Thin-Film Lithium Tantalate (TFLT)

• Thin-Film Lithium Tantalate (TFLT) is a platform using sub-micrometer LiTaO3 films on insulators for integrated photonic, electro-optic, and acousto-optic devices.
• Advanced fabrication methods, including precision lithography and annealing, yield devices with low birefringence, high speed modulation, and stability under demanding power and temperature conditions.
• TFLT platforms enable multi-functional applications—from ultraviolet and telecom modulators to nonlinear and quantum photonics—while process optimizations still target improved loss and integration challenges.

Thin-film lithium tantalate (TFLT) denotes sub-micrometer LiTaO$_3$ films on insulator, typically lithium-tantalate-on-insulator or LTOI, used as integrated photonic, electro-optic, acousto-optic, and nonlinear photonic device layers. Across recent work, TFLT is presented as a complementary platform to thin-film lithium niobate (TFLN), combining a strong Pockels effect with reduced DC drift, high optical power handling, reduced birefringence, and compatibility with wafer-scale processing. Reported TFLT devices now span traveling-wave and resonant modulators from the telecom band to 375 nm ultraviolet, topological acousto-optic modulators, periodically poled frequency converters, three-octave supercontinuum sources, microring-based quantum light sources, polarization controllers, optical phased arrays, and microwave-optical transducers (Powell et al., 2024, Sayem et al., 10 Apr 2026, Chen et al., 15 Apr 2026, Lin et al., 4 May 2026, Fan et al., 1 Jun 2026, Axline et al., 10 Jun 2026).

1. Platform definition and emergence

TFLT is the thin-film implementation of lithium tantalate, usually realized as a single-crystal LiTaO$_3$ membrane bonded to SiO$_2$ and a supporting substrate. Reported stacks include 200 nm x-cut TFLT on 2 $\mu$m SiO$_2$ for near-IR visible modulators, 300 nm x-cut TFLT for ultraviolet modulators, 400 nm x-cut TFLT on 4.7 $\mu$m SiO$_2$ for polarization control and microwave-optical transduction, 500 nm x-cut TFLT-on-insulator for stable Mach-Zehnder modulators and periodically poled devices, and 600 nm X-cut or z-cut films for high-speed modulators, topological acousto-optics, and broadband nonlinear photonics (Powell et al., 1 May 2025, Lin et al., 4 May 2026, Gao et al., 7 Jan 2026, Axline et al., 10 Jun 2026, Powell et al., 2024, Shelton et al., 24 Apr 2025, Chen et al., 15 Apr 2026, Li et al., 16 Apr 2026, Wang et al., 18 Dec 2025).

The recent literature repeatedly frames TFLT as more than a substitution of one ferroelectric crystal for another. In the topological acousto-optic modulator, TFLT is described as “not just a substrate choice” but the enabling material platform for an ultra-compact, high-efficiency, and unusually high-power device (Chen et al., 15 Apr 2026). In ultraviolet photonics, TFLT is presented as the first integrated UV electro-optic platform at 375 nm (Lin et al., 4 May 2026). In quantum photonics, it is used for the first TFLT quantum light source via spontaneous four-wave mixing (Fan et al., 1 Jun 2026). In microwave quantum interconnects, it supports the first integrated electro-optic microwave-optical transducers on TFLT (Axline et al., 10 Jun 2026). This trajectory suggests that TFLT has moved from a materials prospect to a platform with multiple experimentally validated device classes.

Industrial context is central to this emergence. Several papers emphasize that lithium tantalate thin films are already heavily used in commercial 5G RF front-end surface acoustic wave filters, implying mature wafer technology, established process control, and a lower barrier to high-volume manufacturing than is usually associated with newer photonic thin-film ferroelectrics (Chen et al., 15 Apr 2026). That manufacturing emphasis is reinforced by wafer-scale and 4-inch demonstrations, including hundreds of Cu-Damascene TFLT modulators on a wafer and thin-film LT transferred to a 4-inch fused-silica substrate for slow-wave electrodes (Lin et al., 7 May 2025, Li et al., 16 Apr 2026).

2. Material properties and physical basis

Several material attributes recur across the literature. First, LT is described as having a strong Pockels effect, with representative values such as an EO coefficient of order $30\ \text{pm/V}$, $r_{33} \approx 30~\text{pm/V}$, and, in UV modeling at 375 nm, $r_{33} = 35~\text{pm/V}$ (Powell et al., 2024, Powell et al., 1 May 2025, Lin et al., 4 May 2026). Electro-optic operation follows the standard Pockels relation

$_3$0

with corresponding phase shift

$_3$1

and this formalism underlies the TFLT Mach-Zehnder, coupling-resonator, and traveling-wave modulators (Sayem et al., 10 Apr 2026).

Second, LT is repeatedly distinguished by reduced birefringence. One comparison gives TFLT birefringence $_3$2 versus TFLN $_3$3, while another states that LT has $_3$4 lower birefringence than LN at 633 nm (Chen et al., 15 Apr 2026, Powell et al., 1 May 2025). The reported implications are reduced parasitic polarization and mode coupling, mitigation of stray scattering and crosstalk, and increased design flexibility for dense photonic routing and multimode or multipolarization circuits (Chen et al., 15 Apr 2026).

Third, LT is repeatedly associated with reduced photorefractive response and higher optical damage threshold than LN. In one TFLT acousto-optic study, this is attributed to stronger Ta–O bonds, lower density of intrinsic vacancies and defect states, and higher resistance to photon-induced charge displacement (Chen et al., 15 Apr 2026). In visible and near-IR comparisons, LT is described as having a larger bandgap, lower photorefraction, and higher optical damage threshold than LN; one paper cites $_3$5 for LT, while another contrasts LT cutoff $_3$6 with LN cutoff $_3$7 (Powell et al., 2024, Powell et al., 1 May 2025, Sayem et al., 10 Apr 2026). Separate nonlinear work describes LT as transparent from about 280 nm to 5.5 $_3$8m and, in z-cut material around 1560 nm, gives $_3$9 and $_2$0, again highlighting low birefringence (Wang et al., 18 Dec 2025).

These material characteristics have direct device consequences. Stable operation is reported at powers and temperatures that are problematic for many TFLN devices. TFLT modulators show constant switching voltage down to 10 mHz, negligible drift in UV operation at several kW/mm$_2$1, stable quadrature bias over hours to days, and stable operation at 120–125$_2$2C (Li et al., 16 Apr 2026, Lin et al., 4 May 2026, Sayem et al., 30 Apr 2026, Axline et al., 10 Jun 2026). This does not imply that all TFLT devices are automatically drift-free; rather, the literature consistently shows that the intrinsic material trend is toward lower carrier-drift and weaker photorefractive instability than LN, while interfaces, claddings, and process details remain consequential (Yue et al., 24 Mar 2026).

3. Thin-film stacks, fabrication, and integration strategies

TFLT fabrication is diverse but structurally coherent. Starting films are typically commercial x-cut or z-cut wafers from NanoLN or Inno Semiconductor, with LiTaO$_2$3 thicknesses from 200 nm to 600 nm on SiO$_2$4 and Si, although a 600 nm TFLT layer has also been transferred to a 500 $_2$5m fused-silica substrate (Chen et al., 15 Apr 2026, Li et al., 16 Apr 2026). Patterning uses electron-beam lithography or deep ultraviolet stepper lithography, followed by Ar$_2$6 ICP, ion-beam etching, or, in the UV work, a highly anisotropic wet chemical etch in H$_2$7O$_2$8/KOH/citric acid at 85$_2$9C (Powell et al., 2024, Powell et al., 1 May 2025, Li et al., 16 Apr 2026, Lin et al., 4 May 2026).

Annealing is a recurring process lever. Examples include 2 h in O$\mu$0 at 520$\mu$1C to improve sidewall and material quality, 500$\mu$2C for 2 h in air after oxide deposition for resonator stability, and 300$\mu$3C hotplate treatment to reduce trapped charges (Powell et al., 2024, Sayem et al., 31 Jan 2026, Powell et al., 1 May 2025). In oxide-cladded TFLT microrings, annealing improves Q by about $\mu$4 and reduces the resonance shift at 1 W intracavity power from $\mu$5 GHz to $\mu$6 MHz, yielding a stability factor of $\mu$7 (Sayem et al., 31 Jan 2026).

Process variants are already platform-defining. Periodic poling on 500 nm x-cut TFLT is reported to be robust across acoustic-grade and optical-grade film, multiple electrode metals, and the presence or absence of an oxide interlayer, using a single high-voltage pulse with peak voltage time of 10 ms or less and ramp-down time of 90 s (Shelton et al., 24 Apr 2025). Copper-Damascene metallization yields planarized TFLT modulators with Cu electrodes that have microwave losses approximately $\mu$8 lower than conventional Au and are suitable for chip-on-wafer hybrid bonding (Lin et al., 7 May 2025). LT-on-fused-silica fabrication enables a slow-wave electrode on a low-loss SiO$\mu$9 substrate, producing 64 GHz 3-dB EO bandwidth with $_2$0 V in an 18 mm device (Li et al., 16 Apr 2026).

A common misconception is that TFLT processing is only relevant for conventional waveguides and modulators. In fact, the same material has been integrated into non-suspended topological nanobeams, pulley-coupled high-Q microrings, air-clad and oxide-clad resonators, superconducting microwave resonators, and 42$_2$1Y-cut SH-SAW RF filters, indicating that TFLT now supports multiple photonic and electromechanical process modules rather than a single canonical flow (Chen et al., 15 Apr 2026, Fan et al., 1 Jun 2026, Axline et al., 10 Jun 2026, Anusorn et al., 15 May 2026).

4. Electro-optic and acousto-optic devices

Electro-optic operation in TFLT spans traveling-wave, lumped, resonant, and coupling-modulated devices from 1 $_2$2m to 375 nm. Reported architectures include 7 mm traveling-wave MZMs at 1071 nm, 18 mm slow-wave MZMs on fused silica, 5 mm near-IR visible MZMs at 737 nm, compact 2 mm coupling modulators, and 200 $_2$3m ultraviolet MZMs (Sayem et al., 10 Apr 2026, Li et al., 16 Apr 2026, Powell et al., 1 May 2025, Sayem et al., 31 Jan 2026, Lin et al., 4 May 2026).

Domain	Representative result	Source
1 $_2$4m traveling-wave EO modulation	$_2$5 V; less than 2 dB EO roll-off up to 50 GHz	(Sayem et al., 10 Apr 2026)
LT-on-fused-silica slow-wave MZM	64 GHz 3-dB EO bandwidth; $_2$6 V; 440.6 Gbps net PAM8	(Li et al., 16 Apr 2026)
Cu-Damascene TFLT MZM	100 GHz EO bandwidth for 6 mm; 416 and 540 Gbit/s line rates	(Lin et al., 7 May 2025)
Ultraviolet EO modulation	$_2$7 at 375 nm; 22.7 dB ER; 1.3 dB IL	(Lin et al., 4 May 2026)
Topological acousto-optic modulation	Footprint $_2$8; $_2$9; stable to 28 dBm on-chip optical power	(Chen et al., 15 Apr 2026)

The 1 $\mu$0m traveling-wave MZM establishes that TFLT can operate in a short-wavelength regime where LN photorefraction is especially problematic, while preserving low drive voltage and flat EO response to 50 GHz (Sayem et al., 10 Apr 2026). The fused-silica slow-wave device demonstrates the complementary route of combining TFLT with a low-permittivity substrate and segmented slow-wave electrode to recover low voltage and long interaction length without sacrificing bandwidth (Li et al., 16 Apr 2026). The Cu-Damascene work further shows that TFLT modulators can be fabricated with semiconductor-style planarized Cu electrodes and still deliver data transmission on par with modulators using other low-resistivity metals (Lin et al., 7 May 2025).

Visible and UV electro-optics mark a distinct TFLT niche. At 737 nm, a TFLT MZI reaches $\mu$1, extinction ratio 29.6 dB, and detector-limited bandwidth beyond 20 GHz in $\mu$2-equivalent terms, while showing less than 2 dB drift over 16 minutes at 4.3 dBm on-chip power; an equivalent TFLN device under identical conditions drifts by 8 dB (Powell et al., 1 May 2025). At 375 nm, a TFLT UV MZI reaches $\mu$3 V, $\mu$4, extinction ratio 22.7 dB, insertion loss 1.3 dB, and a measured 3-dB bandwidth of 922 MHz limited by photodetector performance; electrical-to-electrical measurements indicate intrinsic potential beyond 67 GHz (Lin et al., 4 May 2026).

Acousto-optics adds another mode of operation. A one-dimensional topological photonic-crystal cavity on X-cut 600 nm TFLT, driven by a nearby interdigital transducer, yields a 120 $\mu$5m interaction length in a $\mu$6 footprint and stable acousto-optic modulation at on-chip optical power up to 28 dBm (630.9 mW) (Chen et al., 15 Apr 2026). The authors explicitly note that this first TFLT acousto-optic device reaches $\mu$7, comparable to top non-suspended LN devices but not yet to the most aggressive suspended LN cavities, which is an important qualification in comparative assessment (Chen et al., 15 Apr 2026).

5. Nonlinear optics, periodic poling, and quantum photonics

TFLT is also emerging as a $\mu$8 and $\mu$9 nonlinear platform. Periodically poled TFLT ridge waveguides demonstrate telecom-band second-harmonic generation from 1550 nm to 775 nm with normalized conversion efficiency $_2$0, in line with a theoretical value of $_2$1 for the measured geometry (Shelton et al., 24 Apr 2025). In that pole-after-etch process, differential etching shows a poled depth of 178.7 nm under the ridge; simulations indicate that full-depth poling of the 500 nm film would raise the normalized efficiency to approximately $_2$2 (Shelton et al., 24 Apr 2025). This suggests that the present limitation is process geometry rather than LT nonlinearity.

Broadband nonlinear optics in TFLT is now established as well. In dispersion-engineered z-cut 600 nm TFLT ridge waveguides pumped at 1560 nm, supercontinuum generation spans from 240 nm in the ultraviolet to beyond 2400 nm in the near-infrared, exceeding three octaves (Wang et al., 18 Dec 2025). The spectral evolution proceeds from low-power second- and third-harmonic generation to soliton-fission-driven dispersive-wave emission, and the visible dispersive wave is reported to overlap the second harmonic near 780 nm, which is relevant to self-referencing scenarios (Wang et al., 18 Dec 2025).

Quantum photonics extends this nonlinear capability to nonclassical states. A TFLT microring with free spectral range 350 GHz and optical quality factor $_2$3 generates correlated photon pairs from 1510 to 1570 nm via spontaneous four-wave mixing, with a pair generation rate of $_2$4 at 1535.04 nm (Fan et al., 1 Jun 2026). The source yields $_2$5 at a heralding rate of 170 kHz, unheralded $_2$6, and raw energy-time two-photon interference visibility $_2$7, all of which place TFLT among integrated platforms capable of high-quality telecom quantum light generation (Fan et al., 1 Jun 2026).

Resonant power handling is a enabling factor for these nonlinear regimes. Oxide-cladded, annealed TFLT microrings withstand $_2$8 W circulating power with only $_2$9 GHz resonance shift and no observable photorefractive effect, while coupling resonators on the same platform achieve effective $30\ \text{pm/V}$0 V over 2 mm with stable bias and phase control (Sayem et al., 31 Jan 2026). In combination, these results indicate that TFLT nonlinear photonics is not restricted to low-power proofs of concept.

6. Stability, system functions, and integrated control

Bias stability is a defining systems-level property of TFLT. In a 1550 nm TFLT Mach-Zehnder modulator biased at quadrature with 12.1 dBm on-chip optical power, the output fluctuates by less than 1 dB over 46 hours, versus 5 dB for a geometrically equivalent TFLN device under the same conditions (Powell et al., 2024). In a separate oxide-cladded platform, coupling bias remains stable over 40 minutes, and direct resonator phase bias from $30\ \text{pm/V}$1 V to $30\ \text{pm/V}$2 V gives drift below 0.1 pm over 25 minutes (Sayem et al., 31 Jan 2026). High-temperature studies extend this stability to 120–125$30\ \text{pm/V}$3C, where TFLT modulators retain bandwidth beyond 50 GHz and even show approximately 10% reduction in $30\ \text{pm/V}$4 (Sayem et al., 30 Apr 2026).

This low-drift behavior enables control functions that are unusually sensitive to phase error. A four-stage TFLT polarization controller exhibits polarization-dependent loss below 0.3 dB, half-wave voltage below 2.5 V, and reset-free polarization tracking at 1 Mrad/s, with transient tracking up to 2 Mrad/s; the device is validated in a dual-polarization 16-QAM self-homodyne 400-Gbps transmission system (Gao et al., 7 Jan 2026). An LT optical phased array keeps the far-field main lobe 8 dB higher than side lobes for over 4 hours and supports arbitrary spatiotemporal beam waveforms with modulation frequency as low as 0.1 Hz (Yue et al., 24 Mar 2026). In both cases, the platform claim is not merely high EO speed, but low static phase drift over timescales relevant to calibration and closed-loop control.

Microwave-optical conversion offers a distinct systems test. TFLT electro-optic transducers coupling C-band photonic-molecule resonators to 4.9–5.5 GHz superconducting microwave resonators demonstrate coherent bidirectional conversion across six devices, with measured on-chip efficiencies on the order of $30\ \text{pm/V}$5 and inferred single-photon coupling rates $30\ \text{pm/V}$6 (Axline et al., 10 Jun 2026). Continuous operation over multiple days is achieved using a static bias field with minimal feedback, and added noise under pulsed pumping is below one photon for 100 $30\ \text{pm/V}$7s pulses at the highest measured efficiencies (Axline et al., 10 Jun 2026). This is a particularly stringent indication that TFLT’s bias stability is relevant not only to classical modulators but also to cryogenic quantum interconnects.

7. Comparative assessment, limitations, and outlook

The present literature positions TFLT as a complementary, not uniformly superior, alternative to TFLN. On one hand, TFLT already has clear comparative advantages in DC stability, power handling, reduced birefringence, ultraviolet transparency, and operation in thermally or optically demanding regimes (Chen et al., 15 Apr 2026, Sayem et al., 30 Apr 2026, Lin et al., 4 May 2026). On the other hand, several reports are explicit that some device metrics remain process-limited or still trail the most optimized LN implementations. The first TFLT topological acousto-optic modulator, for example, is comparable to top non-suspended LN devices but not to the best suspended LN cavities (Chen et al., 15 Apr 2026).

Loss remains a visible limitation in some spectral regimes. The 375 nm TFLT waveguide loss of 7.2 dB/cm is compatible with a 200 $30\ \text{pm/V}$8m modulator but not yet with long ultraviolet routing, and slab measurements at 443 nm indicate 1.32 dB/cm intrinsic attenuation associated with defect-related absorption in ion-sliced films (Lin et al., 4 May 2026). In periodically poled ridge waveguides, incomplete domain penetration in a pole-after-etch geometry presently caps SHG efficiency well below the simulated full-depth value (Shelton et al., 24 Apr 2025). In microwave-optical transducers, on-chip efficiency is still far below the tens-of-percent regime envisioned by system projections, largely because current microwave $30\ \text{pm/V}$9 values are lower than participation-ratio analysis would suggest (Axline et al., 10 Jun 2026).

Another important qualification is that interface engineering remains decisive. A cladded LT optical phased array loses focused-beam dominance after about 16 minutes, whereas the cladding-free version maintains more than 8 dB side-lobe suppression for over 4 hours (Yue et al., 24 Mar 2026). This does not contradict the claim of intrinsically low LT carrier drift; rather, it shows that extrinsic trap landscapes introduced by claddings and interfaces can still dominate unless carefully managed.

Taken together, the literature indicates that TFLT has already established several distinct niches: bias-stable integrated electro-optics, high-power resonant and acousto-optic photonics, ultraviolet modulation, wafer-compatible nonlinear photonics, and scalable classical-quantum co-integration. A plausible implication is that future TFLT development will be driven less by proof-of-concept novelty than by convergence of three engineering themes already visible in the current record: lower-loss films and interfaces, more industrial metallization and packaging schemes, and systematic co-design of photonic, microwave, acoustic, and control subsystems on the same LT thin-film stack (Lin et al., 7 May 2025, Li et al., 16 Apr 2026, Axline et al., 10 Jun 2026).