Low-Loss Phase Change Materials on Mid-Infrared Metamaterials for Non-Volatile Optical Memory and In-Memory Computing

This work pioneers a mid-infrared in-memory computing platform based on a novel high-entropy phase-change material, GeSbSeTe (GSST), developed through synergistic co-doping of light and heavy elements. The material exhibits an exceptionally high refractive index contrast <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\Delta \mathrm{n}=1.98)$</tex> and ultra-low extinction coefficient (k = 0.01), resulting in a record-high optical figure of merit, outperforming conventional GeSbTe (GST) by over two orders of magnitude. Integrated with a CMOS-compatible Mo-AlN-Mo metamaterial absorber, the system enhances light-matter interaction and supports programmable, non-volatile phase transitions. A resonance wavelength shift of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\sim 1\ \mu \mathrm{m}$</tex> -the largest reported in the mid-infrared regime-was achieved with narrow FWHM for precise spectral control. Using a custom-built mid-infrared hyperspectral imaging setup, the platform demonstrated all-optical in-memory computing with 97% accuracy on the MNIST dataset. This work overcomes longstanding trade-offs in phase-change optical memory and establishes a reconfigurable, energy-efficient route toward intelligent infrared vision and neuromorphic photonic architectures.