⚛️ quantum physics

The symplectic rank of non-Gaussian quantum states

This paper introduces the symplectic rank, a novel and computable non-Gaussianity monotone that quantifies the minimal number of modes required to compress non-Gaussian resources, provides rigorous lower bounds on gate and sample complexities, and establishes the irreversibility of non-Gaussianity resource theory under exact Gaussian operations.

Original authors: Francesco Anna Mele, Salvatore Francesco Emanuele Oliviero, Varun Upreti, Ulysse Chabaud

Published 2026-06-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Francesco Anna Mele, Salvatore Francesco Emanuele Oliviero, Varun Upreti, Ulysse Chabaud

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 understand the "complexity" of a quantum machine that uses light waves (bosonic systems) instead of tiny switches (qubits). Some of these machines are very simple and predictable; we call these Gaussian states. They are like a perfectly smooth, calm lake. Other machines are wild, chaotic, and full of surprises; these are non-Gaussian states, like a stormy ocean with crashing waves.

To build powerful quantum computers, we need that "stormy" non-Gaussian behavior. But how do we measure just how stormy a state is? And how do we compare a stormy state on a photonic chip to one on a superconducting circuit?

This paper introduces a new ruler called the Symplectic Rank. Think of it as a "Non-Gaussianity Score."

Here is what the authors discovered about this new ruler:

1. The "Compression" Analogy

Imagine you have a messy room (a complex quantum state) filled with clutter (non-Gaussianity). You want to tidy it up.

Gaussian operations are like a magical cleaning crew that can rearrange furniture, move boxes, and rotate the room, but they cannot throw anything away or create new mess.
The Symplectic Rank tells you the minimum number of shelves you need to keep all the "mess" (the non-Gaussianity) on, after the cleaning crew has done their best to pack everything else away into empty boxes (vacuum states).

If the rank is 1, all the complexity fits on one shelf. If the rank is 10, you need ten shelves. If the rank is 0, the room is perfectly clean (it's a Gaussian state).

2. The "No-Spreading" Rule (The No-Go Theorem)

The paper proves a strict rule about this mess: You cannot spread the mess out.

Imagine you have a single, very messy toy (a non-Gaussian state). You might think, "If I use my Gaussian cleaning tools, I can split this one messy toy into two slightly messy toys and spread the chaos around."

The Paper's Finding: No, you can't. The Symplectic Rank proves that it is mathematically impossible to take one non-Gaussian state and turn it into two non-Gaussian states using only Gaussian tools, even if you are allowed to be lucky and pick only the successful outcomes (post-selection).
The Consequence: This means you cannot "dilute" complexity. If you want to create a complex state, you have to build it from scratch; you can't just stretch a small amount of complexity to cover a larger area.

3. The "Irreversibility" of the Resource

Because you can't spread the mess, the process of making these states is irreversible.

Think of it like baking a cake. You can mix flour and eggs (resources) to make a cake (a complex state). But you cannot take that cake, use a Gaussian "undo" button, and turn it back into the exact same amount of flour and eggs.
The paper shows that the "cost" to make a complex state is higher than the "value" you get back if you try to break it down. The resource theory is broken; you lose efficiency in the process.

4. The "Difficulty" of Measuring and Simulating

The Symplectic Rank also acts as a difficulty meter for two tasks:

Tomography (Taking a picture of the state): The higher the rank, the harder it is to figure out what the state looks like. The paper shows that the number of samples (photos) you need to take grows exponentially with the rank. If the rank is high, you would need more photos than there are atoms in the universe to describe the state perfectly.
Classical Simulation (Computer modeling): If a quantum computer has a low Symplectic Rank, a regular classical computer can easily simulate it. But if the rank is high, the classical computer gets overwhelmed. This explains why some quantum computers are powerful: they operate at a high rank that classical computers cannot handle.

5. Robustness (The "Noise" Factor)

In the real world, everything is noisy. You might worry that a tiny bit of static (noise) would ruin your measurement of the Symplectic Rank.

The Paper's Finding: The Symplectic Rank is robust. It's like a sturdy mountain; a small earthquake (a tiny perturbation) won't change its height. If a state has a high rank, a slightly noisy version of that state will still have a high rank. This makes the measure practical for real experiments where perfect conditions don't exist.

6. The "Approximate" Version

Since real experiments are never perfect, the authors also created an "Approximate Symplectic Rank." This is like asking, "How close is this messy room to a clean one?"

They showed that even with this fuzzy, approximate version, the rules still hold: you still can't spread the mess, and you can still use it to benchmark how good different quantum machines are at creating complexity.

Summary

The paper introduces a new way to count the "complexity" of quantum light states. It proves that:

Complexity is concentrated: You can't spread it out using standard tools.
Complexity is irreversible: You can't easily turn it back into simple resources.
Complexity is hard to measure: The higher the complexity, the exponentially harder it is to describe or simulate with classical computers.
Complexity is stable: Small amounts of noise won't trick you into thinking a complex state is simple.

This tool allows scientists to compare different types of quantum computers (like those using light vs. those using superconductors) on a level playing field, regardless of the specific hardware they use.

Technical Summary: The Symplectic Rank of Non-Gaussian Quantum States

Problem Statement
Non-Gaussianity is a fundamental resource for achieving quantum advantages in bosonic platforms, yet comparing the non-classical capabilities of different physical platforms (e.g., photonics vs. superconducting circuits) remains challenging. Existing measures of non-Gaussianity, such as the negative volume of the Wigner function or the stellar rank, suffer from limitations: they may be sensitive to specific types of non-Gaussianity or lack a clear operational interpretation regarding state complexity. Furthermore, there is a need for a measure that is robust to experimental noise, computable, and capable of bounding the complexity of tasks like quantum state tomography and classical simulation.

Methodology and Definitions
The authors introduce the symplectic rank, a novel non-Gaussianity monotone, extending a concept originally defined for Gaussian states to arbitrary non-Gaussian quantum states.

Definition: For a pure state $\psi$ with a well-defined covariance matrix $V(\psi)$ , the symplectic rank $s(\psi)$ is defined as the number of symplectic eigenvalues of $V(\psi)$ that are strictly greater than 1 (the symplectic eigenvalues of the vacuum). For mixed states, it is defined via a convex roof construction.
Operational Characterization: The paper establishes a structure theorem (Theorem 29) linking the symplectic rank to $t$ -compressibility. A state has symplectic rank $s$ if and only if it can be transformed via a Gaussian unitary into a tensor product of an $s$ -mode state and vacuum states on the remaining modes. Crucially, the compressed state retains a mean photon number upper-bounded by the original state.
Technical Tools: The authors develop several independent technical tools, including:
- A block decomposition of Gaussian unitaries (Lemma 35), showing any Gaussian unitary can be decomposed into a passive unitary and unitaries acting on subsets of modes.
- Perturbation bounds relating the trace distance between states to the norm distance between their first moments and covariance matrices (Theorems 16-18, 62-64). These bounds are derived for general (possibly non-Gaussian) states, extending previous results limited to Gaussian states.
- Approximate measures: The introduction of the $\epsilon$ -approximate symplectic rank and symplectic fidelities to handle experimental imperfections.

Key Contributions and Results

Monotonicity and No-Go Theorems: The symplectic rank is proven to be non-increasing under post-selected Gaussian operations (Theorem 33). This leads to strong no-go theorems for state conversion. Specifically, it is impossible to deterministically or probabilistically convert a single-mode non-Gaussian pure state into a tensor product of two non-Gaussian pure states using Gaussian operations, even with post-selection. This holds regardless of the "distance" of the states from the Gaussian set, a result not detectable by Wigner negativity or stellar rank.
Irreversibility of Resource Theory: The paper proves that the resource theory of non-Gaussianity is irreversible under exact post-selected Gaussian operations (Theorem 40). By comparing the exact distillable rate and exact cost rate between specific Fock states (e.g., $|2\rangle$ and $|1\rangle^{\otimes 2}$ ), the authors show that the conversion rates are not reciprocal, a property that can be arbitrarily strong (Theorem 41).
Complexity Bounds:
- Tomography: The sample complexity of quantum state tomography for bosonic states scales exponentially with the symplectic rank (Theorem 44). If the rank is $O(1)$ , tomography is efficient; otherwise, it becomes intractable.
- Classical Simulation: The authors provide a classical randomized algorithm to estimate the overlap between a low-symplectic-rank state and a Gaussian state (Theorem 45). They demonstrate that bosonic circuits doped with a constant number of non-Gaussian layers (e.g., cubic phase gates or photon additions/subtractions) can be efficiently simulated classically (Theorems 50, 53).
Robustness: The symplectic rank is shown to be robust against small perturbations. Theorem 42 establishes that any pure state with symplectic rank $s$ is surrounded by a ball of states (in trace distance) that all possess a symplectic rank of at least $s$ , provided energy constraints are met. This allows for the definition of an $\epsilon$ -approximate symplectic rank, which serves as a monotone under Gaussian operations (Theorem 55).
Experimental Witnessing: A protocol is proposed to witness the symplectic rank experimentally by estimating symplectic fidelities via optimization over Gaussian unitaries (Lemma 59), making the measure practical for benchmarking different bosonic platforms.

Significance and Claims
The paper claims that the symplectic rank provides a unified, hardware-agnostic measure of non-Gaussianity that bridges resource theory and computational complexity. Its significance lies in:

Operational Interpretation: It quantifies the minimal number of modes required to "compress" all non-Gaussianity, offering a clear geometric and operational meaning.
Strong Constraints: It establishes stricter limitations on Gaussian state conversion than previously known measures, proving the irreversibility of non-Gaussianity resources.
Benchmarking Utility: Due to its robustness and connection to the covariance matrix (which is easier to measure than full state tomography), it serves as a practical tool for benchmarking the non-Gaussian capabilities of diverse bosonic quantum devices.
Complexity Characterization: It identifies the symplectic rank as a critical parameter determining the hardness of tomography and the feasibility of classical simulation for bosonic systems.

The authors position the symplectic rank as a foundational tool for understanding the resource theory of non-Gaussianity in the multi-mode setting, offering new insights into the interplay between non-Gaussianity, entanglement, and computational power.