This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Special function models of indecomposable sl(2)sl(2) representations: The Laguerre Case

Sébastien Bertrand1 Ian Marquette2 Willard Miller Jr 3 Sarah Post1
1 Department of Mathematics, University of Hawaii, Honolulu, HI, 96822, USA
2 School of Mathematics and Physics, The University of Queensland,
Brisbane, QLD 4072, Australia
3 School of Mathematics, University of Minnesota, Mineapolis, MN, 55455, USA
Abstract

In this paper, we point out connections between certain types of indecomposable representations of sl(2)sl(2) and generalizations of well-known orthogonal polynomials. Those representations take the form of infinite dimensional chains of weight or generalised weight spaces, for which the Cartan generator acts in a diagonal way or via Jordan blocks. The other generators of the Lie algebras sl(2)sl(2) act as raising and lowering operators but are now allowed to relate the different chains as well. In addition, we construct generating functions, we calculate the action of the Casimir invariant and present relations to systems of non-homogeneous second-order coupled differential equations. We present different properties as higher-order linear differential equations for building blocks taking the form of one variable polynomials. We also present insight into the zeroes and recurrence relations.

1 Introduction

Lie algebras, their realizations and connections with special functions have attracted a lot of attention in the area of mathematical physics since the end of the 60s [1, 2, 3, 4, 5, 6, 7]. This has been inspired partly by the fruitful connection of special functions and Lie algebras with quantum mechanical problem [8, 9, 10]. In this context, the generators of the Lie algebras can be integrals of motion, ladder operators or other physically relevant operators. Via the notion of hidden symmetry underlying exact and quasi-exact solvability, the sl(2)sl(2) Lie algebra was connected to various models [11]. The realizations related to other higher-rank Lie algebras such as sl(n)sl(n) and so(n)so(n) have lead to profound connection with generalized hypergeometric, Lauricella and Horn functions (see e.g. [2]). Those special functions have appeared in different contexts of mathematical physics.

The cases of the Lie algebras sl(2)sl(2) with one and two variables realizations have been intensively described in regard of unitary and irreducible representations of sl(2)sl(2). This rank-1 algebras has been related to different orthogonal polynomials, in particular using first but also second-order differential operators [12, 13, 14, 15, 16, 17, 18]. An extensive description of Lie algebras and their realizations, and in particular from point of view of the Casimir invariant was developed by Vilenkin and Klimyk [19]. Some connections with wider types of representation beyond irreducible and unitary was pointed out [20], in regard of indecomposable representations [21, 22, 23] and quantum field theory [24].

Among the indecomposable representations, there are different cases that consist in combinations of finite and infinite dimensional representations glued together via the action of the Cartan or other generators and forming interconnected chains. The paper intends to generalize certain indecomposable representations in context of sl(2)sl(2) with arbitrary number of chains from the perspective of differential operators that were obtained in context of the Laguerre case[2]. This will lead to constructing various polynomials where the raising generators of sl(2)sl(2) will act almost as usual ladder operators but allowing interaction between the different chains. The papers will also establish the connection with systems of coupled nonhomogeneous second-order differential equations.

Standard literature on the theory of ODEs and special functions has mostly looked at uncoupled homogeneous ODEs in the hypergeometric class and their generalisation as the Heun equation [25, 26]. The study of systems of non-homogeneous differential equations, their solutions and connections to special functions have been much less studied. Some cases related to resonant problem [28, 29, 30, 31, 32] were studied. However, there is no systematic search for solution of those systems of ODEs. This paper allows to discuss some classes of such systems of ODEs, that naturally extend the ODE of the Laguerre polynomials, and with an underlying Lie algebras such as sl(2) and related indecomposable representations. This makes those polynomials and special functions of interest for applications in mathematical physics. Here the connection between indecomposable representations and sl(2)sl(2) provides a new perspective on this topic.

The papers is organised as follow. In Section 2, we set the foundation and notation of the manuscript while reviewing properties of the (generalized) Laguerre polynomials. In Section 3 we present the case of indecomposable with chains where T, the Cartan generator, act as a Jordan block. We also present explicitly the polynomials solving the systems of nonhomogenous coupled ODEs. We give extra details on the 2 and 3 chains problems. In particular we present second-order differential-recurrence relations, generating functions and formula in terms of linear combination of Laguerre polynomials. In Section 4 we characterise the case of Laguerre polynomials where T is diagonalizable (but the raising/lowering operators are not). In Section 5 we provide further insight into some of the building block taking the form of new polynomials. We present the third and fourth order ODE that they satisfy, further recurrence relations, related family of polynomial and bi-orthogonality type relations and insight into the zeroes of those polynomials. In Section 6, we discuss how those are in fact connected with certain representations of sl(2)sl(2) where the infinite dimensional representation is obtained via induced representation of a finite representation of the Borel subalgebra.

2 Preliminaries

The motivation for the treatment of indecomposable representations presented here takes origin in the large body of work on sl(2)sl(2) and what we would refer as one-chain representations, and studied in the monograph [2] in which the differential recurrence relations for hypergeometric, confluent hypergeometric and Bessel functions were shown to be expressible as representations of the three four-dimensional Lie algebras sl(2)(E)sl(2)\oplus(E), the Heisenberg algebra, and e(2)(E)e(2)\oplus(E). The differential recurrence relations for these three types of hypergeometric functions are linear first-order ordinary differential equations for the functions Fn(z)F_{n}(z). The idea is to add a new variable tt and to replace the ordinary differential equations by partial differential equations in z,tz,t such that in the new system the basis functions are Fn(z)tnF_{n}(z)t^{n}. If this is done correctly the new recurrence operators will close to generate one of the three Lie algebras listed above. The advantage of the addition of a 2nd variable is that the tools of Lie theory become available and we can find new properties of these functions, such as generating functions with relative ease.

In the monograph [2] these realizations were used to classify the finite and infinite dimensional bounded above or below irreducible representatives of sl(2)(E)sl(2)\oplus(E), the Heisenberg algebra, and e(2)(E)e(2)\oplus(E), most of which do not extend to global representations of the corresponding groups. However, local Lie group theory still applies and can be used to derive additional theorems and generating functions where global group representations no longer apply.

In this paper we shall employ the realization of type B (sl(2)sl(2), Laguerre polynomials) of the differential recurrence relations employed in [2], but now used to find realizations of indecomposable representations of sl(2)sl(2).

If we restrict ourselves to one-chain in the type B, the (generalized) Laguerre are known to possess raising and lowering operators forming an sl(2)sl(2) algebra, that is

[E,E+]=2T,[T,E±]=±E±,[E^{-},E^{+}]=2T,\qquad[T,E^{\pm}]=\pm E^{\pm}, (2.1)

where E+E^{+} and EE^{-} are respectively the raising and lowering operators. The 𝔰𝔩2\mathfrak{sl}_{2} algebra possess the following Casimir element:

K=T2+T+EE+.K=T^{2}+T+E^{-}E^{+}. (2.2)

2.1 Generalized Laguerre polynomials

For the generalized Laguerre polynomials, which can be defined using a generalized hypergeometric function,

Ln(α)(z)=Γ(α+n+1)Γ(α+1)F11(nα+1,z),L_{n}^{(\alpha)}(z)=\dfrac{\Gamma(\alpha+n+1)}{\Gamma(\alpha+1)}\hskip 5.69054ptF_{1\hskip 5.69054pt1}\left(\begin{array}[]{c}-n\\ \alpha+1\end{array},z\right), (2.3)

the raising and lowering operators can be expressed as follow

E+=zddz+n+1+αz,E=zddznE^{+}=z\dfrac{d}{dz}+n+1+\alpha-z,\qquad E^{-}=z\dfrac{d}{dz}-n (2.4)

with the additional operator

T=n+1+α2.T=n+\dfrac{1+\alpha}{2}. (2.5)

The raising and lowering operators satisfy the differential recurrence relations

E+Ln(α)(z)=(n+1)Ln+1(α)(z),ELn(α)(z)=(α+n)Ln1(α)(z)E^{+}L_{n}^{(\alpha)}(z)=(n+1)L_{n+1}^{(\alpha)}(z),\qquad E^{-}L_{n}^{(\alpha)}(z)=-(\alpha+n)L_{n-1}^{(\alpha)}(z) (2.6)

and they satisfy the sl(2)sl(2) commutation relations (2.1) together with TT. By combining these two recurrence relations, we can recover the differential equation

(zd2dz2+(1+αz)ddz+n)Ln(α)(z)=0.\left(z\dfrac{d^{2}}{dz^{2}}+(1+\alpha-z)\dfrac{d}{dz}+n\right)L_{n}^{(\alpha)}(z)=0. (2.7)

Using a linear combination of the raising/lowering operators and their actions, we can recreate the recurrence equation

(n+1)Ln+1(α)(z)=(2n+1+αz)Ln(α)(z)(n+α)Ln1(α)(z).(n+1)L_{n+1}^{(\alpha)}(z)=(2n+1+\alpha-z)L_{n}^{(\alpha)}(z)-(n+\alpha)L_{n-1}^{(\alpha)}(z). (2.8)

The generating function for the generalized Laguerre polynomials is given by

(1t)(α+1)exp(zt1t)=n=0Ln(α)(z)tn.(1-t)^{-(\alpha+1)}\exp\left(\dfrac{-zt}{1-t}\right)=\sum_{n=0}^{\infty}L_{n}^{(\alpha)}(z)t^{n}. (2.9)

2.2 Many chains of polynomials

The realizations of the operators E±E^{\pm} and TT for the polynomials (or one chain) can be adapted to generate two or more chains of new polynomials that are linked to the previous chain. That is we introduce a new variable tt to allow a higher number of chains. For convenience, we will construct and orient the possible states on the chains as follow:

|0,N|1,N|n,N|0,β|1,β|n,β|0,2|1,2|n,2|0,1|1,1|n,1\begin{array}[]{ccccc}|0,N\rangle&|1,N\rangle&...&|n,N\rangle&...\\ \vdots&\vdots&&\vdots&\\ |0,\beta\rangle&|1,\beta\rangle&...&|n,\beta\rangle&...\\ \vdots&\vdots&&\vdots&\\ |0,2\rangle&|1,2\rangle&...&|n,2\rangle&...\\ |0,1\rangle&|1,1\rangle&...&|n,1\rangle&...\end{array}

where NN is the number of chains considered. The states on the first chain, |n,N|n,N\rangle for nn\in\mathbb{N}, are linked with the original polynomials, i.e. the auxiliary variable tt to the power nn times the generalized Laguerre polynomials

|n,N=Ln(α)(z)tn.|n,N\rangle=L_{n}^{(\alpha)}(z)t^{n}. (2.10)

The 𝔰𝔩2\mathfrak{sl}_{2} elements can act centrally and map to the closest neighbours, cardinal and ordinal directions. The cardinal direction includes north, south, east and west. The ordinal includes the northeast, southeast, southwest and northwest. Each elements (excluding the top and bottom chains and the initial states on the right) have 8 nearest neighbors. We assume the existence of a highest weight going north, i.e. the original polynomials, and a lowest weight going west. (Depending on the case, the lowest weight in the West direction may not be the state |0,β|0,\beta\rangle.) We will consider infinite-dimensional chains, that it there will be no highest weight in the East direction.

3 Laguerre polynomials with non-central operators

In this case, we will consider that none of the operators act centrally. One representation of such an algebra takes the form

T|n,β\displaystyle T|n,\beta\rangle =\displaystyle= (n+α+12)|n,β+|n,β+1,\displaystyle\left(n+\dfrac{\alpha+1}{2}\right)|n,\beta\rangle+|n,\beta+1\rangle, (3.1)
E+|n,β\displaystyle E^{+}|n,\beta\rangle =\displaystyle= (n+1)|n+1,β,\displaystyle(n+1)|n+1,\beta\rangle, (3.2)
E|n,β\displaystyle E^{-}|n,\beta\rangle =\displaystyle= (n+α)|n1,β2|n1,β+1,\displaystyle-(n+\alpha)|n-1,\beta\rangle-2|n-1,\beta+1\rangle, (3.3)

i.e. TT acts centrally and north, E+E^{+} acts east and EE^{-} acts west and north-west. It is important to note that on the highest chain, all the operators act like the operators in equation (2.4). Let us denote the vanishing of the north coefficients on the highest chain. A realization is of this representation is given by

T\displaystyle T =\displaystyle= tt+1+α2,\displaystyle t\partial_{t}+\dfrac{1+\alpha}{2}, (3.4)
E+\displaystyle E^{+} =\displaystyle= t(zz+tt+(1+αz)),\displaystyle t(z\partial_{z}+t\partial_{t}+(1+\alpha-z)), (3.5)
E\displaystyle E^{-} =\displaystyle= zzttt,\displaystyle\dfrac{z\partial_{z}-t\partial_{t}}{t}, (3.6)

which also satisfy the sl(2)sl(2) commutation relations (2.1).

To generate a new chain, one can use the equation

T|0,N1=(α+12)|0,N1+|0,N.T|0,N-1\rangle=\left(\dfrac{\alpha+1}{2}\right)|0,N-1\rangle+|0,N\rangle. (3.7)

Using the realization of TT and considering |0,N1|0,N-1\rangle as an unknown function of zz and tt, we obtain a linear differential equation that we can solve. We can then generate all the other elements of the chain by successively applying E+E^{+} on each state |n,N1|n,N-1\rangle. Finally, by checking the restriction of sl(2)sl(2), we can construct the whole states on the second chain.

Once the second chain is constructed, we can reapply the same method to construct the third (or lower) chains. As a result, we can express all the states as

|n,β=(j=0Nβωn,j(z)ln(zt)Nβj(Nβ)!)tn,|n,\beta\rangle=\left(\sum_{j=0}^{N-\beta}\dfrac{\omega_{n,j}(z)\ln(zt)^{N-\beta-j}}{(N-\beta)!}\right)t^{n}, (3.8)

where nn is any natural number and β{1,,N}\beta\in\{1,...,N\}. The functions ωn,j(z)\omega_{n,j}(z) are polynomials in zz satisfying the raising and lowering relations

zddzωn,β(z)+(n+α+1z)ωn,β(z)=(n+1)ωn+1,β(z)2ωn,β1(z),\displaystyle z\dfrac{d}{dz}\omega_{n,\beta}(z)+(n+\alpha+1-z)\omega_{n,\beta}(z)=(n+1)\omega_{n+1,\beta}(z)-2\omega_{n,\beta-1}(z), (3.9)
zddzωn,β(z)nωn,β(z)=(α+n)ωn1,β(z)2ωn1,β1(z).\displaystyle z\dfrac{d}{dz}\omega_{n,\beta}(z)-n\omega_{n,\beta}(z)=-(\alpha+n)\omega_{n-1,\beta}(z)-2\omega_{n-1,\beta-1}(z). (3.10)

By combining those two last equations, we can construct a linear second-order differential equation

zd2dz2ωn,β(z)+(1+αz)ddzωn,β(z)+nωn,β(z)=2ddzωn,β1(z),z\dfrac{d^{2}}{dz^{2}}\omega_{n,\beta}(z)+(1+\alpha-z)\dfrac{d}{dz}\omega_{n,\beta}(z)+n\omega_{n,\beta}(z)=-2\dfrac{d}{dz}\omega_{n,\beta-1}(z), (3.11)

which is the generalized Laguerre differential equation (2.7) but with a non-homogeneous term. For the highest chain, we have by construction that ωn,N(z)\omega_{n,N}(z) is the generalized Laguerre polynomials, i.e.

ωn,N(z)=Ln(α)(z).\omega_{n,N}(z)=L_{n}^{(\alpha)}(z). (3.12)

Note that any polynomial ωn,β(z)\omega_{n,\beta}(z) where n<0n<0 or β<N\beta<N is set to zero.

Using the representation above, the Casimir element 𝒦\mathcal{K} in equation (2.2) does not act centrally, that is

𝒦|n,β=α214|n,β+α|n,β+1+|n,β+2.\mathcal{K}|n,\beta\rangle=\dfrac{\alpha^{2}-1}{4}|n,\beta\rangle+\alpha|n,\beta+1\rangle+|n,\beta+2\rangle. (3.13)

By applying E+E^{+} repetitively on a state, we have

(E+)k|n,β=(n+k)!n!|n+k,β.(E^{+})^{k}|n,\beta\rangle=\dfrac{(n+k)!}{n!}|n+k,\beta\rangle. (3.14)

Therefore, if we exponentiate the raising operator E+E^{+} and apply it on a state, we obtain

exp(uE+)|n,β=k=0uk(n+k)!n!k!|n+k,β.\exp(uE^{+})|n,\beta\rangle=\sum_{k=0}^{\infty}u^{k}\dfrac{(n+k)!}{n!k!}|n+k,\beta\rangle. (3.15)

If we take n=0n=0, β=N\beta=N and u=1u=1, then we obtain the equivalent of the generating function of the Laguerre polynomial

exp(E+)|0,N=k=0Lk(α)(z)tk=(1t)(α+1)exp(zt1t).\exp(E^{+})|0,N\rangle=\sum_{k=0}^{\infty}L_{k}^{(\alpha)}(z)t^{k}=(1-t)^{-(\alpha+1)}\exp\left(\dfrac{-zt}{1-t}\right). (3.16)

In fact, applying the exponentiation of E+E^{+} on any locally analytic function f(z,t)f(z,t) (see [2]) , we have that

exp(uE+)f(z,t)=(1ut)(α+1)exp(uzt1ut)f(z1ut,t1ut).\exp(uE^{+})f(z,t)=(1-ut)^{-(\alpha+1)}\exp\left(\dfrac{-uzt}{1-ut}\right)f\left(\dfrac{z}{1-ut},\dfrac{t}{1-ut}\right). (3.17)

3.1 The two-chain example

In the case where we consider only two chains, i.e. N=2N=2, let us define the vector

Vn=(|n,2|n,1)=(Ln(α)(z)tn(Ln(α)(z)ln(tz)+ωn,1(z))tn).V_{n}=\left(\begin{array}[]{c}|n,2\rangle\\ \nobreak\leavevmode\hfil\\ |n,1\rangle\end{array}\right)=\left(\begin{array}[]{c}L_{n}^{(\alpha)}(z)t^{n}\\ \nobreak\leavevmode\hfil\\ (L_{n}^{(\alpha)}(z)\ln(tz)+\omega_{n,1}(z))t^{n}\end{array}\right). (3.18)

Then we have

TVn\displaystyle TV_{n} =\displaystyle= (n+α+1201n+α+12)Vn,\displaystyle\left(\begin{array}[]{cc}n+\dfrac{\alpha+1}{2}&0\\ 1&n+\dfrac{\alpha+1}{2}\end{array}\right)V_{n}, (3.21)
E+Vn\displaystyle E^{+}V_{n} =\displaystyle= (n+100n+1)Vn+1,\displaystyle\left(\begin{array}[]{cc}n+1&0\\ 0&n+1\end{array}\right)V_{n+1}, (3.24)
EVn\displaystyle E^{-}V_{n} =\displaystyle= ((n+α)02(n+α))Vn1.\displaystyle\left(\begin{array}[]{cc}-(n+\alpha)&0\\ -2&-(n+\alpha)\end{array}\right)V_{n-1}. (3.27)

The Casimir 𝒦\mathcal{K} can be represented as the matrix satisfying

𝒦Vn=(α2140αα214)Vn.\mathcal{K}V_{n}=\left(\begin{array}[]{cc}\dfrac{\alpha^{2}-1}{4}&0\\ \alpha&\dfrac{\alpha^{2}-1}{4}\end{array}\right)V_{n}. (3.28)

The polynomial ωn,1(z)\omega_{n,1}(z) can be solved completely in terms of Laguerre polynomials and ω0,1\omega_{0,1} which is a constant, that is

ωn,1(z)=σ1Ln(α)(z)+m=1n2mLnm(α)(z),where ω0,1(z)=σ1.\omega_{n,1}(z)=\sigma_{1}L_{n}^{(\alpha)}(z)+\sum_{m=1}^{n}\dfrac{2}{m}L_{n-m}^{(\alpha)}(z),\qquad\mbox{where\leavevmode\nobreak\ \leavevmode\nobreak\ }\omega_{0,1}(z)=\sigma_{1}. (3.29)

Using the generation function in equation (3.17) with u=1u=1 on the first state of the second chain, i.e. |0,1=σ1+ln(zt)|0,1\rangle=\sigma_{1}+\ln(zt) and substracting the Laguerre generating function (3.16), we get a generating function for ωn,1(z)\omega_{n,1}(z), i.e.

(σ12ln(1t))(1t)(α+1)exp(zt1t)=k=0ωk,1(z)tk.(\sigma_{1}-2\ln(1-t))(1-t)^{-(\alpha+1)}\exp\left(\dfrac{-zt}{1-t}\right)=\sum_{k=0}^{\infty}\omega_{k,1}(z)t^{k}. (3.30)

3.2 The three-chain example

For the case with three chains, i.e. N=3N=3, let us define the vector

Vn=(|n,3|n,2|n,1)=(Ln(α)(z)tn(Ln(α)(z)ln(tz)+ωn,1(z))tn(Ln(α)(z)ln(tz)22+ωn,1(z)ln(tz)+ωn,2)tn).V_{n}=\left(\begin{array}[]{c}|n,3\rangle\\ \nobreak\leavevmode\hfil\\ |n,2\rangle\\ \nobreak\leavevmode\hfil\\ |n,1\rangle\end{array}\right)=\left(\begin{array}[]{c}L_{n}^{(\alpha)}(z)t^{n}\\ \nobreak\leavevmode\hfil\\ (L_{n}^{(\alpha)}(z)\ln(tz)+\omega_{n,1}(z))t^{n}\\ \nobreak\leavevmode\hfil\\ \left(\dfrac{L_{n}^{(\alpha)}(z)\ln(tz)^{2}}{2}+\omega_{n,1}(z)\ln(tz)+\omega_{n,2}\right)t^{n}\end{array}\right). (3.31)

Then we have

TVn\displaystyle TV_{n} =\displaystyle= (n+α+12001n+α+12001n+α+12)Vn,\displaystyle\left(\begin{array}[]{ccc}n+\dfrac{\alpha+1}{2}&0&0\\ 1&n+\dfrac{\alpha+1}{2}&0\\ 0&1&n+\dfrac{\alpha+1}{2}\end{array}\right)V_{n}, (3.35)
E+Vn\displaystyle E^{+}V_{n} =\displaystyle= (n+1000n+1000n+1)Vn+1,\displaystyle\left(\begin{array}[]{ccc}n+1&0&0\\ 0&n+1&0\\ 0&0&n+1\end{array}\right)V_{n+1}, (3.39)
EVn\displaystyle E^{-}V_{n} =\displaystyle= ((n+α)002(n+α)002(n+α))Vn1.\displaystyle\left(\begin{array}[]{ccc}-(n+\alpha)&0&0\\ -2&-(n+\alpha)&0\\ 0&-2&-(n+\alpha)\\ \end{array}\right)V_{n-1}. (3.43)

The Casimir 𝒦\mathcal{K} can be represented as the matrix satisfying

𝒦Vn=(α21400αα21401αα214)Vn.\mathcal{K}V_{n}=\left(\begin{array}[]{cccc}\dfrac{\alpha^{2}-1}{4}&0&0\\ \alpha&\dfrac{\alpha^{2}-1}{4}&0\\ 1&\alpha&\dfrac{\alpha^{2}-1}{4}\end{array}\right)V_{n}. (3.44)

The polynomial ωn,2(z)\omega_{n,2}(z) can be solved completely in terms of Laguerre polynomials and ω0,2,ω0,1\omega_{0,2},\omega_{0,1} which are constants, that is

ωn,2(z)=σ2Ln(α)(z)+m=1n(2σ1+k=1m14kmLnm(α)(z)),where ω0,2(z)=σ2,ω0,1(z)=σ1\omega_{n,2}(z)=\sigma_{2}L_{n}^{(\alpha)}(z)+\sum_{m=1}^{n}\left(\dfrac{2\sigma_{1}+\sum_{k=1}^{m-1}\dfrac{4}{k}}{m}L_{n-m}^{(\alpha)}(z)\right),\quad\mbox{where\leavevmode\nobreak\ \leavevmode\nobreak\ }\omega_{0,2}(z)=\sigma_{2},\quad\omega_{0,1}(z)=\sigma_{1} (3.45)

or

ωn,2(z)=σ2Ln(α)(z)+k=0n1[2nkσ1+4(γ+ψ(0,nk))nk]Lk(α)(z),\omega_{n,2}(z)=\sigma_{2}L_{n}^{(\alpha)}(z)+\sum_{k=0}^{n-1}\left[\frac{2}{n-k}\sigma_{1}+\frac{4(\gamma+\psi(0,n-k))}{n-k}\right]L_{k}^{(\alpha)}(z), (3.46)

where γ\gamma is the Euler gamma and ψ(n,z)\psi(n,z) the polygamma functions.

Using the generation function in equation (3.17) with u=1u=1 on the first state of the third chain, i.e. |0,1=σ2+σ1ln(zt)+ln(zt)2|0,1\rangle=\sigma_{2}+\sigma_{1}\ln(zt)+\ln(zt)^{2}, we obtain, after subtracting the generating functions of Ln(α)L_{n}^{(\alpha)} and ωn,1\omega_{n,1}, that the generating function for ωn,2\omega_{n,2} is

(σ22σ1ln(1t)+2(ln(1t))2)(1t)(α+1)exp(zt1t)=k=0ωk,2(z)tk.\left(\sigma_{2}-2\sigma_{1}\ln(1-t)+2(\ln(1-t))^{2}\right)(1-t)^{-(\alpha+1)}\exp\left(\dfrac{-zt}{1-t}\right)=\sum_{k=0}^{\infty}\omega_{k,2}(z)t^{k}. (3.47)

4 Laguerre polynomials with one central operator

In the case where TT can be diagonalized, we can consider the actions of TT, E+E^{+} and EE^{-} as

T|n,β\displaystyle T|n,\beta\rangle =\displaystyle= (n+α+12)|n,β,\displaystyle\left(n+\dfrac{\alpha+1}{2}\right)|n,\beta\rangle, (4.1)
E+|n,β\displaystyle E^{+}|n,\beta\rangle =\displaystyle= (n+1+βN)|n+1,β,\displaystyle(n+1+\beta-N)|n+1,\beta\rangle, (4.2)
E|n,β\displaystyle E^{-}|n,\beta\rangle =\displaystyle= (βαnN)|n1,β+|n1,β+1,\displaystyle(\beta-\alpha-n-N)|n-1,\beta\rangle+|n-1,\beta+1\rangle, (4.3)

i.e. where TT acts centrally, E+E^{+} acts east and EE^{-} acts west and north-west. A realization of this action is given the same realization as equation (3.5), that is

T\displaystyle T =\displaystyle= tt+1+α2,\displaystyle t\partial_{t}+\dfrac{1+\alpha}{2},
E+\displaystyle E^{+} =\displaystyle= t(zz+tt+(1+αz)),\displaystyle t(z\partial_{z}+t\partial_{t}+(1+\alpha-z)),
E\displaystyle E^{-} =\displaystyle= zzttt.\displaystyle\dfrac{z\partial_{z}-t\partial_{t}}{t}.

Conversely to the previous case, we will consider that the lowest weight going west is not when n=0n=0 but when n=Nβn=N-\beta for a state |n,β|n,\beta\rangle for a system with NN state. For a finite number of chains, this arrangement of states takes the form of an infinite trapezoid.

For this representation, we cannot use TT to solve for the first state of the second chain, i.e. |1,N1|1,N-1\rangle, but we can use can use EE^{-}. We have

E|1,N1=|0,N.E^{-}|1,N-1\rangle=|0,N\rangle. (4.4)

N.B. that the first chain is known and given by |n,N=Lnα(z)tn|n,N\rangle=L_{n}^{\alpha}(z)t^{n}. Once we the state |1,N1|1,N-1\rangle, we can generate all the other states using E+E^{+}. By checking all the commutation relations of sl(2)sl(2), and then generating NN chains, we obtain

|n,β=ωn,β(z)tn,|n,\beta\rangle=\omega_{n,\beta}(z)t^{n}, (4.5)

where ωn,β(z)\omega_{n,\beta}(z) is polynomial in zz. Those polynomials ωn,β(z)\omega_{n,\beta}(z) satisfy the raising and lowering recurrence relations

zddzωn,β(z)+(n+α+1z)ωn,β(z)=(nβ+1)ωn+1,β(z),\displaystyle z\dfrac{d}{dz}\omega_{n,\beta}(z)+(n+\alpha+1-z)\omega_{n,\beta}(z)=(n-\beta+1)\omega_{n+1,\beta}(z), (4.6)
zddzωn,β(z)nωn,β=(α+n+β)ωn1,β(z)+ωn1,β1(z).\displaystyle z\dfrac{d}{dz}\omega_{n,\beta}(z)-n\omega_{n,\beta}=-(\alpha+n+\beta)\omega_{n-1,\beta}(z)+\omega_{n-1,\beta-1}(z). (4.7)

By combining those two last equations, we can construct a linear second-order differential equation

z2d2dz2ωn,β(z)+z(1+αz)ddzωn,β(z)+(nzαββ2)ωn,β(z)=(nβ+1)ωn,β1(z).z^{2}\dfrac{d^{2}}{dz^{2}}\omega_{n,\beta}(z)+z(1+\alpha-z)\dfrac{d}{dz}\omega_{n,\beta}(z)+(nz-\alpha\beta-\beta^{2})\omega_{n,\beta}(z)=(n-\beta+1)\omega_{n,\beta-1}(z). (4.8)

The Casimir acts north and centrally, that is

𝒦|n,β=(2β+α+1)(2β+α1)4|n,β+(n+1β)|n,β+1.\mathcal{K}|n,\beta\rangle=\dfrac{(2\beta+\alpha+1)(2\beta+\alpha-1)}{4}|n,\beta\rangle+(n+1-\beta)|n,\beta+1\rangle. (4.9)

By using E+E^{+}, it is possible to construct a generating function. We have

(E+)k|n,β=(n+k+βN)!(n+βN)!|n+k,β.(E^{+})^{k}|n,\beta\rangle=\dfrac{(n+k+\beta-N)!}{(n+\beta-N)!}|n+k,\beta\rangle. (4.10)

Hence,

exp(uE+)|n,β=k=0ukk!(n+k+βN)!(n+βN)!|n+k,β.\exp(uE^{+})|n,\beta\rangle=\sum_{k=0}^{\infty}\dfrac{u^{k}}{k!}\dfrac{(n+k+\beta-N)!}{(n+\beta-N)!}|n+k,\beta\rangle. (4.11)

By taking u=1u=1, n=0n=0 and β=N\beta=N and using previous results, we get

(1t)(α+1)exp(zt1t)=k=0Lk(α)(z)tk,(1-t)^{-(\alpha+1)}\exp\left(\dfrac{-zt}{1-t}\right)=\sum_{k=0}^{\infty}L_{k}^{(\alpha)}(z)t^{k}, (4.12)

which matches Laguerre’s generating function.

4.1 The two-chain example

Let us define the vector

Vn=(|n,2|n,1)=(Ln(α)(z)tnωn,1(z)tn).V_{n}=\left(\begin{array}[]{c}|n,2\rangle\\ \\ |n,1\rangle\end{array}\right)=\left(\begin{array}[]{c}L_{n}^{(\alpha)}(z)t^{n}\\ \\ \omega_{n,1}(z)t^{n}\end{array}\right). (4.13)

Hence, the elements TT, E+E^{+} and EE^{-} acts as

TVn\displaystyle TV_{n} =\displaystyle= (n+α+1200n+α+12)Vn,\displaystyle\left(\begin{array}[]{cc}n+\dfrac{\alpha+1}{2}&0\\ 0&n+\dfrac{\alpha+1}{2}\end{array}\right)V_{n}, (4.16)
E+Vn\displaystyle E^{+}V_{n} =\displaystyle= (n+100n)Vn+1,\displaystyle\left(\begin{array}[]{cc}n+1&0\\ 0&n\end{array}\right)V_{n+1}, (4.19)
EVn\displaystyle E^{-}V_{n} =\displaystyle= (nα01n1α)Vn1.\displaystyle\left(\begin{array}[]{cc}-n-\alpha&0\\ 1&-n-1-\alpha\end{array}\right)V_{n-1}. (4.22)

The Casimir can be represented as

𝒦Vn=((α+5)(α+3)/20n(α+3)(α+1)/2)Vn.\mathcal{K}V_{n}=\left(\begin{array}[]{cc}(\alpha+5)(\alpha+3)/2&0\\ n&(\alpha+3)(\alpha+1)/2\end{array}\right)V_{n}. (4.23)

The polynomial ωn,1(z)\omega_{n,1}(z) can be solved explicitly as a sum of Laguerre polynomials, i.e.

ωn,1(z)=σ1((α+1))k=0n1Lk(α)(z)nLn(α)(z))k=0n1Lk(α)(z).\omega_{n,1}(z)=\sigma_{1}\left((\alpha+1))\sum_{k=0}^{n-1}L_{k}^{(\alpha)}(z)-nL_{n}^{(\alpha)}(z)\right)-\sum_{k=0}^{n-1}L_{k}^{(\alpha)}(z). (4.24)

Using the lowest state on the second chain, i.e. |1,1=(σ1z1)t|1,1\rangle=(\sigma_{1}z-1)t, the generating function for the second chain takes the form

(σ1z1+t)t(1t)(α+3)exp(zt1t).(\sigma_{1}z-1+t)t(1-t)^{-(\alpha+3)}\exp\left(\dfrac{-zt}{1-t}\right). (4.25)

4.2 The three-chain example

Let us define the vector

Vn=(|n,3|n,2|n,1)=(Ln(α)(z)tnωn,1(z)tnωn,2(z)tn).V_{n}=\left(\begin{array}[]{c}|n,3\rangle\\ \\ |n,2\rangle\\ \\ |n,1\rangle\end{array}\right)=\left(\begin{array}[]{c}L_{n}^{(\alpha)}(z)t^{n}\\ \\ \omega_{n,1}(z)t^{n}\\ \\ \omega_{n,2}(z)t^{n}\end{array}\right). (4.26)

Hence, the elements TT, E+E^{+} and EE^{-} acts as

TVn\displaystyle TV_{n} =\displaystyle= (n+α+12000n+α+12000n+α+12)Vn,\displaystyle\left(\begin{array}[]{ccc}n+\dfrac{\alpha+1}{2}&0&0\\ 0&n+\dfrac{\alpha+1}{2}&0\\ 0&0&n+\dfrac{\alpha+1}{2}\end{array}\right)V_{n}, (4.30)
E+Vn\displaystyle E^{+}V_{n} =\displaystyle= (n+1000n000n1)Vn+1,\displaystyle\left(\begin{array}[]{ccc}n+1&0&0\\ 0&n&0\\ 0&0&n-1\end{array}\right)V_{n+1}, (4.34)
EVn\displaystyle E^{-}V_{n} =\displaystyle= (nα001n1α001n2α)Vn1.\displaystyle\left(\begin{array}[]{ccc}-n-\alpha&0&0\\ 1&-n-1-\alpha&0\\ 0&1&-n-2-\alpha\end{array}\right)V_{n-1}. (4.38)

The Casimir can be represented as

𝒦Vn=((α+7)(α+5)/200n1(α+5)(α+3)/200n(α+3)(α+1)/2)Vn.\mathcal{K}V_{n}=\left(\begin{array}[]{ccc}(\alpha+7)(\alpha+5)/2&0&0\\ n-1&(\alpha+5)(\alpha+3)/2&0\\ 0&n&(\alpha+3)(\alpha+1)/2\end{array}\right)V_{n}. (4.39)

An explicit expression for the polynomials is

ωn,2=n(n1)σ2Ln(α)(z)+(n1)(σ12σ2(2+α))Ln1(α)(z)\omega_{n,2}=n(n-1)\sigma_{2}L_{n}^{(\alpha)}(z)+(n-1)(\sigma_{1}-2\sigma_{2}(2+\alpha))L_{n-1}^{(\alpha)}(z) (4.40)
+k=0n2((nk12)σ1((nk1)α+(n12k))+σ2(2+α)((nk1)α+(n13k)))Lk(α)(z).+\sum_{k=0}^{n-2}((\frac{n-k-1}{2})-\sigma_{1}((n-k-1)\alpha+(n-1-2k))+\sigma_{2}(2+\alpha)((n-k-1)\alpha+(n-1-3k)))L_{k}^{(\alpha)}(z).

Using the lowest state on the second chain, i.e. |2,1=(σ2z2σ1z+1/2)t2|2,1\rangle=(\sigma_{2}z^{2}-\sigma_{1}z+1/2)t^{2}, the generating function for the second chain takes the form

(σ2z2σ1z(1t)+(1t)2/2)t2(1t)(α+5)exp(zt1t).(\sigma_{2}z^{2}-\sigma_{1}z(1-t)+(1-t)^{2}/2)t^{2}(1-t)^{-(\alpha+5)}\exp\left(\dfrac{-zt}{1-t}\right). (4.41)

5 Properties of the polynomials

In the construction of those indecomposable representations, there are various special functions, first at the level of the two variables differential operators satisfying an sl(2)sl(2) Lie algebra which involve the element of VnV_{n} that take the form of two-variables special functions. The ωn,i\omega_{n,i} are building block of the representations for the diagonal and nondiagonal cases. Those ωn,i\omega_{n,i} also satisfy various differential-recurrence relations based on the raising and lowering operators E+E^{+} and EE^{-}. We have also provided generating functions and explicit solutions in terms of linear combinations of Laguerre orthogonal polynomials which allow to get further insight into their properties. This section will be devoted to provide ordinary differential equations, further recurrence relations and properties such as bi-orthogonality type relations with another set of related polynomial and zeroes of those polynomials in the case of the two chains.

5.1 T nondiagonal

It is interesting to note that the functions ωn,i\omega_{n,i} satisfy coupled homegeneous/non-homegenous equations chains as well as recurrence relations but also this differential equations which can be used to study the further properties:

ωn,1(z)wn,1(z)+wn+1,1(z)=0.\omega_{n,1}(z)-w^{\prime}_{n,1}(z)+w^{\prime}_{n+1,1}(z)=0. (5.1)

Another relation obtained directly from the recurrence relation is the following

(α+n)wn1,1(z)+nwn,1(z)+(1+α+nz)wn,1(1+n)wn+1,1=2Ln1α(z)2Lnα(z).-(\alpha+n)w_{n-1,1}(z)+nw_{n,1}(z)+(1+\alpha+n-z)w_{n,1}-(1+n)w_{n+1,1}=2L_{n-1}^{\alpha}(z)-2L_{n}^{\alpha}(z). (5.2)

Taking a sum of the first relation (5.1) gives

ωk+1,1(z)+i=0kωi,1(z)=0.\omega_{k+1,1}^{\prime}(z)+\sum_{i=0}^{k}\omega_{i,1}(z)=0.

Here we will provide some more properties in particular for ωn,1\omega_{n,1}. Using the homogeneous differential equation for the Laguerre and the inhommogeneous differential equation for ωn,1\omega_{n,1}, we see that they satisfy this fourth-order, homogeneous differential equation

z2ωn,1′′′′+((5+2α)z2z2)ωn,1′′′+((2n52α)z+z2)ωn,1′′z^{2}\omega_{n,1}^{\prime\prime\prime\prime}+((5+2\alpha)z-2z^{2})\omega_{n,1}^{\prime\prime\prime}+((2n-5-2\alpha)z+z^{2})\omega_{n,1}^{\prime\prime}
+((n1)(3+2α)(2n2)z)ωn,1+n(n1)ωn,1=0+((n-1)(3+2\alpha)-(2n-2)z)\omega_{n,1}^{\prime}+n(n-1)\omega_{n,1}=0

This equation display some similarities with other context such a multiple Laguerre or Hermite polynomial which are orthogonal. Here the polynomial are not orthogonal under the usual Laguerre measure but satisfy analog of an bi-orthogonality relation of the form

0qm(z)ωn,1xαez𝑑z=0,n<m,\int_{0}^{\infty}q_{m}(z)\omega_{n,1}x^{\alpha}e^{-z}dz=0,\quad\forall n<m,

where qm(z)q_{m}(z) are another set of polynomial which can be explicitly written as

qm(z)=k=0n(nk)(1)2n+kzk(1+α)k.q_{m}(z)=\sum_{k=0}^{n}\binom{n}{k}(-1)^{2n+k}\frac{z^{k}}{(1+\alpha)_{k}}.

Note that for different choices of the parameters and nn the zeroes are on the real axis.

Refer to caption
Figure 1: Zeroes of ωn,1\omega_{n,1} for n=6n=6, σ=2\sigma=2 and α=10\alpha=10
Refer to caption
Figure 2: Zeroes of ωn,1\omega_{n,1} for n=50n=50, σ=1100\sigma=\frac{1}{100} and α=1100\alpha=\frac{1}{100}

5.2 T diagonal

We consider the polynomials ωn,1\omega_{n,1} from the T diagonal related to Laguerre case. Those polynomial can be re-expressed as linear combination of Laguerre and possess generating function. They also satisfy as demonstrated various differential-recurrence relations which are direct consequences of the action of the generator of the sl(2) algebra. We also have the following relation for given ωn,i\omega_{n,i}

(n+α+1z+n)ωn,i=(ni+1)ωn+1,i+(α+n+i)ωn1,iωn1,i1.(n+\alpha+1-z+n)\omega_{n,i}=(n-i+1)\omega_{n+1,i}+(\alpha+n+i)\omega_{n-1,i}-\omega_{n-1,i-1}.

In the case of ωn,1\omega_{n,1}, it can be shown that

(1+σ1+ασ1)zωn,1′′′+((3+α)(1+σ1+σ1α)+(1σ1((n+1)+α))z)ωn,1′′(-1+\sigma_{1}+\alpha\sigma_{1})z\omega_{n,1}^{\prime\prime\prime}+((3+\alpha)(-1+\sigma_{1}+\sigma_{1}\alpha)+(1-\sigma_{1}(-(n+1)+\alpha))z)\omega_{n,1}^{\prime\prime}
+((n1)+(2n2)σ1(1+α)nσ1z)ωn,1+n2σ1ωn,1=0.+(-(n-1)+(2n-2)\sigma_{1}(1+\alpha)-n\sigma_{1}z)\omega_{n,1}^{\prime}+n^{2}\sigma_{1}\omega_{n,1}=0.

Other aspect such as orthogonality/bi-orthogonality can be investigated here. For example, we have the following orthogonality

0qm(z)ωn,1xαez𝑑z=0,nm,\int_{0}^{\infty}q_{m}(z)\omega_{n,1}x^{\alpha}e^{-z}dz=0,\quad\forall n\leq m,

where

qm(z)=k=0n(nk)(1)2n+kzk(1+σ1+ασ1)(1+α)k(1+(k+1)σ1+ασ1).q_{m}(z)=\sum_{k=0}^{n}\binom{n}{k}(-1)^{2n+k}\frac{z^{k}(-1+\sigma_{1}+\alpha\sigma_{1})}{(1+\alpha)_{k}(-1+(k+1)\sigma_{1}+\alpha\sigma_{1})}.

For different choices of the parameters and nn the zeroes are on the real axis, as in the nondiagonal case.

Refer to caption
Figure 3: Zeroes of ωn,1\omega_{n,1} for n=8n=8, σ=2\sigma=2 and α=10\alpha=10
Refer to caption
Figure 4: Zeroes of ωn,1\omega_{n,1} for n=20n=20, σ=11000\sigma=\frac{1}{1000} and α=11000\alpha=\frac{1}{1000}

6 Further algebraic aspects of the solutions

We have described two cases of representations, with T diagonalizable and T nondiagonalible (both in context of Laguerre polynomials). We have obtained analogs of the Rodrigues formula from (3.9) and (4.6), which generalized well known formula for classical orthogonal polynomials.

In those case of sl(2,)sl(2,\mathbb{C}) indecomposable representations we can understand the construction from point of view of induced representation and the modules can be generated from an n-dimensional indecomposable representation of the Borel of sl(2,)sl(2,\mathbb{C}). The those modules can be related via the following transformation of generators (taking E=eE^{-}=-e and E+=fE^{+}=-f) and by multiplicative factor of the module:

[e,f]=h,[e,f]=h, (6.1)
[h,e]=2e,[h,e]=2e, (6.2)
[h,f]=2f,[h,f]=-2f, (6.3)

with further relations

[h,ea]=2aea,[h,e^{a}]=2ae^{a},
[f,ea]=a(h+a1)ea1,[f,e^{a}]=-a(h+a-1)e^{a-1},
[e,fa]=a(h+a1)fa1,[e,f^{a}]=a(h+a-1)f^{a-1},
[h,fa]=2afa.[h,f^{a}]=-2af^{a}.

T nondiagonalizable

We consider first the case where the generator of sl(2)sl(2) is nondiagonalizable. The modules which form the representations of the Borel subalgebra {e,h}\{e,h\} can be denoted v(1)0(n)v_{(1)0}^{(n)} and then satisfy the explicit action

hv(1)0(n)=λv(1)0(n),\displaystyle hv_{(1)0}^{(n)}=\lambda v_{(1)0}^{(n)}, (6.4)
hv(1)0(k)=λv(1)0(k)+v(1)0(k+1),k=1,,n1,\displaystyle hv_{(1)0}^{(k)}=\lambda v_{(1)0}^{(k)}+v_{(1)0}^{(k+1)},\quad k=1,...,n-1,
ev(1)0(i)=0,i=1,,n.\displaystyle ev_{(1)0}^{(i)}=0,\quad i=1,...,n.

Here the index (1)(1) refer to the choice of representations for the Borel subalgebra v(1)0(n)v_{(1)0}^{(n)} which can be seen as the choice of a column ( n×1n\times 1) in the space of v(i)0(j)v_{(i)0}^{(j)} which consist of a lattice (n×Nn\times N) formed by the collection of generalized Weight spaces. They are related via the generator ee. Here we have one of such generalized Weight space of dimension nn.

Then generating the other chains in the following way with modules denoted by v(1)a(i)v_{(1)a}^{(i)} for i=1,..,n1i=1,..,n-1 can be done as follow:

v(1)a(i)=fav(1)0(i),i=1,..,n1,\displaystyle v_{(1)a}^{(i)}=f^{a}v_{(1)0}^{(i)},i=1,..,n-1, (6.5)
v(1)a(n)=fav(1)0(n).\displaystyle v_{(1)a}^{(n)}=f^{a}v_{(1)0}^{(n)}.

Using commutation relations we obtain the following action

hv(1)a(i)=(λ2a)v(1)a(i)+v(1)a(i+1),\displaystyle hv_{(1)a}^{(i)}=(\lambda-2a)v_{(1)a}^{(i)}+v_{(1)a}^{(i+1)}, (6.6)
hv(1)a(n)=(λ2a)v(1)a(n),\displaystyle hv_{(1)a}^{(n)}=(\lambda-2a)v_{(1)a}^{(n)},
fv(1)a(i)=v(1)a+1(i),i=1,,n1,\displaystyle fv_{(1)a}^{(i)}=v_{(1)a+1}^{(i)},\quad i=1,...,n-1,
fv(1)a(n)=v(1)a+1(n),\displaystyle fv_{(1)a}^{(n)}=v_{(1)a+1}^{(n)},
ev(1)a(i)=a(λa+1)v(1)a1(i)+av(1)a1(i+1),\displaystyle ev_{(1)a}^{(i)}=a(\lambda-a+1)v_{(1)a-1}^{(i)}+av_{(1)a-1}^{(i+1)},
ev(1)a(n)=a(λa+1)v(1)a1(n).\displaystyle ev_{(1)a}^{(n)}=a(\lambda-a+1)v_{(1)a-1}^{(n)}.

T diagonalizable

We now consider the case where (TT i.e. the Cartan element h is diagonalizable ). The modules forming the representation of the Borel subalgebra ({e,h}\{e,h\} or {E,T}\{E^{-},T\}) are given by v(k)0(1)v_{(k)0}^{(1)},

hv(k)0(1)=(λ+2(k1))v(k)0(1),\displaystyle hv_{(k)0}^{(1)}=(\lambda+2(k-1))v_{(k)0}^{(1)}, (6.7)
ev(n)0(1)=0,\displaystyle ev_{(n)0}^{(1)}=0,
ev(i)0(1)=v(i+1)0(1),i=1,,n1,\displaystyle ev_{(i)0}^{(1)}=v_{(i+1)0}^{(1)},\quad i=1,...,n-1,

Here the index (1)(1) refer to the choice of representations for the Borel subalgebra v(i)0(1)v_{(i)0}^{(1)} which can be seen as the choice of a row in the space of v(i)0(j)v_{(i)0}^{(j)} which consist of a lattice (n×Nn\times N) formed by collections of generalized Weight spaces (indexed by ii). Here we have nn one-dimensional Weight spaces.

with the modules of the other chains are constructed via

v(i)a(1)=fav(i)0(1),\displaystyle v_{(i)a}^{(1)}=f^{a}v_{(i)0}^{(1)}, (6.8)
v(n)a(1)=fav(n)0(1),\displaystyle v_{(n)a}^{(1)}=f^{a}v_{(n)0}^{(1)},

and we have

hv(i)a(1)=(λ+2(i1)2a)v(i)a(1),\displaystyle hv_{(i)a}^{(1)}=(\lambda+2(i-1)-2a)v_{(i)a}^{(1)}, (6.9)
hv(n)a(1)=(λ+2(n1)2a)v(n)a(1),\displaystyle hv_{(n)a}^{(1)}=(\lambda+2(n-1)-2a)v_{(n)a}^{(1)},
fv(n)a(1)=v(n)a+1(1),\displaystyle fv_{(n)a}^{(1)}=v_{(n)a+1}^{(1)},
fv(i)a(1)=v(i)a+1(1),i=1,n1,\displaystyle fv_{(i)a}^{(1)}=v_{(i)a+1}^{(1)},\quad i=1,...n-1,
ev(i)a(1)=a(λ+2ia1)v(i)a1(1)+v(i+1)a(1),\displaystyle ev_{(i)a}^{(1)}=a(\lambda+2i-a-1)v_{(i)a-1}^{(1)}+v_{(i+1)a}^{(1)},
ev(n)a(1)=a(λ+2na1)v(n)a1(1).\displaystyle ev_{(n)a}^{(1)}=a(\lambda+2n-a-1)v_{(n)a-1}^{(1)}.

Realization of these representations as differential operators has been a so far unexplored subject and the results above represent the first steps of a program in which hypergeometric, Laguerre and other special functions as well as generalization in term of non homogeneous linear differential equations are considered in this context.

7 Conclusion

In this paper we have presented a complete description of two different types of infinite dimensional indecomposable representations of sl(2)sl(2) involving N chains and their related differential operator realizations with two variables. They correspond to two cases for which the action of the generator T is diagonalizable or nondiagonalizable. Those results constitute generalization of problems considered in [2, 19] previously for sl(2)sl(2).

We have presented several results such as second-order differential-recurrence relations for the ωn,i\omega_{n,i} which are polynomials that are part of the chain as building blocks. Those polynomials also satisfy for the polynomials ωn,1\omega_{n,1} of T diagonal case and T nondiagonal case respectively third- and fourth-order ordinary differential equations. We have also presented further recurrence formula, generating functions and related polynomials satisfying bi-orthogonal type of relations. We have also investigated zeroes of those new polynomials and they lie on a segment on the real axis. Let us remark that those polynomials are not orthogonal under Laguerre’s weight, and there are many instances of polynomials beyond classical orthogonal polynomials that are of interest for applications in physics and in particular in quantum physics. Among them the para-Jacobi polynomials that was exploited in context of supersymmetric quantum mechanics. We can think of Bender-Dunne polynomial referred as quasi-orthogonal polynomials.

From a perspective of systems of coupled non-homogeneous ODEs the systems considered in this papers extend those previously introduced [28, 29, 30, 31, 32] and proposed an algebraic setting. As special functions and in particular Laguerre appear in many context and in particular related to solution of different PDEs among them the Schrödinger equations and as solution of ODE obtained from separation of variables [33, 34] this is other context where results may find useful applications.

The case of non-homogeneous systems of ODE as also been seen in context of supersymmetric quantum mechanics and was involving Jordan blocks [35, 36]. Other areas such as pseudo-hermitian Hamiltonian were also known to be connected with representations involving Jordan blocks [37, 38]. The results obtained in this paper and further generalization are likely to find application in context of quantum mechanical problem or other setting of mathematical physics.

Data availability statement

The data that support the findings of this study are available upon reasonable request from the authors.

Acknowledgement

IM was supported by the Australian Research Council Future Fellowship FT180100099.

References

  • [1] R S Ismagilov, linear representation of the group sl(2,R), Mat Sbornik, 3 4 453-469 (1967)
  • [2] W., Miller, Jr., Lie Theory and Special Functions, Academic Press, New York, 1968
  • [3] W Miller Jr., Lie Theory and Some Special Solutions of the Hypergeometric Equations, SIAM J. Math. An. 1. (1970)
  • [4] W.Miller, Lie theory and the Lauricella fucntions FDF_{D}, J. Math.Phys. 13 1393 (1972)
  • [5] W.Miller Jr, Lie algebras and generalizations of the hypergeometric functions, in harmonic analysis on homogeneous space ( proceedings of the symposium on pure mathematics 26) Amer. Math. Soc. , Providence R I 355 (1973)
  • [6] W.Miller, Jr, Lie theory and generalizations of the hypergeometric functions, SIAM J. of Appl. Math. 25 2 226 (1973)
  • [7] E.G. Kalnins, H.L. Manocha and W.Miller, The Lie theory of two-variables hypergeometric fucntions, Stud. in Appl. Math. 62 143 (1980)
  • [8] DM Fradkin, Three-dimensional Isotropic harmonic oscillator and SU3SU_{3}, American Journal of Physics, Journal Volume: Vol: 33 1965
  • [9] A. O. Barut, Dynamics of a Broken SU(N), Symmetry for the Oscillator Phys. Rev. 139, B1433
  • [10] P. Winternitz, Y. A. Smorodinsky, M. Uhlir and I. Fris, Symmetry groups in classical and quantum mechanics, Sov. J. Nucl. Phys. 4 (1967) 444
  • [11] A. V. Turbiner. Quasi-exactly-solvable problems and sl(2, R) algebra. Comm. Math. Phys., 118:467-474, 1988.
  • [12] T H Koornwinder, A global approach to the representation theory of sl(2,R) , Lecture notes, 1978
  • [13] D.Basu and B Wolf, The unitary irreducible representations of sl(2,R) in all subgroup reductions, J.Math.Phys. 23 2 189-205 1982
  • [14] D.Basu and T Bhattacharya, The Gel’fand realization and the exceptional representations of sl(2,R), J.Math.Phys. 26 12 12-17 (1985)
  • [15] T. Koornwinder, Group theoretical interpretation of Askey’s scheme of hypergeometric orthogonal polynomial, Orthogonal polynomials and their application, Segovia Spain (1986) 46-72
  • [16] MA Al-Bassam and HL Manocha, Lie Theory of Solutions of Certain Differintegral Equations, SIAM J. Math. An. 18. (1987)
  • [17] D.Basu, Representations of sl(2,R) in a Hilbert space of analytic functions and a class of integral transforms, J.Math.Phys. 30 1 1-8 (1989)
  • [18] H. Bacry, A unitary representation of sl(2,R), J.Math.Phys. 31 2061-2077 (1990)
  • [19] N.N. Ja Vilenkin and A.U. Klimyk, Representation of Lie Groups and special function Vol1 : Simplest Lie groups, special function and integral transform ,Springer 1991
  • [20] R.Howe and E. C Tan , Non Abelian Harmonic analysis Application of sl(2,R), Springer Verlag New York 1992
  • [21] JE Humphreys, Representations of Semisimple Lie Algebras in the BGG Category 𝒪\mathcal{O} (AMS, New York), 2008
  • [22] V Mazorchhihk Lectures on 𝔰𝔩2()\mathfrak{sl}_{2}(\mathbb{C})-Modules (Imperial College Press) 2009
  • [23] Y.Su, A Classifiction of indecomposable sl(2,c) modules and a conjeture of Kac on irreducible modules over the Virasoro algebra, J. of Algebra 161 33-46 (1993)
  • [24] J Rasmussen, Affine Jordan cells, logarithmic correlators, and Hamiltonian reduction, Nuclear Physics B 736 225–258. 2006
  • [25] E.L.Ince, Ordinary differential equations, Dover New York (1956)
  • [26] E. Koelik, Orthogonal Polynomials and special functions, Lecture notes in mathematics, Springer, Leuven ,2002
  • [27] L.J.S. Slater, Generalized Hypergeometric Functions, Cambridge University press 1966
  • [28] A.W. Babister, Transcendental Functions satisfying non homogeneous linear differential equations, MacMillan Co, New York, 1967
  • [29] W.Miller Jr, Lie theory and special function satisfying second order non homogeneous differential equations, SIAM J. Math. Anal. 1 246 (1970)
  • [30] DW Zachmann, Multiple solutions of coupled Sturm-Liouville systems, J. Math. An. Appl. 54 467–475. 1976
  • [31] N.B. Backhouse, The resonant Legendre equation, J.of Math. Anal. and Appl. 117 310 (1986)
  • [32] N.B. Backhouse, Resonant equations and special functions, J.of Math. Anal. and Appl. 133 163 (2001)
  • [33] E. G. Kalnins, Separation of variables for Riemannian spaces of non curvature, Wiley, 1986
  • [34] W. Miller, Symmetry and separation of variables, Cambridge University press, 1984
  • [35] A Schulze-Halberg, Arbitrary-order Jordan chains associated with quantum-mechanical Hamiltonians: Representations and integral formulas, J. Math. Phys. 57 023521 2006
  • [36] A Contreras-Astorga and A Schulze-Halberg, On integral and differential representations of Jordan chains and the confluent supersymmetry algorithm, J. Phys. A: Math. Theor. 48 315202.
  • [37] Ian Marquette a and Christiane Quesne, Ladder Operators and Hidden Algebras for Shape Invariant Nonseparable and Nondiagonalizable Models with Quadratic Complex Interaction. I. Two-Dimensional Model,SIGMA 18 (2022), 004, 11 pages
  • [38] Ian Marquette a and Christiane Quesne, Ladder Operators and Hidden Algebras for Shape Invariant Nonseparable and Nondiagonalizable Models with Quadratic Complex Interaction. II. Three-Dimensional Model,SIGMA 18 (2022), 005, 24 pages