Source-linked AI summary
Integrated photonics on thin-film lithium niobate
Di Zhu, Linbo Shao, Mengjie Yu, Rebecca Cheng, Boris Desiatov, C. J. Xin, Yaowen Hu, Jeffrey Holzgrafe, Soumya Ghosh, Amirhassan Shams-Ansari, Eric Puma, Neil Sinclair, Christian Reimer, Mian Zhang, Marko Lončar
TL;DR
Existing integrated-photonics platforms do not simultaneously provide ultra-low loss, fast low-loss modulation, and efficient nonlinearities. This Review surveys thin-film LN photonics from materials and passive components to active devices, finding that LN modulators outperform bulk counterparts in power consumption, bandwidth, and size.
Problem
Existing integrated-photonics platforms cannot simultaneously support ultra-low propagation loss, fast low-loss optical modulation, and efficient all-optical nonlinearities.
Source: §1 · p. 2Method
The Review comprehensively surveys thin-film LN photonics, spanning materials, passive components, active devices, heterogeneous integration, and platform challenges.
Source: §8 · p. 73Results
Thin-film LN modulators have significantly outperformed bulk counterparts in power consumption, bandwidth, and size.
Source: §8 · p. 73Takeaways & Limitations
Thin-film LN integrated photonics is rapidly expanding LN applications from telecommunications and nonlinear optics to quantum photonics, cavity electro-optics, and piezo-optomechanics.
Source: §8 · p. 73LN is not compatible with current CMOS or silicon-photonics processes because of material contamination issues, while low-roughness etching remains challenging.
Source: §7.6 · p. 71
Abstract
from arXiv · showhide
Lithium niobate (LN), an outstanding and versatile material, has influenced our daily life for decades: from enabling high-speed optical communications that form the backbone of the Internet to realizing radio-frequency filtering used in our cell phones. This half-century-old material is currently embracing a revolution in thin-film LN integrated photonics. The success of manufacturing wafer-scale, high-quality, thin films of LN on insulator (LNOI), accompanied with breakthroughs in nanofabrication techniques, have made high-performance integrated nanophotonic components possible. With rapid development in the past few years, some of these thin-film LN devices, such as optical modulators and nonlinear wavelength converters, have already outperformed their legacy counterparts realized in bulk LN crystals. Furthermore, the nanophotonic integration enabled ultra-low-loss resonators in LN, which unlocked many novel applications such as optical frequency combs and quantum transducers. In this Review, we cover -- from basic principles to the state of the art -- the diverse aspects of integrated thin-film LN photonics, including the materials, basic passive components, and various active devices based on electro-optics, all-optical nonlinearities, and acousto-optics. We also identify challenges that this platform is currently facing and point out future opportunities. The field of integrated LNOI photonics is advancing rapidly and poised to make critical impacts on a broad range of applications in communication, signal processing, and quantum information.
1. Introduction
Integrated photonics seeks scalable optical solutions, but existing platforms do not simultaneously provide ultra-low loss, fast low-loss modulation, and efficient nonlinearities. Thin-film LN addresses these limitations through high-index-contrast nanophotonics and supports a broad device toolbox for classical and quantum applications, which this Review surveys from fundamentals to opportunities.
Motivation
- Existing integrated-photonic platforms cannot simultaneously support ultra-low propagation loss, fast and low-loss optical modulation, and efficient all-optical nonlinearities.Integrated photonics nevertheless promises low-cost, scalable solutions for communication, sensing, and computation.Source: §1 · p. 2
Lithium-niobate platform
- Bulk-LN devices were historically bulky and inefficient because material-processing difficulties forced low-index-contrast waveguides with weak confinement, large footprints, and reduced nonlinear efficiencies.LN remains attractive for its electro-, nonlinear-, and acousto-optic properties, wide transparency window, and relatively high refractive index.Source: §1 · p. 2
Thin-film LN advances
- Commercially available thin-film LN wafers and fabrication advances enabled ultra-low-loss, high-index-contrast nanophotonic waveguides and components including compact modulators, frequency combs, wavelength converters, and photon-pair sources.The platform can host optical and optomechanical cavities, filters, modulators, transducers, nonlinear sources, detectors, and quantum memories; Figure 1 illustrates this range.Source: §1 · p. 3, Figure 1 · p. 3
Review scope
- The Review covers thin-film LN from material properties and passive components through electro-optics, all-optical nonlinearities, acousto-optics, heterogeneous integration, and future challenges and opportunities.Its scope spans classical and quantum applications, including modulators, frequency combs, quantum transducers, wavelength conversion, Raman lasing, supercontinuum generation, and piezo-optomechanics.Source: §1 · pp. 3–4
2. Material and integrated passive optics
This section reviews the material properties and fabrication of thin-film lithium niobate, together with integrated passive components including waveguides, fiber couplers, resonators, and photonic crystal cavities. It covers their capabilities, fabrication approaches, remaining losses, and roles in enabling optical, nonlinear, electro-optic, and optomechanical interactions.
2.1. Material properties of lithium niobate
- LN offers 350 nm–5 μm transparency, n≈2.2 at 1550 nm, d33=-27 pm/V, r33≈31 pm/V, and supports nonlinear, electro-optic, Kerr-comb, and acousto-optic devices.Ferroelectric domain engineering enables nonlinear optical frequency conversion, while thermo-optic and photorefractive effects provide tuning mechanisms but may limit device operation.Source: §2.1 · pp. 4–5
2.2. Thin-film lithium niobate
- Smart Cut ion slicing, bonding, annealing, and polishing produce single-crystalline LNOI wafers, commercially available up to 6 inches, with direct bonding preferred for property recovery.Most LNOI uses lithium-deficient congruent LN; direct bonding permits higher annealing temperatures than adhesive BCB bonding.Source: §2.2 · pp. 6–7
2.3. Waveguides
- Thin-film ridge waveguides provide roughly an order-of-magnitude larger index contrast than diffused bulk-LN guides, enabling strong confinement and compact integration.Diffused guides have 10–100 μm2 mode areas and millimeter bending radii, whereas optimized thin-film processing reduced propagation loss from >6 dB/cm to 0.027 dB/cm, though sidewall scattering remains limiting.Source: §2.3.1 · p. 8, §2.3.2 · pp. 8–9
2.3.3. Hybrid waveguides
- Hybrid rib-loaded and bonded waveguides integrate LN with SiNx, Si, and other platforms, adding missing electro-optic and nonlinear functions but weakening interactions because only part of the mode overlaps LN.Propagation losses below 1 dB/cm and, in some cases, a few dB/m have been achieved across LNOI waveguide technologies.Source: §2.3.3 · p. 10
2.3.4. Fiber-to-chip coupling
- Fiber coupling remains a packaging bottleneck: two-step tapers reached 1.7 dB per facet and 0.54/0.59 dB per facet for TE/TM, while grating couplers reached -3.5 dB per coupler.End-fire coupling is broadband and polarization-insensitive, whereas grating coupling offers wafer-scale testing but narrower bandwidth and polarization sensitivity.Source: §2.3.4 · p. 12
2.4. Cavities
- Thin-film LN cavities enhance nonlinear-, electro-, and acousto-optic interactions; microdisks reach Q-factors near 10^7, while microrings and racetracks provide cleaner modes and measure waveguide loss.Thin-film microdisk Q-factors remain below pristine bulk-LN disks reaching 2×10^8, likely because of surface scattering and ion-slicing damage.Source: §2.4 · p. 12, §2.4.1 · p. 13, §2.4.2 · p. 13
2.4.2. Microring and racetrack
- Microring and racetrack resonators fabricated by dry etching, rib loading, CMP, and BIC waveguides have achieved intrinsic Q-factors approaching 10^7.These resonators support low-voltage and broadband electro-optic modulation, frequency conversion, and Kerr comb generation.Source: §2.4.2 · p. 14
- Coupled rings with electro-optic modulation form photonic molecules for frequency shifting and microwave-to-optical photon conversion.The same resonator platform has enabled ultra-efficient parametric wavelength conversion and electro-optic frequency comb generation.Source: §2.4.2 · p. 14
2.4.3. Photonic crystal
- Photonic crystal cavities combine high Q-factors and wavelength-scale mode volumes, enhancing nonlinearities while reducing footprint and enabling photon-level resonance tuning and resonant electro-optic modulation.One- and two-dimensional cavities and optomechanical crystals have been demonstrated on LNOI, including structures confining both optical and mechanical modes.Source: §2.4.3 · p. 14, Table 2 · p. 14
3. Electro-optics · 3.1. Basic formulation of electro-optic modulation · 3.2. Non-resonant electro-optic modulators
This section examines electro-optic modulation in thin-film lithium niobate, covering its basic formulation and non-resonant modulators. It addresses the trade-offs among voltage efficiency, bandwidth, loss, impedance, and device length, as well as reported advances and engineering challenges.
3.1. Basic formulation of electro-optic modulation
- The linear electro-optic effect is formulated by the Pockels tensor, with LN symmetry reducing its coefficients to contracted-index form and coupled-mode calculations using the induced permittivity change.The dominant coefficient is r33 ≈31 pm/V, producing Δn ≈−1/2 n_e^3r33E_z when the field and index change align with the Z axis.Source: §3.1 · p. 15
3.2.1. Common modulator configurations
- Common non-resonant configurations include phase modulators, push-pull Mach–Zehnder intensity modulators, IQ modulators, and Michelson interferometers, with X/Y-cut designs prevailing over more difficult Z-cut implementations.Phase modulators generate sidebands for frequency combs and shifting, while IQ modulators control amplitude and phase for coherent transmission.Source: §3.2.1 · p. 16, Figure 10 · p. 16
3.2.2. Performance metrics
- Non-resonant modulator evaluation spans Vπ, bandwidth, insertion loss, extinction ratio, linearity, and dynamic range, with measured LNOI IMD3 SFDRs of 90–100 dBHz^2/3 below reported bulk-LN values above 120 dBHz^2/3.Insertion loss reflects coupling, propagation, splitter, and metal absorption losses, while SFDR also depends on laser, amplifier, detector, and system noise.Source: §3.2.1 · p. 16, §3.2.2 · pp. 17, 21, 22
- Monolithic LNOI MZMs typically achieve 2.1–2.5 V-cm VπL, with 1.8 V-cm demonstrated at a 3.5 μm electrode gap and 1.4 V measured in a 20 mm device.Reducing electrode gaps improves voltage efficiency but increases optical and microwave loss, while wider waveguides and partially etched structures improve field overlap and bandwidth tradeoffs.Source: §3.2.2 · pp. 17–18, Figure 11 · p. 17
- Velocity matching, impedance matching, and microwave loss jointly influence bandwidth and voltage efficiency in high-frequency LNOI modulators.Segmented CPWs reduce typical loss from 0.7–1.1 dB cm-1 GHz-1/2 to 0.26 dB cm-1 GHz-1/2.Source: §3.2.2 · pp. 19–20
3.2.3. State of the art
- Thin-film LN modulators have demonstrated >100 GHz bandwidth at CMOS-compatible voltages and coherent data rates up to 320 Gbit/s, enabled by monolithic, loaded, and heterogeneous integration.Dry-etched LNOI, rib-loaded structures, and bonding to SOI or SiNx support both performance improvements and prospects for scalable multifunction integration.Source: §3.2.3 · p. 23
- A segmented-CWP-electrode modulator showed 1.8 dB EO roll-off at 50 GHz and an extrapolated 3 dB EO bandwidth of approximately 175 GHz.The cited device was not traveling-wave, and all modulators were measured near 1550 nm.Source: §3.2.3 · pp. 23–24
- Another device was measured to 500 GHz without significant roll-off from index mismatch, while some bandwidth values were estimated from electrical bandwidth, calculations, or prior work.Some reported bandwidths were instrument-limited or extrapolated beyond 200 GHz.Source: §3.2.3 · p. 24
- Low Vπ and high bandwidth in non-resonant EO modulators require long devices, with few-volt Vπ typically needing centimeter-scale lengths.This length requirement limits their suitability for applications where compactness is important.Source: §3.2.3 · p. 24
3.3. Resonant electro-optic modulators
Resonant electro-optic modulators enable compact, strongly modulating thin-film LN devices by electrically shifting optical resonances. Their central tradeoff is high modulation efficiency versus reduced bandwidth, with design strategies extending applicability.
3.3. Resonant electro-optic modulators
- Resonant modulators generally mix amplitude and phase modulation, with behavior determined by the coupling ratio η=κ_e/κ_i and detuning from resonance.In the overcoupled, on-resonance regime, transmission is nearly lossless and the device can approximate phase modulation, whereas purely linear phase response requires non-resonant modulators.Source: §3.3 · pp. 24–25
- Resonant modulators provide strong modulation in compact devices because circulating optical fields repeatedly interact with electrode-generated microwave fields.Their effective interaction length can reach approximately 1 m for Q=10^7 resonators near 1.5 μm, about ten times that of typical non-resonant modulators.Source: §3.3 · p. 25
- For Q=10^7 resonators near 1.5 μm, the decay rate is κ=2π×20 MHz, illustrating the bandwidth penalty accompanying compactness and efficiency.The optical field responds only at rates set by the resonator’s total decay rate.Source: §3.3 · p. 25
- Coupled resonators and coupling modulation can tailor or reduce the efficiency–bandwidth tradeoff, while high-Q devices suit optical switching and beam steering despite limited bandwidth.They offer lower power dissipation and higher bandwidth than thermal tuning for these applications.Source: §3.3 · p. 25
- Thin-film LN resonant modulators span monolithic and hybrid platforms, with typical low-frequency tuning rates of 1–20 pm/V shaped by electrode spacing and coverage.Photonic-crystal cavities can provide near-optimal electrode coverage, while Bragg reflectors can form resonant modulators or tunable filters.Source: §3.3 · pp. 25–26
3.4. Electro-optic frequency combs
Integrated LN electro-optic combs use Pockels-effect phase modulation to provide coherent, tunable, spectrally flat combs, while resonant enhancement substantially broadens their bandwidth. These sources support on-chip dual-comb spectroscopy with flexible repetition rates and operating frequencies.
3.4.1. Non-resonant electro-optic comb
- Over 40 spectral elements with 30-GHz spacing were generated from a 2-cm X-cut LNOI phase modulator driven at approximately 4Vπ.The modulation index controls comb-spectrum shape, while tight waveguides, low optical loss, and strong microwave–optical overlap enable broad generation; cascading modulators improved flatness to 0.89 dB for limited lines.Source: §3.4.1 · p. 27
3.4.2. Resonant electro-optic comb
- Injecting a CW laser into a high-Q LNOI racetrack cavity with approximately 10-GHz FSR, modulated at its FSR, generated 900 comb lines in the telecommunications region, with spacing tunable to FSR multiples.Dispersion engineering can help overcome resonant EO-comb bandwidth limitations, while current efficiency and output power are relatively low and higher-Q resonators may improve bandwidth.Source: §3.4.2 · p. 28
3.4.3. Spectroscopy
- LNOI electro-optic combs enabled dual-comb spectroscopy with approximately 10-GHz resolution, 5×10^-3-s coherence time, 10% residual, and 3.4% standard deviation in acetylene measurements.Shared optical pumping and phase-locked microwaves provide mutual coherence; cascading resonant and non-resonant combs further produced a 3-GHz comb and enabled probing regions 6.6 THz apart.Source: §3.4.3 · p. 29
3.5. Coupled-resonator-based modulators · 3.6. Cavity electro-optics · 3.7. Synthetic dimension based on electro-optic modulation
Thin-film LN electro-optic modulation enables coupled-resonator frequency control, cavity-based microwave–optical transduction, and high-dimensional synthetic frequency crystals. These advances support applications in coherent photon control, quantum interconnects, and topological photonics.
3.5. Coupled-resonator-based modulators
- Electro-optically modulated coupled LNOI microrings formed a photonic two-level system exhibiting Rabi oscillation, Ramsey interference, Autler–Townes splitting, and Stark shifts.The photonic molecule could also be programmed into bright–dark mode photon pairs for photon storage and retrieval.Source: §3.5 · p. 30, Figure 18 · p. 30
- ~99% frequency-shift efficiency with 0.45 dB insertion loss was demonstrated, while the coupled-ring device could also operate as a tunable frequency beam splitter.The splitting ratio and frequency are controlled by microwave power and frequency, respectively, and cascaded shifting beyond 100 GHz is proposed with several-tens-of-GHz drive.Source: §3.5 · p. 31
3.6. Cavity electro-optics
- Cavity electro-optics combines optical resonators with low-loss microwave cavities to strengthen microwave–optical interactions, trading enhanced modulation for bandwidth limited by the microwave-resonator linewidth.Microwave resonance is matched to the splitting between hybrid optical modes, and a DC bias capacitor can tune that splitting.Source: §3.6 · p. 31, Figure 19 · p. 32
- Cavity electro-optic transducers directly convert microwave and optical fields, offering a route to connect superconducting qubits with optical networks while avoiding intermediate mechanical excitation.Optical links mitigate the thermal noise and attenuation that constrain microwave quantum networks at room temperature.Source: §3.6 · p. 32
- 1% on-chip microwave-to-optical conversion efficiency was achieved in thin-film LN cavity electro-optic transducers by mitigating photorefractive effects and increasing pump power.Earlier devices demonstrated efficiencies up to 2×10^-5, while near-unity efficiency is expected to be feasible in optimized devices.Source: §3.6 · p. 32
3.7. Synthetic dimension based on electro-optic modulation
- Electro-optic modulation couples neighboring frequency modes into a synthetic lattice, enabling frequency crystals and motivating LNOI studies of topological and other high-dimensional photonic phenomena.Theoretical proposals include Bloch oscillation, photonic Weyl points, synthetic gauge fields, and combinations of frequency with physical or other synthetic dimensions.Source: §3.7 · pp. 33–34
- High-dimensional frequency crystals with dimensions up to four were theoretically investigated and experimentally realized on LNOI, including measured one- to four-dimensional densities of states and coherent random walks.Coherent random walks were measured in one- and two-dimensional crystals.Source: §3.7 · p. 33, Figure 20 · p. 34
4. All-optical nonlinearity
Thin-film LN supports broadband all-optical nonlinear processes through its second- and third-order nonlinearities, low loss, wide transparency, and absent multiphoton absorption above 800 nm. Phase matching, dispersion engineering, domain poling, and cavity integration expand this platform toward classical and quantum applications.
Material platform
- LN’s χ^(2) and χ^(3) nonlinearities, wide ultraviolet-to-mid-infrared transparency, reduced propagation loss, and absent multiphoton absorption above 800 nm enable broadband processes at modest pump powers.The reviewed processes include harmonic generation, sum- and difference-frequency generation, optical parametric generation, and four-wave mixing.Source: §4 · p. 34
Coupled nonlinear processes
- Interplay between χ^(2)- and χ^(3)-based processes can provide cascaded effective third-order nonlinearity, self-referencing, dispersive-wave tailoring, and soliton compression.These capabilities motivate emerging applications in both classical and quantum fields.Source: §4 · pp. 34–35
Phase matching and dispersion
- Engineering group velocity, group-velocity dispersion, modal phase matching, and quasi-phase matching through ferroelectric domain poling enables nonlinear interactions across broad wavelength ranges.Dispersion engineering and cavity integration position periodically poled LN as a platform for emerging nonlinear science and applications.Source: §4 · pp. 34–35
4.1. Dispersion engineering
Thin-film LN enables precise engineering of group velocity, GVD, and higher-order dispersion through waveguide geometry and cladding, supporting phase-matched nonlinear interactions and applications including frequency combs and supercontinuum generation.
4.1. Dispersion engineering
- Thin-film LN waveguide geometry provides a wider and finely tunable GVD range than bulk LN, enabling near-zero or anomalous dispersion.Width, etch depth, film thickness, and etch angle can tailor dispersion for targeted nonlinear applications.Source: Figure 21 · p. 35, §4.1 · p. 36
- Cladding configurations, including oxide-clad and suspended structures, further modify the TE-mode dispersion of a fixed LN waveguide geometry.Suspended structures also enable use of LN’s piezoelectric properties.Source: §4.1 · p. 36, Figure 22 · p. 36
- Anomalous GVD supports positive four-wave-mixing gain, while engineered dispersion enables Kerr microcombs, supercontinuum generation, and electro-optic microcombs in LN.PPLN additionally provides control of effective Kerr nonlinearity through the sign of phase mismatch.Source: §4.1 · p. 36
- Group-velocity and higher-order dispersion engineering improves broadband second-harmonic generation, cascaded χ(2)-based supercontinuum conversion, and control of dispersive-wave emission.Zero group-velocity mismatch benefits broadband SHG, while dispersion-operator engineering controls dispersive-wave positions for visible and mid-infrared generation.Source: §4.1 · p. 37, Figure 23 · p. 37, Figure 24 · p. 37
- Efficient χ(2) three-wave mixing requires phase matching, Δk = k1 + k2 − k3 = 0, to transfer energy from the pump to target wavelengths.Material and waveguide dispersion otherwise create momentum mismatch and cause energy to oscillate between wavelengths.Source: §4.1 · p. 38
4.2. Phase matching
Efficient frequency conversion requires engineered phase matching, achieved through birefringent, intermodal, quasi-phase-matching, cyclic, metasurface-assisted, and periodically grooved approaches. These methods trade polarization or modal overlap, bandwidth, efficiency, fabrication complexity, and loss.
4.2.1. Birefringent phase matching
- Birefringent phase matching breaks the effective-index ordering by assigning ordinary-axis polarization to one or more lower-frequency waves, but is rarely used in integrated LN photonics.Type-I and type-II configurations are common in bulk crystals, whereas type-0 interactions using LN’s largest χ^(2) component require other methods.Source: §4.2.1 · p. 38
4.2.2 Intermodal phase matching
- Intermodal phase matching enables type-0 frequency conversion by using a different spatial profile, typically a higher-order mode at ω3, to break the effective-index inequality.Dispersion engineering via waveguide dimensions targets specific frequencies, and symmetry often favors fundamental and third-order modes. Demonstrations include etched LN waveguides and resonators.Source: §4.2.2 · pp. 38–39
4.2.3. Quasi-phase matching
- Quasi-phase matching restores unidirectional frequency conversion by periodically inverting χ^(2), allowing all interacting waves to share polarization and spatial modes.Periodic poling supplies grating momentum K = 2π/Λ opposite to the intrinsic mismatch, but thin-film LNOI fabrication is more complicated than bulk PPLN.Source: §4.2.3 · pp. 39–40
4.2.4. Other phase-matching approaches
- Cyclic phase matching broadens the phase-matching bandwidth in LN rings or disks, while metasurface and grooved approaches provide additional momentum engineering with efficiency or loss tradeoffs.Cyclic phase matching demonstrated 0.1%/mW normalized SHG in a double-resonant X-cut LN microresonator, while deeper grooves increase conversion but also Bloch scattering loss.Source: §4.2.4 · pp. 42–43
- 1,660% W^-1cm^-2 normalized SHG efficiency was achieved at telecom wavelengths using a dielectric metasurface that enables broadband phase-matching-free conversion.Silicon nanoantennas create a unidirectional phase gradient and couple second-harmonic light into higher-order modes, though collecting those modes is generally inefficient.Source: §4.2.4 · p. 42
4.3. Second-order nonlinear wavelength conversion
Thin-film LN waveguides and resonators support efficient second-order processes including SHG, SFG, DFG, OPO, and SPDC. Periodic poling, nanophotonic confinement, and resonant enhancement have enabled record conversion efficiencies and quantum light generation.
4.3. Second-order nonlinear wavelength conversion
- Thin-film LN implements SHG, SFG, DFG, OPO, and SPDC in periodically poled waveguides and microcavities, extending nonlinear wavelength conversion from bulk crystals to integrated devices.These demonstrations continuously push nonlinear efficiency higher through periodic poling and nanophotonic integration.Source: §4.3 · p. 43, §4.3.1 · p. 43, §4.3.2 · p. 45
4.3.1. Traveling-wave processes
- 4600%/W·cm2 normalized SHG efficiency was achieved in a 300 μm monolithic LN ridge waveguide, while nanophotonic confinement enables shorter devices and broader phase-matching bandwidths.Nanophotonic PPLN waveguides reduce mode areas by more than an order of magnitude relative to bulk PPLN, enhancing conversion efficiency at fixed parameters; type-I modal phase matching offers greater thermal tuning but lower efficiency.Source: §4.3.1 · pp. 43–44
4.3.2. Resonant-enhanced processes
- 5,000%/mW SHG was achieved in a Z-cut PPLN microring, with single-photon nonlinearity approaching 1%, while other resonators demonstrated cascaded nonlinear processes.Resonant conversion requires energy and momentum conservation, strong nonlinear susceptibility and mode overlap, small mode area, high quality factors, and critical coupling; periodic poling can relax momentum matching.Source: §4.3.2 · pp. 45–46
4.3.3. Spontaneous parametric down conversion
- SPDC on thin-film LN produced a coincidence-to-accidental ratio of 668 ± 7 at 12 MHz and generated energy-time-entangled pairs across wavelength channels separated by more than 120 nm.These sources support frequency-domain multiplexing and heralded single photons, but spectral purity of heralded photons has not yet been demonstrated on thin-film LN; filtering or dispersion engineering may be required.Source: §4.3.3 · pp. 47–48
4.4. Cascaded second-order nonlinearity
Cascaded second-order processes use phase-mismatched SHG followed by downconversion to create a tunable effective Kerr nonlinearity in thin-film LN. Quasi-phase matching and dispersion engineering enable low-power nonlinear conversion, broad SHG, supercontinuum generation, and further nonlinear dynamics.
4.4. Cascaded second-order nonlinearity
- Two successive second-order processes—SHG upconversion and downconversion to the FW—produce a nonlinear phase shift proportional to [χ(2)]^2, yielding an effective Kerr nonlinearity.The process results from back-action of the SH wave at 2ω on the FW at ω.Source: §4.4 · p. 48
- Engineering the QPM condition tunes the effective χeff^(3) positive or negative, while near phase matching can produce giant χKerr^(3) for lower-power four-wave mixing.The phase mismatch is Δk = kSH − 2kFW − kQPM.Source: §4.4 · p. 48
- A multi-octave supercontinuum was generated with a few pJ pulse energy using cascaded second-order nonlinearity in an integrated PPLN waveguide.The associated supercontinuum spectra are shown at various pulse energies in Figure 29.Source: §4.4 · p. 48, Figure 29 · p. 49
- Thin-film LN’s wavelength-scale confinement reduces the effective mode area by more than an order of magnitude, enabling χ(3)-like dynamics at extremely low pump powers.The platform supports dispersion-engineered processes including soliton compression, Cherenkov radiation, Kerr microcombs, and controllable Raman-like frequency shifting, but many remain under study.Source: §4.4 · pp. 48–49
- Group-velocity-matched cascaded processes enabled less than 100 fJ SHG pulse energy and a 3-dB bandwidth > 110 nm, an order of magnitude broader than conventional PPLN technologies.The SHG bandwidth is governed by temporal walk-off, with Δω ∼ 1/|Δk′|L.Source: §4.4 · p. 48
4.5. Kerr comb
Thin-film LN enables compact, highly coherent Kerr comb sources with integrated electrooptic tuning and self-referencing, while Raman effects and high-power instabilities remain central challenges. Stable soliton combs exceeding 200 nm and 300 nm have been demonstrated by avoiding Raman scattering, with recent 4/5-octave spanning reported in a-Z-cut LN.
4.5. Kerr comb
- Kerr combs use χ(3)-enabled four-wave mixing at high circulating power, while soliton locking balances nonlinearity, dispersion, cavity loss, and gain.Soliton states can produce octave-spanning combs suitable for f-2f referencing.Source: §4.5 · p. 49
- Thin-film LN can integrate Kerr comb generation with electrooptic resonance tuning and on-chip f-2f self-referencing in a single photonic circuit.These capabilities leverage LN’s χ(2) nonlinearity and electrooptic properties, reducing reliance on bulky off-chip components.Source: §4.5 · p. 49, Figure 30 · p. 50
- Kerr combs spanning 700 nm were demonstrated on X-cut LNOI in both TE and TM modes using dispersion-engineered waveguides with anomalous dispersion.The TE spectrum exhibited strong Raman peaks because its Z-axis polarization uses LN’s large r33 component, and Raman scattering competes with four-wave mixing.Source: §4.5 · p. 49, Figure 30 · p. 50
- >300 nm and >200 nm soliton microcombs were achieved at 2 μm and 1.5 μm, respectively, by avoiding Raman scattering in Z-cut LN.Raman interference can also be mitigated using the non-polar axis or a larger free spectral range.Source: §4.5 · p. 50, Figure 30 · p. 50
- 4/5-octave Kerr combs with 335 GHz line spacing were reported in an a-Z-cut LN microresonator by raising the Raman oscillation threshold and using a large FSR.Annealing and cladding removal may further stabilize cavity resonance at high optical power by controlling photorefractive effects.Source: §4.5 · p. 50
4.6. Raman lasing
Raman oscillation in thin-film LN resonators exhibits polarization- and direction-dependent stimulated scattering, competes with four-wave mixing, and may affect nonlinear photonic circuits. Its strength and spectral features vary across devices, potentially because of fabrication differences.
Raman lasing
- LN Raman scattering comprises A- and E-symmetry phonon modes, with A modes polarized along the crystal Z axis and E modes in the X-Y plane, supporting forward and backward processes.Forward Raman scattering corresponds to a polariton effect involving strong photon–phonon mixing and electromagnetic radiation.Source: §4.6 · pp. 50–51
- Differences in Raman oscillation strength, peak position, and linewidth across thin-film LN devices may arise from annealing, etching, lithography, and wafer Li/Nb concentration variations.Stimulated Raman scattering in thin-film LN waveguides is therefore important for developing LN-based nonlinear photonic circuits.Source: §4.6 · p. 51
- Backward stimulated Raman scattering dominates in monolithic X-cut LN microrings and racetrack cavities, with strongest gain for pump polarization parallel to the crystal axis.This polarization also maximizes electro-optic and second-order nonlinear effects, potentially enabling strong pump-to-Stokes energy transfer at high optical powers.Source: Figure 31 · p. 51, §4.6 · p. 51
- Raman oscillations compete with four-wave mixing in continuously pumped LN microresonators, producing Raman–Kerr combs without observed soliton formation.The associated phonon modes are identified in the optical spectrum of the generated comb.Source: Figure 31 · p. 51, §4.6 · p. 51
4.7. Supercontinuum generation
Integrated LN waveguides enable coherent, broadband supercontinuum generation through strong χ(2) and χ(3) nonlinearities, tight confinement, and dispersion engineering. These capabilities support on-chip f-2f interferometry, ultrabroadband UV-to-mid-IR sources, and low-energy supercontinuum generation.
4.7. Supercontinuum generation
- Coherent two-octave supercontinuum and fCEO detection were achieved in a single dispersion-engineered LN waveguide through simultaneous second-harmonic generation and broadband spectral overlap.The LN platform combines intrinsic χ(2) and χ(3) nonlinearities with engineered dispersion to generate the required spectra on chip.Source: §4.7 · p. 52
- A reduced scalar nonlinear envelope equation models multi-octave LN-waveguide dynamics by combining quartic and cubic nonlinearities while accounting for birefringence, dispersion, and propagation loss [296, 379–382].In a 0.5-cm dispersion-engineered waveguide, 90 fs, 107 pJ pulses generated second- and third-harmonic components, self-phase broadening, and an 860 nm dispersive wave.Source: §4.7 · pp. 52–53, Figure 32 · p. 52
- An integrated LN f-2f interferometer directly detected and stabilized a mode-locked fiber laser’s fCEO, simplifying bulk PPLN interferometry and enabling a path toward fully integrated frequency combs.The waveguide output was sent directly to a silicon avalanche photodiode, with feedback stabilization of the carrier-envelope offset beatnote.Source: §4.7 · pp. 52–53, Figure 33 · p. 53
- 0.35–4.1 µm spectral bandwidth was generated with 240 pJ pulses by combining dispersive-wave, harmonic, and difference-frequency generation in one LN device.SHG, sum-frequency generation, and cascaded SHG extend the spectrum toward the visible and UV, while DFG reaches the mid-infrared for spectroscopy sources.Source: §4.7 · p. 53
- 2.5-octave supercontinuum generation was achieved with pulse energies as low as 10 pJ using cascaded χ(2) nonlinearities in an integrated PPLN waveguide.Quasi-phase-matched SHG provided >110 nm 3-dB bandwidth and effective self-phase modulation 200 times stronger than intrinsic χ(3) nonlinearity.Source: §4.7 · p. 54
5. Piezo-optomechanics
Piezo-optomechanics links microwaves, elastic waves, and light in thin-film piezoelectric materials, with LN offering strong piezoelectricity and diverse acoustic-mode platforms. Recent devices enable microwave-to-optical conversion, frequency shifting, and nonreciprocal transmission, while coupling, fabrication, thermal handling, and material-quality limitations remain.
5.1. Piezoelectricity and photoelasticity
- LN’s piezoelectric coefficient is highest among commonly used ZnO, AlN, GaAs, and LN materials, while the secondary electro-optic effect can exceed one third of total microwave-to-light coupling.Device design must jointly optimize piezoelectric and photoelastic tensors; crystal orientation can be tailored separately for each coupling using acoustically connected regions.Source: §5 · p. 54, §5.1 · pp. 54–55
5.2.2 Surface acoustic wave on LNOI / 5.2.3 LN on Sapphire
- LN-on-oxide confines optical and acoustic modes while tolerating more than 1 W microwave power, whereas LN-on-sapphire supports constructive shear-mode coupling and higher acoustic velocity.LNOI Rayleigh-like modes can suffer strain cancellation and poor thermal conduction; LN-on-sapphire remains limited by wafer bubbles reducing usable area to less than 50%.Source: §5.2.2 · pp. 56–57, §5.2.3 · p. 57
5.2.4 Suspended thin-film LN
- Suspended LN enables higher acoustic frequencies and Q factors, but piezoelectrically driven optomechanical crystals achieve only ~0.1% microwave-to-mechanical conversion, while slab resonators provide GHz operation and return loss below -10 dB.Suspended structures face robustness and power-handling constraints, with about 100 mW RF power applied to acoustic cavities.Source: §5.2.4 · pp. 57–58
5. Piezo-optomechanics / 5.2.1 Surface acoustic wave on bulk LN / 5.3.3 Nonreciprocal optical devices
- Piezo-optomechanics supports microwave-to-optical conversion, optical frequency shifting, and magnetic-field-compatible nonreciprocal devices, but nonreciprocal efficiency remains below state-of-the-art magnetic bulk isolators and circulators.Bulk-LN SAWs are low-loss and efficiently IDT-driven but provide weak high-frequency optical-acoustic overlap, while overlay waveguides improve photoelastic coupling yet scatter acoustic waves.Source: §5.2.1 · p. 56, §5.3 · p. 58, §5.3.3 · p. 60
5.3.1 Microwave-to-optical conversion
- 5% single-microwave-photon-to-optical-photon conversion has been demonstrated using LN piezo-optomechanical crystals, exceeding cavity electro-optic implementations through high mechanical Q factors and small mode volumes.LN devices can also deliver >80% microwave power to suspended slabs without an electromagnetic resonator, but lower photoelasticity limits optomechanical coupling relative to silicon.Source: §5.3.1 · p. 59
5.3.2 Integrated acousto-optic modulation and frequency shifting
- 10-100 times higher frequency-shifting efficiencies than pioneering thin-film AlN devices have been achieved by integrated LN AOFS, although efficiencies remain only a few percent versus commercial bulk AOFS.Integrated devices are limited by ~100 μm interaction regions and about 1 W RF power, compared with centimeter-scale bulk devices handling 10 W.Source: §5.3.2 · p. 59
6. Heterogeneous integration
Heterogeneous integration extends thin-film LN with mature photonic platforms and adds functions absent from LN, including detectors, emitters, rare-earth gain, and prospective III-V components. The section reviews hybrid modulators, detectors, single-photon emitters, rare-earth-ion devices, and III-V integration approaches.
6. Heterogeneous integration
- Heterogeneous integration combines LN’s nonlinear and electro-optic functionality with scalable Si or SiNx platforms and supplies LN circuits with otherwise missing detectors, sources, memories, lasers, amplifiers, and electronics.Integration is pursued through flip-chip or wafer bonding while leveraging established fabrication processes.Source: §6 · p. 60
6.1. Hybrid modulators
- Hybrid LN/Si and LN/SiNx modulators combine silicon scalability with LN electro-optic performance, including traveling-wave devices exceeding 100 GHz 3 dB EO bandwidth.Designs use mode transitions into LN for stronger modulation confinement, while rib loading with SiNx, Ta2O5, chalcogenide glass, or Si avoids LN etching and can approach monolithic performance.Source: §6.1 · pp. 60–61, Figure 36 · p. 61
6.2. Detectors
- Directly integrated SNSPDs and amorphous-silicon detectors demonstrate heterogeneous detection on LNOI, with amorphous-silicon responsivity of 22 mA/W to 37 mA/W from 635 nm to 850 nm.SNSPD integration remains fabrication-challenging, and partially etched LN gives approximately 0.1 dB/μm TE absorption, motivating longer nanowires, resonant enhancement, or flip-chip transfer.Source: §6.2 · pp. 61–62, Figure 37 · p. 62
6.3. Single-photon emitters
- LN photonic circuits have coupled transferred InAs/InP quantum-dot emitters with moderate 40.1% coupling efficiency, supporting future reconfigurable on-demand single-photon sources.Integrated LN modulators, beam splitters, and filters could combine emission, manipulation, and routing, while 2D-material emitters offer another deterministic-source direction.Source: §6.3 · pp. 62–63, Figure 38 · p. 63
6.4. Rare-earth ions
- Thin-film Tm3+:LN preserved bulk-like spectroscopic behavior, including 0.7 nm transition linewidths and 157 μs fluorescence lifetimes, while Er3+:LNOI rings produced Purcell enhancement of 3.Er3+-doped thin-film LN also demonstrated 18 dB net gain in a 3.6 cm spiral, 5 dB net gain in a 5 mm ridge waveguide, and telecom-wavelength lasing, although photorefractive effects require mitigation.Source: §6.4 · pp. 64–65
6.5. III-V materials
- III-V/LN integration could combine LN electro-optics with III-V lasers, amplifiers, and photodetectors for communications, microwave photonics, and LIDAR.Proposed routes include transferring III-V stacks onto LN, transferring LN onto III-V substrates, or pick-and-place placement with taper or butt/end-fire coupling.Source: §6.5 · p. 65
7. Challenges and Opportunities
Thin-film LN offers major opportunities in quantum, nonlinear, nonreciprocal, microwave-photonic, and wafer-scale integration, but charge-carrier stability and scalable manufacturing remain important challenges.
7.1. Charge carrier effects
- Charge-carrier effects cause bias or detuning drift that limits LNOI stability and power handling, although they can enable in situ tuning and strong optical nonlinearities.Photorefractive effects are stronger and faster in LNOI, while dielectric relaxation remains insufficiently studied; annealing, uncladded waveguides, doping, elevated temperature, and thermo-optic control offer mitigation with tradeoffs.Source: §7.1 · pp. 66–67
7.2. Quantum photonics
- Thin-film LN combines low loss, electro-optic control, wavelength flexibility, strong nonlinearities, and switchable components to support integrated quantum photonics across time, frequency, and path encodings.Electro-optic phase shifters reduce power and heat relative to thermal tuning, while fast switches, interferometers, frequency shifters, and PPLN sources support multiplexing and quantum processing.Source: §7.2 · pp. 67–68
7.3. Nonlinear photonics
- Thin-film LN’s strong χ(2), χ(3), and electro-optic effects enable Kerr and Pockels solitons, combs, chimera states, band solitons, and other nonlinear dynamics.Raman gain, scattering, four-wave mixing, and counterpropagating-mode coupling require further understanding and engineering but may enable broad, stable low-repetition-rate combs.Source: §7.3 · pp. 68–69
7.4. Magnetic-free non-reciprocity and optical isolation
- Magnetic-free optical isolation in LN can exploit directional χ(2) wavelength conversion, traveling-wave or interband electro-optic modulation, and acousto-optic momentum transfer.A PPLN difference-frequency-generation isolator has been demonstrated, while acoustic waves may provide stronger suppression of undesired transitions through larger phase mismatch.Source: §7.4 · pp. 69–70, Figure 42 · p. 70
7.5. Microwave photonics
- > 100 GHz bandwidth, few-volt half-wave voltage, < 0.5 dB on-chip insertion loss, and ~ 100 dB Hz2/3 SFDR make LNOI competitive for integrated microwave photonics.LNOI also provides narrow tunable resonant filters, broadband comb sources, and magnetic-free isolators, but native sources, amplifiers, and detectors require heterogeneous III-V integration.Source: §7.5 · pp. 70–71, Figure 43 · p. 71
7.6. Wafer-scale lithium-niobate photonic integration
- LN’s CMOS incompatibility, contamination risks, limited low-roughness etching, and reliance on electron-beam lithography constrain scalable foundry manufacturing.Heterogeneous integration improves process compatibility but may limit LN processing, whereas monolithic processing offers simpler fabrication and potentially higher performance.Source: §7.6 · pp. 71–72
- Below 6 nm etched-film thickness variation across demonstrated 4-inch and 6-inch monolithic LNOI wafers shows that scalable, low-cost manufacturing of high-performance LN chips is possible.These variations exceed current silicon-photonics levels but are already sufficient for large-scale LN photonic manufacturing and could support future foundry services.Source: Figure 44 · p. 72, §7.6 · pp. 72–73
8. Conclusion
This Review surveys integrated thin-film LN photonics across materials, passive and active components, heterogeneous integration, and key challenges, highlighting its rapid expansion from telecommunications to quantum and other emerging applications.
Conclusion
- Integrated thin-film LN photonics encompasses materials, passive components, electro-optic, all-optical nonlinear, piezo-optomechanical, and heterogeneous-integrated devices, while facing challenges needed to realize future opportunities.Promising quantum integration includes superconducting detectors, quantum-dot single-photon emitters, and rare-earth ions.Source: §8 · p. 73
- Thin-film LN photonics is rapidly advancing across telecommunications, nonlinear optics, quantum photonics, cavity electro-optics, and piezo-optomechanics, with industry efforts moving it toward high-yield, wafer-scale production.Its balanced material properties and industry-proven operation could support rapid deployment, volume scaling, and cost reduction.Source: §8 · p. 73
- Thin-film LN modulators outperform bulk counterparts in power consumption, bandwidth, and size, while nonlinear converters and down-conversion sources achieve orders-of-magnitude higher efficiencies than diffusion-based waveguides.The platform also demonstrates frequency combs, supercontinuum sources, Raman lasing, microwave-to-optical transduction, and acousto-optic modulators.Source: §8 · p. 73