Magnetic uncertainty in variable geometry

1. Problem Statement

The paper investigates Hardy-type uncertainty principles and unique continuation properties for linear covariant Schrödinger equations and heat equations in the presence of variable coefficients (variable geometry) and electromagnetic potentials (magnetic and electric fields).

The core equation studied is the Cauchy problem for the variable-coefficient covariant Schrödinger equation:
$$\begin{cases} i\partial_t u + H_{A,G} u = 0, & (t, x) \in \mathbb{R} \times \mathbb{R}^d \\ u(x, 0) = u_0(x) \end{cases}$$
where the operator is defined as:
$$H_{A,G} u = \sum_{j,k=1}^d D_j (g_{jk} D_k u) + V(t, x)u$$
Here, $D_j = \partial_j - iA_j$ is the covariant derivative, $G = (g_{jk})$ is a real-valued, symmetric, uniformly elliptic matrix representing the variable metric, $A$ is the magnetic potential, and $V$ is the electric potential.

The Challenge:
Previous works had established uncertainty principles for:

1. Constant coefficients with magnetic fields (Barceló–Fanelli–Gutiérrez–Ruiz–Vilela).
2. Variable coefficients without magnetic fields (Federico–Li–Yu).

However, combining variable geometry ($G \neq I$) and magnetic effects ($A \neq 0$) creates significant difficulties due to the interaction between the metric and the magnetic structure. Specifically, the loss of translation invariance and the non-trivial commutation relations generated by the magnetic potential prevent the direct application of existing methods. The authors aim to determine if solutions exhibiting rapid exponential decay at two distinct times must vanish identically in this coupled setting.

2. Methodology

The proofs rely on a sophisticated combination of logarithmic convexity arguments and Carleman estimates, adapted to handle the specific difficulties of variable metrics and magnetic fields.

A. Parabolic Regularization

Since the initial data $u_0$ is only assumed to be in $L^2$, the authors introduce a Ginzburg–Landau type parabolic regularization:
$$\partial_t u = (a + ib)(\text{div}_A(G\nabla_A u) + V u + F)$$
This allows them to work with solutions in $H^1(\mathbb{R}^d)$, justifying the necessary weighted commutator computations.

B. Logarithmic Convexity

The authors establish the logarithmic convexity of the weighted $L^2$ norm $H(t) = \|e^{\gamma|x|^2} u(t)\|_{L^2}^2$.

• They decompose the operator into symmetric ($S$) and skew-symmetric ($A$) parts.
• A crucial step is computing the commutator $[S, A]$. In the variable-coefficient case, direct computation is intractable. Instead, they derive the commutator identity via integration by parts and virial identities, carefully tracking terms involving derivatives of the metric $G$ and the magnetic field $B$.
• They prove that under smallness assumptions on the derivatives of $G$ and the magnetic field, the quantity $St + [S, A]$ is bounded from below, ensuring the convexity of $\log H(t)$.

C. Carleman Estimates

To prove uniqueness, they derive Carleman estimates (exponentially weighted inequalities) of the form:
$$\|e^{-\beta \phi} w\|_{L^2} \leq C \|e^{-\beta \phi} P w\|_{L^2}$$

• Weight Functions: A major innovation is the design of the weight function $\phi$.
  • For super-quadratic decay (Theorem 1.3), they use a space-time separated weight: $\phi \sim \mu |x/R|^2 + \mu^{2-\ell}\phi(t)$. This avoids the difficulties of mixed space-time weights in variable metrics.
  • For quadratic decay (Theorem 1.4, 1.5), they introduce a space-time mixed weight: $\phi \sim \mu |x + Rt(1-t)e_1|^2 - \dots$. This weight includes a correction term $N(\varepsilon_0 + \varepsilon_1)\mu^2$ specifically designed to absorb the "bad" terms arising from the interaction between the metric $G$ and the magnetic field $B$.
• Commutator Control: The proofs involve estimating commutator terms generated by the weight and the operator. The authors show that under specific structural assumptions on $G$ (Hypothesis 1.2), many problematic terms vanish or can be absorbed by the coercive parts of the operator.

3. Key Assumptions

The results hold under smallness assumptions on the coefficients:

1. Metric Smallness ($\varepsilon_0$): The derivatives of the metric $G$ must decay sufficiently fast at infinity (specifically, $|\nabla^k G| \lesssim \langle x \rangle^{-3}$).
2. Magnetic Smallness ($\varepsilon_1$): The magnetic field $B$ must be small in a specific weighted norm involving the metric: $\|(Ge_1)^\top B\|_{L^\infty} \leq \varepsilon_1$.
3. Structural Assumption (Hypothesis 1.2): For the sharp quadratic decay result, the metric $G$ must have a specific block structure where the perturbation from the identity matrix only affects the subspace orthogonal to a specific direction $e_1$. This allows for the cancellation of critical commutator terms.

4. Main Results

Theorem 1.3: Super-Quadratic Decay

If a solution $u$ exhibits super-quadratic exponential decay at two distinct times $t=0$ and $t=1$ (i.e., $e^{\sigma|x|^{2+\vartheta}}u \in L^2$ for $\vartheta > 0$), then $u \equiv 0$.

• Significance: This result holds under mild smallness assumptions on both the metric and the magnetic field without requiring the special structural form of $G$.

Theorem 1.4: Quadratic Decay (Schrödinger Case)

Under the structural assumption on $G$ (Hypothesis 1.2) and smallness of the magnetic field, if $u$ exhibits quadratic exponential decay ($e^{\gamma|x|^2}u \in L^2$) with $\gamma > \frac{1}{2} + \frac{N(\varepsilon_0 + \varepsilon_1)}{16}$, then $u \equiv 0$.

• Significance: This recovers the sharp Hardy uncertainty principle (analogous to the constant coefficient case) in the presence of variable geometry and magnetic fields. The threshold for $\gamma$ depends explicitly on the size of the metric and magnetic perturbations.

Theorem 1.5: Quadratic Decay (Heat Equation)

For the variable-coefficient heat equation with magnetic perturbations, if $e^{\gamma|x|^2}u \in L^2$ at $t=0, 1$ with $\gamma > \frac{1}{2} + \frac{N\varepsilon_1}{16}$, then $u \equiv 0$.

• Significance: Notably, the metric variation parameter $\varepsilon_0$ does not appear in the threshold for the heat equation. The dissipative structure of the parabolic operator provides sufficient coercivity to absorb metric-related terms, whereas the Schrödinger equation lacks this property. However, the magnetic term $\varepsilon_1$ remains critical in both cases due to its antisymmetric nature.

5. Significance and Contributions

1. Unification: This is the first work to provide a unified treatment of uncertainty principles for Schrödinger and heat equations with both variable metrics and magnetic fields. It bridges the gap between the constant-coefficient magnetic theory and the variable-coefficient non-magnetic theory.
2. New Commutator Estimates: The paper develops refined commutator estimates that handle the interaction between the variable metric and the magnetic potential. This requires new techniques beyond simple combinations of existing arguments.
3. Adapted Weight Functions: The introduction of specific space-time separated and mixed weight functions allows the authors to overcome obstructions that arise when trying to apply standard Carleman estimates to variable-coefficient covariant flows.
4. Sharpness: The results recover known sharp constants in the limiting cases ($G=I$ or $A=0$) and provide explicit quantitative bounds on how the uncertainty threshold degrades as the geometry and magnetic field become non-trivial.
5. Parabolic vs. Dispersive: The paper highlights a fundamental difference between the Schrödinger and Heat equations: the heat equation's dissipative nature makes it more robust against metric perturbations, while the Schrödinger equation requires stricter control on the metric variation to maintain uniqueness at the quadratic scale.

In conclusion, the paper establishes that Hardy-type uncertainty principles are stable under small geometric and magnetic perturbations, providing a robust framework for studying unique continuation in complex, variable-coefficient dispersive and parabolic systems.