Exceptional Bannai-Ito polynomials (Q1717628)

\textit{E. Bannai} and \textit{T. Ito} originally introduced the Bannai-Ito polynomials in their work [Algebraic Combinatorics. I: Association schemes. Menlo Park, California: The Benjamin/Cummings Publishing Company (1984; Zbl 0555.05019)]. The second author in one of his recently co-authored papers [\textit{S. Tsujimoto} et. al., Dunkl shift operators and Bannai-Ito polynomials, Adv. Math. 229, No. 4, 2123--2158 (2012; Zbl 1248.33022)] classified them as a new kind of classical orthogonal polynomial (COP) as ``the eigenfunctions of the difference operator \(H = \alpha (x)({R -I}) + \beta (x)({TR -I})\) where, \(H\) is a Dunkl shift operator, \(R\) is the reflection operator, \(T\) is the forward shift operator, and \(I\) is the identity operator acting as \(R\left[ {f(x)} \right] = f({-x}),T\left[ {f(x)} \right] = f({x + 1}),I\left[ {f(x)} \right] = f(x)\).'' In this paper the authors give the exceptional orthogonal polynomial (XOP) version of the Bannai-Ito polynomials. By the application of a ``generalized Darboux transformation (DT) on the Dunkl shift operator \(H\)'' and utilizing the intertwining relations \({{\mathcal {F}}_\phi} \circ H = {H^{(1)}} \circ {{\mathcal {F}}_\phi},H \circ {B_\phi} = {B_\phi} \circ {H^{(1)}}\) in which the Dunkl shift operator \({{\mathcal{F}}_\phi}\) satisfies \({{\mathcal {F}}_\phi}\left[ {\phi (x)} \right] = 0\) for the eigenfunction \(\phi (x)\) of \(H\) and by considering the eigenvalue problem \[ H\left[ {\phi (x)} \right] = \alpha (x)({R -I})\left[\phi (x)\right] + \beta (x)({TR -I})\left[ {\phi (x)} \right] = \mu \phi (x) \] in which \(\alpha (x)\) and \(\beta (x)\) are functions of \(x\) and by defining \(\chi (x) = ({I -R})\phi (x) = \phi (x) -\phi ({-x})\), \(\tilde \chi (x) = ({I + TR})\left[ {\beta ({-x -1})\phi (x)} \right] = \beta ({-x -1})\phi (x) + \beta (x)\phi ({-x -1})\) and \({{\mathcal {F}}_\phi} = \frac{1}{{\chi (x)}}({R -I}) + \frac{1}{{\tilde \chi (x)}}({TR + I})\beta ({-x -1})\) the authors deduce: Proposition 2.2. The Dunkl shift operator \({H^{(1)}}\) which satisfies \({{\mathcal {F}}_\phi} \circ H = {H^{(1)}} \circ {{\mathcal {F}}_\phi}\) with \({{\mathcal {F}}_\phi}\) given above is \[ {H^{(1)}} = \frac{{\tilde \chi ({-x})}} {{\chi (x)}}({R -I}) + \frac{{\alpha ({-x -1})\beta (x)\chi ({-x -1})}} {{\tilde \chi (x)}}({TR -I}) + \frac{{\tilde \chi ({-x}) -\tilde \chi (x)}} {{\chi (x)}}. \] For real numbers \(r_1,r_2,\rho_1,\rho_2\) and the choice \(\alpha (x) = \frac{{({x -{\rho_1}})({x -{\rho_2}})}}{{-2x}},\beta (x) = \frac{{({x -{r_1} + 1/2})({x -{r_2} + 1/2})}}{{2x + 1}}\), ``the Dunkl shift operator \(H\) is called the Bannai-Ito operator'' and the 1-step Darboux transformed Dunkl shift operator \({H^{(1)}}\) is ``called the 1-step exceptional Bannai-Ito operator.'' Next by considering the generalized Darboux transformation of the Bannai-Ito operator they give another interesting result concerning the ``eigenfunctions of the \(n\)-step exceptional Bannai-Ito operator \({H^{(n)}}({n \geqslant 2})\)'' as: Theorem 2.7. The eigenfunctions of the \(n\)th step exceptional Bannai-Ito operator \({H^{(n)}}({n \geqslant 2})\) can be expressed as the \(3 \times 3\) determinant \[ \phi_m^{({n + 1})}(x) = \prod_{j = 1}^{n -2} {\frac{{({{\mu_m} -{\mu_j}})}}{{({{\mu_n} -{\mu_j}})}}\left| {\begin{matrix} {{\mu_m}\tilde \chi_m^{(1)}(x)} & {\tilde \chi_m^{(1)}(x)} & {\tilde \chi_m^{(1)}({-x})} \\ {{\mu_n}\tilde \chi_n^{(1)}(x)} & {\tilde \chi_n^{(1)}(x)} & {\tilde \chi_n^{(1)}({-x})} \\ {{\mu_{n -1}}\tilde \chi_{n -1}^{(1)}(x)} & {\tilde \chi_{n -1}^{(1)}(x)} & {\tilde \chi_{n -1}^{(1)}({-x})} \\ \end{matrix}} \right|} \] \[ \cdot {\left[ {({{\mu_n} -{\mu_{n -1}}})\tilde \chi_n^{(1)}(x)\left| {\begin{matrix} {\tilde \chi_n^{(1)}(x)} & {\tilde \chi_n^{(1)}({-x})} \\ {\tilde \chi_{n -1}^{(1)}(x)} & {\tilde \chi_{n -1}^{(1)}({-x})} \\ \end{matrix}} \right|} \right]^{-1}} \] where \(\mu_n\) is the eigenvalue of \(\phi_n^{(1)}(x):H\left[ {\phi_n^{(1)}(x)} \right] = {\mu_n}\phi_n^{(1)}(x)\), and \(\tilde \chi_n^{(1)}(x) = \beta ({-x -1})\phi_n^{(1)}(x) + \beta (x)\phi_n^{(1)}({-x -1})\). In Section 3 of the paper the authors study the quasi-polynomial eigenfunctions of the type \(H\left[ {\xi (x)p(x)} \right] = \lambda \xi (x)p(x)\) in which \(\xi (x)\) is the gauge factor, ``\(p(x)\) is a polynomial in \(x\) and \(\lambda\) is the corresponding eigenvalue''. In Theorem 3.3 it is proven that the Bannai-Ito operator has 8 sequences of quasi-polynomial eigenfunctions corresponding to the 8 gauge factors enlisted in that theorem. In the fourth section of the paper the authors construct the exceptional Bannai-Ito (XBI) polynomials. If \({\mathbf{D}} = \left\{ {1,2, \ldots ,8} \right\} \times {{\mathbb{Z}}_{\geqslant 0}}\), \({\mathbf{d}} = ({d,m}) \in {\mathbf{D}}\) and for the quasi-polynomial eigenfunction \({\phi_{\mathbf{d}}}(x) = {\xi_d}(x)p_m^{(d)}(x)\), they call the Darboux transformed eigenfunction \(B_{{\mathbf{d}},n}^{(1)}(x) = {{\hat {\mathcal{F}}}_{{\phi_{\mathbf{d}}}}}\left[ {{B_n}(x)} \right]\) ``the exceptional Bannai-Ito polynomial'' and in this connection they derive the following most exemplar result there: Lemma 4.1. Assume that the decoupling coefficient \(r(x)\) is given by \(r(x) = \tilde \chi (x)\chi (x)\frac{{{\eta_d}(x)}}{{x{\xi_d}(x)}}\), where \({{\eta_d}(x)}\) is the polynomial of lowest degree such that \(\frac{{{\xi_d}({-x})}}{{{\xi_d}(x)}} = \frac{{{\eta_d}({-x})}}{{{\eta_d}(x)}}\). Then the Darboux transformed eigenfunction \[ B_{{\mathbf{d}},n}^{(1)}(x) = {{\hat {\mathcal{F}}}_{{\phi_{\mathbf{d}}}}}\left[ {{B_n}(x)} \right] = \frac{{r(x)}} {{\chi (x)\tilde \chi (x)}} \] \[ \times \left| {\begin{matrix} {({I -R})\left[ {{\phi_{\mathbf{d}}}(x)} \right]} & {({I -R})\left[ {{B_n}(x)} \right]} \\ {({I + TR})\left[ {\beta ({-x -1}){\phi_{\mathbf{d}}}(x)} \right]} & {({I + TR})\left[\beta ({-x -1}) {{B_n}(x)} \right]} \\ \end{matrix}} \right| \] is a polynomial (where, \({B_n}(x): = {B_n}({x;{\rho_1},{\rho_2},{r_1},{r_2}})\) are the four parameter Bannai-Ito polynomials introduced in [\textit{E. Bannai} and \textit{T. Ito}, loc. cit.]). These XBI polynomials are also shown to satisfy a discrete orthogonality relation (Theorem 4.11) with respect to a weight function \({{\hat \omega}_1}(x)\) proven in the Theorem 4.9. The reviewer feels that as far as the 1-step exceptional Bannai-Ito polynomials are concerned this pioneering work should serve as a torchbearer for the upcoming work in this and allied studies.