Experimental subdiffraction source discrimination enabled by spatial demultiplexing and single-photon detectors

This paper experimentally demonstrates that spatial mode demultiplexing (SPADE) combined with single-photon detectors enables robust, parameter-independent discrimination of faint asymmetric sources beyond the diffraction limit, significantly outperforming direct imaging in photon-starved regimes even under realistic modal crosstalk.

The Gist

Imagine you are trying to spot a tiny, dim firefly hovering very close to a massive, blindingly bright streetlamp. In the world of light and optics, this is incredibly difficult. The "glare" from the streetlamp usually washes out the firefly, making it impossible to tell if the firefly is actually there or just a trick of the light. This is the classic problem of diffraction: light waves naturally spread out, blurring two close objects into a single, messy blob.

This paper presents a clever new way to solve that problem, not by building a better camera, but by changing how we listen to the light.

The Old Way: The Blurry Photo

Think of traditional imaging (like a standard camera) as taking a photo of the streetlamp and firefly. Because of the physics of light, the photo comes out blurry. You see a big bright spot with a tiny, indistinct smudge next to it. To figure out if the smudge is a real firefly, you have to guess based on how much the blur changes. This method is slow and often fails, especially when the firefly is very dim compared to the lamp.

The New Way: The "Sound Mixer" (SPADE)

The researchers used a technique called SPADE (Spatial Mode Demultiplexing). Imagine instead of taking a photo, you have a magical sound mixer that can separate a complex song into its individual instruments.

In this experiment, the "song" is the light coming from the streetlamp and the firefly. The "instruments" are different shapes of light waves (called spatial modes).

• The Streetlamp (Star): Its light mostly fits into one specific shape (let's call it the "Round Shape").
• The Firefly (Exoplanet): Because it's slightly offset, its light creates a tiny bit of a different shape (the "Wobbly Shape").

The SPADE device acts like a prism for shapes. It splits the incoming light into two buckets:

1. Bucket A: Catches the "Round Shape" (mostly the star).
2. Bucket B: Catches the "Wobbly Shape" (where the firefly's presence would show up).

If the firefly is there, some photons (particles of light) will land in Bucket B. If the firefly is not there, Bucket B should be empty. By counting the photons in Bucket B, the researchers can detect the firefly with much higher precision than a blurry photo ever could.

The Real-World Problem: The "Leaky Bucket"

In a perfect world, the buckets would be perfectly sealed. But in real life, the device has a flaw called crosstalk. This is like having a leaky bucket: sometimes a photon meant for the "Round Shape" bucket accidentally leaks into the "Wobbly Shape" bucket.

If the leak is too big, you might think you heard a firefly (a false alarm) when it's just a leak from the streetlamp. Previous theories suggested that if the leak was too big, this fancy new method wouldn't work at all.

The Big Discovery

The team built a tabletop experiment to test this. They simulated a star and a planet using two small lights (LEDs) and ran them through their "shape-separating" device.

They found two major things:

1. It Works Even When Leaky: They discovered a specific "leak threshold" (about 10% leakage). As long as the device is better than 90% accurate (which modern technology easily achieves), the SPADE method still beats the traditional blurry photo method.
2. It's Much More Efficient: Because the method is so sensitive, it needs far fewer photons to make a correct decision. In their experiment, with a small leak of just 1%, the SPADE method needed 10 times fewer photons (or 10 times less time) to detect the "firefly" compared to the traditional camera method to achieve the same level of certainty.

The Bottom Line

The paper proves that you don't need a perfect, flawless machine to see things smaller than the limit of light. Even with realistic imperfections (leaks), using this "shape-separating" trick allows scientists to detect faint objects next to bright ones much faster and more reliably than ever before. It's like being able to hear a whisper in a noisy room not by turning up the volume, but by using a special filter that only lets the whisper through.

Technical Summary

Technical Summary: Experimental Subdiffraction Source Discrimination Enabled by Spatial Demultiplexing and Single-Photon Detectors

Problem Statement
Resolving faint sources in close proximity to bright ones, such as detecting an exoplanet near its host star, poses a fundamental challenge in optical imaging due to the diffraction limit. Conventional approaches, such as coronagraphy and starshades, attempt to physically block starlight to improve contrast. However, recent theoretical work suggests that spatial demultiplexing (SPADE) offers a superior alternative by projecting the incoming field onto optimized transverse spatial modes and measuring them via photon counting. While SPADE has been shown theoretically to achieve quantum-optimal performance for asymmetric hypothesis testing (distinguishing between the absence $H_0$ and presence $H_1$ of a faint source), practical implementation is hindered by modal crosstalk arising from device imperfections and misalignment. Previous theoretical proposals for crosstalk-tolerant tests either required prior knowledge of source parameters or were suboptimal. The core problem addressed is whether SPADE can maintain its advantage over diffraction-limited direct imaging in realistic experimental conditions characterized by non-zero crosstalk and without requiring prior knowledge of the source separation or intensity ratio.

Methodology
The authors present a universal, parameter-independent hypothesis test implemented on a tabletop experimental setup. The methodology relies on the following components:

• Theoretical Framework: The study employs quantum detection theory to model the discrimination of two incoherent sources with an intensity ratio $\epsilon$ and normalized separation $d_a$. The test is designed to minimize the Type II error (false negative) while keeping the Type I error (false positive) bounded, a scenario relevant to exoplanet detection where false negatives are more costly. The theory accounts for arbitrary modal crosstalk using a stochastic matrix to model the mixing of outcome probabilities between the fundamental mode ($HG_{00}$) and the first-order modes ($HG_{01} + HG_{10}$).
• Experimental Setup: The experiment simulates an exoplanet system using two fiber-coupled LEDs at 1550 nm. One LED represents the star ($A$) and the other the exoplanet ($B$). The beams are combined and coupled into a Hermite-Gaussian (HG) mode demultiplexer (PROTEUS-C model). The separation $d$ and intensity ratio $\epsilon$ are varied using a translation stage and a motorized rotation stage, respectively.
• Detection and Analysis: Photons in the $HG_{01}$ and $HG_{10}$ modes are detected using superconducting nanowire single-photon detectors. The hypothesis test rejects $H_0$ if the number of photons detected in these modes ($N_1$) exceeds a threshold $N^*$. Crucially, $N^*$ is determined solely from the statistics of $H_0$ (absence of the planet) and is independent of the physical parameters $d_a$ and $\epsilon$. The performance is evaluated by measuring the Type II error rate ($\beta$) across various separations and intensity ratios.

Key Contributions

1. First Experimental Demonstration with Crosstalk: This work provides the first experimental validation of the advantage of SPADE in the presence of modal crosstalk. It demonstrates that the theoretical benefits of SPADE persist even when the demultiplexer is imperfect.
2. Universal Parameter-Independent Test: The authors implement a test that does not require prior knowledge of the source separation or intensity ratio, addressing a limitation of previous proposals.
3. Identification of a Crosstalk Threshold: The study theoretically and experimentally identifies a critical crosstalk threshold, $C_{th} \simeq 0.1$. Below this value, SPADE outperforms direct imaging. Specifically, for a crosstalk of $C \simeq 0.01$ (measured in their setup), SPADE achieves the same error rate as direct imaging with approximately one order of magnitude fewer photons.
4. Complete Theoretical Modeling: The paper presents a comprehensive theory modeling arbitrary modal crosstalk, including the effects of thermal source statistics and vacuum contributions, which accurately predicts the experimental error rates.

Results

• Error Rate Scaling: The experimental data confirms that the Type II error rate of the SPADE-based test follows the theoretical prediction, achieving an exponential scaling of $\beta \sim \exp(-N D_{SD})$, where $D_{SD}$ is the relative entropy of the SPADE measurement.
• Comparison with Direct Imaging: In the regime of small separations and low intensity ratios, the measured error rates for SPADE are significantly lower than those predicted for an optimal direct imaging test. The direct imaging test suffers from "Rayleigh-cursed" scaling ($\sim \epsilon^2 d_a^4$), whereas SPADE maintains a superior scaling ($\sim \epsilon d_a^2$ in the ideal limit, degrading to $\sim \epsilon^2 d_a^4$ but with a much larger prefactor advantage in the presence of low crosstalk).
• Crosstalk Regime: With a measured crosstalk of $C \simeq 10^{-2}$, the experiment operates well within the favorable regime ($C < 0.1$). The results show that SPADE is approximately 12.5 times more efficient than direct imaging in terms of the photon budget required to achieve a fixed error probability.
• Robustness: The test remains effective despite the statistical fluctuations inherent in thermal sources and the fixed integration time used in the experiment.

Significance
The paper claims that these results demonstrate SPADE offers an effective methodology for subdiffraction asymmetric hypothesis testing under realistic imperfections. By establishing a quantitative threshold for crosstalk ($C_{th} \simeq 0.1$) below which SPADE outperforms direct imaging, the work validates the practical viability of SPADE for photon-starved imaging tasks. The authors emphasize that the approach does not rely on prior knowledge of source parameters, making it particularly relevant for practical scenarios such as exoplanet detection where such information is often unknown. The findings pave the way for the application of SPADE in biosensing, surface metrology, and astronomical observations where distinguishing faint signals from bright backgrounds is critical.