n$, for some n) give elements of $\mathscr {P}_{\infty }^-$ (see [Reference Knop21, §9]) which are expected to play the role of a standard basis for the limit representation. However, these limits of standard basis elements do not span $\mathscr {P}_{\infty }^-$ but rather the smaller space $\mathscr {P}_{\mathrm {as}}^-$.

Theorem 5.3. The almost symmetric module $\mathscr {P}_{\mathrm {as}}^-$ is a $\mathscr {H}^{-}$-submodule of $\mathscr {P}^-_{\infty }$.

We call this representation the standard representation of $\mathscr {H}^-$. We expect this representation to be faithful.

As explained in [Reference Knop21], a sequence on nonsymmetric Macdonald polynomials indexed by the same composition gives rise to an element of $\mathscr {P}_{\mathrm {as}}^-$, and such elements are common eigenfunctions for the action the operators $\mathscr {Y}^{-1}_i$. The limit nonsymmetric Macdonald polynomials do not span $\mathscr {P}_{\mathrm {as}}^-$, and therefore the spectral theory of the operators $\mathscr {Y}^{-1}_i$ acting on $\mathscr {P}_{\mathrm {as}}^-$ is not yet fully understood.

6 The standard representation of the ${}^{+}$stable limit DAHA

6.1

Before delving into this section, which is the technical core of the article, it might be helpful to offer a summary of the difficulties that one encounters in the effort to understand the limiting behaviour of the Cherednik operators (the analysis of all the other generators of $\mathscr {H}_k^+$ is straightforward) in relationship to the inverse system $(\mathscr {P}_k^+)_{k\geq 2}$, as well as provide a less technical description of the new elements that are employed to describe this limiting behaviour.

It has been known to the specialists that the Cherednik operators $Y^{(k)}_i$ are not compatible with the inverse system $(\mathscr {P}_k^+)_{k\geq 2}$. Our first observation is that the family of operators $\textbf {t}^kY^{(k)}_i$, for fixed i, are compatible with inverse system $(x_i\mathscr {P}_k^+)_{k\geq 2}$, which does lead to a limit operator acting on $x_i\mathscr {P}_{\infty }^+$. However, there is no common domain for all limit operators $Y_i$, $i\geq 1$. To address this situation, we consider some modified operators acting on $\mathscr {P}_k^+$, which we denote by $\widetilde {Y}_i^{(k)}$. Like the Cherednik operators, these operators satisfy the relations $\widetilde {Y}^{(k)}_{i+1}=\textbf {t}^{-1} T_i \widetilde {Y}^{(k)}_i T_i$, $1\leq i\leq k-1$, $T_j \widetilde {Y}^{(k)}_i=\widetilde {Y}^{(k)}_i T_j$, $|i-j|>1$, and are therefore fully determined by the operator $\widetilde {Y}^{(k)}_{1}$, which has the following additional properties

  1. i) $\widetilde {Y}_1^{(k)}=\textbf {t}^k Y_1^{(k)}$ on $x_1\mathscr {P}_k^+$;

  2. ii) $\widetilde {Y}_1^{(k)}\cdot \mathscr {P}_k^+\subseteq x_1\mathscr {P}_k^+$.

The first property implies that $\widetilde {Y}_i^{(k)}=\textbf {t}^k Y_i^{(k)}$ on $x_i\mathscr {P}_k^+$ for all $1\leq i \leq k$; the second property implies that $\widetilde {Y}_i^{(k)}\cdot \mathscr {P}_k^+\subseteq T_{i-1}\cdots T_1 x_1\mathscr {P}_k^+$ for all $1\leq i \leq k$. The operator for $\widetilde {Y}_1^{(k)}$ is obtained by projecting the action of $Y_1^{(k)}$ onto the space $x_1\mathscr {P}_k^+$.

The structural properties of the algebra generated by $\widetilde {Y}_i^{(k)}$, $X_i$, $1\leq k$, and $T_i$, $1\leq i\leq k-1$ are used to define an algebraic structure $\widetilde {\mathscr {H}}_k^+$ called the deformed DAHA. $\widetilde {\mathscr {H}}_k^+$ has a natural action $\mathscr {P}_k^+$ (the standard representation), which is related to the standard representation of $\mathscr {H}_k^+$ on $\mathscr {P}_k^+$ in the manner described above.

The modified Cherednik operators $\widetilde {Y}_i^{(k)}$, for fixed i, are not yet compatible with the inverse system $(\mathscr {P}_k^+)_{k\geq 2}$, but they are quite close to satisfying such a property. To be able to point out more precisely what happens, we first remark that they act consistently on constant sequences (i.e., sequences compatible with the canonical inclusions $\iota _k: \mathscr {P}_{k-1}^+\to \mathscr {P}_k^+$) in this inverse system (Lemma 6.10). This fact implies the existence of limit operators defined on

$$ \begin{align*} \displaystyle{\lim_{{\longrightarrow}}} \mathscr{P}_k^+=\mathbb Q(\textbf{t},\textbf{q})[x_1,x_2,\dots], \end{align*} $$

the ring of polynomials in infinitely many variables, but still not on $\mathscr {P}_{\infty }^+$. Interestingly, the image of $\displaystyle {\lim _{{\longrightarrow }}} \mathscr {P}_k^+$ under the limit operators lies in $\mathscr {P}_{\mathrm {as}}^+$ (which is strictly smaller than $\mathscr {P}_{\infty }^+$), showing that the smallest space on which one can hope to acquire an action of an algebra of limit operators is $\mathscr {P}_{\mathrm {as}}^+$.

In order to understand the failure of the diagram

to be commutative, we can consider the difference $\pi _k \widetilde {Y}_{1}^{(k)} - \widetilde {Y}_{1}^{(k-1)} \pi _k: \mathscr {P}_k^+\to \mathscr {P}_{k-1}^+$. On the subspace $x_k\mathscr {P}_k^+$, this difference has no kernel and reduces to $\pi _k \widetilde {Y}_{1}^{(k)} : x_k\mathscr {P}_k^+\to \mathscr {P}_{k-1}^+$. Since for every monomial in $m\in \mathbb Q(\textbf {t},\textbf {q})[x_1,x_2,\dots ]$ there exists a unique k such that $m\in x_k\mathscr {P}_k^+$, we can use the action of $\pi _k \widetilde {Y}_{1}^{(k)}$ on $x_k\mathscr {P}_k^+$ to define an operator

$$ \begin{align*} W_1: \mathbb Q(\textbf{t},\textbf{q})[x_1,x_2,\dots]\to \mathbb Q(\textbf{t},\textbf{q})[x_1,x_2,\dots] \end{align*} $$

such that $W_1=\pi _k \widetilde {Y}_{1}^{(k)}$ on $x_k\mathscr {P}_k^+$. By restricting $W_1$ to $\mathscr {P}_k^+$, we obtain the operator $W_1^{(k)}: \mathscr {P}_k^+\to \mathscr {P}_k^+$. We can regard the operator $W_1$ as collecting some obvious obstructions to the commutativity of the above diagram.

As it turns out, these are all the obstructions to the commutativity of the diagram: If $ \widetilde {Z}_{1}^{(k)}= \widetilde {Y}_{1}^{(k)} - W_1^{(k)}$, then the corresponding diagram for the $ \widetilde {Z}_{1}^{(k)} $ operators is commutative, leading to an operator

$$ \begin{align*} \widetilde{Z}_{1}^{(\infty)}: \mathscr{P}_{\infty}^+\to \mathscr{P}_{\infty}^+. \end{align*} $$

Furthermore, the action of $W_1$ (or $W_1^{(k)}$) on monomials can be explicitly computed; this is the action described in §6.8. Since this action is particularly simple, we were able to compute the action of the operators $W_1^{(k)}$ on sequences in the inverse system $(\mathscr {P}_k^+)_{k\geq 2}$. This computation revealed that if the sequence in the inverse system has limit in $\mathscr {P}_{\mathrm {as}}^+$, then one obtains by applying $W_1^{(k)}$ another sequence in $(\mathscr {P}_k^+)_{k\geq 2}$ which can be written as a finite linear expression of sequences compatible with the inverse system. The coefficients in these linear expressions are sequences in $\mathbb Q(\textbf {t},\textbf {q})$ that are convergent in the $\textbf {t}$-adic topology (i.e., $\textbf {t}$ is treated as a small positive real number). This led us to the concept of limit described in §6.9. We refer to the sequences that have limit in this sense as convergent sequences.

Since both ${W}_i^{(k)}$ and $\widetilde {Z}_i^{(k)}$ have limit in this sense, we obtain that the operators $\widetilde {Y}_i^{(k)}$ have limit (denoted by $\mathscr {Y}_i$). Although, the co-domain for the limit operators ${W}_i^{(\infty )}$ and $\widetilde {Z}_i^{(\infty )}$ is, in general, $\mathscr {P}_{\infty }^+$, it turns out that $\mathscr {Y}_i: \mathscr {P}_{\mathrm {as}}^+\to \mathscr {P}_{\mathrm {as}}^+$. Furthermore, $\mathscr {Y}_i$ satisfies a continuity property: It sends convergent sequences to convergent sequences.

The analysis of the algebraic structure generated by the action of the limit operators $\mathscr {Y}_i$ makes use of the continuity property. Their most remarkable property of the limit operators is their commutativity. The commutativity holds, of course, for the finite rank Cherednik operators, but it does not hold for the modified Cherednik operators, and it was interesting to discover that the commutativity is restored in the limit. Furthermore, the action of $\mathscr {Y}_i$ on $x_i\mathscr {P}_{\mathrm {as}}^+$ matches the action of the corresponding limit Cherednik operator $Y_i$. Ultimately, the action of the limit operators $\mathscr {Y}_i$, $\mathscr {X}_i$, and $\mathscr {T}_i$ define an action of $\mathscr {H}^+$ on $\mathscr {P}_{\mathrm {as}}^+$.

6.2

The map $\pi _k:\mathscr {P}^+_k\rightarrow \mathscr {P}^+_{k-1}$ is also partially compatible with the actions of $\mathscr {A}_k^+$ and $\mathscr {A}_{k-1}^+$. More precisely,

(6.1)$$ \begin{align} \pi_k T_i &= T_i \pi_k,\quad 1\leq i\leq k-2,\nonumber\\ \pi_k T_{k-1}^{-1}\dots T_1^{-1}\widetilde{\omega}_k^{-1} &= 0\nonumber \\ \pi_k \widetilde{\omega}_k^{-1}T_{k-1} & = \widetilde{\omega}_{k-1}^{-1}\pi_k\\ \pi_k X_i &= X_{i}\pi_k,\quad 1\leq i\leq k-1,\nonumber \\ \pi_k X_i &= 0.\nonumber\end{align} $$

As before, these relations show that the operators $T_i$ and $X_i$ have limits (in the sense of §5.3), denoted by $\mathscr {T}_i$ and $\mathscr {X}_i$, that act on $\mathscr {P}_{\infty }^+$.

6.3

On the other hand, for any $i\geq 1$, the actions of the operators $Y^{(k)}_{i}$ and $Y^{(k-1)}_{i}$ are no longer compatible with the map $\pi _k$. One immediate obstruction is the fact that the eigenvalues of $Y_{i}$ on $\mathscr {P}_k^+$ and $\mathscr {P}_{k-1}^+$ no longer match, as is was the case for their action on $\mathscr {P}_k^-$ and $\mathscr {P}_{k-1}^-$. The common spectrum of the action of $Y_i$, $1\leq i\leq k$, on $\mathscr {P}_k^+$ is known from type $A_{k-1}$ Macdonald theory. The common eigenfunctions are nonsymmetric Macdonald polynomials; they are indexed by compositions $\lambda =(\lambda _1,\dots ,\lambda _k)\in \mathbb Z_{\geq 0}^k$. The eigenvalue of $Y^{(k)}_i$ corresponding to the nonsymmetric Macdonald polynomial indexed by $\lambda $ is

$$ \begin{align*} e_{\lambda}(i):=\textbf{q}^{\lambda_i}\textbf{t}^{1-w_{\lambda}(i)}, \end{align*} $$

where $w_{\lambda }(i)=|\{j=1,\dots ,i~|~\lambda _j\leq \lambda _i\}|+|\{j=1,\dots ,i~|~\lambda _j< \lambda _i\}|$. For the composition $(\lambda ,0)\in \mathbb Z_{\geq 0}^{k+1}$, we have

$$ \begin{align*} e_{(\lambda,0)}(i)=e_{\lambda}(i)\textbf{t}^{-1},\quad \text{if } \lambda_i>0, \quad \text{and}\quad e_{(\lambda,0)}(i)=e_{\lambda}(i),\quad \text{if } \lambda_i=0. \end{align*} $$

One way to partially correct this problem is to consider the normalized operators $\textbf {t}^{k}Y^{(k)}_i$. Comparing the eigenvalues corresponding to $\lambda $ and $(\lambda ,0)$ gives

$$ \begin{align*} \textbf{t}^{k+1}e_{(\lambda,0)}(i)=\textbf{t}^k e_{\lambda}(i),\quad \text{if } \lambda_i>0, \quad \text{and}\quad \textbf{t}^{k+1} e_{(\lambda,0)}(i)=\textbf{t}^k e_{\lambda}(i)\cdot \textbf{t},\quad \text{if } \lambda_i=0. \end{align*} $$

Example 6.1. The discrepancy between the eigenvalues is also reflected in the eigenfunctions. For example, the Macdonald polynomial corresponding to the compositions $(0,1,0)$ and $(0,1)$ are, respectively,

$$ \begin{align*} x_2+\frac{\textbf{q}(1-\textbf{t})}{\textbf{q}-\textbf{t}^2}x_1\quad\quad \text{and}\quad\quad x_2+\frac{\textbf{q}(1-\textbf{t})}{\textbf{q}-\textbf{t}}x_1. \end{align*} $$

This shows, in particular, that $\pi _3 Y_1^{(3)}\neq Y_1^{(2)}\pi _3$.

6.4

A crucial observation is that the actions of the operators $\textbf {t}^{k}Y^{(k)}_iX^{(k)}_i\in \mathscr {H}_k^{+}$ on $\mathscr {P}_{k}^{+}$ are compatible with the inverse system.

Proposition 6.2. For any $1\leq i\leq k-1$, we have

$$ \begin{align*} \pi_k\textbf{t}^{k}Y_i X_i=\textbf{t}^{k-1}Y_iX_i\pi_k. \end{align*} $$

Proof. First note that we have

(6.2)$$ \begin{align} \pi_k\omega^{-1}_k T_{k-1}= \omega^{-1}_{k-1}\pi_k, \end{align} $$

which can be verified by direct computation. Hence, by equation 6.1 we have

$$ \begin{align*} \pi_k Y_i^{(k)}X_i^{(k)}&= \textbf{t}^{1-i}T_{i-1} \dots T_{1}\pi_k(\omega^{-1}_k T_{k-1}^{-1}\dots T_{i}^{-1}X_i)\\ &=\textbf{t}^{1-k}T_{i-1} \dots T_{1}\pi_k(\omega^{-1}_k X_kT_{k-1}\dots T_{i})\\ &=\textbf{t}^{1-k}T_{i-1} \dots T_{1}\omega^{-1}_{k-1}\pi_k( T_{k-1}^{-1}X_kT_{k-1}\dots T_{i})\\ &=\textbf{t}^{-k}T_{i-1} \dots T_{1}\omega^{-1}_{k-1}\pi_k( T_{k-1}X_kT_{k-1}\dots T_{i})\\ &=\textbf{t}^{1-k}T_{i-1} \dots T_{1}\omega^{-1}_{k-1}\pi_k( X_{k-1}T_{k-2}\dots T_{i})\\ &=\textbf{t}^{-i}T_{i-1} \dots T_{1}\omega^{-1}_{k-1}T_{k-2}^{-1}\dots T_{i}^{-1}X_i\pi_k\\ &=\textbf{t}^{-1}Y_i^{(k-1)}X_i^{(k-1)}\pi_k \end{align*} $$

Therefore, we have $\pi _k\textbf {t}^{k}Y_i^{(k)}X^{(k)}_i=\textbf {t}^{k-1}Y_i^{(k-1)}X^{(k-1)}_i\pi _k$.

In other words, one can obtain a limit operator corresponding to $Y_i$, acting on the space $x_i\mathscr {P}_{\infty }^+\subset \mathscr {P}_{\infty }^+$. However, this is not satisfactory because it does not produce a nontrivial common domain for all the limit operators $Y_i$, $i\geq 1$. In order to extend the action of the limit operator $Y_i$ to a larger domain, we introduce the deformed double affine Hecke algebras and the concept of limit detailed in §6.9.

6.5 The deformed double affine Hecke algebras

Definition 6.3. The deformed DAHA $\widetilde {\mathscr {H}}_{k}^{+}$, $k\geq 2$, is the $\mathbb {Q}(\textbf {t},\textbf {q})$-algebra generated by the elements $T_1,\dots ,T_{k-1}$, $X_1,\dots ,X_k$ and $\varpi _k$ satisfying equations (3.1a), (3.1b) and (3.1c) and the following relations:

(6.3)$$ \begin{align} \begin{gathered} \varpi_kT_i=T_{i+1}\varpi_k, \quad 1\leq i\leq k-2,\\ \varpi_kX_i=X_{i+1}\varpi_k,\quad 1\leq i\leq k-1, \end{gathered} \end{align} $$
(6.4)$$ \begin{align} \begin{gathered} \gamma_k T_{k-1}=-\textbf{t} \gamma_k,\quad T_1\gamma_k =\gamma_k,\\ \gamma_k\varpi_k^{k-2}\gamma_k=\gamma_k\varpi_k^{k-1}\gamma_k=\gamma_k\varpi_k^{k}=0, \end{gathered} \end{align} $$

where

$$ \begin{align*} \gamma_k=\varpi_k^2 T_{k-1}-T_{1}\varpi_k^2. \end{align*} $$

Remark 6.4. The quotient of $\widetilde {\mathscr {H}}_k^+$ by the ideal generated by $\gamma _k$ is isomorphic to $\mathscr {H}_k^+$. As it is clear from the relations (6.4), $\gamma _k$ is not central in $\widetilde {\mathscr {H}}_k^+$.

Notation 6.5. Let $\widetilde {Y}_i$, $1\leq i\leq k$, be defined as follows

$$ \begin{align*} \widetilde{Y}_1=\textbf{t}^{k}\varpi_kT_{k-1}^{-1}\dots T_1^{-1},\qquad \widetilde{Y}_{i+1}=\textbf{t}^{-1} T_i \widetilde{Y}_i T_i, \quad 1\leq i\leq k-1. \end{align*} $$

Therefore,

$$ \begin{align*} \widetilde{Y}_i = \textbf{t}^{1-i+k}T_{i-1} \dots T_{1}\varpi_k T_{k-1}^{-1}\dots T_{i}^{-1}, \quad 1\leq i\leq k. \end{align*} $$

Remark 6.6. It is important to remark that the relation (3.4b) is also satisfied in $\widetilde {\mathscr {H}}_k$. Indeed

$$ \begin{align*} \widetilde{Y}_{1} T_1 X_1 &=\textbf{t}^{k}\varpi_kT_{k-1}^{-1}\ldots T_2^{-1}X_1 \\ &=\textbf{t}^{k}X_2\varpi_kT_{k-1}^{-1}\ldots T_2^{-1} \\ &=X_2 \widetilde{Y}_{1} T_1. \end{align*} $$

The relations

$$ \begin{align*}T_j \widetilde{Y}_i=\widetilde{Y}_i T_j, \quad |i-j|>1 \end{align*} $$

are also satisfied. The verification is a simple consequence of the braid relations and equation (6.3).

Remark 6.7. On the other hand, the elements $\widetilde {Y}_i$, $1\leq i\leq k$, do not commute. In fact, we have

$$ \begin{align*} [\widetilde{Y}_1, \widetilde{Y}_2]=\textbf{t}^{2k-1}\gamma_k T_{k-1}^{-1}\dots T_1^{-1}T_{k-1}^{-1}\dots T_2^{-1}=\textbf{t}^{2k-1}\gamma_k T_{k-2}^{-1}\dots T_1^{-1}T_{k-1}^{-1}\dots T_1^{-1}. \end{align*} $$

Therefore, $\gamma _k$ can be seen as an obstruction to the commutativity of the elements $\widetilde {Y}_i$.

Since the notation for the generators of $\widetilde {\mathscr {H}}_k^+$ and the elements $\widetilde {Y}_i$ does not specify the integer k, we will use the notation in the fashion indicated in Convention 3.11.

6.6

For any $1\leq i \leq k$, let $\operatorname {\mathrm {pr}}_i:\mathscr {P}_k^+ \to \mathscr {P}_k^+$, be the $\mathbb {Q}(\textbf {t},\textbf {q})$-linear map which acts as identity on monomials divisible by $x_i$ and as the zero map on monomials not divisible by $x_i$. In other words, $\operatorname {\mathrm {pr}}_i$ is the projection onto the subspace $x_i\mathscr {P}_k^+$.

We define the following action of $\widetilde {\mathscr {H}}_{k}^{+}$ on $\mathscr {P}_{k}^{+}$. The action is related to the action of $\mathscr {H}_k$ on $\mathscr {P}_k^+$.

Theorem 6.8. The following formulas define an action of $\widetilde {\mathscr {H}}_{k}^{+}$ on $\mathscr {P}_k^+$:

(6.5)$$ \begin{align} \begin{split} T_i f(x_1,\dots,x_k) &= s_i f(x_1,\dots,x_k) +(1-\textbf{t})x_i\frac{1-s_i}{x_i-x_{i+1}}f(x_1,\dots,x_k), \\ X_i f(x_1,\dots,x_k) &=x_i f(x_1,\dots,x_k),\\ \varpi_k f(x_1,\dots,x_k) &=\operatorname{\mathrm{pr}}_1f(x_2,\dots,x_k, \textbf{q} x_1). \end{split} \end{align} $$

Note that the action of $ \varpi _k $ coincides with the action of $\operatorname {\mathrm {pr}}_1\omega _k^{-1}$.

Proof. We only need to check the relations (6.3) and (6.4). From equation (3.3c), we have

$$ \begin{align*} \varpi_kT_i f&=\operatorname{\mathrm{pr}}_1\omega_k^{-1}T_i f\\ &=\operatorname{\mathrm{pr}}_1(T_{i+1}\omega_k^{-1} f)\\ &=T_{i+1}\operatorname{\mathrm{pr}}_1\omega_k^{-1} f\\ &=T_{i+1}\varpi_k f \end{align*} $$

for $f\in \mathscr {P}_{k}^{+}$ and $1\leq i\leq k-2$. Similarly, from equation (3.3d), we have

$$ \begin{align*} \varpi_k X_i f=X_{i+1}\varpi_k f \end{align*} $$

for $f\in \mathscr {P}_{k}^{+}$ and $1\leq i\leq k-1$.

By linearity it suffices to verify the relation (6.4) on monomials. A straightforward computation leads to the following explicit expression of the action of $\gamma _k$

(6.6)$$ \begin{align} \gamma_k(x_1^{t_1}\dots x_{k-1}^{t_{k-1}}x_{k}^{t_{k}}) = \begin{cases} 0, & \textrm{ if } t_{k-1}\neq 0,t_k\neq 0\\\\ (1-\textbf{t})\textbf{q}^{t_{k-1}}(x_1^{t_{k-1}-1}x_2+x_1^{t_{k-1}-2}x_2^2+\dots \\+x_1 x_2^{t_{k-1}-1})(x_3^{t_1}x_4^{t_2}\dots x_k^{t_{k-2}}), & \textrm{ if } t_{k-1}\neq 0,t_k= 0\\\\ (\textbf{t}-1)\textbf{q}^{t_{k}}(x_1^{t_{k}-1}x_2+x_1^{t_{k}-2}x_2^2+\dots\\+x_1 x_2^{t_{k}-1})(x_3^{t_1}x_4^{t_2}\dots x_k^{t_{k-2}}), &\textrm{ if } t_{k-1}= 0,t_k\neq 0. \end{cases} \end{align} $$

Note that the expression immediately implies that the actions of $\gamma _k T_{k-1}$ and $-\textbf {t} \gamma _k$ coincide and similarly for the actions of $T_1\gamma _k$ and $\gamma _k$. The remaining relations can also be directly verified.

Remark 6.9. The action of $\widetilde {Y}_i X_i$ coincides with the action of $\textbf {t}^{k}Y_i X_i\in \mathscr {H}_k^+$ on $\mathscr {P}_k^+$. Equivalently, the action of $\textbf {t}^{-k}\widetilde {Y}_i$ and the action of $Y_i\in \mathscr {H}_k^+$ coincide on $x_i\mathscr {P}_k^+$. Therefore, $\widetilde {Y}_i$ can be regarded as a modification of the Cherednik operator $Y_i$.

6.7

We are now ready to explore the compatibility between the action of the operators $\widetilde {Y}_i^{(k)}$ and the map $\pi _k: \mathscr {P}_k^+\to \mathscr {P}_{k-1}^+$.

Lemma 6.10. The following diagram is commutative:

Equivalently, we have

$$ \begin{align*} \pi_{k}\widetilde{Y}_i^{(k)} f=\widetilde{Y}_i^{(k-1)} f \end{align*} $$

for all $f\in \mathscr {P}_{k-1}^{+}$.

Proof. Note that $\operatorname {\mathrm {pr}}_1\omega ^{-1}_{k}=\omega ^{-1}_{k}\operatorname {\mathrm {pr}}_k$. For any $f\in \mathscr {P}_{k-1}^{+}$, we have

$$ \begin{align*} \pi_{k}\widetilde{Y}_i^{(k)}\iota_k f&=\pi_{k}\textbf{t}^{1-i+k}T_{i-1} \dots T_{1} \operatorname{\mathrm{pr}}_1\omega^{-1}_{k}T_{k-1}^{-1}\dots T_i^{-1}f\\ &=\textbf{t}^{k-i}T_{i-1} \dots T_{1}\operatorname{\mathrm{pr}}_1\pi_{k}\omega^{-1}_{k}T_{k-1}T_{k-2}^{-1}\dots T_i^{-1}f\\&\quad+\pi_{k}\textbf{t}^{k-i}(\textbf{t}-1)T_{i-1} \dots T_{1}\operatorname{\mathrm{pr}}_1\omega^{-1}_{k}T_{k-2}^{-1}\dots T_i^{-1}f\\ &=\textbf{t}^{k-i}T_{i-1} \dots T_{1}\operatorname{\mathrm{pr}}_1\omega^{-1}_{k-1}\pi_{k}T_{k-2}^{-1}\dots T_i^{-1}f\\&\quad+\pi_{k}\textbf{t}^{k-i}(\textbf{t}-1)T_{i-1} \dots T_{1}\omega^{-1}_{k}\operatorname{\mathrm{pr}}_k(T_{k-2}^{-1}\dots T_i^{-1}f)\\ &=\widetilde{Y}_i^{(k-1)}\pi_{k}f+0. \end{align*} $$

For the third equality, we used equation (6.2).

As an immediate consequence, we obtain the following.

Proposition 6.11. For any $i\geq 1$, the sequence of operators $(\widetilde {Y}_i^{(k)})_{k\geq 2}$ induces a map

$$ \begin{align*} \widetilde{Y}_i^{(\infty)}: \lim_{{\longrightarrow}} \mathscr{P}_k^+\to \mathscr{P}_{\infty}^{+} \end{align*} $$

such that $\Pi _k \widetilde {Y}_i^{(\infty )} I_k=\widetilde {Y}_i^{(k)}$ for all $k\geq i$, $k\geq 2$.

Proof. Fix $n\geq i$, and denote by $\iota _{n,k}: \mathscr {P}_n^+\to \mathscr {P}_k^+$, $n\leq k$, the canonical inclusion. By Lemma 6.10, the sequence of maps $\widetilde {Y}_i^{(k)} \iota _{n,k}: \mathscr {P}_n^+\to \mathscr {P}_k^+$, $k\geq n,i$ is compatible with the structure maps $\pi _k$. Therefore, they induce a morphism

$$ \begin{align*} \mathscr{P}_n^+\to \lim_{\substack{\longrightarrow \\ k\geq n}} \mathscr{P}_k^+\cong \displaystyle \lim_{{\longrightarrow }} \mathscr{P}_k^+. \end{align*} $$

Furthermore, Lemma 6.10 implies that these maps are compatible with the structure maps $\iota _n$ and therefore induce the desired morphism.

Example 6.12. To illustrate the construction of the limit operator $\widetilde {Y}_1^{(\infty )}: \displaystyle {\lim _{{\longrightarrow }}} \mathscr {P}_k^+\to \mathscr {P}_{\infty }^{+}$, let us compute its action on the element $x_2^2\in \displaystyle {\lim _{{\longrightarrow }}} \mathscr {P}_k^+$. A direct computation gives

$$ \begin{align*} \widetilde{Y}_1^{(2)}\cdot x_2^2&=\textbf{q}^2 \textbf{t}(\textbf{t}-1)x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2\\ \widetilde{Y}_1^{(3)}\cdot x_2^2&=\textbf{q}^2 \textbf{t}(\textbf{t}-1)x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2-\textbf{q}\textbf{t}(\textbf{t}-1)^2 x_1x_3\\ \widetilde{Y}_1^{(4)}\cdot x_2^2&=\textbf{q}^2 \textbf{t}(\textbf{t}-1)x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2-\textbf{q}\textbf{t}(\textbf{t}-1)^2 (x_1x_3+x_1x_4) \end{align*} $$

and, more generally, for $k\geq 4$,

$$ \begin{align*} \widetilde{Y}_1^{(k)}\cdot x_2^2=\textbf{q}^2 \textbf{t}(\textbf{t}-1)x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2-\textbf{q}\textbf{t}(\textbf{t}-1)^2 (x_1x_3+\cdots x_1x_k). \end{align*} $$

Therefore, the limit operator acts as

$$ \begin{align*} \widetilde{Y}_1^{(\infty)}\cdot x_2^2=\textbf{q}^2 \textbf{t}(\textbf{t}-1)x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2-\textbf{q}\textbf{t}(\textbf{t}-1)^2 x_1 e_1[\mathbf{X}_2]. \end{align*} $$

On the other hand, the action same operator on a nonconstant sequence (e.g., one that converges to a nontrivial almost symmetric function) shows a slightly different behaviour.

Example 6.13. Let $F[\mathbf {X}]=x_2^2+x_3^2+\cdots =p_2[\mathbf {X}_1]$, which is the inverse limit of the sequence

$$ \begin{align*} F_{k}=\Pi_k F[\mathbf{X}]=x_2^2+\cdots+x_k^2,~ k\geq 2. \end{align*} $$

We have

$$ \begin{align*} \widetilde{Y}_1^{(2)}\cdot F_{2}&=\textbf{q}^2(\textbf{t}^2-\textbf{t})x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)x_1 x_2\\ \widetilde{Y}_1^{(3)}\cdot F_{3}&=\textbf{q}^2(\textbf{t}^3-\textbf{t})x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)(x_1 x_2+x_1 x_3) \end{align*} $$

and, more generally, for $k\geq 2$,

$$ \begin{align*} \widetilde{Y}_1^{(k)}\cdot F_{k}=\textbf{q}^2(\textbf{t}^k-\textbf{t})x_1^2+\textbf{q}\textbf{t}(\textbf{t}-1)(x_1 x_2+x_1 x_3+\cdots+x_1x_k). \end{align*} $$

As it can be clearly seen, the sequence $ \widetilde {Y}_1^{(k)}F_{k}$ fails to have an inverse limit because of the contribution of the term $\textbf {q}^2(\textbf {t}^k-\textbf {t})x_1^2$.

As we will see next, the general behaviour of the action of fixed operator on a sequence with inverse limit is no more complicated than the one exhibited in Example 6.13. The operators $W_i^{(k)}$ introduced in the next section formalize this observation.

6.8

We define the $\mathbb {Q}(\textbf {t},\textbf {q})$-linear operators $\displaystyle W^{(k)}_i: \mathscr {P}_k^+\to \mathscr {P}_{k}^{+}$, $i\geq 1$, as follows. Let $s\geq 0$, and let $i_1<i_2<\dots <i_s\leq k$ and $t_1, \dots , t_s$ be positive integers. We set

(6.7)$$ \begin{align} W^{(k)}_1(x_{i_1}^{t_{1}}\dots x_{i_s}^{t_{s}})=\begin{cases} 0, & \text{ if } 1=i_1 \text{ or }s=0,\\\\ \ \textbf{t}^{i_s}(1-\textbf{t}^{-1})\textbf{q}^{t_{s}}x_{1}^{t_{s}}x_{i_1}^{t_{1}}\dots x_{i_{s-1}}^{t_{s-1}}, & \text{ if } i_1>1. \end{cases} \end{align} $$

Define

$$ \begin{align*} W^{(k)}_{i+1}=\textbf{t}^{-1} T_i W^{(k)}_i T_i,\quad 1\leq i\leq k-1. \end{align*} $$

For $1\leq i\leq k$, we also denote

$$ \begin{align*} \widetilde{Z}_{i}^{(k)}=\widetilde{Y}_{i}^{(k)}-W_{i}^{(k)}. \end{align*} $$

Lemma 6.14. The following diagram is commutative:

Proof. It is enough to prove the statement for $i=1$. For any monomial $f\in \mathscr {P}_{k-1}^{+}$, we have