Memristive State Writing With Nonlinear Reconfiguration Control

The patent introduces selector devices as intermediary elements between the memristor crossbar array and the external circuit. These selectors act as mediators that block leakage currents from propagating through the crossbar array, thereby enabling scalable implementation while preserving data processing functionality. The selectors are strategically placed at intersections or along bitlines/wordlines to prevent harmful leakage paths.

The patent extracts and separates the leakage current blocking function from the memristor elements themselves by introducing dedicated selector devices. This extraction allows the memristors to focus on data processing while the selectors handle the harmful leakage currents, resolving the contradiction between functionality and scalability.

The patent applies preliminary action by performing initialization only once at the beginning of operation, rather than before every reconfiguration. The selector devices are initialized in a specific state that enables them to block leakage currents throughout subsequent operations, eliminating the need for repeated initialization and reducing time loss.

The selector devices are designed to maintain their leakage-blocking function automatically through self-service mechanisms. Once initialized, they continue to prevent leakage currents without requiring external intervention or re-initialization, thereby eliminating the time-consuming initialization step before each reconfiguration.

The patent segments the memristive structure into two distinct layers, each with specific functionality. The first layer is responsible for data storage while the second layer enables predictable reconfiguration through strong nonlinear resistive behavior. This segmentation allows independent optimization of each layer's function, improving state tracking precision while managing complexity through functional decomposition.

The patent employs composite memristive structures combining two different memristive materials or configurations with complementary properties. The first layer uses a material optimized for data storage, while the second layer uses a material with strong nonlinear characteristics for predictable switching. This composite approach leverages the strengths of each material to achieve both precise state tracking and controlled reconfiguration.