Concepts
Multiplier sequences
A real sequence $\gamma = (\gamma_0, \gamma_1, \dotsc)$ is a multiplier sequence if the diagonal operator
$$a_0 + a_1 x + \dotsb + a_n x^n \;\mapsto\; \gamma_0 a_0 + \gamma_1 a_1 x + \dotsb + \gamma_n a_n x^n$$
sends every real-rooted polynomial to a real-rooted polynomial (or zero). It is a PF multiplier sequence if it preserves real-rootedness with nonnegative coefficients. The finite version restricts to inputs of degree at most $n$.
The following hold:
- Jensen polynomials (Pólya–Schur): a nonnegative sequence is a multiplier sequence if and only if every Jensen polynomial $g_n(x) = \sum_k \binom{n}{k} \gamma_k x^k$ has only real nonpositive zeros.
- Laguerre–Pólya classification (Pólya–Schur): under a growth condition, $\gamma$ is a PF multiplier sequence if and only if its exponential generating function $\sum_k \gamma_k z^k/k!$ is a locally uniform limit of PF polynomials, that is, lies in the Laguerre–Pólya class of type I. Without the sign condition, $\gamma$ is a multiplier sequence if and only if one of $\pm f(\pm z)$ is of type I, where $f$ is the generating function; equivalently, one of $\pm\gamma_k$ and $\pm(-1)^k \gamma_k$ is a PF multiplier sequence.
- Log-concavity: PF multiplier sequences are log-concave.
- Laguerre's theorem: if a polynomial $\varphi$ has only real nonpositive zeros, then $\varphi(0), \varphi(1), \varphi(2), \dotsc$ is a multiplier sequence. It is a PF multiplier sequence when the leading coefficient is nonnegative. The basic case $k + r$, for real $r \geq 0$, has Jensen polynomials $(1+x)^{n-1}\bigl((n+r)x+r\bigr)$. More generally, if $f$ is a locally uniform limit of PF polynomials (type I in the Laguerre–Pólya class), then $f(0), f(1), f(2), \dotsc$ is a PF multiplier sequence.
- Examples: for real $\alpha > 0$, the reciprocal rising factorials $1/(\alpha)_k$ form a PF multiplier sequence, and so in particular do $1/k!$. After clearing denominators, their Jensen polynomials are generalized Laguerre polynomials.
References
G. Pólya and J. Schur, “Über zwei Arten von Faktorenfolgen in der Theorie der algebraischen Gleichungen,” J. Reine Angew. Math. 144 (1914), 89–113. The finite Pólya–Schur theorem through the Schur–Szegő composition is on the Hadamard products page.
≔ Definitions
Diagonal operator of a sequence
def diagonalOperator (gamma : ℕ → ℝ) (p : ℝ[X]) : ℝ[X] :=
p.sum fun n a => monomial n (gamma n * a)
Jensen polynomials
def jensenPolynomial (n : ℕ) (gamma : ℕ → ℝ) : ℝ[X] :=
∑ k ∈ Finset.range (n + 1),
monomial k ((Nat.choose n k : ℝ) * gamma k)
Finite multiplier sequence
def IsFiniteMultiplierSequence (n : ℕ) (gamma : ℕ → ℝ) : Prop :=
∀ {p : ℝ[X]},
p.natDegree ≤ n →
p.Splits →
diagonalOperator gamma p = 0 ∨ (diagonalOperator gamma p).Splits
Multiplier sequence
def IsMultiplierSequence (gamma : ℕ → ℝ) : Prop :=
∀ n : ℕ, IsFiniteMultiplierSequence n gamma
PF multiplier sequence
def IsPFMultiplierSequence (gamma : ℕ → ℝ) : Prop :=
∀ n : ℕ, IsFinitePFMultiplierSequence n gamma
Laguerre–Pólya class of type I
def IsLaguerrePolyaTypeI (f : ℂ → ℂ) : Prop :=
∃ p : ℕ → ℝ[X],
(∀ n, IsPFPolynomial (p n)) ∧
TendstoLocallyUniformly
(fun n (z : ℂ) => (p n).map Complex.ofRealHom |>.eval z) f atTop
⊢ Theorems
Pólya–Schur: multiplier sequences via Jensen polynomialsMain result
theorem isMultiplierSequence_iff_jensenPolynomial_isPF
{gamma : ℕ → ℝ} (hgamma : ∀ k, 0 ≤ gamma k) :
IsMultiplierSequence gamma ↔
∀ n : ℕ, IsPFPolynomial (jensenPolynomial n gamma)
Pólya–Schur: PF multiplier sequences are the Laguerre–Pólya type I classMain result
theorem isPFMultiplierSequence_iff_isLaguerrePolyaTypeI_complexExpGeneratingFunction
{gamma : ℕ → ℝ}
(hpositive : ∃ R : NNReal, 0 < R ∧
Summable (fun k => ‖gamma k‖ * (R : ℝ) ^ k / k.factorial)) :
IsPFMultiplierSequence gamma ↔
IsLaguerrePolyaTypeI (complexExpGeneratingFunction gamma)
Pólya–Schur: classification of all multiplier sequencesMain result
theorem isMultiplierSequence_iff_isLaguerrePolyaTypeISigned_complexExpGeneratingFunction
{gamma : ℕ → ℝ}
(hpositive : ∃ R : NNReal, 0 < R ∧
Summable (fun k => ‖gamma k‖ * (R : ℝ) ^ k / k.factorial)) :
IsMultiplierSequence gamma ↔
IsLaguerrePolyaTypeISigned (complexExpGeneratingFunction gamma)
Every multiplier sequence is a PF one up to signs
theorem IsMultiplierSequence.exists_pf_sign_normalization
{gamma : ℕ → ℝ} (hgamma : IsMultiplierSequence gamma) :
(IsPFMultiplierSequence gamma ∨
IsPFMultiplierSequence (fun k => -gamma k)) ∨
(IsPFMultiplierSequence (fun k => (-1 : ℝ) ^ k * gamma k) ∨
IsPFMultiplierSequence (fun k => -((-1 : ℝ) ^ k * gamma k)))
PF multiplier sequences are log-concave
theorem IsPFMultiplierSequence.logConcave {gamma : ℕ → ℝ}
(hgamma : IsPFMultiplierSequence gamma) (k : ℕ) :
gamma k * gamma (k + 2) ≤ gamma (k + 1) ^ 2
Reciprocal rising factorials 1/(α)ₖ form a PF multiplier sequence
theorem isPFMultiplierSequence_inv_ascPochhammer {α : ℝ} (hα : 0 < α) :
IsPFMultiplierSequence (fun k => ((ascPochhammer ℝ k).eval α)⁻¹)
1/k! is a PF multiplier sequence
theorem isPFMultiplierSequence_inv_factorial :
IsPFMultiplierSequence (fun k => ((k.factorial : ℝ))⁻¹)
Laguerre's theorem: φ(0), φ(1), … is a multiplier sequenceMain result
theorem isMultiplierSequence_eval_of_roots_nonpos {φ : ℝ[X]} (hφ : φ.Splits)
(hroots : ∀ r ∈ φ.roots, r ≤ 0) :
IsMultiplierSequence (fun k => φ.eval (k : ℝ))
k + r is a PF multiplier sequence for r ≥ 0
theorem isPFMultiplierSequence_natCast_add {r : ℝ} (hr : 0 ≤ r) :
IsPFMultiplierSequence (fun k => (k : ℝ) + r)
Type I functions sampled at 0, 1, 2, … are PF multiplier sequences
theorem IsLaguerrePolyaTypeI.isPFMultiplierSequence_eval_natCast {f : ℂ → ℂ}
(hf : IsLaguerrePolyaTypeI f) :
IsPFMultiplierSequence (fun k => (f k).re)
Lean source RealRooted/Challenges/MultiplierSequence.lean at revision 5827550b.