Silicon Quantum Photonics

Joshua W. Silverstone; Damien Bonneau; Jeremy L. O'Brien; Mark G. Thompson

Source-linked AI summary

Silicon Quantum Photonics

Joshua W. Silverstone, Damien Bonneau, Jeremy L. O'Brien, Mark G. Thompson

arXiv:1707.02334v1Jul 7, 2017quant-phphysics.optics

TL;DR

Integrated quantum photonics needs scalable, integrated components, but existing approaches and silicon itself face limits in complexity, sources, losses, and control. This review surveys the components and competing platforms, evaluates silicon’s strengths and weaknesses, and identifies challenges and potential solutions. It concludes that silicon’s density and manufacturing advantages could support quantum photonic systems at very large scale, including systems with millions of components.

Problem
Integrated quantum photonic systems require scalable components and integration, while existing platforms and silicon face limitations in complexity, sources, losses, and control.
Method
The review surveys quantum photonic components, compares six on-chip technology platforms across five categories, and discusses silicon’s challenges and potential solutions.
Results
Silicon scores well overall, with strengths in integration density, loss, and passive optics, while weaknesses remain in active optics and photon sources.
Takeaways & Limitations
Silicon may provide the viable route to assembling quantum photonic systems containing millions of components and potentially light-based quantum computers.

Abstract

from arXiv · show

Integrated quantum photonic applications, providing physially guaranteed communications security, sub-shot-noise measurement, and tremendous computational power, are nearly within technological reach. Silicon as a technology platform has proven formibable in establishing the micro-electornics revoltution, and it might do so again in the quantum technology revolution. Silicon has has taken photonics by storm, with its promise of scalable manufacture, integration, and compatibility with CMOS microelectronics. These same properties, and a few others, motivate its use for large-scale quantum optics as well. In this article we provide context to the development of quantum optics in silicon. We review the development of the various components which constitute integrated quantum photonic systems, and we identify the challenges which must be faced and their potential solutions for silicon quantum photonics to make quantum technology a reality.

I. INTRODUCTION

Quantum photonics uses photons and optics to process, transmit, and encode information, while integrated optics has demonstrated increasingly complex quantum operations. Silicon is presented as a scalable platform whose density, functionality, and manufacturability could support larger quantum photonic systems despite material and fabrication challenges.

Quantum information science uses superposition and entanglement to process, transmit, and encode information, with photons and optics at its forefront.
Glass-based integrated quantum photonics has demonstrated interference, logic operations, quantum walks, entanglement manipulation, and boson sampling.
10-cm devices can be required even for simple glass-based circuits, limiting scalability and circuit complexity.
Silicon photonics offers scale, functionality, and manufacturability motivated by its established role in optical interconnects and microelectronics.
A generic silicon quantum photonic device combines photon sources, pump-removal filters, passive and active optics, detectors, and control electronics.
Silicon’s χ(3) nonlinearity enables devices from photon sources to photo-optic switches, but TPA, free-carrier and thermal effects, and sidewall roughness constrain performance.

II. QUANTUM PHOTONIC DEVICES

Silicon quantum photonic devices combine photon-pair sources with passive and active optical components, while source statistics, nonlinear loss, and resonator effects constrain performance. Silicon-based SFWM and resonator sources offer compact, integrable routes to quantum photonic functionality, but scalability and purity remain important challenges.

Photon-pair generation: Silicon sources can generate quantum-correlated signal and idler photons through spontaneous four-wave mixing (SFWM), with Raman noise localized 15.6 THz from the pump.In a monomode waveguide, both photons emerge in the same transverse mode.
Photon-pair generation: SFWM is probabilistic: weak pumping approximates a photon pair, whereas stronger operation produces vacuum, single-pair, or multi-pair events.Multi-pair emission introduces noise, and signal-idler correlations reduce heralded single-photon purity.
Photon-pair generation: The single-pair probability reaches a maximum around 25% for an ideal source emitting into two spectral modes.The pair-number distribution depends on squeezing and the ratio of collection to full emission bandwidth.
Nonlinear effects: SFWM efficiency grows quadratically with pump power at low power, but two-photon absorption causes pair-generation saturation at higher power.TPA also produces cross two-photon absorption and free-carrier absorption, increasing propagation loss.
Resonator sources: Resonators reduce source footprint, enhance brightness, and shape joint spectral emission by confining pump, signal, and idler fields to cavity modes.Silicon resonators require care because TPA and FCA can shift resonances and affect biphoton spectral amplitude and heralded-photon purity.

B. Passive Optics

Silicon passive optics support increasingly dense quantum photonic circuits, from photon-pair generation and interference to programmable processing and on-chip entanglement. These capabilities address the scalability limits of glass-based integrated quantum photonics.

Passive optical processing: Linear optical elements manipulate quantum states for encoding, measurement, and state preparation, with complex applications requiring thousands or millions of passive components.QKD may use only two Mach-Zehnder interferometers, whereas linear optical quantum computation may require thousands or millions of elements.
Silicon scaling: Silicon provides higher component density and semiconductor-enabled functionality than glass-based platforms.A demonstrated silicon processor contained 112 multimode interferometers and 213 silicon-based tuners, substantially exceeding the scale of the cited glass devices.
Demonstrated quantum operations: Silicon quantum optics has demonstrated quantum interference and on-chip multi-qubit logic using integrated passive components.Hong-Ou-Mandel and Mach-Zehnder experiments exceeded 80% visibility, while a later device used 21 passive elements, 16 thermal tuners, and 4 photon-pair sources.
On-chip entanglement: On-chip entanglement circuits generate photon pairs in superposition between resonant sources and reconfigure and analyze them in the path-qubit basis.The architecture combines photon-pair sources, directional couplers, and wavelength-division multiplexing.

C. Single-Photon Detectors

Single-photon detection is essential to integrated quantum photonics but remains constrained by wavelength and integration requirements. SNSPDs provide strong performance at telecom wavelengths, while room-temperature waveguide-coupled detection remains unresolved.

Detection wavelength: Telecom-band silicon photonics benefits from existing lasers, detectors, and optics, but silicon avalanche photodiodes are not sensitive near 1.55 µm.The 1.55-µm region is compatible with silicon waveguides yet creates a single-photon detection challenge.
Detector integration: A yield-improvement strategy fabricates SNSPDs on Si3N4 membranes, selects them by performance, and transfers them onto silicon waveguides.The detector and waveguide are fabricated on different substrates before transfer.
SNSPDs: Cryogenic SNSPDs detect across a wide wavelength range, including 1.55 µm, and offer near-ideal detection characteristics.They typically require only a single electron-beam-lithography patterning step.
Remaining challenge: No waveguide-coupled single-photon detector has yet been demonstrated at or near room temperature.A vertically coupled germanium avalanche photodiode is identified as a possible future solution if it achieves sufficient performance.

D. Optical switches

Optical switching enables feedforward in quantum photonic systems, but silicon must balance switching speed, loss, noise, complexity, and the absence of native χ(2) nonlinearity.

Feedforward operation: Feedforward detects a heralding photon, switches its partner’s path, and stores the heralded photon for τ = τd + τs.The transmitted efficiency is η = ηdηsexp[ατvg], exposing a tradeoff between detector, switch, and delay-line performance.
Switching requirements: Large heat- and MEMS-based phase tuners are versatile but operate at most in the MHz regime, too slowly for rapid time-of-flight feedforward.Feedforward requires reconfiguration within a short time frame.
Electro-optic switching: Electro-optic χ(2) modulators can operate at 40 Gb/s and handle quantum states without additional inherent electro-optic loss.Silicon naturally has χ(2) = 0 because it is a centro-symmetric crystal.
Silicon limitations: Carrier-based silicon switches reach tens of GHz but introduce phase-dependent loss and, in forward bias, noise.All-optical alternatives using free carriers or Kerr-effect cross-phase modulation require adding and removing a bright pump.
Potential solutions: Silicon χ(2) can be induced through strain or a strong DC field, and high-χ(2) materials can be integrated directly with silicon.A reverse-biased PIN modulator demonstrated the DC-field approach using standard silicon manufacturing techniques.

E. High-Extinction Filters

High-extinction filtering must suppress intense pump light before it triggers nearby detectors. Existing filters achieve extreme rejection, but scattering around components has required physically separated filtering stages.

Filtering requirement: A quantum photonic device may need to prevent detectors from responding to approximately 10^10 pump photons, corresponding to about 1 nJ per pulse.The stated example tolerates no more than 1 false detection event in every 100 pulses.
Filter technologies: CROW-based filters have achieved 100 dB extinction, while Bragg-reflector and ring-resonator combinations enabled photon-pair correlation measurements without further external filtering.These approaches combine coherent and incoherent coupling between filtering substructures.
Integration boundary: Existing filtering demonstrations generally use two filtering stages on separate chips connected by optical fibre.The separation addresses pump leakage through cladding, fibre-coupling apparatus, and substrate rather than only the filter design.

F. Fibre-to-Chip Coupling

Silicon fibre-to-chip coupling relies mainly on edge couplers and grating couplers, each offering different practical trade-offs. Both approaches support pump injection and single-photon collection, while two-dimensional gratings also connect path and polarization encodings.

Edge coupling uses lensed fibre and an adiabatic spot-size converter, but its specialty fibre and small mode size hinder simultaneous multi-fibre coupling.
Grating couplers vertically couple cleaved fibre to waveguides and can be densely placed, but their grating structure inherently limits bandwidth.
Both edge and grating couplers are routinely used to inject pump fields and collect single photons.
Polarization-splitting two-dimensional gratings can convert stable on-chip path-encoded qubits into polarization-encoded qubits suited to short fibre and free-space links.

G. Delay Lines

Silicon delay lines provide memory for photonic protocols that need fixed-duration storage, while programmable and tunable alternatives address broader timing requirements. Low-loss delay demonstrations show practical progress, but footprint and control trade-offs remain relevant.

Programmable delay lines can use CROWs, photonic crystals, or sequences of unbalanced MZIs, while plasma-dispersion delays have inherent losses and lack demonstrated single-photon operation.
Many quantum repeaters, QKD systems, and one-way quantum computers require only fixed-duration memory, which can be implemented with a long waveguide.
Low-loss silicon-compatible delay lines achieve 0.024 dB/ns loss and delays up to 130 ns.
Low-loss optical fibre offers high bandwidth, no noise, no active control, and operation from millikelvin to room temperature, despite its potentially large footprint.

III. OTHER PLATFORMS

The review compares six leading on-chip quantum-photonic platforms across integration, optical loss, optical functionality, and photon sources. Silicon performs well overall, with strong passive and density-related properties but weaknesses in active optics and photon-pair sources.

The comparison assesses silicon, etched silica, direct-write, silicon nitride, indium phosphide, and lithium niobate across five technology categories.
Density and loss metrics use published bend radii and propagation losses, with bend footprint and loss per bend indicating density and total loss.
Photon-source assessment covers SFWM, SPDC, true single-photon emitters, and compatibility with application requirements across platforms.
Silicon scores well overall, with strengths in density, loss, and passive optics but weaknesses in active optics and photon sources.

IV. THE FUTURE OF QUANTUM OPTICS IN SILICON

Future silicon quantum photonics must integrate diverse components despite conflicting operating environments and manufacturing requirements. Cryogenic operation offers benefits for some devices, while unresolved material, detector, and source constraints shape the path toward larger architectures.

IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: The central engineering challenge is combining the reviewed components on one die or in one package while satisfying common-environment and mutually compatible manufacturing requirements.
IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: Cryogenic operation is increasingly likely because superconducting nanowire detectors operate around 2 K and no waveguide-coupled room-temperature detector has been demonstrated.
IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: Low-temperature silicon retains compact integrated optics and largely temperature-independent χ(3) effects, allowing SFWM, XPM, and FWM to continue working.
IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: A room-temperature detector is not inconceivable, but developing a waveguide-coupled, CMOS-compatible germanium avalanche photodiode may prove challenging.
IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: Integrated single-photon detection could provide access to previously unimaginable system architectures, including multiplexing of single photons.
IV. THE FUTURE OF QUANTUM OPTICS IN SILICON: SNSPDs, fast low-loss modulators, and engineered-substrate couplers require materials or processes beyond those in a typical CMOS foundry.

B. Two-Photon Absorption

Two-photon absorption is a major limitation for silicon nonlinear quantum photonics, reducing efficiency and heralding performance. Proposed solutions change either the source material or the operating wavelength.

Effect on nonlinear optics: TPA reduces nonlinear-optical efficiency when silicon uses the Kerr effect through SPM, XPM, or FWM.For SFWM, this loss reduces heralding efficiency and can become intolerable over time.
Effect on heralding efficiency: TPA causes extra uncorrelated signal loss in silicon SFWM, lowering the attainable heralding efficiency.The loss is intrinsic to SFWM in silicon at 1.55 µm.
Potential solutions: Two solution categories are replacing silicon with a material having reduced TPA or operating beyond silicon’s two-photon band edge.The review discusses both approaches as responses to the TPA problem.
Larger bandgap or alternative materials: Organic materials, amorphous silicon, and χ(2) materials such as GaN or AlN offer alternative routes to photon-pair generation.These options involve slot-waveguide χ(3) nonlinearities or SPDC instead of SFWM, with Raman-scattering considerations for amorphous materials.

2) Larger bandgap materials:

Longer-wavelength silicon devices can largely eliminate nonlinear absorption while retaining high nonlinearity, but they require new sources and detectors. Manufacturing infrastructure and silicon’s scalability support broader quantum-photonic deployment.

2) Larger bandgap materials:: 2-µm SOI devices maintain high nonlinearity while almost completely eliminating nonlinear absorption near silicon’s 2.2-µm two-photon bandgap.This approach shifts complexity to developing pump sources and detectors for longer wavelengths.
V. CONCLUSION: A fully chip-based QKD system has demonstrated that quantum photonic applications are approaching technological reach.The system used an indium phosphide transmitter and a silicon oxynitride receiver.
V. CONCLUSION: MPW services let small research teams share state-of-the-art silicon manufacturing costs for small-batch and prototype devices.They support incremental testing of designs, integration, and packaging without large-scale investment.
V. CONCLUSION: MPW services could accelerate silicon quantum photonics development in the near and medium terms.They enable the community to address remaining issues and test integration and packaging strategies incrementally.
V. CONCLUSION: Silicon technology is presented as the only viable route to assembling quantum-photonic systems with millions of components.The review connects this scalability claim with rising photonic capability and lower thresholds for linear optical quantum computation.