The initial-to-final-state inverse problem with critically-singular potentials

1. Problem Statement

The paper addresses an inverse problem for the time-independent Schrödinger equation in high dimensions ($n \geq 2$). The equation is given by:
$$\begin{cases} i\partial_t u = -\Delta u + V(x)u & \text{for } (t, x) \in (0, T) \times \mathbb{R}^n, \\ u(0, x) = f(x) & \text{for } x \in \mathbb{R}^n. \end{cases}$$
The central question is whether the initial-to-final-state map $U_T$, which maps an initial state $f \in L^2(\mathbb{R}^n)$ to the solution $u(T, \cdot)$ at a fixed time $T$, uniquely determines the potential $V(x)$.

Key Constraints and Context:

• Data: Only the map $U_T$ is known; no boundary measurements (Dirichlet-to-Neumann maps) are used.
• Potential Class: The authors focus on time-independent potentials $V$ that may exhibit critical singularities. Specifically, they consider $V \in L^1(\mathbb{R}^n) \cap L^q(\mathbb{R}^n)$, where:
  • $q > 1$ if $n=2$.
  • $q \geq n/2$ if $n \geq 3$.
• Comparison to Prior Work: Previous results for time-dependent potentials required super-exponential decay at infinity. Earlier work by the same authors on time-independent potentials allowed only polynomial decay and bounded potentials. This paper significantly relaxes the decay assumptions to $L^1$-integrability at infinity while allowing for $L^q$-type singularities (including the critical case $q=n/2$).

2. Methodology

The proof strategy relies on constructing specific test solutions (stationary states) and extending orthogonality relations beyond standard physical solutions.

A. Stationary State Construction

The authors construct special solutions of the form:
$$\psi(t, x) = e^{-i\lambda^2 t} \left( e^{-i\lambda \omega \cdot x} + w_{\text{cor}}(x) \right),$$
where $\lambda > 0$ is a large energy parameter, $\omega \in S^{n-1}$, and $w_{\text{cor}}$ is a correction term accounting for the potential $V$.

• Mechanism: The correction term is constructed by inverting the operator $(I - P_\lambda \circ V)$, where $P_\lambda$ is the solution operator for the Helmholtz equation $(\Delta + \lambda^2)u = f$.
• Challenge: For the potential to be "small" enough to invert this operator via Neumann series, the operator norm $\|P_\lambda \circ V\|$ must decay as $\lambda \to \infty$.
  • Non-endpoint case ($q > n/2$): Standard Kenig–Ruiz–Sogge (KRS) resolvent estimates provide the necessary decay in $\lambda$.
  • Critical endpoint case ($q = n/2$): Standard KRS estimates fail to provide decay in $\lambda$. To overcome this, the authors introduce a new scale of Banach spaces ($X_\lambda$ and its dual $X^*_\lambda$) that combine $L^p$ and $L^q$ norms with $\lambda$-dependent weights. They prove a refined resolvent estimate within this framework, recovering the necessary decay factor $\lambda^{-1/(n+1)}$.

B. Extended Orthogonality Relation

The core of the inverse problem proof involves an Alessandrini-type identity.

1. Physical Solutions: Initially, the equality $U_T^1 = U_T^2$ implies $\int_\Sigma (V_1 - V_2) u_1 v_2 = 0$ for physical solutions $u, v \in C([0, T]; L^2)$.
2. Extension to Stationary States: Since stationary states (constructed above) are not in $C([0, T]; L^2)$, the authors must extend the orthogonality relation.
   • They utilize integration by parts identities and density arguments (Proposition 4.2) to show the relation holds even when one or both test functions are the constructed stationary states, provided the potentials satisfy the $L^1 \cap L^q$ condition.
   • This allows them to test the orthogonality relation against the highly oscillatory stationary states $\psi_1, \psi_2$.

C. Uniqueness Argument

By substituting the stationary states into the extended orthogonality relation:
$$\int_\Sigma (V_1 - V_2) \psi_1 \psi_2 = 0,$$
and letting $\lambda \to \infty$, the oscillatory terms $e^{-i\lambda(\omega_1 - \omega_2) \cdot x}$ isolate the Fourier transform of the difference $F = V_1 - V_2$.

• The error terms involving the correction functions $w_{\text{cor}}$ are shown to vanish as $\lambda \to \infty$ due to the decay estimates derived from the resolvent bounds.
• This yields $\hat{F}(\xi) = 0$ for all $\xi$, implying $V_1 = V_2$ almost everywhere.

3. Key Contributions

1. Relaxation of Decay Assumptions: The paper proves uniqueness for potentials that only require $L^1$-integrability at infinity (instead of super-exponential decay) and allows for $L^q$ singularities (including the critical case $q=n/2$). This is a substantial improvement over previous results for both time-dependent and time-independent settings.
2. Refined Resolvent Estimates: The authors develop a novel refinement of the Kenig–Ruiz–Sogge resolvent estimate specifically for the critical endpoint $q=n/2$. By constructing a tailored scale of Banach spaces, they recover the necessary decay in the energy parameter $\lambda$ where classical estimates fail.
3. Avoidance of Complex Geometrical Optics (CGO): Unlike time-dependent inverse problems which often rely on CGO solutions (which grow exponentially and necessitate strong decay assumptions), this work utilizes time-harmonic solutions. The purely oscillatory nature of these solutions (unimodular leading terms) allows the authors to dispense with strong decay conditions at infinity.
4. Extension of Orthogonality Relations: The paper rigorously extends the Alessandrini identity to include solutions that are not in the standard energy space $C([0, T]; L^2)$, bridging the gap between physical solutions and the stationary states required for the inverse problem.

4. Main Results

Theorem 1.1: Let $V_1, V_2 \in L^1(\mathbb{R}^n) \cap L^q(\mathbb{R}^n)$ be time-independent potentials, where $q > 1$ if $n=2$ and $q \geq n/2$ if $n \geq 3$. Let $U_T^1$ and $U_T^2$ be the corresponding initial-to-final-state maps. If $U_T^1 = U_T^2$, then $V_1 = V_2$ almost everywhere.

5. Significance

This work represents a major advancement in the theory of inverse problems for the Schrödinger equation.

• Physical Relevance: By allowing $L^1$ decay and critical singularities, the results apply to a much broader class of physical potentials, including those with strong local singularities and slow decay at infinity, which are common in realistic quantum mechanical models.
• Methodological Innovation: The technique of refining resolvent estimates via new function space scales to handle critical endpoints offers a powerful tool that could be applied to other inverse problems involving critical singularities.
• Theoretical Bridge: It demonstrates that the time-independent setting is fundamentally more tractable regarding decay assumptions than the time-dependent setting, as it avoids the need for complex geometrical optics solutions that grow in space.

In summary, the paper establishes that the initial-to-final-state map is a complete invariant for a wide class of singular, time-independent potentials, provided the potential is integrable at infinity, thereby resolving a long-standing gap between bounded/polynomial decay results and the more general singular cases.