Abstract
Classical orthogonal polynomials are known to satisfy seven equivalent properties, namely the Pearson equation for the linear functional, the second-order differential/difference/q-differential/ divided-difference equation, the orthogonality of the derivatives, the Rodrigues formula, two types of structure relations, and the Riccati equation for the formal Stieltjes function. In this work, following previous work by Kil et al. (J Differ Equ Appl 4:145–162, 1998a; Kyungpook Math J 38:259–281, 1998b), we state and prove a non-linear characterization result for classical orthogonal polynomials on non-uniform lattices. Next, we give explicit relations for some families of these classes.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Univariate orthogonal polynomials (or orthogonal polynomials for short) are systems of polynomials \((p_n)_n\) with \(\deg (p_n) = n\), satisfying a certain orthogonality relation. They are very useful in practice in various domains of mathematics, physics, engineering, image processing and so on, because of the many properties and relations they satisfy. As examples of areas where orthogonal polynomials play important roles, we could cite approximation theory (see for example [6, 31]) and also numerical analysis (see [14, 15]).
It is known that any family of orthogonal polynomials \({(p_n)_{n\ge 0}}\) satisfies a three-term recurrence relation of the form
If \( {h_n=\langle {\mathcal {L}},p_n^2\rangle }\), where \({\mathcal {L}}\) is the corresponding linear functional with respect to the sequence \({(p_n)_{n\ge 0}}\) and \(k_n\) is the leading coefficient of \({ p_n(x)}\) (see [17]), then
and we set \(C_0=1\).
The systems of orthogonal polynomials associated with the names of Hermite, Laguerre, Jacobi and Bessel (including the special cases named after Tchebychev, Legendre, and Gegenbauer) are the most extensively and widely applied systems.
An orthogonal polynomial system \((p_n)_{n\ge 0}\) with respect to a weight function \(\rho (x)\) is called classical if it satisfies one of the equivalent assertions (see [17]):
-
\((p_n)_{n\ge 0}\) satisfies a second-order linear differential equation of the Sturm–Liouville type
$$\begin{aligned} \phi (x)y''(x)+\psi (x)y'(x)+\lambda _n y(x)=0, \end{aligned}$$(1.2)where \(\phi (x)\) is a polynomial of degree \(\le 2\) and \(\psi (x)\) is a polynomial of exact degree 1, both independent on n and \(\lambda _n\) is independent on x.
-
The derivatives \((p'_{n+1})_{n\ge 0}\) form an orthogonal polynomial system.
-
The \(p_n\)s have the Rodrigues representation
$$\begin{aligned} p_n(x)={\dfrac{D_n}{\rho (x)}}\left( \phi ^n(x)\rho (x)\right) ^{(n)},\quad n\ge 0. \end{aligned}$$(1.3) -
The weight function \(\rho (x)\) satisfies a Pearson-type equation
$$\begin{aligned} (\phi (x)\rho (x))'=\psi (x)\rho (x). \end{aligned}$$(1.4) -
The \(p_n\)s satisfy a difference-differential equation (or structure relation) of the form
$$\begin{aligned} \pi (x)p'_n(x)=(\alpha _nx+\beta _n)p_n(x)+\gamma _n p_{n-1}(x). \end{aligned}$$(1.5)
In his paper [2], Al-Salam has obtained an expression for the derivative of the product of two consecutive Bessel polynomials and has shown that this expression does, in fact, characterize the Bessel polynomials. Based on this paper, McCarthy in [25] proved that there is an analogous characterization for very classical orthogonal polynomials (Hermite, Laguerre and Jacobi polynomials). This characterization can be stated as
-
\((p_n)_{n\ge 0}\) satisfies a non-linear equation of the form:
$$\begin{aligned} \phi (x)\dfrac{{\text {d}}}{{\text {d}}x}(p_n(x)p_{n-1}(x))=(\alpha _nx+\beta _n)p_n(x) p_{n-1}(x)+\gamma _n p_n^2(x)+\delta _n p^2_{n-1}(x){,} \end{aligned}$$(1.6)where \(\alpha _n\), \(\beta _n\), \(\gamma _n\) and \(\delta _n\) are independent on x.
Note that several other characterizations of classical orthogonal polynomials with respect to the derivative operator can be found in [23].
Very close to the very classical orthogonal polynomials (classical orthogonal polynomials of a continuous variable) are the classical orthogonal polynomials of a discrete variable. An orthogonal polynomial system \((p_n)_{n\ge 0}\) of a discrete variable with respect to a weight function \(\rho (x)\) is called classical if it satisfies one of the equivalent assertions (see [1, 8, 13]):
-
\({(p_n)_{n\ge 0}}\) satisfies a second-order linear difference equation of the Sturm–Liouville type
$$\begin{aligned} \phi (x)\Delta \nabla y(x)+\psi (x)\Delta y(x)+\lambda _n y(x)=0, \end{aligned}$$(1.7)where \(\phi (x)\) is a polynomial of degree \(\le 2\) and \(\psi (x)\) is a polynomial of exact degree 1, both independent on n and \(\lambda _n\) is independent on x.
-
The sequence of difference polynomials \((\Delta p_{n+1})_{n\ge 0}\) form an orthogonal polynomial system of discrete variable.
-
The \(p_n\)s have the Rodrigues representation
$$\begin{aligned} p_n(x)={\dfrac{D_n}{\rho (x)}}\Delta ^n\left( \phi ^n(x)\rho (x)\right) ,\quad n\ge 0. \end{aligned}$$(1.8) -
The weight function \(\rho (x)\) satisfies a Pearson-type equation
$$\begin{aligned} \Delta [\phi (x)\rho (x)]=\psi (x)\rho (x){.} \end{aligned}$$(1.9) -
The \(p_n\)s satisfy a difference equation (or structure relation) of the form
$$\begin{aligned} \pi (x)\nabla p_n(x)=(\alpha _nx+\beta _n)p_n(x)+\gamma _n p_{n-1}(x){,} \end{aligned}$$(1.10)or otherwise stated (see [19])
$$\begin{aligned} \phi (x) \nabla p_{n}(x)=\tilde{\alpha }_n p_{n+1}(x) +\tilde{\beta }_n p_n(x)+\tilde{\gamma }_n p_{n-1}(x){.} \end{aligned}$$(1.11) -
For each \(n\ge 1\), \(p_n\) and \(p_{n-1}\) satisfy a relation of the form (see [21, Theorem 5.2])
$$\begin{aligned}{} & {} \pi (x)\left[ p_n(x)\nabla p_{n-1}(x)+p_{n-1}(x)\nabla p_{n}(x)\right] \\{} & {} \quad =U_n p_n^2(x)+V_n p_{n-1}^2(x) +(W_nx+Y_n)p_n(x)p_{n-1}(x), \end{aligned}$$where the coefficients \(U_n\), \(V_n\), \(W_n\) and \(Y_n\) are independent on x and \(\pi \) is a polynomial of degree less or equal to 2.
It should be noted that the operators \(\Delta \) and \(\nabla \) are respectively defined by
Close to the classical discrete orthogonal polynomials are classical orthogonal polynomials of a q-discrete variable. An orthogonal polynomial system \((p_n)_{n\ge 0}\) of a q-discrete variable with respect to a weight function \(\rho (x)\) is called classical if it satisfies one of the equivalent assertions (see [8, 18, 19]):
-
\({(p_n)_{n\ge 0}}\) satisfies a second-order linear q-difference equation of the Sturm–Liouville type
$$\begin{aligned} \phi (x){\mathcal {D}}_q{\mathcal {D}}_{\frac{1}{q}} y(x)+\psi (x){\mathcal {D}}_q y(x)+\lambda _n y(x)=0, \end{aligned}$$(1.12)where \(\phi (x)\) is a polynomial of degree less than or equal to 2 and \(\psi (x)\) is a polynomial of exact degree 1, both independent on n and \(\lambda _n\) is independent on x.
-
The sequence of q-difference polynomials \(({\mathcal {D}}_q p_{n+1})_{n\ge 0}\) form an orthogonal polynomial system of a q-discrete variable.
-
The \(p_n\)s have the Rodrigues representation
$$\begin{aligned} p_n(x)={\dfrac{D_n}{\rho (x)}}{\mathcal {D}}_q^n\left( \phi ^n(x)\rho (x)\right) ,\quad n\ge 0. \end{aligned}$$(1.13) -
The weight function \(\rho (x)\) satisfies a Pearson-type equation
$$\begin{aligned} {\mathcal {D}}_q[\phi (x)\rho (x)]=\psi (x)\rho (x). \end{aligned}$$(1.14) -
The \(p_n\)s satisfy a q-difference equation (or structure relation) of the form (see [19])
$$\begin{aligned} \phi (x) {\mathcal {D}}_{\frac{1}{q}} p_{n}(x)=\tilde{\alpha }_n p_{n+1}(x) +\tilde{\beta }_n p_n(x)+\tilde{\gamma }_n p_{n-1}(x). \end{aligned}$$(1.15) -
For each \(n\ge 1\), \(p_n\) and \(p_{n-1}\) satisfy a relation of the form (see [22, Theorem 3.5])
$$\begin{aligned}{} & {} \tilde{\pi }(x)\left[ p_n(x){\mathcal {D}}_{\frac{1}{q}} p_{n-1}(x)+p_{n-1}(x) {\mathcal {D}}_{\frac{1}{q}} p_{n}(x)\right] \\{} & {} \quad =\tilde{U}_n p_n^2(x)+\tilde{V}_n p_{n-1}^2(x) +(\tilde{W}_nx +\tilde{Y}_n)p_n(x)p_{n-1}(x), \end{aligned}$$where the coefficients \(\tilde{U}_n\), \(\tilde{V}_n\), \(\tilde{W}_n\) and \(\tilde{Y}_n\) are independent on x and \(\tilde{\pi }\) is a polynomial of degree less than or equal to 2.
It should be noted that the q-derivative \(D_q\) is defined as
The difference operator \(\Delta \) and the q-derivative \(D_q\) are both special cases of the Hahn’s operator \(D_{q,\omega }\) (see [7]) which is defined as
More precisely, \(D_q=D_{q,0}\) and \(\Delta = D_{1,1}\).
In this paper, we prove equivalent non-linear characterization results similar to (1.6) for classical orthogonal polynomials on non-uniform lattices (including Wilson and Askey–Wilson polynomials). Also, we prove such a non-linear characterization for Meixner–Pollaczek and Continuous Hahn polynomials. Indeed, we give explicitly the coefficients of these relations for some families of classical orthogonal polynomials on non-uniform lattices.
2 Preliminaries
This section contains some preliminary definitions and results that are useful for a better reading of this article. The q-hypergeometric series, a fractional q-derivative and fractional q-integral are defined. The reader will consult the reference [18] for more informations about these concepts.
2.1 The Hypergeometric Series
In what follows, the symbol \((a)_n\) denotes the so-called Pochhammer symbol and is defined by
and the hypergeometric series is defined as
2.2 The q-Hypergeometric Series
The basic hypergeometric or q-hypergeometric series \(_r\phi _s\) is defined by the series
where
with
For \(n=\infty \), we set
The notation \((a;q)_n\) is the so-called q-Pochhammer symbol.
2.3 Difference and Divided-Difference Operators
2.3.1 The Operators \({\mathcal {D}}\) and \({\mathcal {S}}\)
We define the difference operator \({\mathcal {D}}\) (see [26, 28]) and its companion operator \({\mathcal {S}}\) as follows:
with \(i^{2}=-1\).
The operator \({\mathcal {D}}\) transforms a polynomial of degree n (\(n\ge 1\)) in x into a polynomial of degree \(n-1\) in x and a polynomial of degree 0 into the zero polynomial. The operator \({\mathcal {S}}\) transforms a polynomial of degree n in x into a polynomial of degree n in x.
The operators \({\mathcal {D}}\) and \({\mathcal {S}}\) fulfill the following properties.
Proposition 2.1
(See [26, 30]) The operators \({\mathcal {D}}\) and \({\mathcal {S}}\) satisfy the following product rules
where \({\textbf{I}}f=f\).
2.3.2 The Operators \({\textbf{D}}\) and \({\textbf{S}}\)
We define the difference operator \({\textbf{D}}\) (see [27]) and its companion operator \({\textbf{S}}\) as follows:
with \(i^{2}=-1\). The operator \({\textbf{D}}\) transforms a polynomial of degree n (\(n\ge 1\)) in \(x^2\) into a polynomial of degree \(n-1\) in \(x^2\) and a polynomial of degree 0 into the zero polynomial. The operator \({\textbf{S}}\) transforms a polynomial of degree n in \(x^2\) into a polynomial of degree n in \(x^2\).
The operators \({\textbf{D}}\) and \({\textbf{S}}\) fulfill the following properties.
Proposition 2.2
(See [27]) The operators \({\textbf{D}}\) and \({\textbf{S}}\) satisfy the following product rules
where \(\textbf{I}f=f\).
2.3.3 The Operators \({\mathbb {D}}_{x}\) and \({\mathbb {S}}_{x}\)
We define the operator \({\mathbb {D}}_{x}\) (called divided-difference operator) and its companion operator \({\mathbb {S}}_{x}\) (called mean operator) as [5, 9, 11, 29]
where x(s) is a non-uniform lattice (see [9]). The operator \({\mathbb {D}}_{x}\) transforms a polynomial of degree n (\(n\ge 1\)) in x(s) into a polynomial of degree \(n-1\) in x(s) and a polynomial of degree 0 into the zero polynomial. The operator \({\mathbb {S}}_{x}\) transforms a polynomial of degree n in x(s) into a polynomial of degree n in x(s).
The operators \({\mathbb {D}}_{x}\) and \({\mathbb {S}}_{x}\) satisfy the product rules
where \(U_2\) is a polynomial of degree 2
and \(\delta _{x}\) is a constant depending on \(\alpha ,\,\beta \) and the initial values x(0) and x(1) of x(s):
and
Note that
where \(F_n(x(s))\) is a function defined in [24]. More properties of the non-uniform lattices x(s), the properties of the divided-difference operator \({\mathbb {D}}_{x}\) and its companion \({\mathbb {S}}_{x}\) can be found in [10,11,12, 16, 24] : x(s) satisfies the conditions
for \(k=0,1, \ldots \), with
and the sequences \((\alpha _k)\), \((\beta _k)\), \((\gamma _k)\) satisfy the following relations
for \(k=0,1, \ldots \).
3 Non-linear Characterization for Meixner–Pollaczek and Continuous Hahn Polynomials
The Meixner–Pollaczek polynomials \(P_n^{(\lambda )}(x;\varphi )\) and the Continuous Hahn polynomials \(p_n(x;a,b,c,d)\), respectively, have the hypergeometric representation (see [18]):
They are known to satisfy the second-order difference equation (see [30])
where \(\phi \) and \(\psi \) are polynomials of degree 2 and 1, respectively, and \(\lambda \) is a constant depending on the degree of the polynomial solution and the parameters involved in the polynomials.
Note that for the Meixner–Pollaczek polynomials, we have (see [30])
and
and for the Continuous Hahn polynomials we have (see [30])
and
Theorem 3.1
(Non-linear characterization) Let \({(P_n)_{n\ge 0}}\) be a sequence of classical orthogonal polynomials on non-uniform lattice. Then, for \(n\ge 1\), \(P_n(x)\) and \(P_{n-1}(x)\) satisfy
Furthermore, if \((Q_n)_{n\in {\mathbb {N}}}\) is a sequence of polynomials such that \(Q_0(x)=P_0(x)\) and, for \(n\ge 1\), \(Q_n(x)\) and \(Q_{n-1}(x)\) satisfy (3.6). Then \(Q_n(x)=P_n(x)\), for all \(n\ge 0\).
Proof
Using the fact that the sequence \((P_n)_{{n\ge 0}}\) is a classical orthogonal polynomial sequence, for all non-negative integer n, \(P_{n+1}(x)\) satisfies (3.3), namely:
with
In (1.1), using the relations (2.1), (2.2), (2.3) and (2.4), we obtain:
and
Using (1.1), (3.8) and (3.9) to replace \({\mathcal {D}}^2 P_{n+1}(x)\), \({\mathcal {S}}{\mathcal {D}}P_{n+1}(x)\) and \(P_{n+1}(x)\) in (3.7), we obtain:
For \(n\ge 2\), we replace n by \(n-1\) in (3.10) and obtain:
We replace again n by \(n-1\) in (1.1) and use the resulting relation to replace \(P_{n-2}(x)\) in (3.11) to obtain:
If we multiply (3.10) by \(P_{n-1}(x)\), (3.12) by \(P_{n}(x)\) and add the resulting expression, we get:
This proves the first part of Theorem 3.1.
Now, we prove the second part.
Let \((Q_n)_{n\in {\mathbb {N}}}\) be a sequence of polynomials of a quadratic variable such that \(Q_0(x)=P_0(x)\) and, for \(n\ge 1\), \(Q_n(x)\) and \(Q_{n-1}(x)\) satisfy
Let \(b_n\) be the leading coefficient of \(Q_n(x)\). We shall firstly show by induction that \(k_n=b_n\) for all \(n\ge 0\). We have \(b_0=k_0\) and we assume that \(n\ge 1\) and \(b_{n-1}=k_{n-1}\). If we compare the coefficients of \(x^{2n}\) in (3.13), we find that we must consider two cases according as the degree of \(\phi \) is less than two or equal to two.
-
If the degree of \(\phi \) is less than two then we have
$$\begin{aligned} 2\psi _1b_nk_{n-1}=\psi _1b_nk_{n-1}+{\psi _1\over A_{n-1}}(b_n)^2 \end{aligned}$$and \(b_n\ne 0\) implies that for the quadratic or q-quadratic variable, we have \(b_n=A_{n-1}k_{n-1}=k_n\).
-
If the degree of \(\phi \) is equal to two, then we have
$$\begin{aligned}&\phi _2\left( (n-1)ib_nk_{n-1}+nib_nk_{n-1}\right) +2\psi _1b_nk_{n-1}\\&\quad =(\psi _1+2i\phi _2)b_nk_{n-1}+{1\over A_{n-1}}\left( \psi _1+(2n-3)i\phi _2\right) b_n^2, \end{aligned}$$and the regularity of the corresponding linear functional with respect to the sequence \({(Q_n)_{n\ge 0}}\) implies that \(\psi _1+(2n-3)i\phi _2\ne 0\) and \(b_n\ne 0\) we have \(b_n=A_{n-1}k_{n-1}=k_n\).
We have by assumption \(Q_0(x)=P_0(x)\). Assume further that \(n\ge 1\) and \(Q_{n-1}(x)=P_{n-1}(x)\) but \(Q_{n}(x)\ne P_{n}(x)\).
Then \(Q_n(x)=P_n(x)+g(x)\) where \(g(x)=c(x^r+\cdots ) \), \(c\ne 0\). Since \(Q_n(x)\) and \(P_n(x)\) have the same degree and the same leading coefficient, we must have \(r<n\). From (3.13), we get
Using the fact that \(P_n(x)\) and \(P_{n-1}(x)\) satisfy (3.6) we obtain
We compare the coefficients of \(x^{n+r}\) in (3.14). Let us consider two cases:
-
1.
If the degree of \(\phi \) is less than two, then we get
$$\begin{aligned} 2\psi _1cb_{n-1}=\psi _1cb_{n-1}+2{b_n\over A_{n-1}}c\psi _1 \end{aligned}$$which is equivalent to
$$\begin{aligned} 2\psi _1cb_{n-1}=\psi _1cb_{n-1}+2c\psi _1b_{n-1}. \end{aligned}$$Then, the fact that \(\psi _1 b_{n-1}\ne 0\) implies that this is impossible if \(c\ne 0\).
-
2.
If the degree of \(\phi \) is equal to two, then we get
$$\begin{aligned}&\phi _2\left( ci(n-1)b_{n-1}+ricb_{n-1}\right) +2c\psi _1b_{n-1}\\&\quad =(\psi _1+2i\phi _2)cb_{n-1}+{2\over A_{n-1}}\left( \psi _1+(2n-3)i\phi _2\right) cb_n, \end{aligned}$$which is equivalent to
$$\begin{aligned}&\phi _2\left( ci(n-1)b_{n-1}+ricb_{n-1}\right) +2c\psi _1b_{n-1}\\&\quad =(\psi _1+2i\phi _2)cb_{n-1}+2\left( \psi _1+(2n-3)i\phi _2\right) cb_{n-1}. \end{aligned}$$The regularity of the corresponding linear functional with respect to the sequence \((P_n)\) implies that \(\psi _1+(3n-3-r)i\phi _2\ne 0\) and the previous equation is impossible if \(c\ne 0\).
The proof is therefore completed. \(\square \)
The following corollaries give explicit coefficients for the non-linear characterization of the Meixner–Pollaczek and the Continuous Dual Hahn polynomials.
Corollary 3.2
The Meixner–Pollaczek polynomials are characterized by the following non-linear difference equation
Corollary 3.3
The Continuous Hahn polynomials are characterized by the following non-linear difference equation
where \(D_n\) depends on n, a, b, c and d.
4 Non-linear Characterization for Wilson and Continuous Dual Hahn Polynomials
The Wilson polynomials \(W_n(x^2;a,b,c,d)\) and Continuous Dual Hahn polynomials \(S_n(x^2;a,b,c)\), respectively, have the hypergeometric representation (see [18]):
They are known to satisfy the second-order divided-difference equation (see [27])
and these two families satisfy the three-term recurrence relation
Theorem 4.1
(Non-linear characterization) Let \((P_n)_{n\in {\mathbb {N}}}\) be a sequence of classical orthogonal polynomials on a non-uniform lattice. Then, for \(n\ge 1\), \(P_n(x^2)\) and \(P_{n-1}(x^2)\) satisfy
Furthermore, if \((Q_n(x^2))_{n\in {\mathbb {N}}}\) is a sequence of polynomials such that \(Q_0(x^2)=P_0(x^2)\) and, for \(n\ge 1\), \(Q_n(x)\) and \(Q_{n-1}(x)\) satisfy (4.5). Then \(Q_n(x^2)=P_n(x^2)\), for all \(n\ge 0\).
Proof
For all integers n, \(P_{n+1}(x^2)\) satisfies (4.3), namely:
with
From (4.4), using the relations (2.5), (2.6), (2.7) and (2.8), we obtain:
and
We use (4.4), (4.7) and (4.8) to replace \({\textbf{D}}^2 P_{n+1}(x^2)\), \({\textbf{S}}{\textbf{D}}P_{n+1}(x^2)\) and \(P_{n+1}(x^2)\) in (4.6); we obtain:
For \(n\ge 2\), we replace n by \(n-1\) in (4.9) and we obtain:
We replace again n by \(n-1\) in (4.4) and we use the resulting relation to replace \(P_{n-2}(x^2)\) in (4.10) to obtain:
If we multiply (4.9) by \(P_{n-1}(x^2)\), (4.11) by \(P_{n}(x^2)\) and add the resulting expressions, we obtain:
This proves the first part of Theorem 4.1.
Now, we prove the second part.
Let \((Q_n)_{n\in {\mathbb {N}}}\) be a sequence of polynomials of a quadratic variable such that \(Q_0(x^2)=P_0(x^2)\) and, for \(n\ge 1\), \(Q_n(x^2)\) and \(Q_{n-1}(x^2)\) satisfy (4.5). Let \(b_n\) be the leading coefficient of \(Q_n(x^2)\). We shall first show by induction that \(k_n=b_n\) for all \(n\ge 0\). We have \(b_0=k_0\) and we assume that \(n\ge 1\) and \(b_{n-1}=k_{n-1}\). If we compare the coefficients of \(x^{4n}\) in (4.5), we find that we must consider two cases whether the degree of \(\phi \) is less than two or equal to two.
-
If the degree of \(\phi \) is less than two then, we have
$$\begin{aligned} 2\psi _1b_nk_{n-1}=\psi _1b_nk_{n-1}+{\psi _1\over A_{n-1}}(b_n)^2 \end{aligned}$$and \(b_n\ne 0\) implies that for the quadratic or q-quadratic variable, we have \(b_n=A_{n-1}k_{n-1}=k_n\).
-
If the degree of \(\phi \) is equal to two then, we have
$$\begin{aligned}&\phi _2\left( (n-1)b_nk_{n-1}+nb_nk_{n-1}\right) +2\psi _1b_nk_{n-1} \\&\quad =(\psi _1+2\phi _2)b_nk_{n-1}+{1\over A_{n-1}}\left( \psi _1+(2n-3)\phi _2\right) b_n^2, \end{aligned}$$and the regularity of the corresponding linear functional with respect to the sequence \((Q_n)_{{n\ge 0}}\) implies that \(\psi _1+(2n-3)\phi _2\ne 0\) and \(b_n\ne 0\) we have \(b_n=A_{n-1}k_{n-1}=k_n\).
We have by assumption \(Q_0(x^2)=P_0(x^2)\). Assume further that \(n\ge 1\) and \(Q_{n-1}(x^2)=P_{n-1}(x^2)\) but \(Q_{n}(x^2)\ne P_{n}(x^2)\).
Then \(Q_n(x^2)=P_n(x^2)+g(x^2)\) where \(g(x^2)=c(x^{2r}+\cdots ) \), \(c\ne 0\). Since \(Q_n(x^2)\) and \(P_n(x^2)\) have the same degree and the same leading coefficient, we must have \(r<n\). From (4.5), we get
Using the fact that \(P_n(x)\) and \(P_{n-1}(x)\) satisfy (4.5), we obtain
We compare the coefficients of \(x^{2n+2r}\) in (4.12) and consider two cases:
-
1.
If the degree of \(\phi \) is less than two, then
$$\begin{aligned} 2\psi _1cb_{n-1}=\psi _1cb_{n-1}+2{b_n\over A_{n-1}}c\psi _1 \end{aligned}$$which is equivalent to
$$\begin{aligned} 2\psi _1cb_{n-1}=\psi _1cb_{n-1}+2c\psi _1b_{n-1}. \end{aligned}$$Then, the fact that \(\psi _1 b_{n-1}\ne 0\) implies that this is impossible if \(c\ne 0\).
-
2.
If the degree of \(\phi \) is equal to two, then
$$\begin{aligned}&\phi _2\left( c(n-1)b_{n-1}+rcb_{n-1}\right) +2c\psi _1b_{n-1}\\&\quad =(\psi _1+2\phi _2)cb_{n-1}+{2\over A_{n-1}}\left( \psi _1+(2n-3)\phi _2\right) cb_n \end{aligned}$$which is equivalent to
$$\begin{aligned}&\phi _2\left( c(n-1)b_{n-1}+rcb_{n-1}\right) +2c\psi _1b_{n-1}\\&\quad =(\psi _1+2\phi _2)cb_{n-1}+2\left( \psi _1+(2n-3)\phi _2\right) cb_{n-1}. \end{aligned}$$The regularity of the corresponding linear functional with respect to the sequence \((P_n)_{{n\ge 0}}\) implies that \(\psi _1+(3n-3-r)\phi _2\ne 0\) and the previous equality is impossible if \(c\ne 0\).
\(\square \)
Corollary 4.2
The Wilson polynomials are characterized by the following non-linear difference equation
where \(D_n\) depends on n, a, b, c and d.
Corollary 4.3
The Continuous Dual Hahn polynomials are characterized by the following non-linear difference equation
where \(D_n\) depends on n, a, b, c and d.
5 Non-linear Characterization for Orthogonal Polynomials on q-Quadratic Lattices
A family \(p_n(x)\) of polynomials of degree n is a family of classical q-quadratic orthogonal polynomials (also known as orthogonal polynomials on non-uniform lattices) if it is the solution of a divided-difference equation of the type (see [8, 9])
where \(\phi \) is a polynomial of maximal degree two and \(\psi \) is a polynomial of exact degree one, \(\lambda _n\) is a constant depending on the integer n and the leading coefficients \(\phi _2\) and \(\psi _1\) of \(\phi \) and \(\psi \):
and x(s) is a non-uniform lattice defined by
and the sequences \((\alpha _n)\) and \((\gamma _n)\) are given explicitly by :
5.1 General Theorem
In this section, we state and prove a non-linear characterization result for classical orthogonal polynomials on non-uniform lattices. The result is stated in the following theorem.
Theorem 5.1
Let \((P_n)_{{n\ge 0}}\) be a sequence of classical orthogonal polynomials on a non-uniform lattice. Then, for \(n\ge 1\), \(P_n(x(s))\) and \(P_{n-1}(x(s))\) satisfy
where
Furthermore, if \((Q_n)_{n\in {\mathbb {N}}}\) is a sequence of polynomials a on non-uniform lattice such that \(Q_0(x)=P_0(x)\) and, for \(n\ge 1\), \(Q_n(x(s))\) and \(Q_{n-1}(x(s))\) satisfy (5.3). Then \(Q_n(x(s))=P_n(x(s))\), for all \(n\ge 0\).
Proof
Using the fact that \({(P_n(x(s))_{n\ge 0}}\) is a classical q-orthogonal polynomial sequence on non-uniform lattice, substituting n by \(n+1\) in (5.1) we obtain
In (1.1), using the product rules given in [12, page 407], in [11, pages 741-742] or in [10, page 4], we obtain:
and
Using (1.1), (5.6) and (5.7) to replace \({\mathbb {D}}_{x}^2 P_{n+1}(x(s))\) , \({\mathbb {S}}_{x}{\mathbb {D}}_{x}P_{n+1}(x(s))\) and \( P_{n+1}(x(s))\) in (5.5), we obtain:
which is equivalent to
where \(D_{n}\), \(E_n\) and \(G_n\) are defined in (5.4). For \(n\ge 2\), we replace n by \(n-1\) in (5.8) and we obtain:
We also replace n by \(n-1\) in (1.1) and use the resulting relation to replace \(P_{n-2}(x(s))\) in (5.9) to obtain:
If we multiply (5.8) by \(P_{n-1}(x(s))\) and (5.10) by \(P_n(x(s))\) and add the resulting expressions, we obtain:
This proves the first part of Theorem 5.1.
Now, we prove the second part.
Let \((Q_n(x(s)))_{n\in {\mathbb {N}}}\) be a sequence of polynomials of a q-quadratic variable such that \(Q_0(x(s))=P_0(x(s))\) and, for \(n\ge 1\), \(Q_n(x(s))\) and \(Q_{n-1}(x(s))\) satisfy
Let \(a_n\) be the leading coefficient of \(Q_n(x(s))\). We shall firstly show by induction that \(k_n=a_n\) for all \(n\ge 0\). We have \(a_0=k_0\) and we assume that \(n\ge 1\) and \(a_{n-1}=k_{n-1}\). If we compare the coefficients of \(F_{2n}(x(s))\) in (5.11), we find that we must consider two cases whether the degree of \(\phi \) is less than two or equal to two.
-
If the degree of \(\phi (x(s))\) is less than two then, we have
$$\begin{aligned} \psi _1((\alpha _{n-1})^2+(\alpha _n)^2)a_nk_{n-1}=(D_n+D_{n-1}-A_{n-1}G_{n-1}) a_nk_{n-1}+G_{n-1}(a_n)^2 \end{aligned}$$and \(a_n\ne 0\) implies that for the q-quadratic variable, we have \(a_n=A_{n-1}k_{n-1}=k_n\).
-
If the degree of \(\phi (x(s))\) is equal to two then we get
$$\begin{aligned}&\psi _1((\alpha _{n-1})^2+(\alpha _n)^2)a_nk_{n-1}+\phi _2(\alpha _{n-1}\gamma _{n-1} +\alpha _{n}\gamma _{n})a_nk_{n-1}\\&\quad =(D_n+D_{n-1}-A_{n-1}G_{n-1})a_nk_{n-1}+G_{n-1}(a_n)^2 \end{aligned}$$and \(a_n\ne 0\) implies that for the quadratic case or the q-quadratic case, we have \(a_n=A_{n-1}k_{n-1}=k_n\).
We have by assumption \(Q_0(x(s))=P_0(x(s))\). Assume further that \(n\ge 1\) and we \(Q_{n-1}(x(s))=P_{n-1}(x(s))\) but \(Q_{n}(x(s))\ne P_{n}(x(s))\).
Then \(Q_n(x(s))=P_n(x(s))+g(x(s))\), where \(g(x(s))=c(F_r(x(s))+\cdots )\), \(c\ne 0\). Since \(Q_n(x(s))\) and \(P_n(x(s))\) have the same degree and the same leading coefficient, we must have \(r<n\). From (5.11), we get
Using the fact that \(P_n(x(s))\) and \(P_{n-1}(x(s))\) satisfy (5.3), we obtain
We compare the coefficients of \(F_{n+r}(x(s))\) in (5.12). Two cases arise:
-
1.
If the degree of \(\phi (x(s))\) is less than two, then we get
$$\begin{aligned} \psi _1\left( (\alpha _{n-1})^2+(\alpha _r)^2\right) c k_{n-1}=(D_n+D_{n-1}-G_{n-1}A_{n-1})c k_{n-1}+2c k_{n}G_{n-1}, \end{aligned}$$which is equivalent to
$$\begin{aligned} \psi _1\left( (\alpha _{n-1})^2+(\alpha _r)^2\right) c k_{n-1}=(D_n+D_{n-1}-G_{n-1}A_{n-1})c k_{n-1}+2c k_{n-1}G_{n-1}A_{n-1}. \end{aligned}$$Then, for the q-quadratic variable, this is impossible if \(c\ne 0\).
-
2.
If the degree of \(\phi (x(s))\) is equal to two, then we get
$$\begin{aligned}&\psi _1\left( (\alpha _{n-1})^2+(\alpha _r)^2\right) c k_{n-1}+\phi _2\left( \alpha _{n-1}\gamma _{n-1}+\alpha _r\gamma _{r}\right) c k_{n-1}\\&\quad =(D_n+D_{n-1}-G_{n-1}A_{n-1})c k_{n-1}+2c k_{n}G_{n-1}, \end{aligned}$$which is equivalent to
$$\begin{aligned}&\psi _1\left( (\alpha _{n-1})^2+(\alpha _r)^2\right) c k_{n-1}+\phi _2\left( \alpha _{n-1}\gamma _{n-1}+\alpha _r\gamma _{r}\right) c k_{n-1}\\&\quad =(D_n+D_{n-1}-G_{n-1}A_{n-1})c k_{n-1}+2c k_{n-1}G_{n-1}A_{n-1}. \end{aligned}$$Again this is impossible if \(c\ne 0\).
\(\square \)
5.2 Special Cases
We can specialize the above result to the various classical orthogonal polynomials on non-uniform lattice, namely Askey–Wilson, q-Racah, Continuous dual q-Hahn, Continuous q-Hahn, Dual q-Hahn, Al-Salam Chihara, q-Meixner–Pollaczek, Continuous q-Jacobi, Dual q-Krawtchouk, Continuous big q-Hermite, Continuous q-Laguerre and Continuous q-Hermite polynomials. Note that the results for the Askey–Wilson and the q-Racah polynomials would be enough since the other families can be obtained by some limit transitions. But here, we would like to provide a complete database for all these polynomials orthogonal on a q-quadratic lattices.
5.2.1 Askey–Wilson Polynomials
The Askey–Wilson polynomials have the q-hypergeometric representation [18, P. 415]
They satisfy the divided-difference equation (5.1) with
The monic Askey–Wilson polynomials are characterized by the following non-linear recurrence relation
where
with
5.2.2 q-Racah Polynomials
The q-Racah polynomials have the q-hypergeometric representation [18, P. 422]
where
and
with N a non-negative integer. They satisfy (5.1) with
The monic q-Racah polynomials are characterized by the following relation
where
with
5.2.3 Continuous Dual q-Hahn Polynomials
The Continuous Dual q-Hahn polynomials have the q-hypergeometric representation [18, P. 429]
They satisfy (5.1) with
The monic continuous dual q-Hahn polynomials are characterized by the following non-linear recurrence relation
where
5.2.4 Continuous q-Hahn Polynomials
The Continuous q-Hahn polynomials have the q-hypergeometric representation [18, P. 415] or [8, P. 75]
here \(\displaystyle {x=\cos (\theta +\varphi )}\). They satisfy the divided-difference equation (5.1) with
where \(t=e^{i\varphi }\).
The monic Continuous q-Hahn polynomials are characterized by the following non-linear recurrence relation
where
with \(t=e^{i\varphi }\),
5.2.5 Dual q-Hahn Polynomials
The Dual q-Hahn polynomials have the q-hypergeometric representation ([18, P. 450] or [8, P. 76]
where \(x(s)=q^{-s}+\gamma \delta q^{s+1}\) and N a non-negative integer. They satisfy (5.1) with
The monic Dual q-Hahn polynomials are characterized by the following non-linear recurrence relation
where
5.2.6 Al-Salam–Chihara Polynomials
The Al-Salam–Chihara polynomials have the q-hypergeometric representation [18, P. 455] or [8, P. 77]
They satisfy the divided-difference equation (5.1) with
The monic Al-Salam–Chihara polynomials are characterized by the following non-linear recurrence relation
where
5.2.7 q-Meixner–Pollaczek Polynomials
The q-Meixner–Pollaczek polynomials [18, P. 460] or [8, P. 78]
They satisfy (5.1) with
The monic q-Meixner–Pollaczek polynomials are characterized by the following non-linear recurrence relation
5.2.8 Continuous q-Jacobi Polynomials
The Continuous q-Jacobi polynomials have the q-hypergeometric representation [18, P. 463] or [8, P. 78]
They satisfy the divided-difference equation (5.1) with
with \(p=q^{2}\).
The monic Continuous q-Jacobi polynomials are characterized by the following non-linear recurrence relation
where
with
5.2.9 Dual q-Krawtchouk Polynomials
The Dual q-Krawtchouk polynomials [18, P. 505] or [8, P. 80]
where \(x(s)=q^{-s}+cq^{s-N}\). They satisfy (5.1) with
The monic Dual q-Krawtchouk polynomials are characterized by the following non-linear recurrence relation
5.2.10 Continuous Big q-Hermite Polynomials
The continuous big q-Hermite polynomials [18, P. 509]
They satisfy (5.1) with
The Continuous big q-Hermite polynomials are characterized by the following non-linear recurrence relation
5.2.11 Continuous q-Laguerre Polynomials
The Continuous q-Laguerre polynomials have the q-hypergeometric representation [18, P. 514] or [8, P. 81]
They satisfy the divided-difference equation (5.1) with
with \(p=q^{2}\).
The monic Continuous q-Laguerre polynomials are characterized by the following non-linear recurrence relation
where
with \(\xi _n=-(q-1)^{2}(p^{2}-1)q^{n+\frac{7}{2}}\),
with \(\nu _n=2(p^{2}-1)q^{\frac{11}{4}}\),
where \(\tau _n=-(q-1)^{2}(p^{2}-1)q^{n+\frac{3}{2}}\),
5.2.12 Continuous q-Hermite Polynomials
The Continuous q-Hermite polynomials have the q-hypergeometric representation [18, P. 540] or [8, P. 82]
They satisfy the divided-difference equation (5.1) with
The monic Continuous q-Hermite polynomials are characterized by the following non-linear recurrence relation
References
Al-Salam, W.A.: Characterization theorems for orthogonal polynomials. In: Nevai, P. (ed.) Orthogonal Polynomials: Theory and Practice, NATO ASI Series C: Math. and Phys. Sciences, vol. 294. Kluwer, Dordrecht, pp. 1–24 (1990)
Al-Salam, W.A.: On the Bessel polynomials. Boll. Un. Mat. Ital. (3) 12, 227–229 (1957)
Al-Salam, W.A.: The Bessel polynomials. Duke Math. J. 24, 529–545 (1957)
Al-Salam, W.A., Chihara, T.S.: Another characterization of the classical orthogonal polynomials. SIAM J. Math. Anal. 3, 65–70 (1972)
Atakishiyev, N.M., Rahman, M., Suslov, S.K.: On classical orthogonal polynomials. Constr. Approx. 11, 181–226 (1995)
Brezinski, C.: Padé-Type Approximation and General Orthogonal Polynomials, International Series of Numerical Mathematics, vol. 50. Birkhäuser, Basel (1980)
Chihara, T.S.: An Introduction to Orthogonal Polynomials. Gordon and Breach, New York (1978)
Foupouagnigni, M.: On difference and differential equations for modifications of classical orthogonal polynomials. Habilitation Thesis, Fachbereich Mathematik der Universität Kassel (2006)
Foupouagnigni, M.: On difference equations for orthogonal polynomials on non-uniform lattices. J. Differ. Equ. Appl. 14, 127–174 (2008)
Foupouagnigni, M., Mboutngam, S.: On the polynomial solution of divided-difference equations of hypergeometric type on nonuniform lattices. Axioms, 8 (2019)
Foupouagnigni, M., Kenfack-Nangho, M., Mboutngam, S.: Characterization theorem of classical orthogonal polynomials on non-uniform lattices: the functional approach. Integr. Transf. Spec. Funct. 22(10), 739–758 (2011)
Foupouagnigni, M., Koepf, W., Kenfack-Nangho, M., Mboutngam, S.: On solutions of holonomic divided-difference equations on non-uniform lattices. Axioms 3, 404–434 (2013)
García, A.G., Marcellán, F., Salto, L.: A distributional study of discrete classical orthogonal polynomials. J. Comput. Appl. Math. 5(7), 147–162 (1995)
Gautschi, W.: Orthogonal Polynomials: Computation and Approximation. Oxford University Press, New York (2004)
Gautschi, W.: Orthogonal polynomials, quadrature, and approximation: computational methods and software (in Matlab). In: Orthogonal Polynomials and Special Functions. Lecture Notes in Mathematics, vol. 1883, pp. 1–77. Springer, Berlin (2006)
Kenfack-Nangho, M.: On Laguerre-Hahn orthogonal polynomials on non-uniform lattices. PhD Thesis, Université de Yaoundé I (2014)
Koekoek, R., Swarttouw, R.: The Askey-scheme of hypergeometric orthogonal polynomials and its \(q\)-analogue, Report no. 98-17 (1998). Faculty of Information technology and Systems, Delft University of Technology
Koekoek, R., Lesky, P.A., Swarttouw, R.F.: Hypergeometric Orthogonal Polynomials and their \(q\)-Analogues. Springer Monographs in Mathematics. Springer, Berlin (2010)
Koepf, W., Schmersau, D.: On a structure formula for classical q-orthogonal polynomials. J. Comput. Appl. Math. 136, 99–107 (2001)
Krall, H.L., Frink, O.: A new class of orthogonal polynomials: the Bessel polynomials. Trans. Am. Math. Soc. 65, 100–115 (1949)
Kwon, K.H., Lee, D.W., Park, S.B., Yoo, B.H.: Discrete classical orthogonal polynomials. J. Differ. Equ. Appl. 4, 145–162 (1998)
Kwon, K.H., Lee, D.W., Park, S.B., Yoo, B.H.: Hahn class orthogonal polynomials. Kyungpook Math. J. 38, 259–281 (1998)
Marcellán, F., Branquinho, A.P., Petronilho, J.: Classical orthogonal polynomials: a functional approach. Acta Appl. Math. 34(3), 283–303 (1994)
Mboutngam, S.: On semi-classical orthogonal polynomials on non-uniform lattices. PhD Thesis, Université de Yaoundé I (2014)
McCarthy, P.J.: Characterizations of classical orthogonal polynomials. Port. Math. 20, 47–52 (1961)
Njionou Sadjang, P.: Moments of classical orthogonal polynomials. PhD thesis. Universität Kassel (2013). http://nbn-resolving.de/urn:nbn:de:hebis:34-2013102244291
Njionou Sadjang, P., Koepf, W., Foupouagnigni, M.: On structure formulas for Wilson polynomials. Integr. Transf. Spec. Funct. 26(12), 1000–1014 (2015)
Njionou Sadjang, P., Koepf, W., Foupouagnigni, M.: On moments of classical orthogonal polynomials. J. Math. Anal. Appl. 424, 122–151 (2015)
Suslov, S.K.: The theory of difference analogues of special functions of hypergeometric type. Russ. Math. Surv. 44, 227–278 (1989)
Tcheutia, D.D., Njionou Sadjang, P., Koepf, W., Foupouagnigni, M.: Divided-difference equation, inversion, connection, multiplication and linearization formulae of the Continuous Hahn and the Meixner-Pollaczek polynomials. Ramanujan J. 45, 33–56 (2018)
Trefethen, L.N.: Approximation Theory and Approximation Practice, Extended Society for Industrial and Applied Mathematics (SIAM), Philadelphia (2020)
Tricomi, F.G.: Vorlesungen über Orthogonalreihen, Die Grundlehren der mathematischen Wissenschaften in Einzeldarstellungen mit besonderer Berücksichtigung der Anwendungsgebiete, Band LXXVI. Springer, Berlin (1955)
Acknowledgements
The authors are grateful to some anonymous reviewers whose valuable comments and remarks have helped to improve our manuscript considerably.
Funding
Open Access funding enabled and organized by Projekt DEAL. M. Foupouagnigni and W. Koepf would like to thank the Alexander von Humboldt Foundation for their steady support.
Author information
Authors and Affiliations
Contributions
Njionou Sadjang and Mboutngam have prepared the introduction and proved the characterization Theorems 3.1, 4.1 and 5.1. Sawalda used these characterization theorems to compute the explicit coefficients in Corollaries 3.2, 3.3, 4.2, 4.3 and the coefficients appearing in Sect. 5. Foupouagnigni and Koepf wrote the Maple package for the computations of all the coefficients appearing in the article. Koepf did the final verification.
Corresponding author
Ethics declarations
Conflict of interest
None of the authors has competing interests of a financial or personal nature.
Ethical approval
Not applicable.
Availability of data and materials
Not applicable.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Njionou Sadjang, P., Kalda Sawalda, D., Mboutngam, S. et al. On Non-linear Characterizations of Classical Orthogonal Polynomials. Mediterr. J. Math. 20, 10 (2023). https://doi.org/10.1007/s00009-022-02207-y
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00009-022-02207-y
Keywords
- Classical orthogonal polynomials on non-uniform lattices
- difference equation
- divided-difference equation
- three-term recurrence relation