Multivariate Hermite–Kakeya–Obreschkoff theorem

Two real multivariate polynomials are in proper position if and only if all their real linear combinations are real stable or zero.

Results

A polynomial $P\in\mathbb C[z_i:i\in\sigma]$ is stable if $P(z)\ne0$ whenever every $z_i$ lies in the open upper half-plane. A polynomial with real coefficients is real stable if it is stable as a complex polynomial; in particular it is nonzero. For $F,G\in\mathbb R[z_i:i\in\sigma]$ we say that $F$ and $G$ are in proper position, written $F\ll G$, if $G+iF$ is stable.

Theorem (Hermite–Kakeya–Obreschkoff). Let $\sigma$ be finite and $F\ll G$. Then for all real $\alpha,\beta$ the polynomial $\alpha F+\beta G$ is zero or real stable.

Theorem (converse) [Obr63]. Let $\sigma$ be any index set, and suppose $\alpha F+\beta G$ is zero or real stable for all real $\alpha,\beta$. Then $F=G=0$, or $F\ll G$, or $G\ll F$.

Theorem (affine pencil). For arbitrary $\sigma$, $F\ll G$ if and only if the polynomial $G+wF$, in the variables $z_i$ and a new variable $w$, is real stable.

Use

These are the multivariate analogues of the classical Hermite–Kakeya–Obreschkoff theorem for interlacing polynomials; see the univariate Obreschkoff page. Proper position is the multivariate counterpart of interlacing through the Hermite–Biehler theorem; see the Hermite–Biehler page.

Proof

For the affine pencil, restrict $G+iF$ to a line $z_i=a_i+b_it$ with $b_i>0$, which gives a univariate Hermite–Biehler statement in $t$. A norm comparison between a point and its complex conjugate then shows that $G+wF$ is stable. Conversely, $w=i$ gives $G+iF$. A stable pencil yields the combinations $\alpha F+\beta G$ by specializing $w$ and, for the direction polynomial $F$, differentiating in $w$; this is where finiteness of $\sigma$ is used.

For the converse, the cases $G=0$ and $F=cG$ are direct. Otherwise, for every $z$ with all $\operatorname{Im}z_i>0$, the quotient $F(z)/G(z)$ is not real, since a real value $c$ would make $F-cG$, which is real stable, vanish at $z$. The set of such points is connected, so $\operatorname{Im}(F/G)$ has constant sign. A negative sign makes $G+iF$ stable, and a positive sign makes $F+iG$ stable.

References

  1. [BB09] Julius Borcea & Petter Brändén. The Lee–Yang and Pólya–Schur programs. I. Linear operators preserving stability. Inventiones mathematicae, 177(3), 541-569, 2009.
    BibTeX
    @article{BorceaBranden2009I,
      doi = {10.1007/s00222-009-0189-3},
      url2 = {https://doi.org/10.1007/s00222-009-0189-3},
      year = {2009},
      month = mar,
      publisher = {Springer Science and Business Media {LLC}},
      volume = {177},
      number = {3},
      pages = {541--569},
      author = {Julius Borcea and Petter Br{\"{a}}nd{\'{e}}n},
      title = {The {L}ee--{Y}ang and {P}{\'{o}}lya--{S}chur programs. {I}. {L}inear operators preserving stability},
      journal = {Inventiones mathematicae}
    }
  2. [BB10] J. Borcea & P. Brändén. Multivariate Pólya--Schur classification problems in the Weyl algebra. Proceedings of the London Mathematical Society, 101(1), 73-104, 2010.
    BibTeX
    @article{BorceaBranden2010,
      eprint = {math/0606360},
      doi = {10.1112/plms/pdp049},
      url = {https://doi.org/10.1112/plms/pdp049},
      year = {2010},
      month = jul,
      publisher = {Wiley},
      volume = {101},
      number = {1},
      pages = {73--104},
      author = {J. Borcea and P. Br{\"{a}}nd{\'{e}}n},
      title = {Multivariate {P}{\'{o}}lya--{S}chur classification problems in the {W}eyl algebra},
      journal = {Proceedings of the London Mathematical Society}
    }
  3. [Obr63] Nikola Obreschkoff. Verteilung und Berechnung der Nullstellen reeller Polynome. Hochschulbücher für Mathematik, vol. 55, VEB Deutscher Verlag der Wissenschaften, 1963.
    BibTeX
    @book{Obreschkoff1963,
      author = {Obreschkoff, Nikola},
      title = {Verteilung und Berechnung der Nullstellen reeller Polynome},
      series = {Hochschulb{\"u}cher f{\"u}r Mathematik},
      volume = {55},
      publisher = {VEB Deutscher Verlag der Wissenschaften},
      address = {Berlin},
      year = {1963}
    }

Definitions

  • Proper position F ≪ G

    RealRooted.MvProperPositionsource
    def MvProperPosition {σ : Type*} (F G : MvPolynomial σ ℝ) : Prop :=
      MvUpperHalfPlaneStable
        (complexifyMv G + MvPolynomial.C Complex.I * complexifyMv F)
  • Stable affine pencil G + wF

    RealRooted.MvStablePencilsource
    def MvStablePencil {σ : Type*} (F G : MvPolynomial σ ℝ) : Prop :=
      MvRealStable
        (MvPolynomial.rename some G + MvPolynomial.X none *
          MvPolynomial.rename some F)
  • Every real combination is real stable or zero

    RealRooted.AllComboMvRealStableOrZerosource
    def AllComboMvRealStableOrZero {σ : Type*}
        (F G : MvPolynomial σ ℝ) : Prop :=
      ∀ α β : ℝ,
        MvRealStableOrZero
          (MvPolynomial.C α * F + MvPolynomial.C β * G)

Theorems

  • Hermite–Kakeya–Obreschkoff converse: a stable real span gives proper positionMain result

    RealRooted.AllComboMvRealStableOrZero.eq_zero_or_properPositionsource
    theorem AllComboMvRealStableOrZero.eq_zero_or_properPosition
        {σ : Type*} {F G : MvPolynomial σ ℝ}
        (hall : AllComboMvRealStableOrZero F G) :
        (F = 0 ∧ G = 0) ∨ MvProperPosition F G ∨ MvProperPosition G F
  • Hermite–Kakeya–Obreschkoff: proper position gives a stable real span

    RealRooted.MvProperPosition.allComboMvRealStableOrZerosource
    theorem MvProperPosition.allComboMvRealStableOrZero
        {σ : Type*} [Finite σ] {F G : MvPolynomial σ ℝ}
        (h : MvProperPosition F G) : AllComboMvRealStableOrZero F G
  • Proper position is stability of the affine pencil G + wF

    RealRooted.mvProperPosition_iff_mvStablePencilsource
    theorem mvProperPosition_iff_mvStablePencil
        {σ : Type*} {F G : MvPolynomial σ ℝ} :
        MvProperPosition F G ↔ MvStablePencil F G

Lean source RealRooted/Challenges/Stability/MultivariateObreschkoff.lean at revision d6670742. Formalized 2026-10-13.