Silicon photonics has enabled large-scale production of integrated optical devices for a vast array of applications. However, extending its use to nonlinear devices is difficult since silicon does not exhibit an intrinsic second-order nonlinearity. While heterogeneous integration of strongly nonlinear materials is possible, it often requires additional procedures since these materials cannot be directly grown on silicon. On the other hand, CMOS-compatible materials often suffer from weaker nonlinearities, compromising efficiency. A promising alternative to current material platforms is scandium-doped aluminum nitride (Al_1−xSc_xN), which maintains the CMOS compatibility of aluminum nitride (AlN) and has been used in electrical devices for its enhanced piezoelectricity. Here, we observe enhancement in optical second-order susceptibility (χ⁽²⁾) in CMOS-compatible Al_1−xSc_xN thin films with varying Sc concentrations. For Al_0.64Sc_0.36N, the χ⁽²⁾ component d₃₃ is enhanced to 62.3 ± 5.6 pm/V, which is 12 times stronger than intrinsic AlN and twice as strong as lithium niobate. Increasing the Sc concentration enhances both χ⁽²⁾ components, but loss increases with a higher Sc concentration as well, with Al_0.64Sc_0.36N exhibiting 17.2 dB/cm propagation loss at 1550 nm and Al_0.80Sc_0.20N exhibiting 8.2 dB/cm at 1550 nm. Since other material properties of this alloy are also affected by Sc, tuning the Sc concentration can balance strong nonlinearity, loss, and other factors depending on the needs of specific applications. As such, Al_1−xSc_xN could facilitate low cost development of nonlinear integrated photonic devices.

The deluge of sensors and data generating devices has driven a paradigm shift in modern computing from arithmetic-logic centric to data-centric processing. Data-centric processing require innovations at the device level to enable novel compute-in-memory (CIM) operations. A key challenge in the construction of CIM architectures is the conflicting trade-off between the performance and their flexibility for various essential data operations. Here, we present a transistor-free CIM architecture that permits storage, search, and neural network operations on sub-50 nm thick Aluminum Scandium Nitride ferroelectric diodes (FeDs). Our circuit designs and devices can be directly integrated on top of Silicon microprocessors in a scalable process. By leveraging the field-programmability, nonvolatility, and nonlinearity of FeDs, search operations are demonstrated with a cell footprint <0.12 μm² when projected onto 45 nm node technology. We further demonstrate neural network operations with 4-bit operation using FeDs. Our results highlight FeDs as candidates for efficient and multifunctional CIM platforms.

Ferroelectric Al1−xScxN has raised much interest in recent years due to its unique ferroelectric properties and complementary metal oxide semiconductor back-end-of-line compatible processing temperatures. Potential applications in embedded nonvolatile memory, however, require ferroelectric materials to switch at relatively low voltages. One approach to achieving a lower switching voltage is to significantly reduce the Al1−xScxN thickness. In this work, ferroelectric behavior in 5–27 nm films of sputter deposited Al0.72Sc0.28N has been studied. We find that the 10 kHz normalized coercive field increases from 4.4 to 7.3 MV/cm when reducing the film thickness from 27.1 to 5.4 nm, while over the same thickness range, the characteristic breakdown field of a 12.5 μm radius capacitor increases from 8.3 to 12.1 MV/cm. The 5.4 nm film demonstrates ferroelectric switching at 5.5 V when excited with a 500 ns pulse and a switching speed of 60 ns.