Reconfigurable Volatility and Switching Modes in Memristors

48 concepts29 papers1 proposal

In traditional computing architectures, memory and processing are strictly segregated. Brain-inspired neuromorphic computing seeks to dismantle this wall by utilizing memristors—two-terminal resistive switching devices that alter their internal resistance based on historical electrical activity. Historically, as reviewed by Zhuoyu Ji et al. (2011), nonvolatile memory devices were engineered exclusively for rigid, long-term data retention.

However, a biological brain requires two fundamentally distinct building blocks: synapses to store long-term weights (non-volatile memory) and neurons to perform transient, short-term integrations and fire pulses (volatile threshold switching). Fabricating separate hardware components for neurons and synapses drastically increases circuit complexity.

To solve this, a powerful paradigm has emerged: dynamically modulating the volatility and switching modes of a single memristive device. By altering external electrical constraints or engineering the internal electrolyte morphology, a single device can seamlessly transition between a volatile sensor or neuron and a non-volatile memory cell.


Controlling Volatility via Compliance Current

The most direct way to toggle a memristor between volatile and non-volatile states is by adjusting the compliance current ($I_{cc}$) during the "set" process. To understand why this works, we must look at the nanoscale dynamics of conductive filament formation.

When a voltage is applied across an electrochemical metallization (ECM) cell, metal ions migrate through a solid electrolyte to form a conductive bridge between the electrodes.

Low Compliance Current (Unstable)      High Compliance Current (Stable)
       [ Top Electrode ]                       [ Top Electrode ]
             \   /                                   |||||
              \ /  <- Thin, fragile                  ||||| <- Thick, robust
               |      filament                       |||||    filament
       [Bottom Electrode]                      [Bottom Electrode]

As demonstrated by Bojun Cheng et al. (2019) and Yong Huang et al. (2022), enforcing a low compliance current restricts the number of ions participating in the bridge. The resulting filament is incredibly thin and fragile. Once the external electrical bias is removed, interfacial energy, thermal relaxation, or surface tension causes this atomic-scale bridge to spontaneously rupture. The device quickly resets to its high-resistance state, acting as a volatile, self-resetting selector or neuron.

Conversely, raising the compliance current allows a larger flux of ions to cross the electrolyte. The authors show that this builds a thick, structurally robust filament. This bridge withstands thermal fluctuations and remains intact long after the power is turned off, providing stable, non-volatile data storage.

This dynamic scaling is highly adaptive. Bojun Cheng et al. (2021) note that volatile memristors with self-adaptive filament growth naturally adjust their filament diameter to match the circuit's current demands, maintaining a constant voltage drop and preventing signal degradation. This behavior mirrors findings by Suhas Kumar et al. (2013) in oxide-based systems like vanadium dioxide ($VO_2$), where the physical width of a metallic filament scales directly with the current load, driving volatile, Joule-heating-induced phase transitions.


Material Architecture: Alloys and Phase Boundaries

While external electrical limits provide a simple control mechanism, researchers also engineer the internal material composition of the electrolyte and electrodes to achieve precise volatility control.

Dual-Ion Alloyed Filaments

Instead of relying on a single metal, Elias Passerini et al. (2025) developed a memristor utilizing an alloyed filament composed of two metals with distinct ionic mobilities (such as silver and tin).

The intuition here is elegant: the lower-mobility metal ions migrate slowly and form a permanent, rigid structural "skeleton" across the device. The higher-mobility metal ions move rapidly through this skeleton. When a low current is applied, only the fast ions move, enabling rapid, volatile neuronal firing. When a higher current is applied, the slower ions lock into place, stabilizing the entire structure into a permanent synaptic weight.

Crystalline-Amorphous Phase Tuning

The physical environment surrounding the ions also dictates their stability. Shijie Wang et al. (2023) engineered an organic electrochemical transistor (OECT) featuring a channel with a mixed crystalline-amorphous morphology.

Amorphous Region (Volatile)             Crystalline Region (Non-Volatile)
    o   o      o   o                        +---+---+---+
  o   o   o  o   o   o                      | o | o | o |  <- Ions physically
    o   o      o   o                        +---+---+---+     trapped in 
  (Fast, temporary diffusion)               | o | o | o |     ordered lattice
                                            +---+---+---+

In the messy, disordered amorphous regions, ions diffuse rapidly and escape just as quickly when the bias drops, facilitating volatile sensing. In contrast, the highly ordered crystalline regions act as physical cages, trapping the ions securely to provide stable, 10-bit non-volatile memory storage.

Similarly, Yaoyuan Wang et al. (2018a) utilized the phase-change material $Ge_2Sb_2Te_5$ (GST) as a solid electrolyte. By electrically toggling the GST matrix itself between amorphous and crystalline phases, the authors dynamically altered the internal diffusion barriers, allowing the cell to switch between volatile buffering and permanent storage modes.

On an atomic scale, this principle extends to two-dimensional materials. Manasa Kaniselvan et al. (2025) demonstrated that the volatility of an "atomristor" depends entirely on the chemical bonding stability between migrated electrode atoms and the vacancies within a 2D host lattice. Metals that form weak, unstable bonds with the lattice yield temporary threshold switching, while metals that form strong, covalent bonds create permanent memory states.


Modulating Modes via Electrical Bias and Environment

Beyond changing whether a device remembers, researchers can modulate how it forgets and resets by adjusting the electrical bias profile or leveraging environmental interactions.

Pulse and Bias Tuning

The duration and magnitude of the applied voltage sweep dictate filament stability. As shown by Yaoyuan Wang et al. (2018a), short or low-voltage pulses generate transient filaments that dissolve spontaneously. Extended or high-voltage pulses drive a massive ion migration that solidifies into a permanent connection.

By precisely adjusting this stoichiometry and operating bias, Kyung Seok Woo et al. (2024) tuned memristors into a "conditionally nonvolatile" state. This intermediate regime allows a single component to act as a stochastic entropy source for encryption under low bias, or as a deterministic logic unit under higher bias.

Toggling Bipolar vs. Unipolar Modes

The choice of electrolyte can even alter the fundamental switching mode of the device. Traditional memristors are often locked into either bipolar switching (requiring a voltage of opposite polarity to reset) or unipolar switching (resetting using the same polarity via thermal disruption).

Yaoyuan Wang et al. (2018b) bypassed this limitation by introducing an anion-supplying active electrode. The authors demonstrated that the choice of electrolyte material dictates whether a conductive bridge is ruptured electrochemically (by reversing voltage polarity) or thermally via Joule heating (using the same polarity). This allows the same electrode framework to select between bipolar and unipolar operation modes.

Environmental and Structural Gating

Switching behavior can also be modulated by external or structural gates:

  • Atmospheric Moisture: Xiaoyue Ji et al. (2021) discovered that ambient humidity acts as a physical switch for metal-oxide memristors. Water molecules facilitate redox reactions and ion migration that are completely restricted in dry conditions, causing the hardware to evolve from a simple capacitor to a full resistive switch.
  • Three-Terminal Gating: To decouple the programming voltage from switching dynamics, Yang Shen et al. (2021) incorporated a graphene back-gate into a three-terminal architecture, tuning the internal potential distribution across an oxide layer. Similarly, Lanqing Zou et al. (2025) utilized a tri-layer $HfO_x$ composition to precisely regulate ion-gating mechanisms.
  • Fluidic Phase Transitions: Moving beyond solid-state devices, Ke Sun et al. (2023) achieved volatile synaptic behavior in fluidic nanopores. By using voltage to drive an electrowetting transition between liquid-filled (conductive) and vapor-filled (resistive) states, they mimicked biological learning and forgetting with extreme energy efficiency.

Quantifying the Thresholds of State Transitions

To design predictable circuits using these multi-modal devices, engineers require precise mathematical models that capture how these internal states shift.

A major challenge is that physical channel growth simultaneously alters both bulk resistance and interface barriers. To capture this dual effect, Dániel Hajtó et al. (2025) modeled self-directed channel memristors as a parallel combination of an Ohmic resistor and a Schottky diode, where both components are scaled by a single, shared internal state variable.

The sensitivity of this state variable to incoming electrical stimuli can be mathematically framed using "window functions." As proven by Panayiotis S. Georgiou et al. (2016), the mathematical flatness of a memristor's window function directly dictates the charge threshold required to initiate a state transition:

Flat Window Function (High Threshold)     Steep Window Function (Low Threshold)
          State Change                              State Change
               ^                                         ^
               |       +-------+                         |    / \
               |       |       |                         |   /   \
               +-------+-------+-->                      +--+-----+-->
                     Integrated Charge                         Integrated Charge
            (Requires high accumulation)               (Highly sensitive/rapid)

A higher flatness parameter means the device must integrate a large amount of input charge before its resistance shifts significantly, providing a mathematical mechanism to tune device sensitivity.

Once this threshold is crossed, the speed of the transition is highly non-linear. Siddharth Gaba et al. (2013) demonstrated that the characteristic wait time for a resistive state change exhibits an exponential sensitivity to the applied bias voltage. A tiny increase in voltage can compress the average switching time by several orders of magnitude. By exploiting this exponential relationship, circuit designers can precisely tune the device's behavior, operating it either as a highly reliable, deterministic memory cell or as a perfectly calibrated, volatile stochastic computing element.

Key Foundational Work

Organic memristor Devices for Logic Elements with Memory (2012) — contributed 1 concept elements to this direction.
Advancements in organic nonvolatile memory devices (2011) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_electrolyte_mediated_switching_mode_sele_ee0aa458,
  title = {electrolyte-mediated switching mode selection},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/electrolyte-mediated-switching-mode-selection/rdr_ee0aa4589e7f49e2adc918dccec0c025},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_electrolyte_mediated_switching_mode_sele_ee0aa458_v1,
  title = {electrolyte-mediated switching mode selection},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/electrolyte-mediated-switching-mode-selection/rdr_ee0aa4589e7f49e2adc918dccec0c025/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake