Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 5.11 — eigenvalues and eigenfunction expansion of the regular Sturm–Liouville problem

Proved
TeschlODE.SturmLiouville.eigenfunction_expansion

by mikedeng1 · Sep 28, 2026 · Mathlib 0df444a (Lean v4.33.1)

eigenfunction-expansionordinary-differential-equationsp2o-batch-books5p2o-gran-per-chapterp2o-plan-bookp2o-v1spectral-theoremsturm-liouville

Assume (5.45): a<ba < ba<b, r,q∈C([a,b],R)r, q \in C([a,b],\mathbb{R})r,q∈C([a,b],R), p∈C1([a,b],R)p \in C^1([a,b],\mathbb{R})p∈C1([a,b],R), p,r>0p, r > 0p,r>0 on [a,b][a,b][a,b], and fix α,β∈R\alpha, \beta \in \mathbb{R}α,β∈R. Let L=r−1(−ddxpddx+q)L = r^{-1}\bigl(-\frac{d}{dx} p \frac{d}{dx} + q\bigr)L=r−1(−dxd​pdxd​+q) on D(L)={f∈C2([a,b],C):BCa(f)=BCb(f)=0}D(L) = \{f \in C^2([a,b],\mathbb{C}) : BC_a(f) = BC_b(f) = 0\}D(L)={f∈C2([a,b],C):BCa​(f)=BCb​(f)=0} and H0=C([a,b],C)H_0 = C([a,b],\mathbb{C})H0​=C([a,b],C) with ⟨f,g⟩=∫abf∗g r dx\langle f, g\rangle = \int_a^b f^* g\, r\, dx⟨f,g⟩=∫ab​f∗grdx.

Then the regular Sturm–Liouville problem has a countable number of discrete and simple eigenvalues EnE_nEn​ which accumulate only at ∞\infty∞, and the corresponding normalized eigenfunctions unu_nun​ can be chosen real-valued and form an orthonormal basis of H0H_0H0​:

f(x)=∑n=0∞⟨un,f⟩ un(x),f∈H0,(5.70)f(x) = \sum_{n=0}^{\infty} \langle u_n, f\rangle\, u_n(x), \qquad f \in H_0, \qquad (5.70)f(x)=n=0∑∞​⟨un​,f⟩un​(x),f∈H0​,(5.70)

with convergence in H0H_0H0​; for f∈D(L)f \in D(L)f∈D(L) the series converges uniformly on [a,b][a,b][a,b].

Precisely: there are E:N→RE : \mathbb{N} \to \mathbb{R}E:N→R and real-valued unu_nun​ such that

  • each unu_nun​ is an eigenfunction of LLL for EnE_nEn​, and the EnE_nEn​ are pairwise distinct;
  • every eigenvalue z∈Cz \in \mathbb{C}z∈C of LLL equals some EnE_nEn​, and every eigenfunction for it is a constant multiple of unu_nun​ on [a,b][a,b][a,b] (simplicity);
  • ∣En∣→∞|E_n| \to \infty∣En​∣→∞ (no finite accumulation point);
  • ⟨um,un⟩=δmn\langle u_m, u_n\rangle = \delta_{mn}⟨um​,un​⟩=δmn​;
  • for every f∈H0f \in H_0f∈H0​, ∥f−∑n<N⟨un,f⟩un∥→0\bigl\| f - \sum_{n<N} \langle u_n, f\rangle u_n \bigr\| \to 0​f−∑n<N​⟨un​,f⟩un​​→0 as N→∞N \to \inftyN→∞;
  • for every f∈D(L)f \in D(L)f∈D(L), ∑n<N⟨un,f⟩un→f\sum_{n<N} \langle u_n, f\rangle u_n \to f∑n<N​⟨un​,f⟩un​→f uniformly on [a,b][a,b][a,b].

This is the eigenfunction expansion theorem for regular Sturm–Liouville problems, the basis of separation of variables for the heat and wave equations on an interval (§5.1).

Formalization Note. "Countable, discrete, accumulating only at ∞\infty∞" is read as: the eigenvalues are exactly the values of an injective sequence EnE_nEn​ with ∣En∣→∞|E_n| \to \infty∣En​∣→∞ (so there are infinitely many, and only finitely many in any bounded set). "Simple" is: every eigenfunction for EnE_nEn​ is c unc\,u_ncun​ on [a,b][a,b][a,b]. "Can be chosen real-valued" is the existential choice of un:R→Ru_n : \mathbb{R} \to \mathbb{R}un​:R→R. The orthonormal basis is the book's (5.32)/(5.34) in the incomplete space H0H_0H0​: norm convergence of the partial sums, written as Re⁡⟨f−SNf,f−SNf⟩→0\operatorname{Re}\langle f - S_N f, f - S_N f\rangle \to 0Re⟨f−SN​f,f−SN​f⟩→0, not Mathlib's HilbertBasis and not L2L^2L2. The sequence EnE_nEn​ is not required to be increasing here (the ordering is Lemma 5.12). α,β\alpha, \betaα,β are arbitrary reals; the boundary conditions depend only on them modulo π\piπ.

Preamble
import Mathlib
import Definitions.Def_TeschlODE_SturmLiouville_RegularSL
import Definitions.Def_TeschlODE_SturmLiouville_SLOp
import Definitions.Def_TeschlODE_SturmLiouville_SLDomain
import Definitions.Def_TeschlODE_SturmLiouville_IsSLEigenfunction
import Definitions.Def_TeschlODE_SturmLiouville_SLInner
Formal statement
namespace TeschlODE.SturmLiouville

/-- Teschl, Theorem 5.11, p. 160: under (5.45), the regular Sturm–Liouville problem
`L = r⁻¹(−(d/dx) p (d/dx) + q)` on `D(L)` (5.53)–(5.55) has countably many discrete, simple
eigenvalues `Eₙ` accumulating only at `∞`: there are real `Eₙ` (pairwise distinct, `|Eₙ| → ∞`)
and real-valued eigenfunctions `uₙ ∈ D(L)`, `L uₙ = Eₙ uₙ`, such that every eigenvalue `z ∈ ℂ` of
`L` is some `Eₙ` and every eigenfunction for it is a multiple of `uₙ`. The `uₙ` are orthonormal
for (5.52) and form an orthonormal basis of `H₀ = C([a, b], ℂ)`: for every `f ∈ H₀` the partial
sums of `Σ ⟨uₙ, f⟩ uₙ` (5.70) converge to `f` in the `H₀`-norm; for `f ∈ D(L)` they converge
uniformly on `[a, b]`. -/
theorem eigenfunction_expansion {p q r : ℝ → ℝ} {a b : ℝ} (α β : ℝ)
    (hreg : RegularSL p q r a b) :
    ∃ (E : ℕ → ℝ) (u : ℕ → ℝ → ℝ),
      (∀ n, IsSLEigenfunction p q r a b α β (E n : ℂ) (fun x => ((u n x : ℝ) : ℂ))) ∧
      Function.Injective E ∧
      (∀ (z : ℂ) (f : ℝ → ℂ), IsSLEigenfunction p q r a b α β z f →
        ∃ n, z = (E n : ℂ) ∧ ∃ c : ℂ, ∀ x ∈ Set.Icc a b, f x = c * (u n x : ℂ)) ∧
      Filter.Tendsto (fun n => |E n|) Filter.atTop Filter.atTop ∧
      (∀ m n, SLInner r a b (fun x => ((u m x : ℝ) : ℂ)) (fun x => ((u n x : ℝ) : ℂ)) =
        if m = n then 1 else 0) ∧
      (∀ f : ℝ → ℂ, ContinuousOn f (Set.Icc a b) →
        Filter.Tendsto
          (fun N => (SLInner r a b
            (f - fun x => ∑ n ∈ Finset.range N,
              SLInner r a b (fun y => ((u n y : ℝ) : ℂ)) f * (u n x : ℂ))
            (f - fun x => ∑ n ∈ Finset.range N,
              SLInner r a b (fun y => ((u n y : ℝ) : ℂ)) f * (u n x : ℂ))).re)
          Filter.atTop (nhds 0)) ∧
      (∀ f : ℝ → ℂ, SLDomain p a b α β f →
        TendstoUniformlyOn
          (fun N x => ∑ n ∈ Finset.range N,
            SLInner r a b (fun y => ((u n y : ℝ) : ℂ)) f * (u n x : ℂ))
          f Filter.atTop (Set.Icc a b)) := by sorry

end TeschlODE.SturmLiouville
Source
Teschl, Ordinary Differential Equations and Dynamical Systems (author's preliminary version of AMS GSM 140, 2012), p. 160, Theorem 5.11
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

The statement is about a quintuple (p,q,r,a,b)(p, q, r, a, b)(p,q,r,a,b) of real data and two further real numbers α,β\alpha, \betaα,β. Its conclusions are phrased using four predicates or functions defined elsewhere in the bundle: RegularSL\mathrm{RegularSL}RegularSL, IsSLEigenfunction\mathrm{IsSLEigenfunction}IsSLEigenfunction, SLDomain\mathrm{SLDomain}SLDomain and SLInner\mathrm{SLInner}SLInner. Their code is not part of the supplied declaration. This read-back therefore gives only the arguments each one takes and how the statement uses it. It cannot unfold what they mean.

Data and hypotheses.

  • p,q,r:R→Rp, q, r : \mathbb{R} \to \mathbb{R}p,q,r:R→R are arbitrary real functions on the whole real line.
  • a,b∈Ra, b \in \mathbb{R}a,b∈R are arbitrary reals. No order such as a<ba < ba<b is assumed in the statement itself.
  • α,β∈R\alpha, \beta \in \mathbb{R}α,β∈R are arbitrary reals. They are not restricted by any hypothesis.
  • The only hypothesis is RegularSL(p,q,r,a,b)\mathrm{RegularSL}(p, q, r, a, b)RegularSL(p,q,r,a,b). It does not involve α\alphaα or β\betaβ.

Notation for the conclusion. Write

⟨g,h⟩:=SLInner(r,a,b,g,h)∈C\langle g, h \rangle := \mathrm{SLInner}(r, a, b, g, h) \in \mathbb{C}⟨g,h⟩:=SLInner(r,a,b,g,h)∈C

for complex-valued functions g,h:R→Cg, h : \mathbb{R} \to \mathbb{C}g,h:R→C. This quantity depends only on rrr, aaa, bbb, ggg and hhh.

Write Eig(z,f)\mathrm{Eig}(z, f)Eig(z,f) for IsSLEigenfunction(p,q,r,a,b,α,β,z,f)\mathrm{IsSLEigenfunction}(p, q, r, a, b, \alpha, \beta, z, f)IsSLEigenfunction(p,q,r,a,b,α,β,z,f), where z∈Cz \in \mathbb{C}z∈C and f:R→Cf : \mathbb{R} \to \mathbb{C}f:R→C.

Write Dom(f)\mathrm{Dom}(f)Dom(f) for SLDomain(p,a,b,α,β,f)\mathrm{SLDomain}(p, a, b, \alpha, \beta, f)SLDomain(p,a,b,α,β,f). This predicate does not involve qqq or rrr.

Conclusion. There exist a sequence of reals (En)n∈N(E_n)_{n \in \mathbb{N}}(En​)n∈N​ and a sequence of real-valued functions un:R→Ru_n : \mathbb{R} \to \mathbb{R}un​:R→R with the seven properties below. Here N\mathbb{N}N starts at 000. Each unu_nun​ is regarded as a complex-valued function through the inclusion R⊂C\mathbb{R} \subset \mathbb{C}R⊂C.

  1. Each unu_nun​ is an eigenfunction for EnE_nEn​. For every nnn, Eig(En,un)\mathrm{Eig}(E_n, u_n)Eig(En​,un​) holds, with EnE_nEn​ viewed as a complex number.
  2. The EnE_nEn​ are pairwise distinct. That is, n↦Enn \mapsto E_nn↦En​ is injective.
  3. Every eigenpair is one of the listed ones. Take any z∈Cz \in \mathbb{C}z∈C and any f:R→Cf : \mathbb{R} \to \mathbb{C}f:R→C with Eig(z,f)\mathrm{Eig}(z, f)Eig(z,f). Then there is an index nnn with z=Enz = E_nz=En​, so in particular zzz is real. For that nnn there is a constant c∈Cc \in \mathbb{C}c∈C with
f(x)=c un(x)for all x∈[a,b].f(x) = c\, u_n(x) \quad \text{for all } x \in [a, b].f(x)=cun​(x)for all x∈[a,b].

Nothing is asserted about fff outside [a,b][a, b][a,b], and c=0c = 0c=0 is not excluded. 4. The eigenvalues grow without bound in absolute value. ∣En∣→∞|E_n| \to \infty∣En​∣→∞ as n→∞n \to \inftyn→∞. 5. Orthonormality. For all m,nm, nm,n,

⟨um,un⟩={1m=n,0m≠n.\langle u_m, u_n \rangle = \begin{cases} 1 & m = n, \\ 0 & m \neq n. \end{cases}⟨um​,un​⟩={10​m=n,m=n.​
  1. Expansion of continuous functions. Let f:R→Cf : \mathbb{R} \to \mathbb{C}f:R→C be continuous on [a,b][a, b][a,b], with no condition on fff elsewhere. For each NNN put
SNf(x):=∑n=0N−1⟨un,f⟩ un(x)(x∈R).S_N f(x) := \sum_{n=0}^{N-1} \langle u_n, f \rangle\, u_n(x) \quad (x \in \mathbb{R}).SN​f(x):=n=0∑N−1​⟨un​,f⟩un​(x)(x∈R).

In each coefficient, unu_nun​ is the first argument of SLInner\mathrm{SLInner}SLInner and fff the second. The statement asserts

Re⁡⟨f−SNf, f−SNf⟩⟶0(N→∞).\operatorname{Re} \langle f - S_N f,\ f - S_N f \rangle \longrightarrow 0 \quad (N \to \infty).Re⟨f−SN​f, f−SN​f⟩⟶0(N→∞).

Only the real part is claimed to tend to 000. Nothing is said about the imaginary part. 7. Uniform convergence on the domain. For every f:R→Cf : \mathbb{R} \to \mathbb{C}f:R→C with Dom(f)\mathrm{Dom}(f)Dom(f), the partial sums SNfS_N fSN​f converge to fff uniformly on [a,b][a, b][a,b] as N→∞N \to \inftyN→∞.

Degenerate cases.

  • Whether the hypothesis is satisfiable. Whether RegularSL(p,q,r,a,b)\mathrm{RegularSL}(p, q, r, a, b)RegularSL(p,q,r,a,b) forces a<ba < ba<b, or can hold at all, is determined by that definition, which is not shown here.
    • If it permits a>ba > ba>b, then [a,b][a, b][a,b] is empty. The pointwise equality in clause 3 then holds trivially, with any ccc, and clause 7 holds trivially.
    • If it permits a=ba = ba=b, then [a,b]={a}[a, b] = \{a\}[a,b]={a} is a single point.
    • In either case the remaining clauses still demand infinitely many pairwise distinct EnE_nEn​ with ∣En∣→∞|E_n| \to \infty∣En​∣→∞, each satisfying Eig\mathrm{Eig}Eig and orthonormal for SLInner\mathrm{SLInner}SLInner.
  • Infinitely many eigenfunctions are always asserted. The statement asserts the existence of infinitely many such unu_nun​ for every admissible p,q,r,a,bp, q, r, a, bp,q,r,a,b and every choice of α,β\alpha, \betaα,β. If Eig\mathrm{Eig}Eig admitted no solutions for some admissible data, the asserted conclusion would fail for those data. The statement makes no exception for them.
  • The case N=0N = 0N=0. Here S0fS_0 fS0​f is the empty sum, which is the zero function.
  • Values of unu_nun​ outside [a,b][a, b][a,b]. These are constrained only insofar as IsSLEigenfunction\mathrm{IsSLEigenfunction}IsSLEigenfunction and SLInner\mathrm{SLInner}SLInner depend on them. The same holds for the values of fff outside [a,b][a, b][a,b] in clause 6.
  • Default values. The statement contains no division, natural-number subtraction, or explicit integral or supremum. Any default values that arise, for example from an integral of a non-integrable function, would sit inside SLInner\mathrm{SLInner}SLInner, IsSLEigenfunction\mathrm{IsSLEigenfunction}IsSLEigenfunction, SLDomain\mathrm{SLDomain}SLDomain or RegularSL\mathrm{RegularSL}RegularSL, whose code is not supplied here.
Human review
  • Endorsed by Shuze Chen · Oct 2, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 2, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me