Thin-film lithium tantalate (TFLT, or lithium tantalate on insulator) is an electro-optic photonic platform that keeps the strong Pockels effect of lithium niobate while shedding its photorefractive, DC-drift and power-handling weaknesses, and in 2026 it produced working modulators from 375 nm to telecom wavelengths, including a 110 GHz device carrying 460 Gbit/s.
Summary
Integrated photonics needs a material that changes refractive index instantly under an applied voltage. For a decade that material has been thin-film lithium niobate (TFLN), which is bonded as a sub-micron layer onto an oxidised silicon wafer and etched into waveguides. TFLN is fast and efficient, but it drifts under DC bias, suffers photorefractive damage at short wavelengths and high optical powers, and has strong birefringence. TFLT is the same architecture built from lithium tantalate (LiTaO3), a crystal isomorphic to lithium niobate: a thin single-crystal film on insulator, etched into rib or strip waveguides with metal electrodes alongside. The sources consistently attribute to it low photorefractive noise, a high optical damage threshold, reduced birefringence, reduced DC drift and a wide transparency window.
One mechanism behind the practical difference has been measured directly. In a one-to-one comparison of TFLT and TFLN modulators under identical conditions, TFLN produced distorted optical pulses with long temporal tails caused by charge activation and transport, while TFLT gave sharp, distortion-free edges because its defect-related charge carriers have a substantially larger activation energy, suppressing leakage-current charging. The same underlying stability shows up as DC-bias stability at 120 degrees C with no loss of modulation efficiency or bandwidth, and as an hour of stable operation at 5 dBm output at 532 nm, which the authors report TFLN counterparts of similar structure could not achieve.
The parameters that decide the platform are the usual photonic ones: propagation loss (reported below 0.1 dB/cm, and 0.05 to 0.06 dB/cm in one non-volatile demonstration), modulation efficiency expressed as the voltage-length product VpiL, electro-optic bandwidth, fibre-to-chip coupling loss, and the resonator quality factor for nonlinear and quantum work (loaded Q of 2.48 x 10^6 in a microdisk, 10^6 in a microring). A platform-specific constraint is the silicon handle wafer: its high dielectric constant spoils microwave-optical velocity matching and adds RF loss, which one group solved by undercutting the substrate to suspend the electrode region, reaching 110 GHz 3 dB electro-optic bandwidth at Vpi of 5.1 V and a 460 Gbit/s net data rate.
Beyond high-speed telecom modulation, the same films support ultraviolet modulation at 375 nm, visible-light links, cascaded Raman and parametric frequency conversion, polarisation control, non-volatile ferroelectric-domain phase trimming and correlated photon-pair generation. That breadth, all reported within a few months of 2026, is the strongest single signal about the platform.
Evidence base
- 21 July 2026: a suspended-electrode LTOI Mach-Zehnder modulator reached 110 GHz 3 dB electro-optic bandwidth, Vpi of 5.1 V and 460 Gbit/s net data rate, after undercutting the silicon handle to cut microwave loss and remove parasitic surface conductance.
- 10 August 2026: a direct one-to-one comparison under identical conditions showed distortion-free high-extinction pulses in TFLT versus charge-transport-induced distortion in TFLN, attributed to TFLT’s larger defect-carrier activation energy.
- 5 May 2026: the first integrated UV electro-optic modulator on TFLT achieved VpiL of 85 mV.cm at 375 nm, 22.7 dB extinction ratio, 1.3 dB insertion loss and Vpi of 4.2 V, claimed as up to four orders of magnitude better bandwidth per VpiL than bulk technologies.
- 14 May 2026: 532 nm TFLT modulators flat to about 50 GHz held 5 dBm modulated output for an hour where comparable TFLN devices could not, and drove 112 Gb/s over a 3 m and 64 Gb/s over a 9 m underwater link.
- 5 June 2026 (submitted 30 April 2026): TFLT waveguide and resonant modulators remained DC-bias stable at 120 degrees C with unchanged modulation and bandwidth, and showed a 10 per cent Vpi reduction at higher temperatures.
- 28 July 2026: monolithic non-volatile LTOI photonics demonstrated 0.05 to 0.06 dB/cm propagation loss, 137 resolvable phase positions across pi (about 0.007pi resolution), state retention through 10^6 write cycles and zero-static-power bias control of a modulator beyond 110 GHz.
- 2 June 2026: the first TFLT quantum light source produced telecom photon pairs at 24 MHz/mW^2 with heralded g2(0) of 0.071 plus or minus 0.004 at a 170 kHz heralding rate, in a microring with Q of 10^6.
Open questions
- What is LTOI wafer availability, size, film-thickness uniformity and cost relative to lithium niobate on insulator, and does any commercial foundry offer a qualified process?
- Can the 110 GHz suspended-electrode geometry and the 0.9 dB 3D laser-written couplers be replaced by volume-manufacturable equivalents without giving back the bandwidth and coupling loss?
- How large is the electro-optic efficiency penalty versus lithium niobate at equal footprint, given reported Vpi of 4.2 to 5.1 V, and does it force longer devices or higher drive power?
- Do the non-volatile ferroelectric-domain phase states survive the elevated temperatures and long dwell times of uncooled co-packaged optics, or do the 120 degrees C bias-stability and non-volatile tuning advantages conflict?
Assessment drafted 2026-08-31 from up to 11 KB sources using the technology-scorecard framework; scores are a draft read pending review.