⚛️ quantum physics

A mathematical foundation for self-testing: Lifting common assumptions

This paper establishes a rigorous mathematical foundation for self-testing by proving a general theorem that removes common assumptions like projective measurements and pure states, while simultaneously identifying specific counterexamples where such assumptions are indispensable and revealing new limitations on quantum correlations.

Original authors: Pedro Baptista, Ranyiliu Chen, Jędrzej Kaniewski, David Rasmussen Lolck, Laura Mančinska, Thor Gabelgaard Nielsen, Simon Schmidt

Published 2026-04-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Pedro Baptista, Ranyiliu Chen, Jędrzej Kaniewski, David Rasmussen Lolck, Laura Mančinska, Thor Gabelgaard Nielsen, Simon Schmidt

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 a detective trying to figure out how a mysterious, locked box works. You can't open it, and you can't see inside. All you can do is ask the box questions (inputs) and listen to the answers it gives back (outputs).

This is the world of Self-Testing in quantum physics.

In this field, scientists want to know: "If a black-box quantum device gives us the perfect answers to a specific game, can we be 100% sure it is using a specific quantum state and specific measurements inside?"

For a long time, the detectives (scientists) had to make some big guesses to solve the case. They assumed:

The box contains a perfectly pure quantum state (like a pristine, unblemished crystal).
The box is full-sized (it uses all the space available inside).
The box uses projective measurements (a very specific, rigid type of measurement, like checking if a coin is Heads or Tails).

The problem? In the real world, quantum devices are messy. They might be mixed with "dirt" (noise), they might be smaller than the box allows, and they might use fuzzy, non-rigid measurements. The old rules said, "If your device isn't perfect, we can't prove what's inside."

This paper, "A mathematical foundation for self-testing: Lifting common assumptions," is like a master detective rewriting the rulebook. They ask: "Do we really need those perfect assumptions? Or can we prove what's inside even if the device is messy?"

Here is the breakdown of their findings using simple analogies:

1. The "Magic Elevator" (Lifting Assumptions)

The authors discovered a "magic elevator" (a mathematical tool) that allows them to take a messy, imperfect device and mathematically "lift" it up to a perfect one to analyze it, and then bring it back down.

The Old Way: "We can only prove the box works if it's a perfect crystal."
The New Way: "Even if the box is a muddy, half-empty, fuzzy-crystal, if it gives the right answers, we can mathematically prove it must be equivalent to the perfect crystal we know."

They proved that for most existing self-testing results, you don't need to assume the device is perfect. You can drop the assumptions about "purity" and "full size," and the proof still holds. This is a huge deal because it makes self-testing much more practical for real-world quantum computers.

2. The "One-Way Street" (Where Assumptions Stick)

However, the detectives also found a "One-Way Street." There are some specific, tricky situations where you cannot drop the assumptions.

They found a specific quantum "fingerprint" (a correlation) that acts like a chameleon.

If you look at it through a "perfect lens" (assuming the device is perfect), it looks like one thing.
If you look at it through a "messy lens" (allowing for real-world imperfections), it looks like something else entirely.

This is the first time anyone has found a quantum fingerprint that cannot be created by a perfect, full-sized device using rigid measurements. It's like finding a shadow that can only be cast by a specific, weirdly shaped object, but if you try to recreate that shadow with a perfect sphere, it's impossible. This proves that sometimes, you do need to be careful about what you assume.

3. The "Dictionary" (Unifying Definitions)

For years, different scientists used slightly different definitions for what counts as "proving" a device works. It was like everyone speaking a slightly different dialect of the same language.

Some said, "It's a match if the states look the same."
Others said, "It's a match if the measurements look the same."

This paper acts as a universal translator. They showed that in most natural situations, these different definitions are actually saying the exact same thing. They cleaned up the dictionary so everyone agrees on what "Self-Testing" means.

Why Does This Matter?

Imagine you are building a quantum internet. You need to be sure that the quantum devices you buy from a shady vendor are actually doing what they claim.

Before this paper: You had to trust the vendor that their device was "perfect" (pure, full-size, rigid). If it wasn't, your security check failed.
After this paper: You can check the device even if it's imperfect, noisy, or weird. You can say, "I don't care if your device is dirty or small; as long as it plays the game correctly, I know exactly what's inside."

The Big Takeaway

The authors essentially said: "We can stop making so many guesses."

They showed that for the vast majority of quantum self-testing scenarios, we can assume the quantum devices are as powerful and messy as possible (using mixed states and fuzzy measurements), and we can still prove exactly what they are doing. This makes the technology more robust, secure, and ready for the real world.

However, they also warned: "Be careful with the weird exceptions." There are rare, exotic quantum fingerprints where the old, strict assumptions are actually necessary, and ignoring them leads to wrong conclusions.

In short, they built a stronger, more flexible foundation for quantum certification, allowing us to trust our quantum devices even when they aren't perfect.

1. Problem Statement

Self-testing is a protocol in quantum information that allows a classical verifier to certify the internal state and measurements of untrusted quantum devices (provers) based solely on the statistical correlations (input-output behavior) they produce, typically within a nonlocal game framework.

Despite its widespread application in device-independent cryptography, delegated computation, and complexity theory (e.g., $MIP^* = RE$ ), the mathematical formalism of self-testing has historically relied on restrictive assumptions about the untrusted devices. Specifically, most existing self-testing theorems assume that the arbitrary strategy employed by the provers satisfies at least one of the following:

Purity: The shared quantum state is pure.
Full-Rank: The shared state has full Schmidt rank (the local supports span the entire Hilbert space).
Projectivity: The measurements are Projective Valued Measures (PVMs) rather than general Positive Operator-Valued Measures (POVMs).

The authors argue that these assumptions weaken the concept of self-testing. In a truly "black-box" scenario, an adversary could hold the purifying space of a mixed state or perform non-projective measurements. The paper addresses the following core questions:

Can these assumptions be lifted without loss of generality?
Are existing definitions of self-testing equivalent?
Which strategies (states and measurements) are fundamentally self-testable?

2. Methodology

The authors approach the problem from first principles, establishing a rigorous mathematical framework to analyze the relationship between arbitrary strategies and canonical (reference) strategies.

Key Mathematical Tools:

Local Dilations: The core concept of self-testing is formalized via local isometries. A strategy $S$ is a local dilation of $\tilde{S}$ if $S$ can be mapped to $\tilde{S}$ (plus an auxiliary state) via local isometries.
Support-Preserving Strategies: The authors introduce the concept of a strategy being "support-preserving," meaning the measurement operators map the state's support subspace to itself. They prove this property is invariant under local dilation.
Restriction and Naimark Dilation:
- Restriction: Mapping a strategy to the subspace spanned by the state's support (converting a non-full-rank strategy to a full-rank one).
- Naimark Dilation: Embedding a POVM strategy into a larger Hilbert space to convert it into a PVM strategy.
Invariance Analysis: The paper rigorously proves how properties like "projectivity" and "support-preserving" behave under these transformations (restriction and dilation).

Logical Structure:
The authors construct a hierarchy of self-testing definitions based on the assumptions made on the arbitrary strategy (the one being tested) and the canonical strategy (the reference). They use these tools to prove implications between different classes of self-tests (e.g., does a "Pure PVM Self-Test" imply an "Assumption-Free Self-Test"?).

3. Key Contributions

A. Lifting Common Assumptions (The Main Theorem)

The paper's primary contribution is Theorem 4.1, which establishes conditions under which common assumptions can be removed.

Result: If a canonical strategy $\tilde{S}$ $\tilde{S}$ is Full-Rank and Projective, then:
- A "Pure PVM Self-Test" implies an "Assumption-Free Self-Test."
- A "Pure Full-Rank Self-Test" implies an "Assumption-Free Self-Test."
Significance: This means that for the vast majority of existing self-testing results (which typically target pure, full-rank, projective strategies like the CHSH game), the proofs can be automatically upgraded to be assumption-free. The verifier does not need to assume the provers are using pure states or projective measurements; the self-test holds regardless.

B. Limits of Self-Testing (Counter-Examples)

The authors demonstrate that assumptions cannot always be lifted, specifically when the canonical strategy itself lacks certain properties.

Theorem 4.2: If a game is an assumption-free self-test, the canonical strategy must be 0-projective (projective on its support) and support-preserving.
Theorem B.2 (Counter-Example): The authors construct a specific quantum correlation $\tilde{p}$ $\tilde{p}$ that serves as a "Pure Full-Rank PVM Self-Test" for a specific strategy but fails to be an assumption-free self-test.
- This correlation can be realized by a full-rank strategy with non-projective measurements.
- It can also be realized by a non-full-rank strategy with projective measurements.
- Crucially, it cannot be realized by projective measurements on a full-rank state.
Implication: This is the first example of a quantum correlation that cannot be implemented by projective measurements on a full-rank bipartite state. It proves that one cannot generally restrict arbitrary strategies to be both full-rank and projective without losing generality.

C. Equivalence of Definitions

The paper compares various definitions of self-testing found in the literature (e.g., vector-form vs. matrix-form dilations, extraction vs. standard dilations).

Theorem 5.4 & 5.6: They prove that under natural settings (specifically for full-rank strategies), different definitions of local dilation are mathematically equivalent.
Separation: They show that for strategies that are not full-rank or not projective, these definitions diverge, leading to weaker notions of self-testing.

4. Key Results Summary

Scenario	Canonical Strategy $\tilde{S}$	Arbitrary Strategy Assumption	Result
Standard Case	Pure, Full-Rank, Projective	None (Assumption-Free)	Self-Testing Holds. (Theorem 4.1)
Restricted Case	Pure, Full-Rank, Projective	Pure + PVM	Self-Testing Holds. (Implies Assumption-Free)
Restricted Case	Pure, Full-Rank, Projective	Pure + Full-Rank	Self-Testing Holds. (Implies Assumption-Free)
General Case	Non-Projective or Non-Full-Rank	None	Self-Testing Fails. (Theorem 4.2 & B.2)
General Case	Non-Projective	None	Cannot be Self-Tested. (Theorem C)

Specific Findings:

Purity: Can be lifted if the canonical state is full-rank.
Projectivity: Can be lifted if the canonical strategy is full-rank.
Full-Rank: Can be lifted if the canonical strategy is projective.
Non-Projective Measurements: Cannot be self-tested in an assumption-free manner. If the canonical strategy uses non-projective measurements, no assumption-free self-test exists.
Non-Full-Rank States: Cannot be self-tested in an assumption-free manner unless the canonical strategy is restricted to its support (which effectively makes it full-rank).

5. Significance and Impact

Rigorous Foundation: The paper places self-testing on a solid mathematical footing, clarifying exactly what can and cannot be certified without prior knowledge of the device's internal structure.
Simplification of Proofs: By proving that "Pure PVM" or "Pure Full-Rank" self-tests automatically imply "Assumption-Free" self-tests (for standard canonical strategies), the authors allow researchers to bypass the complex task of proving self-testing for general mixed states and POVMs. Existing proofs can be "upgraded" with minimal effort.
Security Implications: In device-independent cryptography, security often relies on the assumption that an adversary cannot hold the purifying space of a mixed state. This work validates that for standard protocols (like CHSH-based randomness generation), the security holds even if the adversary holds the purifying space, provided the canonical strategy is full-rank and projective.
New Limits: The identification of a correlation that is a self-test only under specific assumptions, and one that cannot be realized by full-rank projective measurements, sets new boundaries for the field. It highlights that "assumption-free" is not a universal property but depends heavily on the target canonical strategy.

In conclusion, Baptista et al. provide a comprehensive "user manual" for self-testing, delineating the precise conditions under which the powerful tool of self-testing can be applied without restrictive assumptions, while simultaneously identifying the fundamental limits of the technique.