Positive functionals and Hessenberg matrices

Jean B. Lasserre, Mihai Putinar

Abstract

Not every positive functional defined on bi-variate polynomials of a prescribed degree bound is represented by the integration against a positive measure. We isolate a couple of conditions filling this gap, either by restricting the class of polynomials to harmonic ones, or imposing the vanishing of a defect indicator. Both criteria offer constructive cubature formulas and they are obtained via well known matrix analysis techniques involving either the dilation of a contractive matrix to a unitary one or the specific structure of the Hessenberg matrix associated to the multiplier by the underlying complex variable.

Introduction

In mathematical analysis, the Riesz Representation Theorem provides a fundamental link between linear functionals and measures. A "positive functional" is a linear map that returns a non-negative value for any non-negative function. Ideally, such a functional can be represented as an integral against a positive measure, meaning the functional acts as a weighted sum of function values at specific points, known as nodes.

However, when we restrict our scope to polynomials of a fixed maximum degree—a scenario known as the truncated moment problem—this representation is not guaranteed. There exists a structural gap where a functional may appear positive on the space of polynomials but cannot be realized by a physical measure. This paper addresses this gap for bivariate polynomials by providing constructive conditions—mathematical recipes—to identify when and how integration nodes and weights (cubature formulas) can be found.

Methodology: From Real Variables to Complex Operators

To bridge the gap between abstract functionals and concrete numerical integration, the authors shift the perspective from real-variable quadratic forms to complex variables ($z = x + iy$). This transition allows the application of powerful tools from operator theory and matrix analysis.

1. Sz.-Nagy Unitary Dilation

The authors employ the Sz.-Nagy dilation theorem, which demonstrates that a contractive multiplication operator can be "dilated" into a larger unitary operator. The spectral resolution of this resulting unitary operator provides the necessary nodes and weights for a cubature formula, effectively transforming an operator-theoretic problem into a numerical integration solution.

2. Hessenberg Matrix Representation

By representing the multiplication-by-$z$ operator ($M$) within an orthonormal basis of polynomials, the authors generate a Hessenberg matrix. This matrix is nearly upper-triangular, featuring non-zero entries only on the main diagonal, the upper triangles, and the first sub-diagonal. The structural properties of this matrix are central to determining the existence of a measure.

3. Normality and the Defect Indicator

The existence of a cubature formula is intimately tied to whether the Hessenberg matrix is "normal" (i.e., it commutes with its adjoint, $MM^* = M^*M$). The paper proves that these matrices are always "almost normal," specifically normal up to a rank-one perturbation. The authors identify a specific entry in the matrix, $a_{d, d+1}$, which acts as a "defect indicator." If this value vanishes, the functional is guaranteed to have a representation as a measure on a specific subspace.

Key Insights

The "Price" of Cubature

A primary insight of this work is that obtaining a cubature formula with positive weights often requires a trade-off. One must typically "pay a price" by narrowing the requirement: the formula may not work for all polynomials of a certain degree, but rather for a distinguished subspace—such as harmonic polynomials—which comprises roughly half the original dimension.

Harmonic Simplicity

For harmonic polynomials (those satisfying the Laplace equation $\Delta f = 0$), the paper demonstrates that a cubature formula always exists. Furthermore, the nodes for these formulas are located predictably on a circle, simplifying the search for integration points.

Operator Theory Connection

The research elegantly links the truncated moment problem to the study of hyponormal operators. These are operators where the self-commutator $[M^*, M]$ is positive semi-definite. The authors show that the multiplication operator associated with the functional is normal after a rank-one perturbation, meaning the self-commutator has at most one negative eigenvalue.

Principal Results

  • Theorem 3.1: Establishes that for any real functional non-negative on Hermitian squares, there exists a cubature formula for all harmonic polynomials of degree $d$ using at most $(d+1)^2$ nodes situated on a circle of radius $R$.
  • Theorem 4.1: Proves that the multiplication operator $M$ is a rank-one perturbation of a normal operator, providing a structural limit on the "non-normality" of the system.
  • Corollary 4.2: Provides a quantitative criterion for existence. A cubature formula exists for polynomials of degree $d+1$ (within certain constraints) if and only if the defect indicator $a_{d, d+1}$ in the Hessenberg matrix vanishes.
  • Example 5.1: Demonstrates an unexpected application where a Dirichlet-type inner product, which involves derivatives and does not initially resemble a standard integration functional, can be represented by complex quadrature nodes.

Conclusion and Future Directions

While the paper provides a robust constructive method for finding cubature formulas, it highlights the inherent limitations of the truncated moment problem. The most powerful results currently apply to specific subspaces rather than the entire space of bivariate polynomials.

Future work may involve exploring a hierarchy of stronger, non-equivalent conditions, such as $k$-hyponormality. These conditions, derived from the positivity of the functional on different types of polynomial squares, could lead to even more refined numerical certificates for the existence of measures in multi-variable calculus.

Summary updated

Abstract

Not every positive functional defined on bi-variate polynomials of a prescribed degree bound is represented by the integration against a positive measure. We isolate a couple of conditions filling this gap, either by restricting the class of polynomials to harmonic ones, or imposing the vanishing of a defect indicator. Both criteria offer constructive cubature formulas and they are obtained via well-known matrix analysis techniques involving either the dilation of a contractive matrix to a unitary one or the specific structure of the Hessenberg matrix associated to the multiplier by the underlying complex variable.
BibTeX
@inproceedings{lasserre2018positive,
  author = {Jean B. Lasserre and Mihai Putinar},
  title = {{Positive functionals and Hessenberg matrices}},
  booktitle = {Proceedings of the American Mathematical Society},
  year = {2018},
  doi = {10.1090/proc/14266},
}
Figure