Pulsed Laser Deposition-Based Growth and Parameter Optimization of Thin Films for Next-Generation In-Memory Computing

Abstract

This research demonstrates precise control over the structural, compositional, and electronic properties of epitaxial antimony selenide (Sb2Se3) thin films by meticulously tailoring the background deposition pressure. Elevating the deposition pressure to 100 mTorr induces a microstructural transition from a highly oriented growth to a polycrystalline and textured morphology. X-ray photoelectron spectroscopy analysis reveals that this defect-rich morphology significantly increases the effective surface area and the resulting fraction of oxidized antimony states, while intrinsic binding energies and baseline selenium desorption remain consistent. Crucially, this structurally engineered device delivers an exceptional On/Off ratio of 2.6 x 105, robust switching endurance exceeding 1,000 cycles, and highly stable data retention, validating the efficacy of defect engineering for reliable artificial synapses. As a parallel strategy, non-volatile parasitic memcapacitors were successfully demonstrated for the first time by exploiting the robust ferroelectric properties and defect chemistry of epitaxial bismuth ferrite (BFO) thin films. Utilizing an elevated growth temperature of 710 °C intentionally promotes bismuth volatilization, creating a vacancy- mediated parasitic effect that dramatically modulates the interface charge profiles. This mechanism yields a record-high zero-bias capacitance modulation ratio of 160% at 10 kHz while establishing ten distinct, programmable capacitance states. Operational stability evaluations conducted at 85 °C for 24 hours reveal excellent non-volatile retention with negligible degradation. Furthermore, statistical assessments across physically fabricated devices confirm tight distribution profiles and outstanding device-to-device reproducibility. To further optimize memory performance, defect engineering frameworks were investigated through the systematic aliovalent substitution of divalent barium ions into the trivalent bismuth sites of the BFO lattice. Capacitance-voltage characterizations establish that a 1% barium doping level provides the ideal defect configuration to maximize the zero-bias memory window. Moving toward array-level implementation, heavily doped 20% barium- substituted heterostructures were successfully integrated to construct a fully functional 11x11 selector-free memristive crossbar array. Individual cells within this hardware array flawlessly preserve excellent synaptic plasticity with near-zero nonlinearity during weight updates, and the array demonstrates outstanding electrical robustness, successfully sustaining over 107 switching cycles without noticeable conductance degradation. By establishing a precise equilibrium between thermodynamic growth parameters, chemical doping, and targeted defect generation, this research delivers a highly reliable, scalable, and energy-efficient material platform utilizing chalcogenide and perovskite oxide thin films, providing vital pathways for the hardware realization of high-density artificial synaptic networks.