---
title: 'Threshold Switching: Abrupt Transitions'
url: https://www.emergentmind.com/topics/threshold-switching
type: topic
---

# Threshold Switching: Abrupt Transitions

Threshold switching denotes a class of phenomena in which a system remains in one regime until a control variable crosses a critical value, after which an abrupt transition occurs. In the electronic and ionic literature, the term commonly refers to a sharp resistance collapse above a critical field or voltage; in spintronics, it denotes the minimum current needed to destabilize or reverse magnetization; in all-optical magnetism, it denotes the minimum incident laser fluence required to produce switching; and in stochastic or algorithmic settings it denotes threshold criteria or threshold-triggered mode changes rather than a material phase transition [1602.01885] [1210.3442] [2110.12063] [2110.15435].

## 1. Terminological scope and threshold variables

Across the cited literatures, the common structure is a critical point separating qualitatively distinct dynamical regimes. What changes from field to field is the controlled variable and the meaning of the post-threshold state. In phase-change and selector devices, the post-threshold state is a low-resistance conductive state; in magnetic switching it is a reversed or destabilized magnetization state; in optical switching it is a switched magnetic domain; in epidemic theory it is persistence rather than extinction; and in online learning it is a threshold-triggered action reconsideration event [1810.09666] [2403.03235].

| Domain | Threshold quantity | Post-threshold change |
|---|---|---|
| Phase-change / OTS / memristors | Critical electric field or voltage | Abrupt resistance collapse |
| Spintronics | Threshold current density | Magnetization instability or reversal |
| All-optical magnetism | Threshold fluence | Switched magnetic domain |
| Photovoltaics | Critical voltage or field | Conductive path to buried electrode |
| Online learning | Threshold event \(E^t\) | Action reconsideration or resampling |
| Epidemic switching models | \(\mathcal R_0\) or averaged threshold condition | Extinction or persistence |

This breadth of usage matters because the phrase does not refer to a single microscopic mechanism. A phase-change threshold switch, a spin-orbit-torque threshold current, and a semi-Markov epidemic threshold are formally analogous only at the level of regime transition. This suggests that “threshold switching” functions as a cross-disciplinary descriptor for abrupt regime selection, while each literature supplies its own state variables, observables, and stability criteria.

## 2. Electronic and ionic threshold switching in materials

In amorphous chalcogenides, threshold switching is the abrupt, field-induced collapse of the high electrical resistance of an amorphous glass into a much more conductive state once a critical electric field is exceeded. In amorphous \(\mathrm{Ag_4In_3Sb_{67}Te_{26}}\) (AIST), few-picosecond THz excitation produced reversible conductivity changes and crystallization under repeated pulsing, with the threshold field under THz excitation found to be \(230 \pm 60~\mathrm{kV/cm}\), while actual crystallization began at \(340~\mathrm{kV/cm}\). The onset of the nonlinear conductivity increase was extracted to be sub-picosecond, which the authors interpret as evidence for a purely electronic threshold event preceding any slower thermal crystallization [1602.01885].

A complementary device-level picture appears in electrothermal modeling of amorphous \(\mathrm{Ge_2Sb_2Te_5}\) (a-GST), where the conductivity law is written as
\[
\sigma_a(T,\vec{E})=\min\!\left(\sigma_T(T)+\sigma_E(E),\ \sigma_{\max}\right).
\]
Within that framework, threshold switching and snap-back arise from positive feedback between Joule heating, temperature-dependent conductivity increase, and field-enhanced conductivity, and the on-state is filamentary rather than spatially uniform. The same model reproduces switching fields spanning roughly \(5\) to \(42.5~\mathrm{MV/m}\) and captures selector-like holding behavior in isolated devices and in OTS+PCM series structures [1906.09316].

Other material systems exhibit distinct microscopic realizations of the same abrupt transition. In 1T-TaS\(_2\), threshold switching corresponds to a voltage-driven transition between nearly commensurate and incommensurate charge-density-wave states, with representative room-temperature thresholds \(V_H = 0.94~\text{V}\) and \(V_L = 0.89~\text{V}\). The switching remains operative under \(10\)-keV X-ray irradiation up to \(1~\text{Mrad(SiO}_2)\), with threshold-voltage shifts of only about \(2\%\), which the paper attributes to the high intrinsic carrier concentration of both CDW phases [1712.01354]. In amorphous Nb\(_2\)O\(_5\)-based devices, reactive top electrodes create metal-oxide interlayers whose conductivity and oxygen-exchange behavior determine whether the device behaves as a symmetric volatile threshold switch, an asymmetric threshold switch, or an integrated selector-memory element; under negative bias the mean threshold and hold voltages are \(V_{\mathrm{TH}} \approx -2.2 \pm 0.25~\text{V}\) and \(V_{\mathrm{H}} \approx -1.6 \pm 0.3~\text{V}\) [1912.08373].

Two-dimensional systems add explicit structural confinement. In Ni/h-BN/Ni cross-point memristors with an h-BN thickness of \(5 \pm 0.5\) nm and active area \(34~\mu\text{m}^2\), volatile threshold switching occurs at \(V_{\text{th,on}} = 2.0 \pm 0.1~\text{V}\) and \(V_{\text{th,off}} = 0.3 \pm 0.1~\text{V}\), with an On/Off ratio of \(10^7\), off-current around \(0.13~\text{pA}\) at \(0.1~\text{V}\), and cycle-to-cycle variability of \(5\%\). Temperature-dependent transport and TEM/EFTEM support a mechanism of Ni filament formation and retraction along boron-defect pathways [2301.10158]. In vertical Ag/SiO\(_x\)/VAMoS\(_2\)/Au heterostructures, volatile threshold switching occurs at low voltages of \(\sim 0.63~\text{V}\), with pulsed on-state currents over \(200~\mu\text{A}\), stable switching exceeding \(10^4\) cycles, on-switching time of \(311~\text{ns}\), and spontaneous relaxation in \(233~\text{ns}\) [2504.18546].

## 3. Modeling, interface control, and variability engineering

Threshold-switching devices are often modeled as state-dependent nonlinear conductors with an internal variable capturing delayed formation of the conductive state. A compact model for OTS devices expresses the terminal current as
\[
i= C \cdot \frac {d}{dt}v_R + (1- S) \cdot i_\text{off}(v)+ S \cdot i_\text{on}(v),
\]
where the internal state variable controls branch selection between off-state, snapback, and on-state behavior. The paper develops this model from a transistor-based LTSPICE macro model and then reinterprets it physically as a delay structure adjacent to a dominant internal junction, with threshold \(v_{th} = 2.4~\text{V}\), \(i_{th} = 1~\mu\text{A}\), \(R = 1~\text{M}\Omega\), \(C = 1~\text{nF}\), and \(R_b = 5~\text{k}\Omega\) in a fit to a Western Digital Research OTS device [2308.02017].

A different modeling tradition emphasizes electrothermal continuum descriptions. In a-GST selector and PCM simulations, threshold switching is reproduced without committing to a unique microscopic carrier mechanism, by coupling current continuity and the heat equation and allowing conductivity to rise through additive temperature and field terms. This yields abrupt snap-back, holding behavior, and filamentary on-state localization, and it explains why device thickness, ambient temperature, and pulse rise time materially alter the observed switching field [1906.09316].

Recent heterostructure work uses variability-aware ECM modeling to make threshold switching more reproducible. In Ag/SiO\(_x\)/VAMoS\(_2\)/Au versus Ag/SiO\(_x\)/Au, the heterostructure exhibits \(V_{\mathrm{hold}} = 0.4 \pm 0.02~\text{V}\) instead of \(0.1 \pm 0.04~\text{V}\), switching times of \(620 \pm 160~\text{ns}\) instead of \(940 \pm 450~\text{ns}\), and cycle-to-cycle on-current variability of \(3.0\%\) instead of \(46.5\%\). The proposed mechanism is vdW-gap-confined Ag migration in vertically aligned MoS\(_2\), which forms ultrathin seed filaments that guide subsequent growth in SiO\(_x\) and reduce stochastic tunneling-gap fluctuations [2511.09561].

A broader implication is that threshold switching can be engineered not only by changing the bulk active medium but also by controlling interfaces, filament nucleation geometry, and the state variable that governs transition delay. This is especially explicit in Nb\(_2\)O\(_5\)-based devices, where reactive top-electrode chemistry tunes the system continuously from selector-only behavior to integrated threshold-memory behavior through the formation of a metal-oxide interlayer [1912.08373].

## 4. Magnetic and optomagnetic switching thresholds

In spintronics, threshold switching is usually expressed as a current threshold for destabilizing a magnetic equilibrium. For spin-Hall-effect switching of a perpendicular ferromagnetic layer in current-in-plane geometry, the threshold current is the minimum current density needed to drive the magnetization from one perpendicular state to the opposite one. In the high-damping regime, the threshold is independent of the damping constant and is almost linearly proportional to both effective perpendicular magnetic anisotropy field and external in-plane field. For \(H_x \ll H_{K,\mathrm{eff}}\), the useful design expression is
\[
J_{C,\perp}^{\mathrm{SH}}
=
\frac{2e}{\hbar}\frac{M_s t_F}{\theta_{\mathrm{SH}}}
\left(
\frac{H_{K,\mathrm{eff}}}{2}
-
\frac{H_x}{\sqrt{2}}
\right),
\]
and the damping-independent regime is found numerically for roughly \(\alpha \gtrsim 0.03\) [1210.3442].

Field-free spin-orbit-torque switching introduces a different threshold structure. In anomalous-SOT-driven perpendicular switching, an out-of-plane spin-polarization component \(\sigma_z=\sin\beta\) breaks the symmetry that ordinarily prevents deterministic field-free reversal. The analytical threshold current
\[
J_c =
\frac{e\mu_0 M_s H_K t}{\hbar \theta_{SH}}
\frac{4\alpha}{\sqrt{\sin^2\beta+16\alpha^2\cos^2\beta}+\sin\beta}
\]
shows negative correlation with the out-of-plane torque proportion and positive correlation with anisotropy, free-layer thickness, and damping. In the same paper, sub-threshold current produces an effective exchange-bias field,
\[
H_{\mathrm{bias}}
=
\frac{\theta_{SH}\hbar \sin\beta}{2 e t \mu_0 M_s \alpha }\, J_{\mathrm{SOT}},
\]
which provides a direct route to estimating anomalous-SOT efficiency [2304.02248].

For field-free switching of in-plane magnetization in three-terminal MTJs, the threshold current is derived by linearizing the LLG dynamics around the initial equilibrium and applying the Routh–Hurwitz criterion. Two cases are treated: a Type-x device switched by combined SOT+STT and a Type-x device with a canted easy axis. The theory predicts that the threshold is linear in damping for the hybrid SOT+STT case and decreases with cant angle \(\phi\) in the canted-axis case, with good agreement between analytical formulas and numerical integration [2305.03388].

All-optical switching uses a different threshold variable: incident fluence. In dielectric-coated Co/Gd synthetic ferrimagnets, the threshold fluence is modeled by the proportionality
\[
F_{\mathrm{th}} \propto \frac{1}{A_{\mathrm{loc}}},
\]
and thin-film interference yields the quarter-wave condition
\[
t_d = (2m+1)\frac{\lambda}{4n}.
\]
Experimentally, dielectric engineering reduces the single-pulse threshold fluence from \(3.2~\mathrm{mJ/cm^2}\) in an uncoated film to \(1.2~\mathrm{mJ/cm^2}\) with a bottom SiO\(_2\) layer and to \(0.6~\mathrm{mJ/cm^2}\) in the optimized double-coated structure, corresponding to \(15~\mathrm{fJ}\) for a \(50 \times 50~\mathrm{nm}^2\) cell [2110.12063].

## 5. Photovoltaic and circuit-level exploitation

In CdTe photovoltaics, threshold switching is investigated as a manufacturing-enabling conductive-path formation process rather than as a memory state. The 2021 study reports a snap-back I–V feature in completed thin-film CdTe solar cells, with threshold voltage \(V_{th}\sim 10~\text{V}\), and interprets the post-threshold state as a conductive filament through the solar cell reaching the buried TCO. Using a sheet-resistance model, the total resistance of such a filamentary contact is written as
\[
R_{\rm tot}=R+(s/2\pi)\ln (L/a),
\]
and the inferred filament radius is \(a\sim 1~\mu\text{m}\). The paper proposes that electrically written vias could replace laser-scribed interconnects in a “no-scribe” photovoltaic architecture, while also acknowledging unresolved issues of mechanism, reproducibility, and integration [2108.03756].

The 2023 CdTe study frames threshold switching as a sharp drop in resistance when the applied electric field or voltage exceeds a critical threshold and uses a more direct laboratory protocol: \(10~\text{cm} \times 10~\text{cm}\) CdTe PV samples, a small scribe about \(3.5~\text{mm}\) long, and fixed DC biases applied for \(10~\text{s}\). Positive-bias examples include \(1.00 \times 10^6~\Omega \rightarrow 3.49 \times 10^3~\Omega\) at \(15~\text{V}\); negative-bias examples include \(1.01 \times 10^6~\Omega \rightarrow 1.63 \times 10^3~\Omega\) at \(-20~\text{V}\). The switched state persists for at least \(48~\text{h}\) in several samples. At the same time, the paper explicitly notes that the evidence is based on before/after resistance measurements rather than full swept I–V switching curves and that the demonstration still uses a manually created scribe region [2310.18487].

Threshold-switching devices have also been used as circuit primitives for hysteretic rectification. In the Hyperdiode concept, a threshold-switching phase-transition material is placed in series with a diode-connected MOSFET. For a 3-stage Cockcroft–Walton multiplier, the no-load relation is written as
\[
V_{OUT:N} = 2 \times N \times V_P - N \times V_{DROP},
\]
and the hysteretic selector lowers the effective forward drop because the device remains ON until the lower turn-off threshold. In the reported simulations, a standalone Ag-HfO\(_2\)-Pt threshold switch and an Ag-HfO\(_2\)-Pt-based Hyperdiode yield settling times of about \(5~\mu\text{s}\) and \(120~\mu\text{s}\), respectively, compared with \(2.5~\text{ms}\) for a MOS-diode, corresponding to approximately \(500\times\) and \(20\times\) lower settling time. In the same work, optimum Hyperdiode parameters were selected from a stability sweep over \(-0.2 < V_{th} < 0.8\): \(V_{th}=-0.2~\text{V}\), \(n_{fin}=2\) for Ag-HfO\(_2\)-Pt, and \(V_{th}=0~\text{V}\), \(n_{fin}=8\) for VO\(_2\) [2310.09292].

These application papers collectively shift threshold switching from a microscopic transport phenomenon to a system function: buried-electrode access in photovoltaics, nonlinear selectivity in multipliers, and low-leakage hysteretic rectification. A plausible implication is that the technological importance of threshold switching now lies as much in how abruptness and hysteresis are exploited at architecture level as in the underlying material mechanism.

## 6. Threshold-triggered switching in algorithms and dynamical systems

Outside condensed-matter and device physics, the phrase reappears in formally related but mechanistically different settings. In adversarial online learning with feedback graphs and switching costs, Threshold Based EXP3 keeps the current action until a threshold event \(E^t = E_1^t \cup E_2^t \cup E_3^t\) occurs. The time threshold is
\[
E_{2}^{t}=\{r>\gamma_{t}\},\qquad
\gamma_t = T^{1/3}c^{2/3}/\mathrm{mas}(G_{(T)})^{1/3},
\]
and the loss-threshold event \(E_3^t\) uses accumulated importance-weighted losses to decide whether switching is worth the penalty. The resulting expected regret in symmetric partial information and MAB is order-optimal at \(T^{2/3}\)-type scale, and the expected number of switches is bounded by
\[
\mathbf{E}\bigg[\sum_{t=2}^{T}1(i_{t-1}\neq i_{t})\bigg]\leq 2T^{2/3}c^{-2/3}\mathrm{mas}^{1/3}(G_{(T)}).
\]
Here threshold switching means threshold-triggered action persistence rather than a physical state transition [1810.09666].

In stochastic epidemic theory, threshold switching appears as a semi-Markov regime-switching problem. For an SAIRS model with vaccination and switching transmission rates, the long-run regime occupation is weighted by \(\pi_r m_r\), not only by the stationary distribution of the embedded chain. In the symmetric case \(\beta_A(r)=\beta_I(r)=\beta(r)\) and \(\delta_A=\delta_I=\delta\), the paper defines
\[
\mathcal R_0
=
\frac{\sum_{r \in \mathcal M} \pi_r m_r \beta(r) S_0}{\sum_{r \in \mathcal M}\pi_r m_r (\delta+\mu)},
\qquad
S_0=\frac{\mu+\gamma}{\mu+\nu+\gamma},
\]
and proves that \(\mathcal R_0<1\) implies almost sure extinction while \(\mathcal R_0>1\) implies persistence in time mean. In this literature, threshold switching refers to environmental switching and threshold dynamics between extinction and persistence rather than to an abrupt conductive transition [2110.15435].

Digital circuit timing theory supplies yet another abstraction. A thresholded hybrid system is defined so that the active ODE mode is determined solely by externally supplied digital input signals, while digital outputs are produced by comparing an internal analog state to a threshold:
\[
\Theta_\xi(x)(t)=
\begin{cases}
0 & \text{if } x(t) \le \xi,\\
1 & \text{if } x(t) > \xi.
\end{cases}
\]
For a large class of such systems, the mapping from digital input signals to digital output signals is proved continuous, and this continuity is preserved under strictly causal composition. The paper applies the result to inverters and two-input CMOS NOR gates and provides a refined thresholded hybrid NOR model that captures multi-input switching effects [2403.03235].

Taken together, these generalized usages show that threshold switching has become a structural concept: a threshold separates modes, and the transition law is analyzed for continuity, stability, regret, or persistence rather than for transport or magnetization alone.

Threshold switching is therefore best understood as a family of threshold-governed abrupt transitions with domain-specific realizations. In resistive and selector devices it is often volatile, filamentary, and hysteretic; in spintronics it is a current-driven stability boundary; in optomagnetism it is a fluence threshold governed in part by absorption engineering; in photovoltaics it can be exploited as an electrically written contact path; and in mathematics and algorithms it becomes a rule for threshold-triggered mode change or threshold-determined phase behavior. Across these usages, the central invariant is the same: a critical control parameter partitions qualitatively different dynamical regimes.

Source: https://www.emergentmind.com/topics/threshold-switching