Hermite polynomials
The probabilists' Hermite polynomials are real-rooted with simple roots, consecutive ones interlace, and they are orthogonal for the Gaussian weight.
Results
The probabilists' Hermite polynomials satisfy $H_0=1$, $H_1=x$ and $$H_{n+2}(x)=x\,H_{n+1}(x)-(n+1)\,H_n(x).$$
Theorem. For all $n$:
- $H_n$ is monic of degree $n$ and has only real, simple roots;
- $H_n\ll H_{n+1}$, and $H_n,H_{n-1},\dots,H_0$ is a Sturm sequence, that is, each consecutive pair interlaces;
- the recurrence is a Favard recurrence with diagonal $0$ and subdiagonal $n$;
- $\displaystyle\int_{-\infty}^{\infty}H_i(x)H_j(x)\,e^{-x^2/2}\,dx=\delta_{ij}\sqrt{2\pi}\,i!$.
Proof
The root and interlacing statements follow from the general Favard theory, because the subdiagonal is positive.
References
[Sze39].
- [Sze39] G. Szegő. Orthogonal Polynomials. Colloquium Publications, American Mathematical Society, 1939.
BibTeX
@book{Szego1939,
author = {Szegő, G.},
title = {Orthogonal {P}olynomials},
year = {1939},
series = {Colloquium Publications},
publisher = {American Mathematical Society},
isbn = {9781470431716},
doi = {10.1090/coll/023},
url = {https://doi.org/10.1090/coll/023}
}
≔ Definitions
Probabilists' Hermite polynomials
abbrev hermiteReal (n : ℕ) : ℝ[X] :=
(Polynomial.hermite n).map (Int.castRingHom ℝ)
Gaussian weight
def hermiteGaussianWeight (x : ℝ) : ℝ :=
Real.exp (-(x ^ 2 / 2))
⊢ Theorems
Hermite polynomials are real-rooted
theorem hermiteReal_isRealRooted (n : ℕ) :
hermiteReal n ≠ 0 ∧ (hermiteReal n).Splits
Hermite polynomials have simple roots
theorem hermiteReal_hasSimpleRoots (n : ℕ) :
HasSimpleRoots (hermiteReal n)
Consecutive Hermite polynomials interlaceMain result
theorem hermiteReal_strictInterl_succ (n : ℕ) :
StrictInterl (hermiteReal n) (hermiteReal (n + 1))
Hermite polynomials form a Sturm sequence
theorem hermiteReal_isSturmSeq (n : ℕ) :
IsSturmSeq ((List.range (n + 1)).reverse.map hermiteReal)
Hermite polynomials satisfy a Favard recurrence
theorem hermiteReal_satisfiesFavardRecurrence :
SatisfiesFavardRecurrence hermiteReal
(fun _ ↦ (0 : ℝ)) (fun n ↦ (n : ℝ))
Hermite polynomials are orthogonal for the Gaussian weight
theorem hermiteReal_integral_orthogonal (i j : ℕ) :
(∫ x : ℝ,
(hermiteReal i).eval x * (hermiteReal j).eval x *
hermiteGaussianWeight x) =
if i = j then
Real.sqrt (2 * Real.pi) * (i.factorial : ℝ)
else 0
Lean source RealRooted/Challenges/Orthogonal/HermitePolynomials.lean at revision d6670742. Formalized 2026-09-07.