betaFeedback

WorksMathematics

Bilateral Christoffel-Geronimus-Uvarov formulas for mixed-type orthogonality on algebraic supports

Abstract

Christoffel-type determinant formulas construct both normalized polynomial families and their reproducing kernel for mixed-type orthogonality in several variables after independent rational matrix transformations on the two arguments. The pairing may have a rectangular matrix of functionals, with separate degree bounds on each component, and may be supported on an algebraic variety. The reduced matrices are calculated from the original biorthogonal polynomials, rational moments, polynomial division and the masses left free by division; their nonvanishing determinants characterize existence and uniqueness. For measure pairings the rational data are expressed by Cauchy and multivariate Fantappie transforms. Exact dimension counts incorporate the four matrix factors and the relations of the support. For normal degree-ordered original families and fixed factors, when the component degrees on each side differ by a bounded amount and the normalization dimension grows at most like one degree layer, the resulting orders are bounded on curves and grow at most linearly on surfaces. For pairings allowing scalar factors to move between arguments, called balanced pairings, an exact obstruction determines whether a bilateral mass decomposes into unilateral masses when the denominators act injectively. Residue duality provides coordinates for division and mass quotients on integral curves, including singular points, and on the finite surface quotients specified here. Explicit singular-curve families retain nondegenerate residue pairings as denominator zeros coalesce. Positive congruences admit a further reduction by orthogonal projection, while balanced pairings give simultaneous intertwining identities for all coordinate multiplication matrices.

Dedicated to Professor Francisco Marcellán on the occasion of his 75th birthday

Introduction

The objective is to calculate orthogonal polynomials after independent rational matrix transformations on the two arguments of a pairing. The calculation must preserve the prescribed degrees of each polynomial component and the equations defining its algebraic support. The results give the transformed polynomials, their reproducing kernel, the complete freedom of added masses, and the sizes of the determinants required to compute them.

In mixed-type multiple orthogonality, a polynomial row is required to be orthogonal to prescribed polynomial columns. A typical pairing is

βμ(f,g):=∫Zf(x) dμ(x) g(x),\beta_{\mu}(f,g) := \int_{Z} f(x)\,\mathrm{d}\mu(x)\,g(x),

where ZZ is the specified integration support, μ\mu is a qq-by-pp matrix of measures, ff is a polynomial row with qq components, and gg is a polynomial column with pp components. The product under the integral is a scalar. Each component of ff and gg is assigned its own degree bound. To construct a row polynomial, the column space supplies the orthogonality tests; to construct a column polynomial, the row space supplies them. In either case, the unknown polynomial is required to pair to zero with every element of the opposite test space, and its remaining coefficients are fixed by a prescribed normalization.

More generally, moments can be prescribed by linear functionals instead of measures. The algebraic formulation by a matrix of linear forms belongs to the vector orthogonality framework of Sorokin and Van Iseghem [33]; the mixed-type orthogonality conditions extend types I and II as in Daems and Kuijlaars [34]. This notation is bilinear, hence linear in both arguments. When a sesquilinear pairing is used, the second argument is conjugated explicitly.

Polynomial coefficients and matrix entries may be complex even when the integration variables are real. In the real-support examples, the integration domain is specified separately from the complex coordinate algebra used for the polynomial calculations. Added evaluation or derivative functionals may instead be supported at complex points; their support is stated when they are introduced.

The transformations of Christoffel and Geronimus multiply or divide the measure by polynomial factors, and Uvarov modifications add masses [29, 30, 31]. The transformations considered here use independent rational matrix factors on both sides of the pairing. The order of the matrix factors matters: the coordinate algebra is commutative, but these matrices generally do not commute. The row and column may have different component counts and component degree bounds. The variables satisfy the equations of an algebraic support of any dimension. Reducible supports, algebraic multiplicities and pairings depending on two independent support points are included.

We construct the perturbed mixed-type multiple orthogonal polynomials and their reproducing kernel, a finite sum of polynomial column–row products in two sets of variables. Pairing in one set of variables reproduces polynomials from the chosen space in the other. Division leaves homogeneous masses, terms vanishing after both denominator actions. Our Christoffel–Geronimus–Uvarov problem is to determine all these masses and compute the transformed polynomials and kernels while retaining the prescribed component degree bounds and the defining relations of the support.

More precisely, the data are two qq-by-qq polynomial matrices L0L_0, L1L_1 and two pp-by-pp polynomial matrices R0R_0, R1R_1. The new pairing β^\widehat{\beta} is required to satisfy

β^(fL1,R1g)=βμ(fL0,R0g)\widehat{\beta}(fL_1,R_1g)=\beta_{\mu}(fL_0,R_0g)

for every polynomial row ff and column gg.

The identity is formulated using multiplication. Under the injectivity hypotheses stated below, the rational factors are SL:=L1−1L0S_L:=L_1^{-1}L_0 and SR:=R0R1−1S_R:=R_0R_1^{-1}. Evaluating the original pairing on fSL,SRgfS_L,S_Rg gives a particular solution; its rational moments are additional input whenever they are not fixed by the original polynomial moments. Every other solution adds a mass that vanishes on fL1,R1gfL_1,R_1g. Section 2.1 proves these assertions and also treats the consistency question when the matrices are not injective.

The calculation begins with finite polynomial spaces and an original biorthogonal block on which the pairing matrix is invertible. Multiply the target rows and columns by the rational factors and enlarge the original spaces until all these products can be expressed in them. Their coordinate matrices determine linear equations whose solutions are exactly the admissible images. These equations impose both polynomial divisibility and the required component degrees. Pair the added functions with the original polynomials, insert the chosen mass, and impose the normalization. Eliminating the coefficients already determined by the original orthogonality gives the reduced matrix of the construction. The scalar calculation in Subsection 3.1 carries out every step before the general matrix formula is introduced.

The final step is block elimination. Its inputs and admissible subspaces are supplied by the construction: the ordered rational matrix actions, the complete degree and divisibility equations on the algebraic support, and the mass coordinates. In particular, the constraints can be computed by polynomial reduction followed by a nullspace calculation. Values and derivatives, and residues for the finite quotients treated below, give alternative coordinates for the divisibility conditions. The degree or section-space conditions are imposed alongside them. Pairings with the added rational functions play the role of functions of the second kind; for measures, Subsubsection 3.2.3 gives their Cauchy and Fantappiè integral representations.

The algebraic support affects this calculation through its polynomial relations. Two expressions that differ by a defining relation represent the same function, so they must not be counted as independent directions. At singular points or on supports with multiplicities, these relations also restrict which derivative data can occur. We retain them throughout the construction. Subsubsection 3.2.6 calculates all the blocks on a curve, with noncommuting matrix factors on both sides and a mass that cannot be split into unilateral masses, and verifies the resulting polynomial and kernel directly.

The main construction can be summarized as follows. A square pairing is called normal when its matrix in bases of the chosen spaces is invertible. A normalized orthogonal block consists of polynomial vectors orthogonal to the opposite test space and satisfying specified linear normalization conditions.

Main construction. Let A=C[x1,…,xM]/IA=\mathbb{C}[x_1,\ldots,x_M]/I be the coordinate algebra of the support. Assume that the four polynomial matrices L0L_0, L1L_1, R0R_0, R1R_1 have regular determinants, meaning that multiplication by each determinant is injective on AA. Fix a rational extension of the original pairing, a homogeneous mass, and finite polynomial row and column spaces. For a normalized problem, take k>0k>0, a candidate space of dimension r+kr+k, an opposite test space of dimension rr, and a prescribed CkC^k-valued linear normalization. Choose an original normal block and finite rational enlargements as in Section 3.

The original biorthogonal data determine reduced square matrices for the right and left normalized problems. Each problem has a unique solution if and only if its reduced determinant is nonzero. Every entry of the solution is a bordered determinant divided by that reduced determinant. For equal-dimensional target spaces, the same reduced matrix that tests normality also gives their reproducing kernel.

The orders of these matrices are computed from dimensions of polynomial images, complete constraint spaces and the restricted mass. For fixed factors and original degree-ordered blocks normal at the levels considered, take target degree-ordered flags or component bounds of bounded dispersion on each side. If the normalization dimension is of order at most ns−1n^{s-1}, the degree-based construction has order O(ns−1)O(n^{s-1}) as n→∞n \to\infty, where s=dim⁡A≥1s = \dim A \ge1 and nn is the largest target component degree.

The determinant identities and solution formulas are proved in Theorem 3.3, Corollary 3.4 and the kernel construction (24); Theorem 3.6 and Proposition 3.9 give the size statements. A tau determinant, denoted by τ\tau, is the determinant of the reduced system. Bordering means adding one row and one column to that matrix. The prescribed component degree bounds and normalization are imposed directly, so the construction applies without normality assumptions on intermediate transformations. On each finite space, Proposition 3.2 also supplies a separating list of value and derivative conditions for division; their concrete coordinates are computed in the examples and specializations.

The order of the determinant is part of the output. Subsection 3.3 calculates the orders for specified constructions, including all four matrix factors. Theorem 3.6 gives an explicit degree count and a reduced count obtained from two polynomial coefficient matrices. Corollary 3.8 gives the exact stable count on curves, including the row and column component counts, masses and the position within the last degree block. The bounds of Proposition 3.9 explain the change from bounded orders on curves to at most linear growth on surfaces. The comparison with formulas using a fixed range of consecutive polynomial degrees states which determinant representation is being counted. It also explains how matrix division and component bounds contribute information beyond the degree of a matrix determinant.

The mass equation supplies the remaining input to these matrices. Theorem 2.2 describes every solution when numerator and denominator actions are injective: no nonzero polynomial row or column on the support is sent to zero. Proposition 2.1 also decides existence when this fails. Cancelling common rational factors can discard masses. For pairings allowing polynomial factors to move between arguments, Theorem 2.4 decides whether a bilateral mass splits into left and right masses. Shared pieces of the denominator zero sets can prevent splitting. At their intersections, a mass killed by each denominator separately can be added to one summand and subtracted from the other without changing the sum.

The geometric part of the article calculates particular inputs to the reduced matrices. A divisibility condition is a condition on a remainder class. A mass also evaluates such classes, because it vanishes on the image of the denominator action. Finite quotient coordinates therefore provide a common method for both parts, with the quotient chosen for the matrix division or denominator mass being calculated. These quotients can differ: a Christoffel numerator imposes divisibility conditions even when the denominators are identity matrices and there is no homogeneous mass. Section 4 constructs dual coordinates for the relevant quotients using residues. On an integral curve, a geometric Cauchy kernel gives a matrix divided difference that reproduces the remainder. At a singular point the residue is interpreted using the module of dualizing differentials, which retains the relations between the branches. On surfaces the construction concerns finite quotients obtained by imposing a matrix equation and an additional scalar equation. These coordinates supply the linear tests and mass entries of the Christoffel-type determinant formulas. The rational moments and the freely chosen mass parameters remain part of the input; polynomial moments alone need not determine them.

These residue coordinates are then used in an explicit family. The curve family in Section 5 gives the factors, residue functionals and finite Grams explicitly. Its normality is verified on the stated component bounds by exploiting the triangular matrix structure. The residue quotients remain perfectly paired when denominator zeros reach a non-Gorenstein singular point. A constant nonzero minor also proves that the matrix mass keeps its full rank, so no mass direction is lost at the collision. Once this property and invertibility of the finite Grams are established, the inverse Gram formulas give regular limits of the normalized polynomials and kernels.

The same construction also yields the standard specializations. On the line, Theorem 6.1 and Proposition 6.2 express the division conditions through principal parts, complete Jordan chains and forbidden coefficients at infinity. The Cauchy data include shared poles and their confluent limits. On Laurent supports, conjugation changes the second argument and the degree bounds depend on the chosen embedding; the finite construction retains both features. For symmetric congruences, Proposition 7.1 describes the complete mass quotient and classifies positive balanced masses on real curves. The examples distinguish these evaluation masses from positive bivariate pairings that couple distinct points. Theorem 7.2 eliminates the common space by orthogonal projection and gives the normalized polynomials and kernel from two positive matrices of the same explicit order. Only the mass on the overlap increases that order. Under the coprimality and degree conditions of Corollary 7.3, every positive homogeneous mass leaves the order unchanged. Proposition 7.4 constructs a noncommutative matrix family on a nodal curve, with a positive Sobolev mass involving values and derivatives at the node and exact counts for unequal component bounds. This derivative mass belongs to the positive, generally nonbalanced theory. The coordinate identities below apply when both pairings are balanced, as happens for the evaluation-mass specialization.

Once the polynomial bases have been constructed, they determine coordinate multiplication matrices. For balanced pairings on nested normal spaces that together contain every polynomial vector, multiplication by a coordinate has a matrix in the orthogonal polynomial bases. The four factors give connection matrices between the original and transformed bases. Theorem 8.1 proves four-factor intertwining identities simultaneously for all coordinates. The matrix identities preserve the equations defining the support. For the stated positive congruences, orthonormal bases make the multiplication matrices symmetric.

Relation to previous work and organization

The scalar rational transformation problem has an established spectral formulation in [32]. The algebraic theory of vector orthogonality with a matrix of linear forms is developed by Sorokin and Van Iseghem [33]; the mixed-type formulation and its Christoffel–Darboux formula are treated by Daems and Kuijlaars [34]. Matrix rational transformations and bivariate pairings on the line are developed in [1], including noncommutative factors and bilateral specializations; the matrix Christoffel case appears in [2]. Christoffel, Geronimus and Uvarov transformations for mixed-type orthogonality on the step-line are treated in [3, 4, 5]. Kozhan and Vaktnäs [6, Theorems 3.1 and 3.3] give rational perturbation formulas for types I and II on the line with general multi-indices and componentwise scalar factors; their companion paper [44] treats Christoffel transformations and their relation to nearest-neighbour recurrences. Bueno and Marcellán [42] formulate scalar polynomial perturbations through separate factors in the two arguments of a bilinear functional. Their detailed factorization results concern one-sided modifications and the same modification in both arguments. The construction here retains independent matrix actions on both polynomial spaces over the coordinate algebra of the support, with the prescribed component bounds.

For equal component bounds and a square positive pairing on the line, Corollary 7.5 gives the monic matrix orthogonal polynomials, norm matrices and Christoffel–Darboux kernel after a rational congruence of the measure. This connects the construction with the matrix weights of Durán and Grünbaum [26, 27, 28]. Subsection 7.2 recovers their two Hermite matrix weights by polynomial triangular factors, computes the reduced orders and displays a monic formula in scalar Hermite polynomials. Their differential equations impose additional conditions on the weights; the rational construction supplies the polynomial and kernel transformation.

In several variables, Fernández, Pérez, Piñar and Xu [43] give polynomial and reproducing-kernel formulas for the addition of a Dirac mass to a positive definite functional. The formulas of [35] treat finitely many point-mass additions. Christoffel modifications of total degree two and Uvarov modifications are treated in [36]; scalar Geronimus transformations in two variables, their connection formulas and block Jacobi factorizations appear in [37]. Multivariate rational formulas in [7] use scalar polynomial factors. Related multivariate Christoffel constructions appear in [8, 9]. Here the support relations, the full bilateral mass equation and the finite component degree bounds enter one determinant construction. Polynomial classes on singular or nonreduced supports are included, subject to the stated regularity and finite normality assumptions. Thus the extension is the joint construction with independent matrix actions on the two polynomial spaces over a quotient algebra, together with complete mass data and determinant counts for those spaces.

Fasondini, Olver and Xu [10, 11] construct orthogonal bases and computational recurrences on classes of planar algebraic curves, including singular cubics. The present problem starts from an original biorthogonal system and computes the systems produced by independent rational matrix factors on its two arguments. The scalar calculation in Subsection 3.1 and the full mixed-type example in Subsubsection 3.2.6 illustrate the construction of the finite matrices from the original data.

The geometric ingredients have their own established background. Bertola [12] relates scalar section biorthogonality on smooth curves to matrix biorthogonality through a finite projection. His Proposition 3.8 also gives rational left and right changes induced by changes of geometric data. The curve residues below use the associated Cauchy-kernel framework. On smooth complex curves, the geometric Szegő kernel has the classical theta-function description of Fay [46] and the bundle formulation of Ben-Zvi and Biswas [18]. For smooth projective curves over finite fields, Hu and Yu [13 Theorems 6.2 and 6.3] identify the finite residual Casimir tensor with the intrinsic principal part of a normalized Szegő kernel and prove its reproduction property. This is the direct scalar, smooth antecedent of the residue-quotient construction below. Here that construction is carried out on integral curves with singular points, including non-Gorenstein points, and for matrix quotients with the explicit order of multiplication required by the bilateral formulas. Rosenlicht’s residue description of regular differentials, presented in [45 Chapter IV], supplies the dualizing module at those points; see also [39]. The finite quotient construction on surfaces is stated separately with its own hypotheses. The recurrence interpretation is connected with the functional and Darboux transformations of [38] and with the Hessenberg-matrix factorizations for polynomial perturbations of bilinear functionals in [42]. Here the same four connection matrices intertwine every coordinate action and preserve the polynomial relations of the support.

Section 2 develops the functional equation and its mass space. Section 3 proves the determinant formulas and explains their Gram realization. Section 4 provides the curve and surface residue coordinates. Section 5 develops an explicit bilateral family on singular curves and proves regularity of its normalized polynomials. Sections 6 and 7 treat standard, Laurent and positive specializations. Section 8 proves the simultaneous coordinate Darboux identities.

Functional equation and homogeneous masses

The functional equation prescribes the new pairing only on arguments multiplied by the two denominator matrices. The first task is to decide whether these prescribed values are consistent. The second is to describe every value that remains free. These free values are the homogeneous masses. On a curve their description reduces to finitely many polynomial remainders, although a bivariate mass can still involve arbitrary functionals in the opposite argument. The last part of the section decides when a mass can be written as the sum of a left mass and a right mass, each annihilated by just one denominator.

The functional equation and all its masses

The functional equation and all its masses. We encode the support equations in the coordinate algebra A:=C[x1,…,xM]/IA := \mathbb{C}[x_1,\ldots,x_M]/I, where II is a proper ideal (1∉I1 \notin I). Thus two polynomials represent the same element of AA when their difference belongs to II; addition and multiplication are performed subject to these relations. Write ZoZ^o for this algebraic support, retaining multiplicities.1 Here dim⁡A\dim A is the dimension of the support, whereas vector-space dimensions are over C\mathbb{C}. For instance, C[x]\mathbb{C}[x] has support dimension one but infinitely many linearly independent powers of xx.

The number MM counts coordinate variables. The row and column sizes qq, pp introduced below count polynomial components, not variables.

A scalar means an element a∈Aa \in A, possibly a nonconstant polynomial, as opposed to a matrix over AA. On a vector it multiplies every component by aa. It is regular if ab=0ab = 0 forces b=0b = 0. Thus multiplication by a regular scalar is injective, and cancellation by that scalar is valid. Let KK consist of fractions with regular denominators. For integral AA this is the usual fraction field; on A=C[x,y]/(xy)A = \mathbb{C}[x,y]/(xy), xx is not regular because xy=0xy = 0 with y≠0y \ne0, whereas 1+x+y1 + x + y is regular.

A linear functional is a linear rule assigning a complex number to each vector; evaluations and moments are examples. We write W∗:=Hom⁡C(W,C)W^* := \operatorname{Hom}_{\mathbb{C}}(W,\mathbb{C}) for the space of all such rules on WW, called its algebraic dual. Fix positive integers qq, pp and write Mat⁡a×b(B)\operatorname{Mat}_{a\times b}(B) for aa-by-bb matrices with entries in BB, with Mat⁡a(B):=Mat⁡a×a(B)\operatorname{Mat}_a(B) := \operatorname{Mat}_{a\times a}(B). Matrix entries have comma-separated indices, as in H=[Hi,j]H = [H_{i,j}]. The superscript T\mathrm{T} denotes transpose; semicolons separate block rows in inline matrices.

We keep track of which argument a polynomial belongs to. For an arbitrary bivariate pairing use A:=A⊗CA\mathcal{A} := A \otimes_{\mathbb{C}} A, the algebra of two independent copies of the support. The maps ιL\iota_L and ιR\iota_R place a polynomial in the left or right copy: ιL(a):=a⊗1\iota_L(a) := a \otimes1 and ιR(a):=1⊗a\iota_R(a) := 1 \otimes a. For A=C[t]A = \mathbb{C}[t], this means A=C[x,y]\mathcal{A} = \mathbb{C}[x,y], ιL(a)=a(x)\iota_L(a) = a(x) and ιR(a)=a(y)\iota_R(a) = a(y), so ιL(a)ιR(b)=a(x)b(y)\iota_L(a)\iota_R(b) = a(x)b(y). A balanced pairing allows a polynomial factor to be moved from one argument to the other. For this case one copy suffices: A:=A\mathcal{A} := A and ιL(a):=ιR(a):=a\iota_L(a) := \iota_R(a) := a. Both maps act entry by entry on matrices.

For a row f∈A1×qf \in A^{1 \times q} and a column g∈Ap×1g \in A^{p \times1}, the basic scalar tests are the products ιL(fi)ιR(gj)\iota_L(f_i)\iota_R(g_j). In the bivariate convention these are fi(P)gj(Q)f_i(P)g_j(Q), with independent support points P,QP,Q; in the balanced convention they are products in the same algebra AA. Choose a scalar functional μi,j\mu_{i,j} for each component pair and define

βμ(f,g):=∑i=1q∑j=1pμi,j[ιL(fi)ιR(gj)].\beta_\mu(f,g) := \sum_{i=1}^{q}\sum_{j=1}^{p}\mu_{i,j}[\iota_L(f_i)\iota_R(g_j)].

For a balanced matrix measure this is the familiar integral ∫f dμ g\int f\,d\mu\,g. The balanced identity βμ(af,g)=βμ(f,ag)\beta_\mu(af,g) = \beta_\mu(f,ag) follows because multiplication of scalar entries in AA is commutative. For A=C[x]A = \mathbb{C}[x], the matrix of linear forms in this balanced pairing is the framework of Sorokin and Van Iseghem [33]. The coordinate algebra and the two-copy construction above allow the same functional notation to express support relations and pairings with independent arguments.

It is useful to collect the qpqp tests in a matrix. Put V:=Mat⁡q×p(A)V := \operatorname{Mat}_{q \times p}(\mathcal{A}) and write

Hf,g:=[ιL(fi)ιR(gj)]i∈{1,…,q}, j∈{1,…,p}=ιL(f)TιR(g)T,μ[H]:=∑i,jμi,j(Hi,j).H_{f,g} := [\iota_L(f_i)\iota_R(g_j)]_{i \in\{1,\ldots,q\},\,j \in\{1,\ldots,p\}} = \iota_L(f)^{\mathsf{T}}\iota_R(g)^{\mathsf{T}}, \qquad\mu[H] := \sum_{i,j}\mu_{i,j}(H_{i,j}).

References

  1. [1]C. Álvarez-Fernández, G. Ariznabarreta, J. C. García-Ardila, M. Mañas, and F. Marcellán, Transformation theory and Christoffel formulas for matrix biorthogonal polynomials on the real line, arXiv:1605.04617v7 (2016).
  2. [2]C. Álvarez-Fernández, G. Ariznabarreta, J. C. García-Ardila, M. Mañas, and F. Marcellán, Christoffel transformations for matrix orthogonal polynomials in the real line and the non-Abelian 2D Toda lattice hierarchy, Int. Math. Res. Not. IMRN 2017 (2017), no. 5, 1285–1341. DOI.
  3. [3]M. Mañas and M. Rojas, General Christoffel perturbations for multiple orthogonal polynomials of the mixed type on the step-line, Numer. Algorithms 102 (2026), 1829–1868. DOI.DOI
  4. [4]M. Mañas and M. Rojas, General Geronimus perturbations for mixed multiple orthogonal polynomials, Anal. Math. Phys. 15 (2025), article 50. DOI.arxiv.org/abs/2411.16022
  5. [5]M. Mañas and M. Rojas, Uvarov perturbations for multiple orthogonal polynomials of the mixed type on the step-line, arXiv:2510.13754v1 (2025).arxiv.org/abs/2510.13754
  6. [7]G. Ariznabarreta and M. Mañas, Linear spectral transformations for multivariate orthogonal polynomials and multispectral Toda hierarchies, arXiv:1511.09129v5 (2016).arxiv.org/abs/1511.09129
  7. [8]G. Ariznabarreta and M. Mañas, Multivariate orthogonal polynomials and integrable systems, Adv. Math. 302 (2016), 628–739. DOI.DOI
  8. [9]G. Ariznabarreta and M. Mañas, Christoffel transformations for multivariate orthogonal polynomials, J. Approx. Theory 225 (2018), 242–283. DOI.DOI
  9. [10]M. Fasondini, S. Olver, and Y. Xu, Orthogonal polynomials on planar cubic curves, Found. Comput. Math. 23 (2023), 1–31. DOI.
  10. [11]M. Fasondini, S. Olver, and Y. Xu, Orthogonal polynomials on a class of planar algebraic curves, Stud. Appl. Math. 151 (2023), no. 1, 369–405. DOI.arxiv.org/abs/2211.06999
  11. [12]M. Bertola, Abelianization of matrix orthogonal polynomials, Int. Math. Res. Not. IMRN 2023 (2023), no. 10, 8544–8595. DOI.arxiv.org/abs/2107.12998
  12. [13]C. Hu and L. Yu, Geometric realization of finite residue Casimirs and Weil operators via Szegő kernels, arXiv:2608.23996v1 (2026). arXiv.arxiv.org/abs/2608.23996
  13. [14]D. Eisenbud, Commutative Algebra: With a View Toward Algebraic Geometry, Graduate Texts in Mathematics, vol. 150, Springer, New York, 1995. DOI.
  14. [15]F. M. Dopico and V. Noferini, Root polynomials and their role in the theory of matrix polynomials, Linear Algebra Appl. 584 (2020), 37–78. DOI.DOI
  15. [16]Y. Cid-Ruiz and B. Sturmfels, Primary decomposition with differential operators, Int. Math. Res. Not. IMRN 2023 (2023), no. 14, 12119–12147. DOI.
  16. [17]J. E. McCarthy and M. Putinar, Positivity aspects of the Fantappiè transform, J. Anal. Math. 97 (2005), 57–82. DOI.
  17. [18]D. Ben-Zvi and I. Biswas, Theta functions and Szegő kernels, Int. Math. Res. Not. IMRN 2003 (2003), no. 24, 1305–1340. DOI.
  18. [19]C. A. Berenstein and A. Yger, Residue calculus and effective Nullstellensatz, Amer. J. Math. 121 (1999), no. 4, 723–796. DOI.DOI
  19. [20]T. Brazelton, S. McKean, and S. Pauli, Bézoutians and the A¹-degree, Algebra Number Theory 17 (2023), no. 11, 1985–2012. DOI.DOI
  20. [21]M. Andersson and E. Wulcan, Residue currents with prescribed annihilator ideals, Ann. Sci. École Norm. Sup. (4) 40 (2007), no. 6, 985–1007. DOI.arxiv.org/abs/math/0511237
  21. [22]J.-P. Bézivin and P. Robba, A new p-adic method for proving irrationality and transcendence results, Ann. of Math. (2) 129 (1989), no. 1, 151–160. DOI.DOI
  22. [23]J.-R. Wu and X.-B. Hu, Discrete integrable systems associated with orthogonal polynomials on algebraic curves, Math. Phys. Anal. Geom. 29 (2026), article 34. DOI.DOI
  23. [24]G. Ariznabarreta and M. Mañas, Multivariate orthogonal Laurent polynomials and integrable systems, Publ. Res. Inst. Math. Sci. 58 (2022), no. 1, 79–185. DOI.arxiv.org/abs/1506.08708
  24. [25]J. C. García-Ardila, L. E. Garza, and F. Marcellán, An extension of the Geronimus transformation for orthogonal matrix polynomials on the real line, Mediterr. J. Math. 13 (2016), 5009–5032. DOI.
  25. [26]A. J. Durán and F. A. Grünbaum, Orthogonal matrix polynomials satisfying second-order differential equations, Int. Math. Res. Not. 2004 (2004), no. 10, 461–484. DOI.
  26. [27]A. J. Durán and F. A. Grünbaum, Structural formulas for orthogonal matrix polynomials satisfying second-order differential equations, I, Constr. Approx. 22 (2005), no. 2, 255–271. DOI.
  27. [28]A. J. Durán and F. A. Grünbaum, Matrix orthogonal polynomials satisfying second-order differential equations: Coping without help from group representation theory, J. Approx. Theory 148 (2007), no. 1, 35–48. DOI.DOI
  28. [29]E. B. Christoffel, Über die Gaußische Quadratur und eine Verallgemeinerung derselben, J. Reine Angew. Math. 55 (1858), 61–82. DOI.DOI
  29. [30]J. Geronimus, Sur les polynômes orthogonaux relatifs à une suite des nombres donnée et sur le théorème de W. Hahn, Izv. Akad. Nauk SSSR Ser. Mat. 4 (1940), no. 2, 215–228 (in Russian).
  30. [31]V. B. Uvarov, The connection between systems of polynomials that are orthogonal with respect to different distribution functions, U.S.S.R. Comput. Math. Math. Phys. 9 (1969), no. 6, 25–36. DOI.DOI
  31. [32]A. Zhedanov, Rational spectral transformations and orthogonal polynomials, J. Comput. Appl. Math. 85 (1997), no. 1, 67–86. DOI.DOI
  32. [33]V. N. Sorokin and J. Van Iseghem, Algebraic aspects of matrix orthogonality for vector polynomials, J. Approx. Theory 90 (1997), no. 1, 97–116. DOI.DOI
  33. [34]E. Daems and A. B. J. Kuijlaars, Multiple orthogonal polynomials of mixed type and non-intersecting Brownian motions, J. Approx. Theory 146 (2007), 91–114. DOI.arxiv.org/abs/math/0511470
  34. [35]A. M. Delgado, L. Fernández, T. E. Pérez, M. A. Piñar, and Y. Xu, Orthogonal polynomials in several variables for measures with mass points, Numer. Algorithms 55 (2010), 245–264. DOI.
  35. [36]A. M. Delgado, L. Fernández, T. E. Pérez, and M. A. Piñar, Multivariate orthogonal polynomials and modified moment functionals, SIGMA 12 (2016), 090, 25 pp. DOI.DOI
  36. [37]F. Marcellán, M. E. Marriaga, T. E. Pérez, and M. A. Piñar, Geronimus transformations of bivariate linear functionals, J. Math. Anal. Appl. 484 (2020), no. 2, 123736. DOI.DOI
  37. [38]M. I. Bueno and F. Marcellán, Darboux transformation and perturbation of linear functionals, Linear Algebra Appl. 384 (2004), 215–242. DOI.DOI
  38. [39]J. Lipman, Dualizing sheaves, differentials and residues on algebraic varieties, Astérisque, vol. 117, Société Mathématique de France, Paris, 1984. Publisher’s text.
  39. [40]A. Baker, Transcendental Number Theory, Cambridge University Press, Cambridge, 1975. DOI.DOI
  40. [41]E. J. Huertas and M. Mañas, Mixed-type multiple orthogonal Laurent polynomials on the unit circle, J. Comput. Appl. Math. 475 (2026), 117037. DOI.DOI
  41. [42]M. I. Bueno and F. Marcellán, Polynomial perturbations of bilinear functionals and Hessenberg matrices, Linear Algebra Appl. 414 (2006), no. 1, 64–83. DOI.DOI
  42. [43]L. Fernández, T. E. Pérez, M. A. Piñar, and Y. Xu, Krall-type orthogonal polynomials in several variables, J. Comput. Appl. Math. 233 (2010), no. 6, 1519–1524. DOI.arxiv.org/abs/0801.0065
  43. [44]R. Kozhan and M. Vaktnäs, Christoffel transform and multiple orthogonal polynomials, J. Comput. Appl. Math. 476 (2026), 117121. DOI.DOI
  44. [45]J.-P. Serre, Algebraic Groups and Class Fields, Graduate Texts in Mathematics, vol. 117, Springer, New York, 1988. DOI.DOI
  45. [46]J. D. Fay, Theta Functions on Riemann Surfaces, Lecture Notes in Mathematics, vol. 352, Springer, Berlin–Heidelberg, 1973. DOI.DOI

Paper details

vlnc:W_fn2n3gsv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

ChatGPT, Codex, Claude

View statement
Looking for related papers…
Contents