🔬 atomic physics

Universal Fidelity Law for Linear-Optical Entangling Gates

This paper derives and experimentally validates a universal fidelity law for linear-optical entangling gates that predicts performance degradation based solely on Hong-Ou-Mandel visibility and second-order coherence, regardless of the specific signal-noise overlap or physical platform.

Original authors: Haim Nakav, Ofer Firstenberg

Published 2026-09-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Haim Nakav, Ofer Firstenberg

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

In the quest to build a quantum computer, scientists often turn to light. Photons, the tiny particles that make up light, are excellent carriers of information because they rarely interact with their surroundings, allowing them to hold their quantum state for long periods. They travel at the speed of light and can be manipulated using simple, inexpensive glass and mirrors. However, to perform calculations, these photons must be able to talk to one another, a process that requires them to be perfectly identical and to arrive at the exact same moment. In the real world, creating such perfect photons is incredibly difficult. The sources that generate them are often imperfect, occasionally spitting out extra unwanted photons or sending out photons that are slightly different from one another. These flaws introduce errors that can ruin the delicate calculations a quantum computer is trying to perform.

For years, researchers have measured the quality of these light sources using two standard tests. One test checks how well two photons can interfere with each other, a phenomenon where they either merge or cancel out depending on their timing. The other test measures how likely a source is to accidentally emit two photons instead of one. While these measurements are routine, scientists have struggled to combine them into a single, clear prediction of how well a quantum logic gate—a basic building block of a quantum computer—will actually work. Different types of light sources, such as those based on crystals or tiny semiconductor dots, produce these errors in different physical ways, leading to the belief that a single rule could not possibly apply to all of them.

A team of physicists at the Weizmann Institute of Science in Israel has now derived a universal law that connects these source imperfections directly to the performance of a quantum gate. They focused on a specific type of gate called a controlled-NOT, or CNOT, which is essential for entangling two photons. By analyzing the mathematics of how photons interfere, the researchers discovered that the specific physical origin of the errors—whether the extra noise comes from the same source as the signal or a different one—does not matter when looking at the final result. Instead, the errors are absorbed into the standard measurements of the light source. This means that regardless of whether the photons come from a crystal, a laser, or a solid-state device, the same simple relationship predicts the gate's accuracy.

The researchers tested this theory using their own experimental setup, which uses a series of beam splitters and mirrors to guide photons into a controlled collision. They also compared their findings against published results from a wide variety of other experiments using different technologies. In every case, the predicted performance matched the measured performance without needing to adjust any variables or add extra assumptions. The study revealed that while the physical nature of the noise does change the specific way errors occur inside the machine, the final outcome is always the same when viewed through the lens of standard measurements. This finding effectively turns the standard characterization of a light source into a unified benchmark for the entire field of photonic quantum computing.

One of the most significant insights from this work is the relative importance of different types of errors. The team found that errors caused by photons being slightly distinguishable from one another and errors caused by the accidental emission of extra photons are currently of comparable size in the best experiments. In the past, researchers often treated the emission of extra photons as a minor correction, but this new analysis shows it is a fundamental source of error that must be suppressed just as rigorously as the indistinguishability of the photons. The researchers also identified that once these source-related errors are accounted for, the remaining inaccuracies in a quantum gate come from the optical circuit itself—the mirrors, beam splitters, and detectors. By isolating these two sources of error, the framework allows engineers to see exactly where improvements are needed.

The study also clarified a long-standing confusion regarding how different noise types affect gate performance. When the researchers looked at the raw physics, they found that if the extra noise photons are distinguishable from the signal photons, they cause nearly twice as much error as if they were identical. However, because standard measurements of the light source naturally account for this difference, the final prediction for the gate's fidelity remains consistent across all platforms. This means that a scientist can measure the quality of their light source using standard tools and immediately know the theoretical limit of their quantum gate, regardless of the hardware they are using.

This work provides a clear path forward for the development of quantum networks and computers. By establishing a universal rule that links source quality to gate performance, the researchers have removed a layer of uncertainty that previously complicated the comparison of different technologies. The framework suggests that to build a functional quantum computer, engineers must focus on improving both the purity of the photons and the precision of the optical circuits that guide them. The constant residual error found in the experiments, which did not change even when the source quality varied, points directly to the need for better optical components. With this unified understanding, the field can move toward a more standardized approach to building and evaluating the machines of the future.

Technical Summary: Universal Fidelity Law for Linear-Optical Entangling Gates

Problem Statement
The performance of interference-based linear-optical entangling gates, such as the controlled-NOT (CNOT) gate, is fundamentally limited by the quality of single-photon inputs. While single-photon sources (e.g., SPDC, FWM, quantum dots) are characterized by standard metrics—Hong-Ou-Mandel (HOM) visibility ( $V$ ) and second-order coherence ( $g^{(2)}(0)$ )—these metrics originate from different physical mechanisms across platforms. Specifically, multiphoton emission in parametric sources (SPDC/FWM) often occurs in the same temporal mode as the signal ("identical noise"), whereas in solid-state emitters (quantum dots), it often arises from re-excitation or leakage in a distinct temporal mode ("distinguishable noise"). Existing fidelity characterizations typically treat $g^{(2)}(0)$ as a platform-independent error, failing to account for how the physical origin of multiphoton noise affects gate performance. This lack of a unified framework makes it difficult to benchmark and compare gate fidelities across different photonic platforms.

Methodology
The authors develop an analytical framework for multiphoton-emission errors in linear-optical CNOT gates. The methodology involves:

Modeling the Gate: A detailed derivation of a CNOT gate using partially polarizing beam splitters (PPBS), accounting for all coincidence events, including those where photons exit unmonitored ports.
Noise Parameterization: Introducing a normalized signal-noise overlap parameter, $M_{sn} \in [0, 1]$ . $M_{sn}=1$ represents "identical noise" (indistinguishable from the signal), while $M_{sn}=0$ represents "distinguishable noise" (orthogonal mode).
Photon Number Decomposition: The gate output is analyzed by decomposing input states into pure single-photon events ( $|1,1\rangle$ $∣1, 1 ⟩$ ), pair-emission events ( $|2,0\rangle, |0,2\rangle$ $∣2, 0 ⟩, ∣0, 2 ⟩$ ), and three-photon events ( $|2,1\rangle, |1,2\rangle$ $∣2, 1 ⟩, ∣1, 2 ⟩$ ).
- Pure single-photon events depend on intrinsic indistinguishability $\eta$ .
- Pair events contribute only to the normalization (denominator) of the fidelity.
- Three-photon events introduce a non-trivial error term where the noise photon interferes with the signal. The authors derive an effective indistinguishability parameter, $\eta_{eff}$ , which depends on both $\eta$ and $M_{sn}$ .
Derivation of Universal Law: The authors express the gate fidelity first in terms of intrinsic parameters ( $\eta, g^{(2)}$ ) and then transform these into experimentally measurable quantities ( $V, g^{(2)}$ ). They demonstrate that the dependence on the specific noise model ( $M_{sn}$ ) is absorbed into the measured HOM visibility $V$ .

Key Contributions

Universal Fidelity Law: The paper derives a closed-form expression for gate fidelity (Eq. 26) that depends only on the measured HOM visibility ( $V$ ) and second-order coherence ( $g^{(2)}$ ). This law holds for any noise model (identical or distinguishable) without free parameters.
Noise Model Distinction: The analysis reveals that while the intrinsic error contribution of distinguishable noise is nearly double that of identical noise at state-of-the-art operating points (due to the suppression of bosonic enhancement in three-photon events), this difference vanishes when fidelity is expressed via measured visibility.
Unified Benchmark: The framework converts standard source characterization ( $V$ and $g^{(2)}$ ) into a unified benchmark for photonic quantum computing, allowing direct comparison of gate performance across disparate platforms (SPDC, atomic ensembles, quantum dots).

Results

Theoretical Validation: The derived universal law (Eq. 26) predicts gate fidelity based solely on $V$ and $g^{(2)}$ . The authors show that for $g^{(2)} \ll 1$ , the fidelity is $F \simeq \frac{1+V}{4-2V} - \frac{g^{(2)}}{2(2-V)}$ .
Experimental Verification: The model was tested against:
1. The authors' own actively synchronized CNOT gate experiment across a broad range of source parameters.
2. Published results from SPDC, atomic-ensemble, and quantum-dot platforms.
- All experimental data points (including those from platforms with distinguishable noise) fall on or below the theoretical curve defined by Eq. 26.
- The "inset" of Figure 3 shows a strong agreement between measured and predicted fidelities, with no decoding step required for the noise model.
Error Hierarchy: The analysis separates total error into indistinguishability imperfections and multiphoton errors. The authors find these two sources are of comparable magnitude across reported experiments.
Circuit vs. Source Limits: By analyzing the residual error between measured and predicted fidelity, the authors identify a constant residual error ( $\approx 2.3 \times 10^{-2}$ ) that is independent of $g^{(2)}$ . This indicates that for high-performance gates, circuit imperfections (optical alignment, stability) become the limiting factor once source quality is optimized.

Significance
The paper establishes that multiphoton emission is a fundamental error source in photonic quantum computing, comparable in impact to indistinguishability imperfections. By deriving a universal fidelity law that relies only on standard, directly measurable source metrics, the work provides a rigorous method to benchmark and compare linear-optical gates across different physical implementations. The framework demonstrates that the specific physical origin of multiphoton noise (identical vs. distinguishable) does not alter the fundamental relationship between source quality and gate fidelity when viewed through the lens of measured visibility, thereby unifying the characterization of photonic quantum resources.

More like this