Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Theory of ultrafast conductance modulation in electrochemical protonic synapses by multiphase polarization

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An electric field across a phase-separating tungsten-oxide channel forms a conductive filament at the gate, which the authors argue is what allows 5-nanosecond linear conductance updates despite slow ion diffusion.

desk verdict A plausible, well-transparent theory paper that explains an eight-order-of-magnitude switching-time gap via field-induced phase separation, but the central mechanism rides on two unverified material properties and needs explicit experimental confirmation. read the letter →

arxiv 2507.23576 v1 pith:AZYNAOTJ submitted 2025-07-31 cond-mat.mes-hall

Michael L. Li , Dingyu Shen , Jesus A. del Alamo , Martin Z. Bazant This is my paper · ORCID
classification cond-mat.mes-hall
keywords electrochemicalionicsynapseprotonicmemoryconductancemodulationphaseseparationmultiphasepolarizationtungstenoxidephase-fieldmodeldiffusion-limitedswitching
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that the mysterious nanosecond switching speed of a protonic electrochemical synapse is not a violation of diffusion limits but a consequence of the channel material's thermodynamics. The authors claim that a polycrystalline tungsten-oxide channel, when polarized, forms a thin high-concentration conductive filament at the electrolyte interface even though the average proton concentration lies in a thermodynamically stable single-phase region. This filament pins the interfacial reaction environment, so repeated voltage pulses produce linear and symmetric conductance changes despite bulk diffusion being slow. If correct, the explanation reconciles two experiments that switch eight orders of magnitude apart in pulse duration and points to phase-separating materials as a design route for fast analog memory.

What carries the argument

The load-bearing object is a phase-field model of multiphase polarization: transport of protons and electrons in a lattice-constrained intercalation material with a Cahn-Hilliard gradient penalty and a concentration-dependent enthalpy. The enthalpy is fitted separately for amorphous and polycrystalline WO3 from open-circuit voltage curves, so the polycrystalline free energy develops three wells corresponding to monoclinic and tetragonal phases, creating thermodynamically unstable binodal gaps. Under the condition that electron diffusivity far exceeds proton diffusivity, the two-species model reduces to a single-species Cahn-Hilliard equation, and the electric field can push the local concentration into a binodal gap to nucleate a metastable filament; the high electron conductivity of that filament suppresses further polarization. The model's two governing ratios, reaction-to-diffusion and pulse-to-diffusion timescales, organize a kinetic phase diagram with four operating regimes, placing the fast polycrystalline device in the metastable filament regime.

What would settle it

Measure the crystal structure and phase behavior of the actual 30 nm annealed WO3 film, for example by transmission electron diffraction or glancing-angle X-ray diffraction, and look operando for a high-concentration filament at the gate during a 5 ns pulse; absence of polycrystallinity or of any filament would settle the mechanism against the paper.

Watch

Extended reading notes

Core claim

The paper's central claim is that ultrafast (5 ns) linear conductance modulation in the WO3 electrochemical ionic synapse is made possible by electric-field-induced metastable phase separation. In the fast device, the WO3 channel is polycrystalline and has a multi-well free energy; a large applied field polarizes the low-concentration bulk phase and drives the local proton filling fraction into a binodal gap, nucleating a high-concentration filament at the gate. Because that filament is roughly an order of magnitude more electron-conductive than the surrounding phase, it screens the field and keeps the electrolyte-WO3 interfacial potential nearly constant from pulse to pulse, even though the device is diffusion-limited. The conductance then reads as a weighted sum of low- and high-conductance phase contributions whose filament length grows proportionally with pulse number, giving linearity, with symmetry set by an asymmetric reaction charge-transfer coefficient. The same framework reproduces the slower, reaction-limited behavior of an amorphous solid-solution WO3 device, where concentration polarization is negligible and linearity comes from symmetric interfacial kinetics.

Load-bearing premise

The paper assumes the fast device's 30 nm WO3 film really is polycrystalline and phase-separating because it was annealed at 400 °C, even though the original experiment never measured the film's structure; if the film is amorphous or single-phase, the filament mechanism cannot occur.

Editorial extensions

If this is right

  • If the mechanism is right, diffusion-limited operation is not an obstacle: a phase-separating channel can deliver nanosecond-scale linear conductance updates by exploiting field-induced filaments rather than fast bulk transport.
  • Annealing WO3 to a polycrystalline, phase-separating structure becomes a concrete lever for speeding up EIoS devices, since crystallinity controls whether the multi-well free energy exists.
  • The conductance relaxation of a phase-separating device should show a transient whose sign is opposite to the pulse direction, arising from phase-boundary relaxation on a timescale set by the gradient-penalty coefficient.
  • Pulse train and relaxation responses can serve as a cheap fingerprint of whether a fabricated film is solid-solution or phase-separating, avoiding direct microstructure characterization.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An implication the authors leave implicit: any mixed ion-electron conductor with a multi-well free energy and a large electron-to-proton diffusivity contrast could show the same ultrafast filament-pinned switching, so the design principle may transfer to lithionic or other protonic electrolytes.
  • A testable extension: measuring the proton diffusion coefficient in 30 nm annealed or hydrated films, rather than bulk samples, could remove the orders-of-magnitude discrepancy the paper flags and either support or weaken the pseudo-capacitor hypothesis.
  • The model predicts that if the pulse duration is reduced below the phase-boundary formation timescale, linearity should abruptly degrade; that threshold is a concrete experimental target.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The manuscript proposes a phase-field theory for the ultrafast (5 ns) linear conductance modulation observed in protonic electrochemical ionic synapses with WO3 channels. It models two experimental studies from the Onen group: a slow (1 s) amorphous WO3 device and a fast (5 ns) device argued to be polycrystalline after 400 °C annealing. Proton and electron intercalation are described by a multiphase-polarization model with a fitted excess chemical potential calibrated to literature open-circuit voltage data, and electron/proton diffusivities and interfacial kinetics are parameterized to reproduce both conductance trains. The central mechanism is field-induced metastable phase separation in polycrystalline WO3: a high-concentration filament forms near the gate, screens the bulk field, and pins the interfacial reaction environment, giving linear/symmetric conductance modulation even under diffusion limitation. The paper also constructs a kinetic phase diagram in the (τrxn/τdiff, τpulse/τdiff) plane to delineate regimes and predicts qualitatively different relaxation signatures for solid-solution versus phase-separating materials.

Significance. If the mechanism were validated, the work would be important: it would overturn the common assumption that EIoS switching time is bounded by bulk ion diffusion and would give a concrete materials-design principle (engineering phase-separating thermodynamics) for nanosecond analog memory. The paper has real strengths: the thermodynamic free energy is calibrated to independent OCV data rather than to the target conductance traces; the model reproduces both experimental conductance trains with a parameter set per device; simulation results are openly available; and the kinetic phase diagram plus distinct relaxation predictions are falsifiable. The main caveat is that the fast-device explanation rests on two unverified inputs—the polycrystallinity/phase-separation of the 30 nm film in study 2 and a proton diffusivity 100–1000 times larger than bulk values—and the paper itself concedes that the fitted diffusion timescale should not respond to a 5 ns pulse. These issues are not resolved by the current manuscript.

major comments (4)
  1. [II.A, Fig. 2] The assignment of phase-separating polycrystalline thermodynamics to study 2 is an inference: the text states that 'the physical evidence of this change was not explicitly measured in that study,' and the multi-well free energy is fitted to bulk polycrystalline OCV data (Fig. 2b) rather than to properties of the 30 nm film. Because the entire mechanism in Fig. 3d requires the channel to be phase-separating in the operating filling range, this missing characterization is load-bearing. The paper should either provide direct structural/thermodynamic evidence for the device film (e.g., TEM/XRD or in situ OCV on the actual film) or explicitly reframe the central claim as a hypothesis to be tested, with the experiments that would distinguish it from a single-phase fast pathway.
  2. [III, Table I] The model parameters give τdiff = 3.4×10^-4 s and τpulse = 5×10^-9 s, so τpulse/τdiff ≈ 1.5×10^-5; in a diffusion-limited system, no concentration response is expected on this pulse. The text itself says that 'applying these parameters results in a diffusion timescale of approximately 0.1 seconds. This duration should not exhibit any response to a 5 nanosecond pulse' (this sentence also conflicts with Table I's value of 3.4×10^-4 s). The central claim that the polycrystalline device responds in 5 ns notwithstanding its diffusion-limited character therefore requires a quantitative fast pathway—such as the PSG pseudocapacitor or hydrated WO3 mentioned later—rather than the speculative paragraph currently given.
  3. [III, Table I and Fig. 4] The diffusivity Dp = 2.6×10^-8 cm2 s^-1 is 100–1000 times larger than bulk proton diffusivities cited from Refs. [28–30]. This parameter, together with τrxn, α = 0.4, and Rpsg, determines the placement of the polycrystalline device in the diffusion-limited, filament-forming regime (region 2 of Fig. 4). Because these values are not independently measured, the kinetic-phase-diagram assignment is at risk of circular calibration: the model is placed in the regime that produces the required filament. Please add independent constraints on Dp and a sensitivity analysis showing that the qualitative mechanism is robust to parameter uncertainty.
  4. [III.A, Fig. 5] The claim that phase-separating thermodynamics, rather than the fitted kinetics, is responsible for the experimentally reproduced linearity and symmetry is not supported by an identifiability analysis. The polycrystalline fit uses an asymmetric α = 0.4 and a specific Rpsg to obtain the experimental up/down asymmetry, and Fig. 5 shows the phase-separating model's performance is insensitive to τrxn/τdiff for the chosen parameters. I would like to see either a parameter-perturbation study demonstrating that the filament-pinning mechanism is required to match the experimental conductance trains, or a clear statement of which aspects of the data are uniquely attributable to multiphase polarization.
minor comments (5)
  1. [Abstract] There is a typo in the abstract: 'polarizatino' should be 'polarization'.
  2. [II.A, Eq. (3)] The notation µp,ex is used before its Legendre-polynomial definition is given; please define the fitting range and the coefficients an explicitly to improve reproducibility.
  3. [III, Table I] The sentence 'applying these parameters results in a diffusion timescale of approximately 0.1 seconds' is numerically inconsistent with Table I, where τdiff = 3.4×10^-4 s; please correct the text or clarify which parameter set is being discussed.
  4. [References] Reference [21] is incomplete (no authors or title), and reference [30] appears to be an unpublished preprint without journal or volume information; please supply complete citations.
  5. [Conclusions] The conclusion that crystallinity was 'identified' as a key factor overstates the evidence presented in the manuscript; 'hypothesized' or 'proposed' would be more accurate.

Circularity Check

3 steps flagged · score 6.0 of 10

Partial circularity: the symmetry of the fast-switching response is imposed by the fitted charge-transfer coefficient α=0.4 rather than predicted, and the kinetic phase diagram 'validation' reuses the same fitted parameters; the central phase-separation premise is admitted unmeasured.

  1. fitted input called prediction [Section III, paragraph following Eq. (10b) and the weighted-sum conductance expression (discussion of Fig. 3c,d)]
    "Symmetry of the conductance modulation was achieved in these simulations by utilizing an asymmetric reaction charge-transfer coefficient, which accounted for the different magnitude of the up and down pulses used experimentally."

    The paper presents symmetric conductance modulation as an emergent prediction of the phase-separating mechanism, but the symmetry is built into the model by choosing α=0.4 for the polycrystalline system (Table I). This parameter is explicitly selected to account for the experimental up/down pulse asymmetry, so the 'prediction' of symmetry is a fitted input rather than a derived consequence. The mechanism explains linearity only through the assumed weighted-sum conductance rule; the symmetry axis is constructed by the fit.

  2. fitted input called prediction [Section III.A, paragraph following the enumeration of the four kinetic phase-diagram regions]
    "The parameters governing the polycrystalline WO3 system from Study 2 were compared against the kinetic phase diagram, which indicated operation in Region (2) and was validated by the simulation results."

    Region (2) of the kinetic phase diagram is defined by the timescale inequalities τdiff ≫ τrxn and short pulses, and the diagram is generated from the same model with the same fitted parameters. Placing Study 2 in Region (2) therefore restates the chosen Table I values (τdiff=3.4e-4 s, τrxn=3.2e-7 s, τpulse=5e-9 s) rather than testing them. Calling the simulations a 'validation' is circular because the simulations use those exact parameters; there is no independent experimental determination of the dynamic parameters or of the phase-separation regime.

1 more flagged steps
  1. renaming known result [Section II.A, paragraph describing Figure 2 and the thermodynamic fitting]
    "The crystallinity-specific enthalpy was fitted to open-circuit voltage curves found in the literature... Doing so revealed thermodynamically unstable regions in the polycrystalline WO3 which led to voltage plateaus predicted at the binodal concentrations."

    The homogeneous free energy is least-squares fitted to OCV data that already exhibit the voltage plateaus (Fig. 2b). The 'predicted' binodal voltage plateaus are therefore a restatement of the fitted enthalpy, not an independent prediction of the thermodynamic model. This is a minor renaming of calibration output as prediction; it does not by itself drive the device mechanism, but it contributes to the paper's pattern of presenting fitted outputs as results.

full rationale

The paper has real independent content: the thermodynamic free energies are calibrated to external OCV literature data (Crandall et al., Jarman & Dickens), and the multiphase-polarization transport framework is a prior theoretical model (Tian & Bazant), not a self-citation loaded premise or a uniqueness argument. No machine-checked or externally falsified claim is being replaced by a self-citation chain. However, the presentation is partially circular in three places. First, the symmetry of the polycrystalline conductance modulation is not derived from the phase-separation mechanism; it is imposed by choosing α=0.4 to match the experimental up/down pulse asymmetry, so the headline 'linearity and symmetry arise' is partly a fitted-input-as-prediction. Second, the 'validation' of the kinetic phase diagram location for Study 2 is circular because the phase diagram and the validating simulations share the same fitted τdiff/τrxn/τpulse parameters. Third, the binodal plateaus are presented as predictions when they are the outputs of the OCV fit. The central mechanistic claim—that a metastable high-concentration filament pins the gate interface—remains a genuine hypothesis, but the paper itself flags two severe empirical weaknesses that are correctness risks rather than circularity: the polycrystallinity/phase separation of the actual 30 nm film was not measured ('the physical evidence of this change was not explicitly measured in that study'), and the fitted proton diffusivity yields a diffusion timescale that 'should not exhibit any response to a 5 nanosecond pulse.' These admissions mean the explanation is conditional on unverified inputs, but they do not by themselves make the derivation equivalent to its inputs. Overall, the symmetry and phase-diagram validations reduce by construction, warranting a partial-circularity score of 6.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central mechanism depends on a calibrated phase-field model. The free energy is fitted to external OCV data, which is independent support, but the dynamic parameters Dp, tau_rxn, alpha, and Rpsg are not independently pinned down, and the paper admits a 100 to 1000 times discrepancy in proton diffusivity. The most fragile input is the unmeasured polycrystallinity and phase separation of the fast device. No new physical entities are postulated.

free parameters (6)
  • Legendre coefficients for proton excess chemical potential (mu_p,ex) = not tabulated; least-squares fit to literature OCV curves
    The phase-separating multi-well free energy, which drives the filament mechanism, comes from fitting these coefficients to bulk OCV data rather than to the device itself.
  • Proton diffusivity Dp = 2.6e-8 cm2/s
    Described in the paper as 100 to 1000 times larger than bulk literature values; it is load-bearing for the timescale ratios that put the polycrystalline system in the diffusion-limited regime.
  • Charge-transfer coefficient alpha for polycrystalline WO3 = 0.4
    The paper states that symmetry of the conductance modulation was achieved by using an asymmetric transfer coefficient to account for the different up and down pulse voltages; this is effectively a symmetry-fitting parameter.
  • Exchange current densities encoded in tau_rxn = tau_rxn = 7.0e4 s (amorphous), 3.2e-7 s (polycrystalline)
    No independent measurement is reported for these values; they set the reaction-limited versus diffusion-limited regime classification.
  • PSG solid electrolyte resistance Rpsg = 1.63 and 4.18 Ohm m2
    The solid electrolyte is reduced to a series resistance that dominates the applied potential; the values are not derived from independent transport data in this paper.
  • Gradient penalty kappa for polycrystalline WO3 = 1e-4 in the full model, 1e-3 in the simplified model
    No independent measurement is given; it controls the phase boundary width and the predicted relaxation signature in Figure 5d.
assumptions (6)
  • domain assumption Local electroneutrality cp = cn in the WO3 bulk
    Justified by a Debye length of about 0.4 angstrom, smaller than the 5.2 angstrom lattice constant; this reduces two-species transport to one concentration field.
  • domain assumption Protons and electrons intercalate on distinct lattices with only electrostatic coupling
    The model treats gate proton intercalation and source/drain electron intercalation as spatially decoupled reactions; the validity for nanoscale WO3 films is not directly tested.
  • domain assumption The PSG electrolyte can be represented as a purely resistive series element
    The paper itself raises the possibility of a proton pseudo-capacitor in the electrolyte that is not included in the model, so the resistance-only representation may miss a load-bearing effect.
  • ad hoc to paper Annealing at 400 degrees Celsius makes the WO3 channel polycrystalline and phase-separating
    Not directly measured in the device; inferred from literature on WO3 annealing, and the multi-well free energy is fitted to bulk OCV data from different samples.
  • domain assumption Natural boundary condition n dot grad cp = 0 at phase boundaries neglects surface free energy effects
    The authors state these considerations were neglected; surface energy could change filament nucleation and growth at the gate interface.
  • domain assumption Continuum phase-field dynamics remain valid at 5 ns pulse durations in a 30 nm film
    The diffusion length over 5 ns is roughly 0.1 nm, comparable to the lattice constant, so the continuum approximation is strained at the shortest timescales considered.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Theory of ultrafast conductance modulation in electrochemical protonic synapses by multiphase polarization." pith.science (2026). https://pith.science/paper/AZYNAOTJ

@misc{pith2026250723576,
  author       = {Pith},
  title        = {Pith review of: Theory of ultrafast conductance modulation in electrochemical protonic synapses by multiphase polarization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AZYNAOTJ}},
  note         = {Machine review of arXiv:2507.23576}
}
read the original abstract

Three-terminal electrochemical ionic synapses (EIoS) have recently attracted interest for in-memory computing applications. These devices utilize electrochemical ion intercalation to modulate the ion concentration in the channel material. The electrical conductance, which is concentration dependent, can be read separately and mapped to a non-volatile memory state. To compete with other random access memory technologies, linear and symmetric conductance modulation is often sought after, properties typically thought to be limited by the slow ion diffusion timescale. A recent study by Onen et al.[1] examining protonic EIoS with a tungsten oxide (WO3) channel revealed that this limiting timescale seemed irrelevant, and linear conductance modulation was achieved over nanosecond timescales, much faster than the bulk ion diffusion. This contrasts with previous studies that have shown similar conductance modulation with pulse timescales of milliseconds to seconds. Understanding the phenomena behind these conductance modulation properties in EIoS systems remains a crucial question gating technological improvements to these devices. Here, we provide a theoretical explanation that demonstrates how linearity and symmetry arise from consistent control over the electrolyte-WO3 interface. Comparing these past works, changes in the WO3 channel crystallinity were identified, affecting material thermodynamics and revealing that the device achieving nanosecond pulse timescales underwent phase separation. Coupling of electric field polarizatino and increased electron conductivity in high-concentration filaments, the reaction environment at the gate electrode can be controlled, resulting in ideal conductance modulation within the diffusion-limited regime. This work highlights the potential for phase-separating systems to overcome the traditional diffusion barriers that limit EIoS performance.

Figures

Figures reproduced from arXiv: 2507.23576 by the authors.

Figure 1
Figure 1. FIG. 1. Conductance modulation of protonic electrochemical ionic synapse with a WO [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Experimental and fitted open circuit voltage vs. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Simulated conductance modulation over pulse-train parameters used in (a) study 1 with amorphous WO [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Conductance modulation shown for varying [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 36 canonical work pages

  1. [1]

    Above a critical pulse timescale, the system interca- lated enough protons during a single pulse so that the average concentration in the system lay within the two-phase coexistence region, resulting in phase separation of the bulk

    work page

  2. [2]

    The average pro- ton concentration in WO3 stayed within the single- phase region

    System was diffusion limited τdiff ≫ τrxn, and pulses were short τpulse ≪ 1. The average pro- ton concentration in WO3 stayed within the single- phase region. However, concentration polarization in the low concentration phase led to the formation of metastable filaments near the electrolyte-WO 3 interface

    work page

  3. [3]

    System was not severly diffusion limited τdiff ≯ τrxn, and pulses were short τpulse ≪ 1. The aver- age proton concentration in WO 3 remained in the single-phase region; because the reaction timescale was large, the bulk WO 3 does not exhibit large enough electric fields to induce local phase sepa- ration in the material

    work page

  4. [4]

    The length of this phase bound- ary is given by Lboundary = Lchannel √ ˜κ

    τpulse is much smaller than the τdiff associated with a phase boundary. The length of this phase bound- ary is given by Lboundary = Lchannel √ ˜κ. The WO 3 film does not have sufficient time to respond to the pulse. The parameters governing the polycrystalline WO 3 system from Study 2 were compared against the kinetic phase diagram, which indicated operat...

    work page

  5. [5]

    Huang, M

    M. Huang, M. Schwacke, M. Onen, J. Del Alamo, J. Li, and B. Yildiz, Advanced Materials 35, 10.1002/adma.202205169 (2023), publisher: Wiley

    work page doi:10.1002/adma.202205169 2023

  6. [6]

    M. Onen, N. Emond, B. Wang, D. Zhang, F. M. Ross, J. Li, B. Yildiz, and J. A. Del Alamo, Science 377, 539 (2022)

    work page 2022

  7. [7]

    Malik, IBM Journal of Research and Development 57, 1:1 (2013), publisher: IBM

    P. Malik, IBM Journal of Research and Development 57, 1:1 (2013), publisher: IBM

    work page 2013

  8. [8]

    H. Kang, J. Seo, H. Kim, H. W. Kim, E. R. Hong, N. Kim, D. Lee, and J. Woo, Micromachines 13, 453 (2022), publisher: MDPI AG

    work page 2022

Show all 37 references
  1. [9]

    A. A. Talin, J. Meyer, J. Li, M. Huang, M. Schwacke, H. W. Chung, L. Xu, E. J. Fuller, Y. Li, and B. Yildiz, Chemical Reviews 125, 1962 (2025), publisher: Ameri- can Chemical Society (ACS)

    2025

  2. [10]

    M. Onen, N. Emond, J. Li, B. Yildiz, and J. A. Del Alamo, Nano Letters 21, 6111 (2021)

    2021

  3. [11]

    Yu, Proceedings of the IEEE 106, 260 (2018), pub- lisher: Institute of Electrical and Electronics Engineers (IEEE)

    S. Yu, Proceedings of the IEEE 106, 260 (2018), pub- lisher: Institute of Electrical and Electronics Engineers (IEEE)

    2018

  4. [12]

    Jeong and L

    H. Jeong and L. Shi, Journal of Physics D: Applied Physics 52, 023003 (2019), publisher: IOP Publishing

    2019

  5. [13]

    Sebastian, M

    A. Sebastian, M. Le Gallo, R. Khaddam-Aljameh, and E. Eleftheriou, Nature Nanotechnology 15, 529 (2020), publisher: Springer Science and Business Media LLC

    2020

  6. [14]

    H. Tian, J. Li, and M. Z. Bazant, Advanced Functional Materials 33, 10.1002/adfm.202213621 (2023), publisher: Wiley

    2023 doi

  7. [15]

    S. K. Deb, Philosophical Magazine 27, 801 (1973), pub- lisher: Informa UK Limited

    1973

  8. [16]

    J. Tang, D. Bishop, S. Kim, M. Copel, T. Gokmen, T. Todorov, S. Shin, K.-T. Lee, P. Solomon, K. Chan, W. Haensch, and J. Rozen, in 2018 IEEE International Electron Devices Meeting (IEDM)(IEEE, San Francisco, CA, 2018) pp. 13.1.1–13.1.4

    2018

  9. [17]

    Y. Xi, B. Gao, J. Tang, A. Chen, M.-F. Chang, X. S. Hu, J. Van Der Spiegel, H. Qian, and H. Wu, Proceedings of the IEEE 109, 14 (2020)

    2020

  10. [18]

    X. Yao, K. Klyukin, W. Lu, M. Onen, S. Ryu, D. Kim, N. Emond, I. Waluyo, A. Hunt, J. A. Del Alamo, J. Li, and B. Yildiz, Nature Communications 11, 10.1038/s41467-020-16866-6 (2020), publisher: Springer Science and Business Media LLC

    2020 doi

  11. [19]

    and (b) polycrystalline WO 3 [20]. (c) Learned homo- geneous free energy curve for the protons in amorphous and polycrystalline WO3; the latter shows a multi-welled structure where comparisons have been made with literature-identified phases.[21] µp,ex = P n anPn(c). Doing so ...

  12. [20]

    Kamal, A

    H. Kamal, A. Akl, and K. Abdel-Hady, Physica B: Con- densed Matter 349, 192 (2004), publisher: Elsevier BV

    2004

  13. [21]

    S. K. Deb, Solar Energy Materials and Solar Cells 92, 245 (2008), publisher: Elsevier BV

    2008

  14. [22]

    A. Yu, N. Kumagai, Z. Liu, and J. Y. Lee, Journal of Solid State Electrochemistry 2, 394 (1998), publisher: Springer Science and Business Media LLC

    1998

  15. [23]

    M. Z. Bazant, Faraday Discussions 246, 60 (2023), pub- lisher: Royal Society of Chemistry (RSC)

    2023

  16. [24]

    Crandall, P

    R. Crandall, P. Wojtowicz, and B. Faughnan, Solid State Communications 18, 1409 (1976)

    1976

  17. [25]

    R. H. Jarman and P. G. Dickens, Journal of The Elec- trochemical Society 129, 2276 (1982)

    1982

  18. [26]

    23–36, iSSN: 0065-3276

    in Advances in Quantum Chemistry(Elsevier, 2004) pp. 23–36, iSSN: 0065-3276

    2004

  19. [27]

    Tian and M

    H. Tian and M. Z. Bazant, Nano Letters 22, 5866 (2022), publisher: American Chemical Society (ACS)

    2022

  20. [28]

    A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, and K. A. Persson, APL Materials 1, 10.1063/1.4812323 (2013), publisher: AIP Publishing

    2013 doi

  21. [29]

    Y. Zou, Y. Zhang, D. Lou, H. Wang, L. Gu, Y. Dong, K. Dou, X. Song, and H. Zeng, Journal of Alloys and Compounds 583, 465 (2014), publisher: Elsevier BV

    2014

  22. [30]

    G. R. Jafari, A. A. Saberi, R. Azimirad, A. Z. Moshfegh, and S. Rouhani, Journal of Statistical Mechanics: Theory and Experiment 2006, P09017 (2006), publisher: IOP Publishing

    2006

  23. [31]

    Raistrick, A

    I. Raistrick, A. Mark, and R. Huggins, Solid State Ionics 5, 351 (1981), publisher: Elsevier BV

    1981

  24. [32]

    de la Gorce, Sparse automatique differentiation (2025), gitHub

    M. de la Gorce, Sparse automatique differentiation (2025), gitHub. Retrieved July 16, 2025

    2025

  25. [33]

    Randin and R

    J. Randin and R. Viennet, Journal of The Electrochem- ical Society 129, 2349 (1982), publisher: The Electro- chemical Society

    1982

  26. [34]

    P. G. Dickens, S. J. Hibble, and R. H. Jarman, Journal of Electronic Materials 10, 999 (1981), publisher: Springer Science and Business Media LLC

    1981

  27. [35]

    Matsuo, T

    R. Matsuo, T. Yamasaki, and T. Omata, High Proton Conductivity of HxWO3 at Intermediate Temperatures: Unlocking Its Application as a Mixed Ionic-Electronic Conductor (2025), version Number: 1

    2025

  28. [36]

    J. B. Mitchell, R. Wang, J. S. Ko, J. W. Long, and V. Augustyn, Journal of The Electrochemical Society 169, 030534 (2022)

    2022

  29. [37]

    M. L. Li, 10.5281/zenodo.16637595 (2025)

    2025 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.