Abstract
This chapter explores the dynamic behavior of dual flux coupled memristor circuits in order to explore the uncharted territory of the fundamental theory of memristor circuits. Neuromorphic computing anticipates highly dense systems of memristive networks, and with nanoscale devices within such close proximity to one another, it is anticipated that flux and charge coupling between adjacent memristors will have a bearing upon their operation. Using the constitutive relations of memristors, various cases of flux coupling are mathematically modeled. This involves analyzing two memristors connected in composite, both serially and in parallel in various polarity configurations. The new behavior of two coupled memristors is characterized based on memristive state equations, and memductance variation represented in terms of voltage, current, charge and flux. The rigorous mathematical analysis based on the fundamental circuit equations of ideal memristors affirms the memristor closure theorem, where coupled memristor circuits behave as different types of memristors with higher complexity.
Introduction
As the field of neuromorphic computing advances toward extreme integration densities—targeting upwards of 100 billion synapses per square centimeter—the physical proximity of components becomes a critical design factor. In proposed 3D architectures, memristors may be packed within 30nm of each other. At this scale, electromagnetic crosstalk, specifically in the form of flux and charge coupling, is no longer a negligible parasitic effect but a fundamental driver of device behavior.
Traditional circuit design often treats memristors as isolated, discrete components. However, in high-density crossbar arrays, the state of one memristor can significantly influence its neighbor. This research addresses the lack of a fundamental theoretical framework for these "composite" behaviors, investigating whether such coupling is merely a source of noise or a potential feature that can be harnessed to create sophisticated multi-level logic and synaptic models for artificial neural networks.
Theoretical Methodology
The study employs an analytical approach grounded in the constitutive relations of ideal memristors. For charge-controlled devices, the relationship is defined as $V=M(q)i$, while for flux-controlled devices, it is $i=W(\phi)v$. To model the switching behavior, the authors utilize a three-part piecewise linear relationship between flux ($\phi$) and charge ($q$), representing a device that transitions between high and low conductance states.
The core of the model introduces first-order mathematical coupling where the memductance ($W$) of a device is a function of both its own internal flux and the flux of an adjacent neighbor, scaled by a coupling coefficient ($\kappa$). The authors derive differential equations for four primary electrical configurations: serial and parallel connections, each evaluated with identical and opposite (anti-serial/anti-parallel) polarities.
Key Insights: From Binary to Multi-State Systems
A primary finding of this modeling is the transition from binary to "infinite" states. While a single ideal memristor typically switches between two discrete states (ON and OFF), coupling two such devices creates a composite system with a continuous range of states. This increased complexity is highly desirable for neuromorphic applications, as it more closely mimics the nuanced plasticity and weight-tuning capabilities of biological synapses.
The "Tug-of-War" and Tunability
In serial configurations, the initial state or bias of one memristor acts as a control parameter for the other. By pre-biasing one device, researchers can precisely modulate the peak-to-peak flux amplitude in its neighbor. This "tug-of-war" effect allows for fine-grained control over the composite system's behavior.
The I-V characteristics (Figure 6) demonstrate that by varying the initial flux conditions, the response of the composite system can be shifted and scaled. This provides a mechanism for "weight tuning" in neural networks without necessitating changes to the global driving voltage, offering a path toward more energy-efficient hardware.
Memristor Closure and Design Constraints
The analysis affirms the Memristor Closure Theorem, which states that a network of coupled memristors still behaves as a single memristor, albeit one with significantly higher complexity. This is a crucial finding for system-level modeling, as it allows complex clusters to be abstracted as single, high-order components.
However, the research also identifies critical risks. In anti-serial configurations, certain conditions can lead to a non-equilibrium state where flux increases indefinitely. In physical hardware, this would manifest as device "burn-out" or a transition into a simple resistive state, highlighting a vital constraint for engineers designing high-density memristive arrays.
Conclusion and Future Directions
This theoretical study provides a rigorous foundation for understanding flux coupling in memristive architectures. While the current model focuses on identical devices under idealized conditions, it sets the stage for future work quantifying these effects in specific fabrication processes (e.g., TiO₂ or ReRAM). Developing "coupling-aware" design rules will be essential to either mitigate interference in standard logic or intentionally exploit these multi-state behaviors to realize high-density, quantized neural weights in next-generation neuromorphic hardware.
Summary updated