Skip to content

Schrödinger Equation in Hilbert Space

The equation iℏψ˙=Hψi\hbar\dot\psi=H\psi can mean several different analytic statements. For a self-adjoint autonomous Hamiltonian, every Hilbert-space initial vector has a continuous unitary evolution, while only domain vectors have a Hilbert-space derivative. The distinction determines what an initial-value problem asserts. The general physical formulation is developed in Time-Dependent Schrödinger Equation; here the operator HH is independent of time.

Helpful background. Stone’s Theorem supplies the unitary group and its exact derivative domain. Quadratic Forms distinguishes finite form energy from membership in the operator domain.

Let HH be self-adjoint on D(H)D(H) in a complex Hilbert space, and set U(t)=e−itH/ℏU(t)=e^{-itH/\hbar}. A strong solution on an interval II containing zero is a function

ψ∈C1(I;H),ψ(t)∈D(H),\psi\in C^1(I;\mathcal H),\qquad \psi(t)\in D(H),

satisfying the equation in H\mathcal H at every time and the prescribed initial condition. The equation then also makes t↦Hψ(t)t\mapsto H\psi(t) continuous.

For ψ0∈D(H)\psi_0\in D(H), Stone’s theorem gives the strong solution ψ(t)=U(t)ψ0\psi(t)=U(t)\psi_0. Functional calculus gives

U(t)D(H)=D(H),HU(t)ψ0=U(t)Hψ0.U(t)D(H)=D(H),\qquad HU(t)\psi_0=U(t)H\psi_0.

In particular, the solution is continuous in the graph norm as well as in the Hilbert norm. These assertions include the boundary conditions encoded in D(H)D(H); a formal pointwise derivative of a wavefunction cannot replace them.

For an initial vector outside D(H)D(H), the orbit has no strong derivative at any time. Indeed, a derivative at time t∗t_* would, after multiplication by U(−t∗)U(-t_*), give a derivative at zero. This does not mean that the orbit has no physical or mathematical meaning.

Weak solutions test the equation against domain vectors

Section titled “Weak solutions test the equation against domain vectors”

A weak solution here means a norm-continuous ψ:I→H\psi:I\to\mathcal H such that, for every fixed ϕ∈D(H)\phi\in D(H), the scalar function ⟨ϕ,ψ(t)⟩\langle\phi,\psi(t)\rangle is continuously differentiable and

iℏddt⟨ϕ,ψ(t)⟩=⟨Hϕ,ψ(t)⟩.i\hbar\frac{d}{dt}\langle\phi,\psi(t)\rangle =\langle H\phi,\psi(t)\rangle.

The inner product is antilinear in its first argument. The Hamiltonian acts on the test vector, so the right side is meaningful without assuming ψ(t)∈D(H)\psi(t)\in D(H).

Every unitary orbit is a weak solution. To check this, write

⟨ϕ,U(t)ψ0⟩=⟨U(−t)ϕ,ψ0⟩.\langle\phi,U(t)\psi_0\rangle =\langle U(-t)\phi,\psi_0\rangle.

The vector in the first slot is differentiable because ϕ∈D(H)\phi\in D(H). Differentiating and using antilinearity produces exactly the displayed weak equation. No derivative of U(t)ψ0U(t)\psi_0 is needed.

The test domain matters. Testing only against functions supported away from a boundary can lose the boundary condition that selects the Hamiltonian. Different self-adjoint extensions may solve the same interior differential equation. The weak formulation above refers to the complete specified operator domain, not just to its formal expression.

The integrated equation has a domain statement too

Section titled “The integrated equation has a domain statement too”

For every ψ0∈H\psi_0\in\mathcal H, the norm-continuous orbit has a Bochner integral

w(t)=∫0tU(s)ψ0 ds.w(t)=\int_0^t U(s)\psi_0\,ds.

Although U(s)ψ0U(s)\psi_0 need not belong to D(H)D(H), the integral does. Spectral calculus proves this using the bounded multiplier

gt(λ)=∫0te−isλ/ℏ ds,λgt(λ)=iℏ(e−itλ/ℏ−1).g_t(\lambda)=\int_0^t e^{-is\lambda/\hbar}\,ds, \qquad \lambda g_t(\lambda) =i\hbar(e^{-it\lambda/\hbar}-1).

The second multiplier has absolute value at most 2ℏ2\hbar, so gt(H)ψ0∈D(H)g_t(H)\psi_0\in D(H). Consequently

ψ(t)=ψ0−iℏH∫0tψ(s) ds.\psi(t)=\psi_0-\frac{i}{\hbar}H\int_0^t\psi(s)\,ds.

This is an integrated solution identity. It is generally invalid to replace H∫ψH\int\psi by ∫Hψ\int H\psi without ensuring that the latter integrand exists. The autonomous orbit is also called the mild solution. Existence and Uniqueness of Time Evolution develops the corresponding uniqueness and forced-evolution statements.

A discontinuous initial state still evolves

Section titled “A discontinuous initial state still evolves”

Use dimensionless units and let H=−d2/dx2H=-d^2/dx^2 on D(H)=H2(R)D(H)=H^2(\mathbb R). The normalized square pulse ψ0=1[0,1]\psi_0=\mathbf1_{[0,1]} belongs to L2(R)L^2(\mathbb R) but not to H1(R)H^1(\mathbb R). With the unitary Fourier convention its momentum-space wavefunction is

ψ^0(k)=12π1−e−ikik,\widehat\psi_0(k) =\frac{1}{\sqrt{2\pi}}\frac{1-e^{-ik}}{ik},

with its continuous value at k=0k=0. Multiplying this function by k2k^2 does not give an L2L^2 function, so ψ0∉D(H)\psi_0\notin D(H). In fact its mean kinetic energy is infinite. The square pulse is therefore an idealized state, while smooth approximations can have finite energies that grow as the edges sharpen.

Its free evolution is a perfectly defined weak and mild solution but is never a strong solution in L2L^2. Unitary Schrödinger evolution preserves the spectral Sobolev regularity; it does not acquire the global smoothing of a heat semigroup. Pointwise or local dispersive regularity statements are different questions and do not contradict this domain invariance.

Norm is conserved for every initial vector. Likewise, its entire energy spectral probability distribution is conserved, so every initially finite energy moment remains finite and unchanged. Energy conservation should not be written as an inner product involving HψH\psi when only the form expectation exists. See Conserved Quantities for the spectral formulation.

  1. Determine gt(0)g_t(0) in the integrated equation and explain what happens to a vector in ker⁡H\ker H.
Solution

The integral definition gives gt(0)=tg_t(0)=t, also the continuous limit of the quotient away from zero. For Hψ0=0H\psi_0=0, w(t)=tψ0w(t)=t\psi_0 and Hw(t)=0Hw(t)=0, so the integrated equation gives the stationary vector ψ(t)=ψ0\psi(t)=\psi_0.

  1. For ψ0∈D(H2)\psi_0\in D(H^2), compute the second time derivative and show that ∥Hψ(t)∥\|H\psi(t)\| is conserved.
Solution

Domain invariance permits a second differentiation, giving ψ¨(t)=−H2U(t)ψ0/ℏ2\ddot\psi(t)=-H^2U(t)\psi_0/\hbar^2. Since HU(t)ψ0=U(t)Hψ0HU(t)\psi_0=U(t)H\psi_0 and U(t)U(t) is unitary, ∥Hψ(t)∥=∥Hψ0∥\|H\psi(t)\|=\|H\psi_0\|. The latter equality only requires ψ0∈D(H)\psi_0\in D(H); the second derivative needs the stronger hypothesis.

  • B. C. Hall, Quantum Theory for Mathematicians, Graduate Texts in Mathematics 267, Springer, 2013, publisher edition — unbounded operators and quantum evolution.
  • G. Teschl, Mathematical Methods in Quantum Mechanics: With Applications to Schrödinger Operators, 2nd ed., Graduate Studies in Mathematics 157, American Mathematical Society, 2014, Section 5.1 and Chapter 7, author’s text — strong autonomous solutions and the free Schrödinger operator.