QuantumPave: Chip-Scale Quantum Interface

• QuantumPave is a chip-scale free-space quantum platform that leverages a quantum phased array for coherent mode selection and optical state preservation.
• It integrates silicon photonics, a large-area free-space-to-chip interface, and RF-domain coherent processing to mitigate geometric loss.
• Demonstrations include 32-pixel squeezed-light imaging, reconfigurable quantum links, and entanglement generation, enabling advances in quantum sensing, communications, and computing.

Searching arXiv for the cited foundation paper and a closely related photonic interconnect paper for context. QuantumPave is a designation applied to the free-space quantum information platform on a chip centered on the quantum phased array (QPA) introduced in "Free-space quantum information platform on a chip" (Gurses et al., 2024). The platform combines a free-space-to-chip interface, integrated silicon photonics, and coherent radio-frequency readout so that quantum optical states received from free space can be preserved, downconverted by homodyne detection, and coherently processed in a standalone, compact form factor. In the reported realization, it provides the first quantum phased array technology demonstration and, to the authors’ knowledge, the first free-space-to-chip interface for quantum links, with demonstrations spanning 32-pixel squeezed-light imaging for quantum sensing, reconfigurable free-space links for quantum communications, and proof-of-concept entanglement generation for measurement-based quantum computing (Gurses et al., 2024).

1. Central concept and problem setting

The central idea of QuantumPave is a wireless, chip-scale quantum node that can receive quantum optical states from free space, preserve and coherently process them on chip, and reconfigure the spatial mode it selects or transmits (Gurses et al., 2024). The motivating systems problem is that free-space links are highly useful for quantum systems because they offer connectivity, parallelization, and reconfigurability, yet they are limited by geometric loss: a diverging free-space beam often mismatches the receiving aperture, so much of the quantum state is lost before detection.

The reported solution is a large-area metamaterial aperture plus coherent RF combining. Rather than simply absorbing a small portion of the field, the receiver collects, reshapes, and processes the incoming quantum field. This design choice is foundational to the platform’s interpretation: the chip is not only a free-space optical front end, but an integrated photonic-electronic stack for programmable quantum reception and processing.

A recurrent misconception is to treat the platform as a passive aperture or a detector array. The paper’s architecture and processing model instead present it as a system that performs coherent mode selection on quantized optical fields. This distinction is central to the term quantum phased array and to the broader claim that free-space quantum links can be made dynamically adjustable rather than statically aligned.

2. Quantum phased array formalism

The quantum phased array generalizes classical phased arrays and wavefront engineering to quantized electromagnetic fields (Gurses et al., 2024). Classically, phased arrays form beams by applying amplitude and phase weights across many elements so that fields interfere constructively in a desired direction. In the QPA formulation, the same logic is applied to quantum optical modes: many spatially distributed pixel-like channels measure local quadratures, and the outputs are then gain- and phase-weighted in RF so that the desired quantum mode is selected coherently.

The input free-space field is written as

$$\hat{a}_{\text{in}}(\rho)=\sum_n \hat{a}_n u_n(\rho),$$

and the field coupled into the $j$-th antenna or pixel is

$$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$

where $U_{jn}$ is the overlap between the antenna mode and the input spatial mode. After homodyne detection, the RF-combined output is

$$\hat{a}_{\text{out}}=\sum_j g_j e^{i\phi_j}\hat{a}_j \approx \sum_n \hat{a}_n\int \mathcal{A}(\rho)u_n(\rho)\,d\rho,$$

with programmable array mode function

$\mathcal{A}(\rho)=g(\rho)e^{i\phi(\rho)}.$

This relation is the key wavefunction-engineering statement of the platform. By choosing $\mathcal{A}(\rho)$, the receiver can match a desired optical mode $u_n(\rho)$, suppress vacuum contributions, and recover the selected quantum signal. The corresponding free-space-to-RF processing model is

$\vec{a}_{\text{out}}=M D U \vec{a}_{\text{in}},$

where $U$ is the free-space change-of-basis transformation from spatial modes to pixel modes, $j$0 is the programmable per-channel gain/phase layer, and $j$1 is the post-detection RF-domain linear processing. In this sense, QuantumPave does not merely detect light; it implements programmable coherent mode transformations on quantum fields.

3. Integrated architecture and free-space-to-chip interface

The reported device is a silicon photonic integrated circuit with more than 1000 functional components on a $j$2 footprint (Gurses et al., 2024). Its architecture includes 32 optical input channels, 32 quantum coherent receivers (QRXs), phase shifters and splitters, photodiodes, transimpedance amplifiers, and RF combining and digitization electronics. The system is explicitly photonic-electronic co-designed rather than purely photonic: optical quantum signals are downconverted via homodyne detection, and the resulting signals are processed coherently in the RF domain.

The free-space-to-chip interface is a $j$3 fully filled aperture containing 32 metamaterial antennas. Each antenna has footprint $j$4, and the aperture contains more than 500,000 sub-wavelength engineered nanophotonic elements. The antennas are designed to couple a collimated free-space beam into a single-mode waveguide with reduced mismatch. This large-area design addresses a known limitation of conventional nanophotonic antennas, whose effective apertures are often too small to efficiently collect a collimated beam.

The reported geometric loss figures are 4.85 dB for 32 apertures with a 200 $j$5 beam and a minimum geometric loss with amplitude weighting of 1.14 dB. These values are presented as low enough to support quantum operation. The electronic side comprises an interposer and motherboard, 32 TIAs, RF outputs, a 32-to-1 power combiner, and digitizers or spectrum analyzer readout. The resulting platform is therefore an end-to-end free-space quantum interface rather than an isolated nanophotonic coupler.

4. Homodyne reception and RF-domain coherent processing

Each QRX consists of a tunable Mach–Zehnder interferometer, balanced Ge photodiodes, and a transimpedance amplifier (Gurses et al., 2024). The MZI mixes the received signal with a strong local oscillator, enabling homodyne detection of the selected quadrature. The LO is split into 32 paths and phase-tuned individually using thermo-optic phase shifters. In balanced homodyne detection, the incoming quantum field interferes with the LO in the MZI, the two photodiode currents are subtracted, and the result is proportional to a quadrature of the optical field.

The measured quadrature in the downconverted RF frame is written as

$j$6

This equation makes the optoelectronic bridge explicit: quantum optical information is converted into an RF quadrature signal that can then be combined, filtered, digitized, and emulated with standard RF hardware.

Two reported receiver metrics are especially important. The shot noise clearance is 30.3 dB in the high-shot-noise-clearance configuration, and the common-mode rejection ratio is 90.2 dB at 1.1 MHz. The former indicates that the measurement is strongly shot-noise-limited rather than electronics-limited; the latter indicates strong rejection of unwanted classical noise in balanced homodyne subtraction. The system also uses a CMRR auto-correction loop that feeds back into the MZI to continuously stabilize balance. This is presented as a major reason the platform can operate robustly across 32 channels.

5. Demonstrated capabilities

The platform was demonstrated in three distinct operating modes (Gurses et al., 2024).

Demonstration	Mechanism	Reported result
32-pixel squeezed-light imaging	Multipixel homodyne readout of 32 spatial pixel modes	spatial distribution of squeezed light across 32 pixels
Reconfigurable free-space quantum links	Beamforming by per-channel phase settings	beamwidth 0.41° with 8 channels and 0.20° with 32 channels; 3-dB loss FoV 2.3° with 8 channels and 2.7° with 32 channels
Proof-of-concept entanglement generation	RF-domain emulation of an optical network generating a two-mode Gaussian cluster state	minimum inseparability $j$7

In the multipixel quantum imaging experiment, a squeezed vacuum state generated off-chip is sent over free space to the QPA chip. The 32 antennas define 32 spatial pixel modes $j$8, and each pixel measures a phase-dependent quadrature

$j$9

For squeezed vacuum, the quadrature variance is

$$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$0

where $$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$1 is the squeezing parameter and $$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$2 is the effective efficiency of channel $$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$3. The experiment shows spatially resolved squeezed-light measurements and therefore establishes a quantum sensing use case.

In the free-space communication demonstration, appropriate phase settings across the 32 channels steer receiver sensitivity toward the transmitter and away from empty space. The receiver can be beamformed to reject the source, producing no detectable quantum signal, and then electronically reconfigured to lock onto the source and recover the squeezed state. The reported beamwidth improves as more channels are combined: 0.41° for 8 channels and 0.20° for 32 channels. The reported 3-dB loss field of view is 2.3° for 8 channels and 2.7° for 32 channels. These results define the device as a dynamic spatial filter for quantum communication rather than a fixed optical receiver.

In the measurement-based quantum computing demonstration, the QPA chip plus RF-domain processing emulates an optical network that generates a two-mode Gaussian cluster state. The entanglement criterion is

$$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$4

and the reported minimum inseparability is

$$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$5

which violates the classical bound by three standard deviations. The composite processing model is

$$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$6

where the RF combiner and digital or analog processing implement the beamsplitter-like transformation $$\hat{a}_j=\int \mathcal{E}_j(\rho)\hat{a}_{\text{in}}(\rho)\,d\rho=\sum_n U_{jn}\hat{a}_n,$$7. The result is explicitly described as proof-of-concept, but it shows that entangled resource states can be generated after optical downconversion.

6. Significance, interpretation, and relation to adjacent platforms

QuantumPave is significant because it addresses several bottlenecks simultaneously (Gurses et al., 2024). In the reported implementation, scalability appears in the 32 integrated channels, more than 500,000 nanophotonic elements in the aperture, and end-to-end photonic-electronic integration on a compact footprint. Robustness appears in room-temperature operation, silicon photonic fabrication, electronic stabilization of homodyne receivers, and the reported high CMRR and shot-noise clearance. Wireless quantum connectivity appears in the use of a free-space interface rather than fiber-only connectivity, in reconfigurable beamforming, and in spatial filtering that mitigates diffraction-induced geometric loss. Quantum functionality extends beyond detection to sensing via squeezed-light imaging, communications via programmable free-space links, and computation via entanglement generation.

These points clarify what QuantumPave is not. It is not merely a free-space optical coupler, because the mode-transformation formalism is programmable. It is not merely a detector array, because the outputs are coherently combined in RF to select a desired quantum mode. It is not a purely photonic receiver, because downconversion, RF combining, and stabilization are part of the essential operating principle. A plausible implication is that the platform defines a system paradigm in which quantum information can be received, steered, filtered, and processed in free space using a compact integrated chip with coherent RF readout.

Within the broader landscape of integrated quantum hardware, QPA-based free-space interfacing occupies a different position from platforms that focus on photonic interfaces to quantum memories. For example, "A scalable photonic quantum interconnect platform" describes a heterogeneous platform based on thin-film diamond membranes, photonic crystal cavities, and silicon-vacancy quantum memories for modular photonic quantum interconnects (Riedel et al., 8 Aug 2025). This suggests a complementary rather than interchangeable relationship: QuantumPave emphasizes wireless quantum links, free-space mode selectivity, and RF-domain coherent processing, whereas memory-centric photonic interconnect platforms emphasize cavity-enhanced memory-photon interfaces and modular node assembly.