QNU: Quantum Networking Unit

Overview

The function of the QNU is to implement any two-node photonic Bell-state measurement (BSM) required by the Entanglement Fabric. At a high level, the Entanglement Fabric requires:

Sparse long-range connectivity, with each QPI entangled sequentially to ~10 other QPIs.
These ~10 entanglement events are required to happen within one QEC cycle (~1ms)

At the hardware-level, this requires:

Active routing using a low-loss, fast (>MHz) ~1x6 non-blocking switch matrix to demultiplex the photons generated by the QPI.
An efficient photonic BSM, using a beam-splitter and Superconducting detectors.
A low latency distributed control system to coordinate entanglement attempts, implementing a repeat-until-success protocol for establishing remote entanglement.

‍Nu Quantum is developing the QNU along two parallel tracks to accelerate availability to near-term distributed quantum computing demonstrators and testbeds, while also developing new technology that will provide the performance and scalability needed by utility-scale quantum computers.

Gen1 QNU Prototype: The near-term development of a productised, modular QNU uses commercially available optical components (COTS) to allow faster development. Building understanding of these components, and how they interact, allows us to achieve performance that is sufficient for early quantum network demonstrators. However, the scale-up to utility-scale quantum computing will start to become limited by the disadvantages of a system based on discrete components (size, cost, loss and speed). The Gen1 QNU is now operational and meeting specifications, and on-track to launch to customers, alongside a Test Harness, in March 2025.
Technology Platform development: We are also developing an optical module based on quantum PIC technology, which will provide low-loss, ultrafast switching and ultra-high efficiency photon detection to meet the needs of utility-scale QC. Our modular design will allow this to work as a drop-in upgrade to later Generations of the QNU.

A scalable distributed control stack sits on top of both the near-term and qPIC hardware, providing ultra-low-latency feed-forward and network-level orchestration.

Gen-1 Quantum Networking Unit

‍

Control & Orchestration

Within a Quantum Networking Unit (QNU) the result of an entanglement attempt is known as a heralding event. A successful herald event signifies that the two qubits have successfully been entangled and the control algorithm should shift from the attempt cycle to some process that will exploit the entangled state (e.g. QEC check or verifying entanglement Fidelity).

There is a requirement on the control system to both minimise the latency between successive attempt loops, in order to maximise the attempt rate, while also providing a fast feedforward path to jump out of the attempt loop and branch to the alternate operating mode. The critical point being that starting the next attempt loop would destroy the qubit state that we want to use. This leads to a requirement on the control system to minimise the communication latency of a successful herald such that we can maximise the attempt rate.

In an initial demonstration, we implemented a <250 ns latency between two separate Control Units—a key advance towards a multi-node quantum network demonstrator. This demonstration uses ARTIQ and customises its DRTIO point to point link. Utility-scale Distributed control will require an ethernet based network. To this end, we are developing a communication protocol over White Rabbit. The latency is likely to be an order of magnitude larger unless we again customise the lower levels of the network stack. In other words, the Entanglement Fabric requires diverging from classical networking standards to deploy higher speed communication—Nu Quantum is best positioned to develop and own this new standard.

Figure: **Demonstration of a low-latency (<250ns) software branching between two control systems**. The Yellow trace is a TTL output from the QNU Control module, indicating a successful herald event. The blue trace is the output of a QPU Control module, toggling every 100ns. The toggling is interrupted after 250 ns, which is the ‘distributed-control’ latency, for an entanglement event to be detected at the QNU, communicated, and changing the course of real-time events at the QPU.

‍

Optical Module & Photonics

COTS Architecture and Results

Nu Quantum has built a first-generation QNU, using COTS-switches and detectors for the optical module. This QNU-gen1 is designed to operate a four-node quantum network test-bed, and features a 4x2 switch matrix. The system-level performance have been measured: the QNU-gen1 will contribute to ~10^-3 entanglement errors and reduce the entanglement rate by 7.5dB on average. The switch settling time will also contribute to a 1 μs deadtime when setting a new network configuration. With a robust quantum-optical modelling of the QNU-performance, NuQ can readily optimise their switch architecture. Better performance can only be unlocked by the development of quantum PIC.

Figure: **Schematics of the QNU-gen1 optical module.** The QNU-gen1 is composed of three 2x2 optical switches and a BSM-stage with high-efficiency superconducting detectors. The switches are arranged such that any pairwise entangled state can be created within a four node quantum network.

‍

Figure: **Quantum Optics model of QNU performance.** Impact on Fidelity of Real-world hardware imperfections such as Cross-talk (CT), Polarisation Extinction Ratio (PER) and Beam-splitter Imbalance (R ≠ 0.5) are all taken into account to arrive at an optimum switch architecture.

‍

Quantum Photonic Integrated Circuit (qPIC)

Nu Quantum is developing a quantum photonic integrated circuit (qPIC) for the next generation of QNUs. We have chosen to work at the native wavelength of the qubits, giving us lower photon losses and higher fidelity than approaches based on frequency conversion. We have proven high component-level performance at this wavelength using in-house workflows for component design, characterization and optimization, and this experience will also allow us to design for new qubit technologies as they emerge.

Series A will see us consolidate our nanofabrication workflows and develop low-loss optical packaging solutions.

‍

Figure: **Measurement of PIC switches** is performed on a probe station, allowing fast characterisation of our PIC components. The data panel on the bottom-right displays optical transmission at each output as of function of applied voltage, from which we can deduce switch cross-talk.

Nu Quantum is also developing waveguide-integrated SNSPDs on the same PIC platform. We have already significantly de-risked device design & fabrication, as well as waveguide integration, but more iterations are needed to optimise the quantum efficiency of our SNSPDs.