🔢 mathematics

Observable sets for Schrödinger equations on combinatorial graphs

This paper establishes that observability for Schrödinger and heat equations on combinatorial graphs is governed by specific local arithmetic conditions rather than geometric thickness, proving that such sets are observable on one-dimensional lattices if and only if they satisfy these criteria, while demonstrating that positive density alone is insufficient for observability on discrete tori and that the complement of any finite set suffices in higher dimensions.

Original authors: Zhiqiang Wan, Heng Zhang

Published 2026-05-12

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zhiqiang Wan, Heng Zhang

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

Imagine you are trying to figure out what is happening inside a giant, invisible maze. The maze is made of a grid of points (like a checkerboard that goes on forever), and at every point, there is a tiny "wave" of energy dancing around. This is the Schrödinger equation on a combinatorial graph.

Your goal is observability: Can you place a few sensors on the grid to watch these waves, and from those limited views, reconstruct the entire state of the maze? If you can, the set of sensors is an "observable set."

This paper by Zhiqiang Wan and Heng Zhang explores a surprising twist: In the digital world of grids, how many sensors you have matters less than where they are placed relative to the numbers.

Here is the breakdown of their findings using everyday analogies:

1. The Digital vs. The Analog World

In the real, continuous world (like a smooth sheet of rubber), if you want to see the whole picture, you just need your sensors to be "thick" enough—meaning they cover a decent percentage of the area everywhere. If you have a net with holes, but the holes are small enough, you catch the fish.

But in the digital world (a grid of integers like 1, 2, 3...), the rules change. The authors found that even if your sensors cover a huge amount of space, you might still be blind if the sensors are arranged in a specific "rhythmic" pattern.

The Analogy:
Imagine trying to listen to a song by only hearing every second note.

Continuous World: If you listen to 50% of the song, you can usually figure out the rest.
Digital World: If you listen to every even number (2, 4, 6...), you might miss the entire melody if the song is built on a pattern that repeats every two beats. It doesn't matter how loud the even numbers are; the pattern itself hides the truth.

2. The "Local Arithmetic Condition" (The Secret Code)

The paper's biggest discovery is a rule called the Local Arithmetic Condition (LAC).

To be able to see the whole system, your sensors cannot just be "dense." They must be arithmetically diverse in every small neighborhood.

The Rule: If you look at any small block of numbers (say, from 100 to 110), the distances between your sensors in that block must be able to generate every integer when you add and subtract them.
The Failure: If your sensors are all multiples of 3 (3, 6, 9...), the "greatest common divisor" of their spacing is 3. This creates a blind spot. The system can hide a wave that vibrates perfectly between the multiples of 3, making it invisible to your sensors.

The Metaphor:
Think of the sensors as keys to a lock.

In the continuous world, having a lot of keys (density) is enough.
In the digital world, you need keys that can open every type of lock. If all your keys are "even-numbered" keys, you can never open an "odd-numbered" door, no matter how many even keys you have. The paper proves that if your sensors don't have the right "mathematical mix" locally, the system can hide a secret wave that vanishes exactly where your sensors are.

3. Heat vs. Waves

The authors also looked at heat spreading on the grid (the heat equation).

Surprise: The same "mathematical mix" rule applies to heat as it does to waves.
Why it matters: Usually, heat and waves behave very differently. But on a digital grid, the "arithmetic obstruction" (the bad arrangement of sensors) stops both of them from being observed. If your sensors are arranged in a "bad rhythm," you can't track the heat either.

4. High Dimensions: The "Diagonal" Trap

What happens if the grid is 2D or 3D?

The Bad News: You can have a set of sensors that covers 99% of the grid and still be blind.
The Example: Imagine a 2D grid. If you remove all the points where the x-coordinate equals the y-coordinate (the diagonal line), you have removed a "thin" line, but the remaining sensors cover almost everything.
The Result: The authors proved that for a free wave equation, this "almost everything" set is not observable. A wave can travel perfectly along that missing diagonal line, and your sensors (which are everywhere except the diagonal) will see nothing.
The Contrast: In the continuous world, removing a line usually doesn't stop you from seeing the whole picture. In the digital world, that missing line is a highway for invisible waves.

5. The Finite World (Discrete Tori)

Finally, they looked at grids that wrap around like a video game screen (a torus).

The Finding: Even if you have a huge number of sensors (positive density), you can still fail to observe the system if the grid size and sensor positions share a specific arithmetic relationship.
The Takeaway: As the grid gets bigger and bigger, you can still find "large" sets of sensors that are mathematically useless because of how the numbers align.

Summary

This paper tells us that in the digital realm of grids, geometry is not enough. You cannot just rely on having "enough" sensors. You must ensure that the numbering pattern of your sensors is complex enough to break any hidden rhythms the system might use to hide itself.

Continuous World: "Cover enough ground."
Digital World: "Cover the ground with the right mathematical rhythm."

If your sensors follow a simple, repeating arithmetic pattern (like only even numbers), the system can hide a wave that is perfectly invisible to you, no matter how long you watch.

Technical Summary: Observable Sets for Schrödinger Equations on Combinatorial Graphs

Problem Statement
This work investigates the observability of Schrödinger equations on combinatorial graphs, specifically focusing on the discrete setting where the underlying space is a graph $G = (V, E)$ . The dynamics are governed by the equation $i\partial_t u = Hu$ , where $H = -\Delta_{\text{disc}} + V$ is a discrete Schrödinger operator, and $\Delta_{\text{disc}}$ is the graph Laplacian. The central question is to characterize the subsets $E \subset V$ (observable sets) from which the initial state $u_0$ can be uniquely determined (up to a constant) via measurements of the solution restricted to $E$ over a time interval $[0, T]$ .

The authors contrast this discrete problem with the well-established theory in continuous Euclidean spaces ( $\mathbb{R}^d$ and $\mathbb{T}^d$ ). In the continuous setting, observability is primarily governed by geometric density conditions, specifically the "thickness" of the observation set (a quantitative version of the Logvinenko–Sereda uncertainty principle). The paper seeks to determine whether similar geometric criteria suffice in the discrete setting or if new, arithmetic obstructions arise.

Methodology
The authors employ a combination of spectral theory, functional analysis, and arithmetic combinatorics. Key methodological components include:

Hautus-Type Resolvent Criterion: The authors utilize the equivalence between observability and a resolvent estimate (Theorem 2.1, based on Miller [27]). Specifically, observability at some time is equivalent to the existence of constants $M, m > 0$ such that for all $\lambda \in \mathbb{R}$ and $f \in \ell^2(V)$ :
$\|f\|^2 \leq M \|(H - \lambda)f\|^2 + m \|f|_E\|^2.$
This reduces the dynamical problem to a static spectral problem involving the operator $H$ and the observation set $E$ .
Arithmetic Obstruction Analysis (1D Lattice): For the one-dimensional lattice $\mathbb{Z}$ , the authors introduce the Local Arithmetic Condition (LAC). They analyze the greatest common divisor (gcd) of differences between points in local neighborhoods of $E$ .
- Sufficiency: They prove that if $E$ satisfies LAC, no non-zero eigenfunction (or approximate eigenfunction) can vanish on $E$ . This is established via a uniform local estimate (Lemma 2.1) and a recurrence argument for the discrete Schrödinger equation.
- Necessity: If LAC fails, the authors construct "arithmetic quasimodes" (Lemma 2.3)—sequences of functions that are localized on the complement of $E$ and approximate eigenfunctions of the operator. These quasimodes demonstrate that the resolvent estimate fails, implying non-observability.
Comparison of Schrödinger and Heat Dynamics: The paper establishes a link between the observability of the Schrödinger equation and the heat equation ( $\partial_t u = -Hu$ ). Using interpolation inequalities and analytic properties of the semigroup generated by bounded self-adjoint operators, they show that for asymptotically constant potentials, the observability criteria for both equations are identical.
Finite Graph and Torus Analysis: For finite graphs (specifically discrete tori $\mathbb{T}^d_N$ ), the problem reduces to the non-existence of eigenfunctions vanishing on $E$ . The authors utilize the spectral structure of the discrete Laplacian on tori (characters) and group homomorphisms to construct high-density sets that fail observability due to arithmetic symmetries.
Exterior Observability: For higher-dimensional lattices $\mathbb{Z}^d$ , the authors use a $TT^*$ operator-norm argument combined with the finite rank of the projection onto the complement of $E$ to prove observability when the unobserved set is finite.

Key Contributions and Results

Characterization on $\mathbb{Z}$ (Theorem 1.1): For the one-dimensional lattice with asymptotically constant potentials, a set $E$ is observable at some (and thus any) time if and only if it satisfies the Local Arithmetic Condition (LAC).
- Definition: $E$ satisfies LAC if there exists $L \in \mathbb{N}^+$ such that for all $m \in \mathbb{Z}$ , the set of differences $\{s - s' : s, s' \in E \cap [m-L, m+L]\}$ has a greatest common divisor of 1.
- Significance: This reveals an arithmetic obstruction absent in Euclidean theory. While thickness is necessary, it is not sufficient. For instance, the set $9\mathbb{Z} \cup (9\mathbb{Z} + 3)$ has the same thickness as $9\mathbb{Z} \cup (9\mathbb{Z} + 1)$ but is not observable because its local differences generate a proper subgroup of $\mathbb{Z}$ .
Heat Equation Equivalence (Theorem 1.2): The same LAC criterion characterizes observability for the heat equation on $\mathbb{Z}$ . This contrasts with continuous settings where heat observability is often linked to different geometric properties, though here the boundedness of the generator allows the transfer of the Schrödinger result.
Higher-Dimensional Lattices (Theorem 1.3 & Example 1.1):
- In dimensions $d > 1$ , a complete arithmetic characterization analogous to LAC is not established.
- The authors show that high thickness does not imply observability in higher dimensions. They construct a set $E = \mathbb{Z}^2 \setminus \{(m, n) : m=n\}$ which is $\gamma$ -thick for any $\gamma < 1$ but is not observable for the free Schrödinger equation.
- However, they prove that the complement of any finite set is observable at any time in $\mathbb{Z}^d$ (Theorem 1.3).
Discrete Tori (Theorem 1.4):
- On the 1D discrete torus $\mathbb{T}^1_N$ , observability is characterized by a gcd condition involving the modulus $N$ and the differences of points in $E$ .
- In higher dimensions, the authors construct subsets of positive density (specifically, density approaching $1 - 2^{-d}$ ) that are not observable at any time. This contrasts sharply with continuous tori, where every non-empty open set is observable.

Significance and Claims
The paper claims to reveal a fundamental qualitative distinction between observability in continuous and discrete settings. While continuous observability is governed by geometric density (thickness), discrete observability is heavily constrained by the local arithmetic structure of the observation set.

Arithmetic vs. Geometric: The authors emphasize that in the discrete setting, "positive density alone does not ensure observability." The failure of observability can occur even for sets with high density if the set's structure aligns with the arithmetic symmetries of the graph Laplacian's spectrum.
Limitations of Euclidean Intuition: The results demonstrate that techniques relying solely on geometric uncertainty principles (like the Logvinenko–Sereda principle) are insufficient for discrete graphs. The discrete setting requires tools that account for the recurrence structure of the operator and the number-theoretic properties of the observation set.
Scope: The paper provides a complete characterization for 1D lattices and discrete tori in 1D, partial results for higher-dimensional lattices (exterior observability), and counterexamples showing the failure of simple density-based criteria in higher dimensions.

The authors do not propose new applications or experimental setups but frame their work as a theoretical advancement in the control theory of discrete systems, highlighting the unique role of arithmetic in discrete spectral analysis.