ConceptComplete

Spectral Theory - Examples and Constructions

The resolvent and resolvent equation provide powerful tools for studying spectra and developing perturbation theory.

DefinitionResolvent Operator

For an operator TT and λρ(T)\lambda \in \rho(T) (the resolvent set), the resolvent is Rλ(T)=(TλI)1R_\lambda(T) = (T - \lambda I)^{-1}

The resolvent is a bounded operator for all λρ(T)\lambda \in \rho(T).

TheoremResolvent Equation

For λ,μρ(T)\lambda, \mu \in \rho(T), RλRμ=(μλ)RλRμR_\lambda - R_\mu = (\mu - \lambda) R_\lambda R_\mu

This identity is fundamental to analytic function theory of operators.

Proof

RλRμ=Rλ[(TμI)(TλI)]Rμ=(μλ)RλRμR_\lambda - R_\mu = R_\lambda[(T - \mu I) - (T - \lambda I)]R_\mu = (\mu - \lambda)R_\lambda R_\mu

TheoremAnalytic Dependence

The resolvent λRλ(T)\lambda \mapsto R_\lambda(T) is analytic on ρ(T)\rho(T) as an operator-valued function. Near any point λ0ρ(T)\lambda_0 \in \rho(T), it has a convergent power series expansion.

ExampleComputing Resolvents
  1. Multiplication Operator: For (Mϕf)(x)=ϕ(x)f(x)(M_\phi f)(x) = \phi(x) f(x) on L2L^2, Rλ(Mϕ)=M1/(ϕλ)R_\lambda(M_\phi) = M_{1/(\phi - \lambda)} defined for λRange(ϕ)\lambda \notin \text{Range}(\phi)

  2. Shift Operator: On 2\ell^2, for (Sx)n=xn+1(Sx)_n = x_{n+1}, Rλ(S)=n=0Snλn+1R_\lambda(S) = \sum_{n=0}^\infty \frac{S^n}{\lambda^{n+1}} for λ>1|\lambda| > 1

  3. Harmonic Oscillator: For H=d2dx2+x2H = -\frac{d^2}{dx^2} + x^2, the resolvent can be expressed using Hermite functions

DefinitionResolvent Set and Spectrum

The resolvent set ρ(T)={λC:(TλI) is bijective with bounded inverse}\rho(T) = \{\lambda \in \mathbb{C} : (T - \lambda I) \text{ is bijective with bounded inverse}\}.

The spectrum σ(T)=Cρ(T)\sigma(T) = \mathbb{C} \setminus \rho(T) decomposes into:

  • Point spectrum σp(T)\sigma_p(T): eigenvalues
  • Continuous spectrum σc(T)\sigma_c(T): TλIT - \lambda I injective, range dense but not closed
  • Residual spectrum σr(T)\sigma_r(T): TλIT - \lambda I injective, range not dense
TheoremStone's Formula

The spectral projections can be recovered from the resolvent via E((a,b])=limε0+12πiab[Rλ+iεRλiε]dλE((a,b]) = \lim_{\varepsilon \to 0^+} \frac{1}{2\pi i} \int_a^b [R_{\lambda + i\varepsilon} - R_{\lambda - i\varepsilon}] \, d\lambda

This formula connects the resolvent (analytic object) to the spectral measure (geometric object).

Remark

The resolvent approach to spectral theory is particularly powerful for perturbation theory. If T0T_0 is well-understood and VV is a "small" perturbation, one can analyze σ(T0+V)\sigma(T_0 + V) by studying how the resolvent (T0+VλI)1(T_0 + V - \lambda I)^{-1} differs from Rλ(T0)R_\lambda(T_0).

Resolvent methods are indispensable in scattering theory, where one analyzes continuous spectra of Schrödinger operators using analytic continuation of resolvents.