🔢 mathematics

An extension theory for fully fractional Schrödinger equations with memory

This paper develops an extension theory for the fully fractional Schrödinger operator with memory, constructing a local boundary value problem in an additional spatial variable to establish a well-posedness framework for nonlinear Schrödinger equations driven by prescribed past histories.

Original authors: Nicola Garofalo, Gigliola Staffilani

Published 2026-09-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Nicola Garofalo, Gigliola Staffilani

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the physical world, many systems do not simply react to the present moment; they carry the weight of their entire past. A piece of rubber that has been stretched for hours does not snap back with the same snap as one stretched for a second; its future behavior is dictated by its history. In mathematics, scientists use special tools called fractional operators to model these hereditary effects, where the current state is a blend of the present and everything that came before. Usually, when solving equations that describe how things change over time, scientists only need to know the starting point, like the position of a ball at the exact moment it is thrown. However, for systems with deep memory, a single starting point is not enough. One must prescribe the entire history of the system, the values it held at every moment in the past, to predict what happens next.

This paper tackles a particularly difficult version of this problem involving quantum waves, known as Schrödinger equations. While standard quantum equations describe how a wave evolves from a single initial moment, the version studied here is "fully fractional," meaning it is nonlocal in both space and time. This makes the equation oscillate in a complex way that defies the standard methods used for heat or diffusion, which rely on smooth, positive spreading. The researchers, Nicola Garofalo and Gigliola Staffilani, faced a challenge: how to solve a nonlinear equation where the future depends on a prescribed past, without getting lost in the infinite complexity of nonlocal interactions. They developed a new mathematical bridge that turns this difficult, memory-laden problem into a more manageable local one.

The core of their achievement is a technique called an extension theory. Instead of trying to solve the equation directly in the familiar four dimensions of space and time, the authors invented a way to lift the problem into a higher dimension. They added an extra spatial variable, effectively creating a fifth dimension, and showed that the complicated, memory-filled equation on the boundary of this new space is equivalent to a simpler, local equation inside the bulk of the space. In this new setting, the "memory" of the system is encoded as an initial condition in this extra dimension. The researchers constructed a specific tool, an oscillatory kernel, which acts like a translator. It takes the prescribed history of the system from the past and lifts it into this new space, creating a starting point for the local equation.

What makes this work distinct is the nature of the waves involved. In many physical models, like heat spreading, the mathematical tools used are positive and smooth. Here, because the underlying physics involves quantum waves that oscillate and change sign, the new tools the authors built are also oscillatory. They are not smooth hills of probability but rather rippling patterns that can be positive or negative. The authors proved that this oscillatory lifting preserves the energy of the system in a precise way. They showed that the energy contained in the lifted history in the higher-dimensional space is exactly equal to a specific energy measure calculated directly from the past history itself. This identity allowed them to define a natural space of all possible histories, ensuring that the mathematical framework is solid and complete.

With this linear foundation established, the team applied their method to the nonlinear version of the problem, where the wave interacts with itself. This is the scenario where the equation becomes significantly harder, as the wave's own shape influences its future evolution. By using the extension theory, they were able to transfer the problem into the higher-dimensional space where powerful existing tools for local equations could be used. They proved that for a wide range of conditions, a unique solution exists for the future evolution of the system, provided the past history is well-behaved. The solution is defined as the trace, or the shadow, of the higher-dimensional wave on the boundary, which corresponds to the physical world.

The paper explicitly clarifies the limits of this approach. The solution is not a single function that magically satisfies the equation at every point in time and space in the traditional sense. Instead, it is a "mild" solution, defined through this extension process. The authors are careful to state that while the solution behaves correctly for positive times, it does not necessarily satisfy the original fractional equation in a strict, pointwise sense unless the history possesses a certain extra level of smoothness. They also ruled out the idea that any arbitrary solution to the equation must come from this specific lifting process; their uniqueness result applies specifically to the class of solutions generated by their method.

The findings provide a rigorous well-posedness theory for these memory-driven quantum equations. This means that for a given past, there is one and only one future evolution that fits the model, and small changes in the past lead to small changes in the future. The authors identified specific thresholds for the strength of the nonlinear interaction where this stability holds. They also noted that the behavior of the system changes depending on the specific parameters of the fractional order, with a distinct shift in mathematical behavior occurring at a specific midpoint. This work does not just solve a specific equation; it establishes a new framework for understanding how systems with deep, nonlocal memory evolve, offering a clear path forward for analyzing complex dispersive phenomena that were previously out of reach.

Technical Summary: An Extension Theory for Fully Fractional Schrödinger Equations with Memory

Problem Statement
The paper addresses the well-posedness of nonlinear Schrödinger equations with memory, governed by the fully fractional operator $L_s = (\partial_t - i\Delta_x)^s$ for $0 < s < 1$ . Unlike classical fractional Schrödinger equations where the Cauchy problem is posed with an initial datum at a single time $t=0$ , the fully fractional operator is nonlocal in both space and time. Consequently, the natural Cauchy problem requires the prescription of a "history" $u_0(x, t)$ defined on the entire past interval $t \leq 0$ . The authors aim to develop a rigorous extension theory for $L_s$ to analyze the nonlinear problem:
$\begin{cases} L_s u(x, t) = \mu |u(x, t)|^{p-1}u(x, t), & x \in \mathbb{R}^d, t > 0, \\ u(x, t) = u_0(x, t), & x \in \mathbb{R}^d, t \leq 0. \end{cases}$
The primary challenge lies in the oscillatory nature of the operator. The Fourier symbol of $L_s$ changes sign across the characteristic paraboloid $\tau = -2\pi|\xi|^2$ , and the underlying Schrödinger semigroup is unitary rather than contractive. This precludes the use of standard positivity-based techniques (comparison principles, maximum principles) common in parabolic extension theories (e.g., Caffarelli-Silvestre for the fractional Laplacian or fractional heat equations).

Methodology
The authors employ an oscillatory extension theory that realizes the nonlocal operator $L_s$ as a Dirichlet-to-Neumann map of a local problem in an additional spatial variable $z > 0$ . The methodology proceeds through three main stages:

Operator Definition and Semigroup Representation:
The operator $L_s$ is defined via a semigroup representation analogous to Balakrishnan's formula for sectorial operators, adapted for the unitary Schrödinger group $S(t) = e^{it\Delta_x}$ :
$L_s u = -\frac{s}{\Gamma(1-s)} \int_0^\infty \frac{P^L_\sigma u - u}{\sigma^{1+s}} d\sigma,$
where $P^L_\sigma = e^{-\sigma L}$ . The authors prove this definition coincides with the Fourier multiplier definition $(ip(\xi, \tau))^s$ , where $p(\xi, \tau) = 2\pi\tau + 4\pi^2|\xi|^2$ . Crucially, the fractional power involves two distinct branches depending on the sign of $p(\xi, \tau)$ , necessitating a careful treatment of the phase in the extension.
Construction of the Oscillatory Extension:
The authors construct a local extension problem in the upper half-space $\mathbb{R}^{d+1}_+ \times \mathbb{R}_t$ governed by a singular Schrödinger equation involving a Bessel operator $B^{(a)}_z = \partial_{zz} + \frac{a}{z}\partial_z$ with $a = 1-2s$ :
$\partial_t U - i(\Delta_x U + B^{(a)}_z U) = 0, \quad U(x, 0, t) = u(x, t).$
Unlike elliptic or parabolic extensions, the associated Poisson kernel $P_a(X, Y_0, t)$ is oscillatory. The authors explicitly compute this kernel in phase space using modified Bessel functions $K_s$ and derive its physical space representation:
$P_a(X, Y_0, t) = C_a z^{1-a} t^{-(3-a)/2} e^{i z^2/4t} S(x, y, t),$
where $S(x, y, t)$ is the classical Schrödinger propagator. This kernel is shown to be oscillatory rather than positive, reflecting the dispersive nature of the underlying dynamics.
History Lifting and Intrinsic Energy Space:
A central contribution is the construction of a "Poisson lifting" for the prescribed history $u_0$ . The authors define an intrinsic Hilbert space $\mathcal{H}_s$ of admissible histories, characterized by an intrinsic quadratic form $Q_s(u_0)$ that represents the weighted bulk energy of the lifted history. They prove that the Poisson lifting operator $P_s: \mathcal{H}_s \to L^2_a(\mathbb{R}^{d+1}_+)$ is an isometry. This establishes a direct link between the energy of the past history and the energy of the initial data for the local extension problem.

Key Contributions and Results

Extension Theory for $L_s$ : The paper provides the first extension theory for the fully fractional Schrödinger operator, overcoming the lack of positivity by utilizing oscillatory techniques and Bessel function asymptotics.
Explicit Poisson Kernel: An explicit formula for the oscillatory Poisson kernel is derived, which serves as the bridge between the boundary history and the bulk solution.
Dispersive Transition at $s=1/2$ : The analysis reveals a qualitative dichotomy in the behavior of the extension operator depending on whether $0 < s \leq 1/2$ $0 < s \leq 1/2$ or $1/2 < s < 1$ $1/2 < s < 1$ .
- For $s \leq 1/2$ , the extension operator is bounded on $L^2(\mathbb{R}^{d+1})$ .
- For $s > 1/2$ , the operator is naturally defined on the graph domain $\text{Dom}(L_s)$ .
  This transition mirrors the dispersive behavior of the Bessel-Schrödinger propagator identified in the authors' previous works.
Well-Posedness of the Nonlinear Memory Problem: By lifting the history $u_0 \in \mathcal{H}_s$ $u_{0} \in H_{s}$ to the bulk initial datum $U_0 = P_s u_0$ $U_{0} = P_{s} u_{0}$ , the authors transfer the well-posedness theory for singular Schrödinger equations with nonlinear Neumann boundary interactions (developed in their companion paper [10]) to the nonlocal memory problem.
- Mild Solutions: The solution to the memory problem is defined as the strong boundary trace of the unique mild solution to the local extension problem.
- Regimes: Well-posedness is established in the mass-critical and subcritical regimes ( $1 < p \leq p_c$ ), where $p_c$ depends on the dimension $d$ and the parameter $a$ .
- Recovery of the Equation: Under additional graph-domain regularity hypotheses, the fractional equation $L_s u = \mu |u|^{p-1}u$ is recovered in the $L^2$ sense for $t > 0$ .

Significance and Claims
The paper claims to establish a rigorous framework for studying nonlinear dispersive equations with memory, a class of problems previously lacking a local extension theory due to the non-sectorial and oscillatory nature of the Schrödinger operator.

Methodological Shift: The work demonstrates that extension techniques can be successfully adapted to dispersive operators by replacing positivity arguments with oscillatory integral estimates and Bessel intertwining properties.
Intrinsic History Space: The identification of the intrinsic Hilbert space $\mathcal{H}_s$ provides a natural functional setting for history data, independent of the extension construction, characterized by a specific quadratic form.
Connection to Nonlocal Potentials: The authors show that the nonlinear boundary term in the extension formulation corresponds exactly to the fractional Duhamel potential associated with the fundamental solution of $L_s$ , thereby unifying the local and nonlocal perspectives.

The authors are modest about the scope of their uniqueness claims: uniqueness is asserted within the class of solutions generated by the extension lifting. They do not claim that all possible weak solutions to the fractional equation must arise from this construction, nor do they assert that the glued past-future function is continuous across $t=0$ without further compatibility conditions. The well-posedness is established in an "extension-induced mild sense," which is the natural solution concept given the nonlocal nature of the problem.

More like this