Certification of the genuine resolution of photon number resolving detectors

This paper introduces an operational framework and a scalable protocol based on coherent state probes to certify the genuine photon-number resolution of detectors, which is demonstrated by achieving four-outcome resolution on a 28-pixel superconducting nanowire single-photon detector.

The Gist

Imagine you have a high-tech camera that claims it can count exactly how many tiny particles of light (photons) hit it at once. The manufacturer says, "This camera can tell the difference between 1 photon, 2 photons, 3 photons, and so on, up to 10!"

But here's the problem: How do you know they aren't just lying? Or worse, how do you know they aren't using a cheap, simple camera that just says "Yes, I saw something" or "No, I saw nothing," and then using a computer to guess how many photons there were?

This paper introduces a new way to test these light-counting cameras to see if they are genuinely good at counting, or if they are just faking it with a simple trick.

The Core Problem: The "Fake Expert"

Think of a "Photon-Number-Resolving" (PNR) detector like a judge in a game show.

• The Real Expert: Can look at a pile of apples and say, "That's exactly 4 apples."
• The Fake Expert: Can only tell if there are any apples or no apples. But, they have a cheat sheet (classical post-processing). If they see any apples, they flip a coin and guess, "I think it's 4!"

If the Fake Expert gets lucky enough times, they might look like a Real Expert. The paper asks: How can we prove the detector is actually doing the hard work of counting, rather than just guessing?

The Solution: A "Guessing Game"

The authors created a simple test, like a game of "20 Questions," to catch the fakes.

1. The Setup: A "referee" (the light source) sends a specific amount of light to the detector. The referee knows exactly how much light they sent (like sending 1, 2, or 3 photons).
2. The Challenge: The detector looks at the light and gives an answer (e.g., "I see 2 photons!").
3. The Score: The referee checks: "Did the detector guess the right amount?"
   • If the detector is a Real Expert, it will get the answer right most of the time.
   • If the detector is a Fake Expert (just guessing or using a simple binary "on/off" sensor), it will fail to distinguish between the different amounts of light often enough.

The paper proves mathematically that if a detector scores high enough in this game, it must be genuinely capable of distinguishing between different numbers of photons. It cannot be faked by a simpler device.

The "Efficiency" Hurdle

The paper also discovered a crucial rule: You can't be a good counter if you are too lazy (or too lossy).

Imagine the detector is a person trying to count apples in a dark room. If the room is very dark (low efficiency), they might miss half the apples. Even if they are a genius, they can't count accurately if they can't see the apples.

The authors calculated that to count up to a certain number of photons accurately, the detector needs to be extremely efficient (catching almost every photon). If the detector loses too many photons, it physically cannot tell the difference between, say, 4 photons and 5 photons, no matter how smart its software is.

The Real-World Test

The team tested this theory on a real, cutting-edge detector made of 28 tiny superconducting wires (think of it as a 28-pixel camera).

• The Claim: The device could distinguish between different numbers of photons.
• The Test: They shone different amounts of laser light at it and ran the "Guessing Game."
• The Result: They proved the detector could genuinely distinguish between 4 different outcomes (e.g., telling the difference between 0, 1, 2, or 3+ photons) with high confidence. They also calculated that the detector was about 77% to 85% efficient, meaning it was catching most of the photons, which is why it passed the test.

Why This Matters

Before this paper, there was no standard, simple way to verify if a fancy light-counting device was actually doing what it claimed. Manufacturers could claim "10-photon resolution," but buyers had no practical way to check if it was true or just a software trick.

This new method is like a driver's license test for these detectors. It doesn't require taking the engine apart (which is complex and expensive); it just requires a simple driving test (the guessing game with light) to prove the driver (the detector) actually knows how to drive (count photons).

In short: The paper gives us a simple, reliable way to ask, "Are you really counting the light, or are you just guessing?" and provides the math to prove the answer.

Technical Summary

Technical Summary: Certification of the Genuine Resolution of Photon Number Resolving Detectors

Problem Statement
Photon-number-resolving (PNR) detectors are critical components for photonic quantum technologies, including quantum communication, sensing, and computing. Despite their importance, there is currently no practical metric to certify how many photons a detector can genuinely resolve in a single measurement cycle. Existing characterization methods, such as quantum measurement tomography, are often impractical due to the prohibitively large number of probe states and complex setups required. Furthermore, a critical distinction exists between a detector that intrinsically resolves photon numbers and a low-resolution device (e.g., a binary click/no-click detector) followed by complex classical post-processing. Current literature lacks a practical figure of merit to distinguish between these two scenarios, particularly in the low photon-number regime essential for quantum applications.

Methodology
The authors introduce an operational framework to quantify the "genuine resolution" of a PNR detector. This concept is defined independently of the number of output labels a device produces. Instead, it relies on the theory of quantum measurements to determine if a device's output can be simulated by a coarser detector (with fewer outcomes) followed by classical post-processing.

• Definition of Genuine Resolution: A measurement has a genuine resolution of at least $R+1$ if it cannot be simulated by any measurement with at most $R$ outcomes. This is formalized using Positive Operator-Valued Measures (POVMs) and tested via linear programming.
• Subspace Resolution: Recognizing that applications often focus on low photon numbers, the authors define "genuine subspace resolution" within a subspace $H_m$ spanned by Fock states $|0\rangle$ to $|m\rangle$.
• Theoretical Bounds: The authors derive lower bounds on the detection efficiency ($\eta_{th}$) required for an ideal detector to achieve a specific genuine resolution $R$ within a subspace $m$. They demonstrate that loss severely limits genuine resolution, with required efficiency increasing as the ratio of resolution to subspace dimension increases.
• Certification Protocol: To avoid full tomography, the authors propose a scalable, prepare-and-measure protocol based on a "guessing game."
  • Setup: A source prepares coherent states with varying intensities $\mu_x$ (chosen from a set of $N$ inputs).
  • Task: The detector produces an outcome $b$, and the goal is to guess the input intensity $x$.
  • Metric: The score is the average guessing probability, $P_{guess}$.
  • Witness: If the observed $P_{guess}$ exceeds a theoretical bound derived for an $R$-outcome simulatable detector, the device is certified to have a genuine resolution of at least $R+1$.
  • Models: The protocol is analyzed under two models: a "trusted" scenario (where source intensities are perfectly known) and an "untrusted" scenario (where only an upper bound on intensity is known), making the method robust against source calibration errors.

Key Contributions

1. Operational Definition: The paper establishes a rigorous, device-agnostic definition of genuine resolution based on simulatability, distinguishing intrinsic resolution from post-processed coarse measurements.
2. Efficiency Thresholds: The authors derive and tabulate the threshold efficiencies required to achieve specific genuine resolutions in various photon-number subspaces, highlighting the fundamental scaling challenges imposed by optical loss.
3. Scalable Certification Protocol: A practical protocol is developed that requires only $O(m)$ coherent state probes (compared to $O(m^2)$ for tomography) to certify lower bounds on genuine resolution.
4. Effective Efficiency Benchmark: The protocol allows for the extraction of an "effective efficiency" ($\eta^*_m$), which subsumes all device-specific imperfections (including combinatorial ambiguity in multiplexed arrays) into a single, subspace-resolved figure of merit.

Experimental Results
The method was experimentally demonstrated using a 28-pixel photon-number-resolving superconducting nanowire single-photon detector (PNR-SNSPD).

• Setup: The detector was probed with $N=7$ coherent states with mean photon numbers ranging from 0 to $\approx 8$.
• Data Acquisition: The system recorded $\approx 2.5 \times 10^5$ non-zero detection events per state, integrating pulse areas to distinguish outcomes.
• Certification:
  • In the subspace $m=1$ (0 and 1 photon), the device was certified to have a genuine resolution of $R=2$.
  • In the subspace $m=3$, a genuine resolution of $R=3$ was certified.
  • In the subspace $m=8$, the device achieved a genuine resolution of $R=4$ in the trusted model (statistical significance $\sim 21\sigma$). In the untrusted model, the observed score did not exceed the bound for $R=4$, though it exceeded the bound for $R=3$.
• Effective Efficiency: The certified effective efficiency was found to be between 82% and 85% for lower subspaces ($m \le 5$), dropping to approximately 75% for higher subspaces. This is consistent with the detector's nominal single-photon efficiency of $\sim 85\%$ and expected losses in the multiplexed architecture.

Significance
The paper claims to fill a crucial gap in the characterization of photonic quantum devices by providing an operational benchmark for PNR detectors. The approach is significant because it:

• Offers a practical, scalable alternative to full measurement tomography.
• Provides a clear distinction between genuine high-resolution detection and classical post-processing of low-resolution data.
• Highlights the critical role of detector efficiency in achieving high genuine resolution, revealing that loss is a fundamental limiting factor.
• Enables direct comparison between different detector architectures (e.g., multiplexed SNSPDs vs. transition-edge sensors) via the effective efficiency metric.

The authors emphasize that their work provides a necessary tool for vendors and clients to verify claims regarding photon counting capabilities, ensuring that reported resolutions reflect intrinsic device performance rather than algorithmic post-processing.