Thin films and coatings are engineered layers from a few to a few hundred nanometres thick whose composition, crystal phase and thickness set the performance ceiling of RF filters, integrated photonics, superconducting qubits, displays and sensors, and the June 2026 literature shows the field’s centre of gravity moving from finding new film materials to controlling their growth process precisely enough to hit device specifications.
Summary
A thin film is a layer of material deposited onto a substrate, typically between one and a few hundred nanometres thick, grown by methods such as molecular beam epitaxy (MBE), sputtering, chemical vapour deposition or bonding-and-thinning of bulk crystals. Because the layer is thinner than most relevant physical length scales, its thickness becomes a design variable in its own right: it sets the resonant frequency of an acoustic device, the phase response of an optical meta-atom, the charge distribution in a transistor channel, or the impedance match of a microwave absorber. The material itself may be identical to a bulk crystal, but confined to a film it acquires different phases, strain states, defect populations and interface physics.
The supplied sources show this playing out across four device families. In photonics, thin-film lithium niobate (TFLN) and the emerging lithium tantalate on insulator (LTOI) platform provide a strong Pockels effect in a waveguide geometry: an 800 nm X-cut TFLN chip was poled at a 1425 nm period to demonstrate backward-wave difference-frequency generation with idler output from 1244 to 1290 nm, while LTOI waveguides showed propagation losses below 0.1 dB/cm with 3D-printed polymer couplers at 0.9 dB per interface over an 850 to 1740 nm band. In RF, 500 nm 128-degree Y-cut lithium niobate on silicon supports higher-order XBAR resonators whose quality factor turns out to depend on in-plane orientation, not only on coupling coefficient. In quantum hardware, superconducting circuit performance is described as fundamentally limited by material losses, with tantalum films attracting attention after strong resonator and qubit results, complicated by the fact that tantalum is polymorphic and the phase that grows depends on the substrate. In electronics, a dual-layer a-IGZO/a-IZO oxide transistor model identifies an optimal top-layer thickness of 9 to 12 nm from the competition between charge confinement and oxygen-vacancy trap density.
The parameters that decide outcomes are therefore process parameters, not just chemistry: substrate temperature, flux ratio, growth rate, thickness, crystallographic orientation and the sequence of fields applied during processing. The measured outputs used as proxies are X-ray rocking-curve full width at half maximum and atomic-force-microscope RMS roughness. A machine-learning study of GaSe MBE found growth rate dominates rocking-curve FWHM while the Se/Ga flux ratio dominates roughness, and that in situ RHEED patterns correlate strongly with FWHM but weakly with roughness. Growth windows can be wide: Mn intercalation to form Mn2In2Se5 worked across substrate temperatures of 250 to 450 degrees C and Se:Mn flux ratios of 1.1 to 3.1, yielding 1.5 nm RMS films and, with longer deposition, 8 nm MnSe layers with 0.35 degree rocking-curve FWHM.
Running alongside this is an automation layer. Closed-loop workflows now search over entire processing histories rather than end-state compositions: an evolutionary search coupled to uncertainty-aware deep kernel learning discovered scanning-probe bias waveforms that enhanced the nonlinear electromechanical response of ferroelectric thin films by de-aging them, and the NIMO platform packages twelve AI algorithms for self-driving laboratories, with thin-film exploration among six demonstrated applications.
CNT conductive coatings and primers (added 12 Oct 2026)
SWCNT conductive primers for electrostatic painting of plastic car parts (OCSiAl with Kanat) were “undergoing validation” with carmakers in 2021; no OEM approval found since ref. OCSiAl claims static-dissipative coatings from 0.01-0.1% SWCNT against c.100x more carbon black, with the advantage being colour and low loading (vendor figures, internally inconsistent) ref. See Carbon Nanotubes.
Evidence base
- 16 June 2026: LTOI photonic circuits reported propagation loss below 0.1 dB/cm with 3D-printed polymer couplers at 0.9 dB (strip) and 1.25 dB (rib) per interface, over 850 to 1740 nm and stable at 1 W input.
- 16 June 2026: an integrated terahertz photonic receiving frontend delivered undegraded link noise across 140 to 450 GHz and outperformed electronic frontends, which had previously been the better option on noise.
- 15 June 2026: a review of tantalum for superconducting integrated circuits attributes recent qubit and resonator gains to the film, while identifying polymorphism and substrate-dependent phase selection as the central control problem.
- 16 June 2026: backward-wave difference-frequency generation was demonstrated in 800 nm X-cut TFLN using a 1425 nm poling period achieved by poling before etching, with idler output from 1244 to 1290 nm.
- 15 June 2026: Mn intercalation produced single-crystalline Mn2In2Se5 across substrate temperatures of 250 to 450 degrees C and Se:Mn flux ratios of 1.1 to 3.1, giving 1.5 nm RMS films and 8 nm MnSe with 0.35 degree rocking-curve FWHM.
- 15 June 2026: a two-equation electrostatic model of dual-layer a-IGZO/a-IZO transistors identified an optimal a-IGZO thickness of 9 to 12 nm from the trade-off between charge confinement and oxygen-vacancy trap density.
Open questions
- Do the thin-film lithium niobate and lithium tantalate loss and coupling figures (below 0.1 dB/cm, 0.9 dB per interface) hold across full wafers and after packaging, or only on selected devices?
- Can tantalum phase selection be made deterministic across substrates, given that its polymorphism is what produces the variation in superconducting properties and coherence?
- Does the weak correlation between in situ RHEED embeddings and ex situ RMS roughness mean surface morphology cannot be controlled in real time, and if so which additional in situ probe closes that gap?
- Is defect-enabled ferroelectricity of the Bi2SeO5 type stable under repeated cycling, or does the vacancy population that enables switching also cause the degradation that vacancies normally cause?
Assessment drafted 2026-08-31 from up to 18 KB sources using the technology-scorecard framework; scores are a draft read pending review.