Semiconductor devices for entangled photon pair generation: a review

1. Introduction

• Further entanglement tests address the quantum–classical boundary and relativistic non-locality, including proposed satellite experiments.The authors note that experiments have closed all significant Bell-test loopholes, while these additional motivations remain.Source: §1 · p. 3
• Photons are suited to long-distance quantum communication because they travel at light speed and are almost immune to decoherence.They are described as flying qubits in a prospective quantum internet of links and processing nodes.Source: §1 · p. 3
• Large-scale quantum networks require practical, reliable, cost-effective entangled-photon sources, and semiconductors offer miniaturized integration of quantum components.The platform could support on-chip generation, manipulation, and detection of quantum light.Source: §1 · p. 4
• The review covers semiconductor devices for discrete two-level entanglement, including generation processes, measurement methods, recent results, applications, and prospects.Continuous-variable entanglement is explicitly outside its scope.Source: §1 · p. 4

2.1. Photonic Qubits

• Photonic qubits encode information in polarization, spatial or temporal modes, orbital angular momentum, or frequency, while several of these degrees of freedom also support higher-dimensional qudits.For fibre transmission, photons should use the second or third telecommunication window, around 1319 or 1555 nm; OAM entanglement has been demonstrated, while frequency qubits remain less developed because arbitrary measurement bases are difficult to choose.Source: §2.1 · pp. 5, 7

2.2. Two-qubit entanglement

• Entangled photon pairs arise when two indistinguishable creation processes produce a coherent superposition, enabling polarization or momentum entanglement with nonlinear-crystal sources.Examples include orthogonally oriented type-I crystals or a type-II crystal in a Sagnac interferometer for polarization entanglement, and selection of opposing emission directions for momentum entanglement.Source: §2.2 · pp. 8–10
• Time-bin and energy-time entanglement are more robust to fibre-propagation decoherence than polarization and momentum entanglement, with Franson experiments requiring τc, τdet << ∆t << τp.Time-bin pairs are generated by pumping a nonlinear crystal with two separated pulses; continuous-wave pumping yields energy-time entanglement through an undetermined pair-emission time, with interference between indistinguishable short-short and long-long paths.Source: §2.2 · pp. 10–11, Figure 7 · p. 11

2.3.1. Parametric processes in nonlinear materials

• Parametric down-conversion and four-wave mixing generate photon pairs without net energy transfer to the material, and efficient conversion requires energy and momentum conservation.Momentum conservation is phase matching, generally requiring refractive-index dispersion engineering; generated photons may have the same or orthogonal polarizations and may propagate collinearly or non-collinearly.Source: §2.3.1 · pp. 11–12

2.3.2. Biexciton-exciton cascade in optically active quantum dots

• Quantum dots can emit on-demand single photons under resonant excitation, and their biexciton-exciton cascade emits a biexciton photon followed by an exciton photon for photon-pair generation.The biexciton recombines to a bright exciton state before the remaining exciton recombines; the cascade is one of the schemes used to generate entangled pairs from quantum dots.Source: §2.3.2 · pp. 13–14

2.4. Indistinguishability of photons – HOM effect for path entanglement

• Completely indistinguishable photons entering opposite ports of a 50/50 beam splitter exit together, forming a path-entangled N00N state through Hong-Ou-Mandel interference.Indistinguishability requires the photons to match in all degrees of freedom except their input directions; HOM interference also underpins entanglement swapping and linear-optical CNOT gates.Source: §2.4 · pp. 14–16
• AlGaAs waveguides showed raw same-pair HOM visibilities of 0.85 and 0.861 ± 0.027, limited by facet reflections; antireflection coatings could in principle enable near-perfect visibility.The reflections arise from the large refractive-index difference between AlGaAs and air.Source: §2.4 · p. 16
• Silicon waveguides demonstrated raw HOM visibility of 0.73 for photons from independent sources and generated path entanglement at 0.945 ± 0.003 raw visibility (1.000 ± 0.004 net).For heralded photons from different pair sources, narrow spectral filtering or engineering frequency-uncorrelated pair emission is needed to ensure indistinguishability.Source: §2.4 · p. 16
• Quantum-dot photons reached indistinguishabilities of 0.995 ± 0.007 and 0.9956±0.0045 from successive emissions, while 0.921 ± 0.005 was achieved across a 14.7 µs interval.The section also reports 0.81 for photons from subsequent excitations, 0.97 ± 0.02 under resonant excitation, and entangled-pair values of 0.86 ± 0.03 for XX photons and 0.71 ± 0.04 for X photons. Indistinguishability decreases with longer emission intervals because of dephasing, including charge fluctuations near the dot.Source: §2.4 · pp. 16–17
• For separate quantum dots, HOM indistinguishabilities rose from 0.181 ± 0.004 and 0.33 ± 0.01 to 0.51±0.05 [87], while remote-QD interference reached 0.82 ± 0.02 with coherently scattered photons and 0.87 ± 0.04 with Raman photons.Separate dots are harder to match because even one elementary charge can significantly shift emitted-photon frequency; reported values use varied analyses and corrections and should be compared with care.Source: §2.4 · p. 17
• Fattal et al. created polarization entanglement by assigning opposite polarizations to indistinguishable photons, interfering them at a beam splitter, and postselecting detections at opposite output ports.Indistinguishability can enable entanglement generation through quantum interference beyond the HOM effect itself.Source: §2.4 · p. 17

• Parallel polarization analyzers yield total correlation between the two photons’ measurement results, while the EPR argument treats predictable outcomes as evidence of pre-existing hidden variables.The section frames locality as the assumption that one side’s analyzer setting and outcome cannot influence the other side’s outcome.Source: §3 · pp. 17–19
• Bell inequalities bound correlations predicted by local realistic models; a Bell state can reach S = 2√2 and violate the bound of 2 for specific analyzer directions.Violation is a standard test for strongly entangled states, though not all entangled states violate Bell inequalities.Source: §3 · pp. 19–20
• Quantum tomography reconstructs a density matrix from joint-projection statistics, providing a complete description of the quantum state.Coherent superpositions have off-diagonal density-matrix terms, unlike statistical mixtures.Source: §3 · p. 20
• Fidelity measures overlap with a known state, Bell-state fidelity above 1/2 witnesses entanglement, and entropy quantifies state mixedness.The section also introduces concurrence, negativity, and tangle as measures of two-qubit entanglement; the measures described range from 0 for separable states to 1 for maximally entangled states.Source: §3 · pp. 20–21

4.1.1. Phase-matching techniques

• AlGaAs lacks natural birefringence, so its second-order frequency conversion requires engineered phase matching; approaches include form-birefringence, modal matching, and quasi-phase matching.Modal phase matching compensates phase-velocity mismatch through multimode waveguide dispersion; unlike form-birefringence matching, it requires no oxidation and is compatible with electrical injection.Source: §4.1 · p. 21, §4.1.1 · pp. 22–23

4.1. AlGaAs-based sources using three-wave mixing

• The AlGaAs platform combines large second-order susceptibility, established processing, a direct band gap, and potential monolithic integration with superconducting nanowire detectors.The review’s scope is entanglement between discrete two-level systems, not continuous-variable entanglement.Source: §4.1 · p. 21

4.1.1. Phase-matching techniques

• QWI periodically intermixes quantum wells to suppress χ(2), avoiding etch-and-regrowth; super-lattice designs enabled SPDC photon-pair generation under continuous-wave pumping.Disk resonators can obtain quasi-phase matching through AlGaAs crystal symmetry; second-harmonic generation was reported, while SPDC remained under investigation.Source: §4.1.1 · pp. 24–25

4.1.2. Entanglement generation

• Counterpropagating sources reached raw Bell-state fidelity 0.83 ± 0.04 and concurrence 0.68 ± 0.07; pump shaping can tailor frequency correlations.Difference-frequency generation enabled high-resolution reconstruction of the two-photon joint spectrum.Source: §4.1.2 · pp. 25–26
• Bragg-modal sources demonstrated polarization entanglement with S = 2.61 ± 0.16 in a Bell-type experiment, while a separate measurement reported concurrence C = 0.52.Energy-time entanglement was also tested with a Franson experiment, demonstrating simultaneous emission of indistinguishable and entangled photon pairs.Source: §4.1.2 · pp. 26–27
• QWI waveguides generated energy-time-entangled photons by a type-I process with Bell parameter S = 2.687 ± 0.013.The review notes that recent fabrication advances reduced optical losses that had hindered this approach.Source: §4.1.2 · p. 27

4.2. Silicon-based sources using four-wave mixing

• Silicon’s strong third-order nonlinearities support four-wave mixing, while two-photon absorption constrains operation when photon energies exceed half the bandgap energy.The section defines Fn = n2/(λβ) as a figure of merit and introduces a comparison of nonlinear optical coefficients at 1.5 µm.Source: §4.2 · pp. 27–28, Table 2 · p. 28
• A theoretical study identified silicon SFWM as a source of highly correlated photon pairs with high spectral brightness, and later work demonstrated correlated pairs in silicon nanowires.The silicon Raman spectrum lies 15.6 THz from the pump and is about 100 GHz wide, allowing SpRS photons to be suppressed by choosing signal and idler frequencies away from the Raman peak.Source: §4.2 · p. 28

4.2.1. Straight nanowire waveguides

• Time-bin interference visibility in silicon-wire sources rose from larger than 0.73 to 0.95 with mode size converters that reduced outcoupling losses.The first silicon-wire entanglement demonstration used a centimetre-long waveguide pumped by a CW telecom laser modulated into double pulses at 100 MHz.Source: §4.2.1 · p. 28
• The integrated nanowire source generated polarization entanglement with concurrence C = 0.88 ± 0.02 after a polarization rotator made TE and TM photon pairs indistinguishable in time.The source uses two nonlinear nanowires with a silicon-wire polarization rotator between them.Source: §4.2.1 · p. 29, Figure 15 · p. 29

4.2.2. Microcavities

• A silicon-on-silica ring resonator produced energy-time entangled photons and violated Bell’s inequality by more than seven standard deviations, with an internal pair generation rate exceeding 10^7 Hz.The device occupied 300 µm^2; a CROW source reached 420 µm in length, with visibility limited by noise photons from optical losses.Source: §4.2.2 · p. 30

4.2.3. Path-entanglement generation

• Olislager and coworkers converted path entanglement from two silicon waveguides into polarization entanglement using a 2D grating coupler.Four-wave mixing produces horizontally polarized pairs in each waveguide before the coupler maps the two paths to polarization states.Source: §4.2.3 · pp. 30–31, Figure 17 · p. 31
• A chip combining two silicon FWM sources with a reconfigurable phase shifter generated degenerate and non-degenerate path-entangled pairs, with on-chip visibility of 0.945 ± 0.003 without background noise subtraction.The sources are interferometrically combined, and the phase shift controls bunching or splitting at the output.Source: §4.2.3 · p. 31, Figure 18 · p. 31

4.3.1. Entanglement generation via the biexciton-exciton cascade

• The biexciton–exciton cascade produces polarization-entangled photons through two recombination paths with opposite circular polarizations; on-demand generation was later achieved with resonant two-photon excitation.The cascade was presented by Benson et al., and on-demand generation was reported using a two-photon excitation scheme.Source: §4.3.1 · pp. 32–33
• Fine-structure splitting drives exciton-spin precession, so time-integrated measurements cancel opposite-phase states; tuning or feedback can mitigate this constraint.Approaches include symmetric growth, annealing, electric or magnetic fields, and strain tuning.Source: §4.3.1 · pp. 33–34
• Zero biexciton binding energy makes biexciton and exciton photons overlap across pathways, but time reordering must erase arrival-time information; filtering can raise entanglement at an efficiency cost.The complete scheme, including tuning the binding energy and reordering photons, had not been experimentally demonstrated.Source: §4.3.1 · pp. 35–36
• Time-bin entanglement uses early and late excitation pulses with controlled probabilities; resonant two-photon excitation demonstrated the approach but leaves double-excitation multiphoton emission unsuppressed.An alternative converts polarization entanglement into time-bin entanglement and allows on-demand pair generation.Source: §4.3.1 · p. 37
• Leapfrog emission bypasses the exciton and can be enhanced by a cavity, but competes with cascaded recombination and no entangled-pair demonstration had been reported.A narrower cavity can favor the direct two-photon transition, at the expense of brightness.Source: §4.3.1 · p. 38

4.3.2. Quantum dots in photonic microstructures

• A double-micropillar source achieved a Purcell factor of 3–5 and 0.12 collection efficiency per pair; concurrence was 0.267 without temporal selection and 0.387 with it.The measured fidelity to a maximally entangled state was 0.63 without temporal selection.Source: §4.3.2 · pp. 39–40
• An electrically driven entangled-light-emitting diode reported a maximum time-gated fidelity of 0.87 ± 0.04; piezoelectric strain tuning can remove fine-structure splitting in quantum dots.The strain-tuned device was designed so almost any quantum dot in it could be used for entangled-photon generation.Source: §4.3.2 · p. 40, Figure 23 · p. 41
• Position-controlled structures yielded high-fidelity pairs: pyramid dots reached 0.721 ± 0.043 [162], droplet-epitaxy dots 0.86 ± 0.02, and nanowire dots 0.817 ± 0.002 without temporal selection [220].The nanowire source had a brightness of 0.0025 photon pairs per excitation into the first lens.Source: §4.3.2 · pp. 41–42

4.4. Comparison of performance

• PPKTP crystal sources in Sagnac loops reach Bell fidelities and purities close to unity, the highest entanglement quality reported among the compared sources.The review notes that AlGaAs and Si parametric down-conversion sources also demonstrate Bell-inequality violations, with S up to approximately 2.7.Source: §4.4 · pp. 42–43
• GaAs quantum dots grown by droplet etching reached Bell-state fidelity 0.94±0.01 and Bell parameter 2.64±0.01, with measurements limited by fabrication and detector imperfections.The review identifies technological developments as a prospect for further progress.Source: §4.4 · p. 43
• Quantum-dot sources can in principle approach unit brightness, whereas parametric down-conversion is practically limited to 0.001 to 0.01 pairs per excitation pulse to maintain high entanglement quality.The brightest source reported in the review achieved 0.12 pairs per excitation pulse and fidelity 0.63.Source: §4.4 · p. 43
• Nonlinear semiconductor sources commonly target the Telecom C-band, while quantum-dot entanglement has been reported from 780 nm to the Telecom O-band around 1300 nm.Parametric down-conversion operates at room temperature; the highest reported quantum-dot operating temperatures were 53 K with temporal post-selection and 30 K without it.Source: §4.4 · p. 43

5. Applications and prospects

• Entangled photon pairs with quantum memories enable teleportation between arbitrarily distant locations, overcoming propagation losses that limit direct-link transmission.They are also needed for device-independent quantum cryptography, although this application remains too demanding for present detector technologies.Source: §5 · pp. 43–44
• An AlGaAs source's broadband polarization-entangled emission has supported proof-of-principle quantum-key distribution among multiple user pairs using standard telecom components.The review notes that semiconductor entangled-photon sources have so far seen limited use in quantum communication protocols.Source: §5 · p. 44
• Integrated silicon devices have demonstrated path-polarization entanglement conversion and high-fidelity entanglement distribution between two quantum chips linked by optical fibre.Other integration examples include a tunable AlGaAs Mach-Zehnder interferometer and integrated pump filtering for silicon sources.Source: §5 · p. 44
• Quantum dots have supported hyper-entanglement in polarization and time-bin, while mode-entanglement from indistinguishable photons has been used for quantum sensing and boson sampling.The review also identifies spin-photon interfaces as relevant to quantum computing and quantum repeaters.Source: §5 · p. 45
• The review concludes that semiconductor optical technologies, including quantum-light sources, guided-wave circuits, and detectors, offer promising paths for communication, sensing, and simulation.It presents these as prospects for future photonic quantum-information technologies.Source: §5 · p. 45