Photonic Quantum Computers
Photonic quantum computers use individual photons as qubits, encoding quantum information in polarization, time-bin, or other degrees of freedom. Linear optical quantum computing (LOQC) uses beam splitters and single-photon detectors; measurement-based schemes use cluster states of entangled photons. Gaussian boson sampling machines from Xanadu demonstrated quantum advantage claims. PsiQuantum targets fault-tolerant photonic computing using silicon photonics manufacturing. No large-scale photonic gate-based system is yet commercially available.
GATE SPEED
Photons travel at ~c
CONNECTIVITY
Reconfigurable via integrated photonic s…
ENVIRONMENT
Room temperature (for many linear optic schemes)
How Photonic Qubits Work
Photonic quantum computers use individual photons as the physical carriers of quantum information. A photon's quantum state can be encoded in several degrees of freedom — polarization (horizontal/vertical), time-bin (early/late arrival), path (which waveguide carries it), or the continuous variables of the optical mode (quadrature amplitudes of the electromagnetic field). The choice of encoding determines the computational model and the optical components needed.
Linear optical quantum computing (LOQC), formalized by Knill, Laflamme, and Milburn in 2001, showed that universal quantum computation is achievable using only passive optical elements (beam splitters, phase shifters) plus photon-number-resolving detectors and single-photon sources. The catch: two-photon entangling gates (like CNOT) implemented with linear optics are inherently probabilistic, succeeding only a fraction of the time. This requires large resource overhead: many single photons are consumed to herald one successful gate. Measurement-based (cluster state) approaches and fusion-based quantum computing (FBQC) are alternative architectures that avoid the need for deterministic two-photon gates.
Silicon photonics offers a promising manufacturing platform: waveguides, beam splitters, ring resonators, and phase modulators can all be fabricated on a silicon chip using CMOS-compatible processes, enabling integration at chip scale. PsiQuantum is pursuing this route to large-scale fault-tolerant photonic quantum computing. Xanadu's Gaussian boson sampling (GBS) machines use squeezed light states rather than single photons, and have claimed quantum computational advantage for certain sampling problems.
Physical Implementation
Connectivity and Gate Operations
Connectivity
Photonic circuits implement arbitrary unitary transformations on multiple modes via networks of beam splitters and phase shifters — an N-mode optical circuit implements any N×N unitary. Integrated photonic switches allow reconfigurable routing. However, two-photon interactions (needed for universal quantum gates) are probabilistic in linear optics, requiring post-selection or feed-forward from detectors.
Native Gate Set
Linear optical networks: arbitrary single-qubit rotations (via beam splitters and phase shifts). Probabilistic two-qubit gates: type-I and type-II fusion gates used in FBQC. For continuous-variable (CV) computing: squeezing operations, displacement, homodyne/heterodyne measurement. Optical switches for feed-forward.
Gate Speed
Photons travel at ~c; operations limited by optical path length and detector timing (~ns)
Potential Advantages
- ✓Many optical components operate at room temperature — no dilution refrigerator for the photonic circuit itself
- ✓Photons do not suffer from thermal decoherence — the dominant noise source is photon loss, not thermalization
- ✓Natural compatibility with optical fiber for quantum networking, quantum key distribution, and distributed quantum computing
- ✓Silicon photonics manufacturing is compatible with existing semiconductor fabs, enabling chip-scale integration at wafer scale
- ✓High-speed operation — photons travel at c, and optical operations can be performed at GHz rates in principle
- ✓Gaussian boson sampling demonstrated claims of quantum computational advantage (Xanadu, 2020–2022)
Limitations & Challenges
- —Two-photon entangling gates are inherently probabilistic in linear optics — resource overhead scales badly with circuit depth
- —High-efficiency single-photon sources (>99%) remain an unsolved materials challenge; current best demonstrated ~99% in limited setups
- —Photon loss in waveguides (typically 0.1–1 dB/cm in Si) is the dominant error source and accumulates with circuit depth
- —No large-scale gate-based photonic quantum computer exists today; current systems are proof-of-concept or analog
- —Photon-number-resolving detectors still require cryogenic cooling (~1 K), partially eliminating the "room temperature" advantage
- —Classical feed-forward (measuring an output photon and conditioning subsequent gates on the result) requires ultrafast electronics within photon transit time
Scaling Considerations
PsiQuantum's FBQC (Fusion-Based Quantum Computing) architecture is the most detailed published scalability roadmap for photonic quantum computing. It uses resource states of a few entangled photons, fused together by two-photon gates ("fusion gates"), to build a large fault-tolerant cluster state. The key advantage is that resource state generation can be parallelized across many chips, and a small fraction of photon loss is tolerable with appropriate codes (e.g., foliated surface codes). The critical thresholds are: single-photon source efficiency ~99%, photon loss ~0.1 dB per component, and detector efficiency ~99%. Reaching these simultaneously is a major engineering challenge.
Coherence: Photons do not naturally decohere (no thermal noise), but photon loss and detector inefficiency are the dominant error sources.
Available QPUs
View all →Leading Companies
Workload Suitability
Full guide →Simulating molecular electronic structure to compute ground-state energies, reac…
Simulating solid-state physics phenomena including high-temperature superconduct…
Solving NP-hard combinatorial problems including vehicle routing, supply chain s…
Monte Carlo simulation acceleration via quantum amplitude estimation, derivative…
Training and inference on quantum neural networks (QNNs), kernel methods using q…
Shor's algorithm can factor large integers and solve discrete logarithms exponen…
WELL SUITED FOR
- ›Quantum communication and quantum key distribution (QKD) protocols
- ›Quantum network nodes and quantum repeater stations
- ›Boson sampling and molecular vibronic spectra simulation (GBS)
- ›Continuous-variable quantum machine learning
- ›Quantum-enhanced sensing using squeezed light
LESS SUITED FOR
- —General-purpose digital quantum computing at scale (not yet available)
- —Applications requiring deterministic two-qubit gates without resource overhead
- —Problems that benefit from the long coherence times of matter-based qubits
Frequently Asked Questions
Key References
- [1]A scheme for efficient quantum computation with linear optics — Nature / Knill, Laflamme, Milburn, 2001
- [2]Fusion-based quantum computation — Nature Communications / PsiQuantum, 2023
- [3]Quantum computational advantage with a programmable photonic processor — Nature / Xanadu, 2022