Barium titanate modulator (BTO)

Silicon has no linear electro-optic effect, so silicon modulators change index through free carriers (plasma dispersion), which also absorbs and is nonlinear in drive. Thin-film lithium niobate supplies a clean Pockels effect but must be bonded as a separate material. Barium titanate, a ferroelectric perovskite, offers a Pockels coefficient far larger than lithium niobate's, and it can be grown epitaxially on silicon with a strontium titanate buffer and then bonded onto silicon photonic wafers.

The key number is the coefficient. Abel et al. measured $r_{42}$ = 923 pm/V in BTO films integrated on silicon, approaching the bulk value; the largest coefficient of lithium niobate, $r_{33}$, is 30.8 pm/V, about thirty times smaller. In a modulator the index change is proportional to $r$ times the field, so a larger coefficient translates directly into a shorter device or a lower half-wave voltage: silicon-BTO Mach-Zehnder modulators are quoted at $V_\pi L$ of a few tenths of a V·cm, against 2 to 5 V·cm for thin-film lithium niobate. BTO also works well in resonant devices, where its linear, lossless phase shift tunes a ring without the absorption of carrier injection.

The coefficient comes with the complications of a ferroelectric. The effect is a tensor with several components, and which one a device uses depends on crystal orientation and the direction of the applied field, so the film's domain structure must be controlled and poled. The permittivity is very large, which concentrates the field in the surrounding materials and complicates high-speed electrode design. The coefficient also depends strongly on frequency, unlike lithium niobate's: measurements from megahertz to sub-terahertz show the dispersion, and show that device geometry can be designed to give a flat electro-optic response despite it (Chelladurai et al. 2025). Long-term stability over time and temperature is the other open question for data modulation. Commercial silicon-BTO modulators and foundry offerings exist, but the platform is younger than lithium niobate and its reliability record is shorter.

References: S. Abel et al., Nat. Mater. 18, 42 (2019); D. Chelladurai et al., Nat. Mater. 24, 868 (2025).