Doping in garnet-type electrolytes: Kinetic and thermodynamic effects from molecular dynamics simulations

Mottet, Matthieu; Marcolongo, Aris; Laino, Teodoro; Tavernelli, Ivano

doi:10.1103/physrevmaterials.3.035403

Cited by 14 publications

(17 citation statements)

References 58 publications

(87 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The activation energies are very stable with values ranging from 0.2 to 0.22 eV. Previous computational studies of LLZO report values between 0.18 and 0.27 eV . Our results are in very good agreement with the experimental activation energy of cubic LLZO.…”

Section: Resultssupporting

confidence: 88%

“…The DeepMD model is Hamiltonian in nature and is therefore suited to evaluate transport statistical mechanical properties like diffusion coefficients. In this work, we develop training strategies and test the power of the DeePMD model for the prediction of diffusion properties of three well‐known SSEs: Li 10 GeP 2 S 12 , Li 7 La 3 Zr 2 O 12 , and Na 3 Zr 2 Si 2 PO 12 . Overall, the deepMD model is able to reproduce the superionic behavior of SSEs and we reach with our training protocol an uncertainty on the computed diffusion coefficients of about 20 % or smaller, which is precise enough for applications.…”

Section: Introductionmentioning

confidence: 83%

“…The recently developed open-source DeePMD model, based on a deep neural network (DNN) architecture 1,18 , has already been tested for its transferability properties in Al-Mg systems 19 , and it is therefore a good candidate model for such a purpose. In this work, we develop training strategies and test the power of the DeePMD model for the prediction of diffusion properties of three well-known SSEs: Li 10 GeP 2 S 12 (LGPS) [20][21][22] , Li 7 La 3 Zr 2 O 12 (LLZO) [23][24][25][26] , and Na 3 Zr 2 Si 2 PO 12 (NASICON) 27,28 . Overall, the deepMD model is able to reproduce the superionic behavior of SSEs and we reach with our training protocol an uncertainty on the computed diffusion coefficients of about 20% or smaller, which is precise enough for applications.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Simulating Diffusion Properties of Solid‐State Electrolytes via a Neural Network Potential: Performance and Training Scheme

et al. 2019

Self Cite

View full text Add to dashboard Cite

The recently published DeePMD model, based on a deep neural network architecture, [1] brings the hope of solving the timescale issue which often prevents the application of first principle molecular dynamics to physical systems. With this contribution we assess the performance of the DeePMD potential on a reallife application and model diffusion of ions in solid-state electrolytes. We consider as test cases the well known PO 12 . We develop and test a training protocol suitable for the computation of diffusion coefficients, which is one of the key properties to be optimized for battery applications, and we find good agreement with previous computations. Our results show that the DeePMD model may be a successful component of a framework to identify novel solid-state electrolytes. ChemSystemsChemArticles . Arrhenius plots for Li diffusion in LLZO, with S 0 sampled from a short (5 ps) FPMD trajectory and a simulations cell of 192 atoms. The color coding refers to the step number during the temperature ramp, counting as first the training step at 300 K. Error bars indicate one standard deviation. Plots (a) and (b) report diffusion coefficients evaluated during the second and third traning loop, respectively. Figure 5. Arrhenius plots for Li diffusion in LLZO, with S 0 sampled from long PFF-MD simulations and a simulation cell of 192 atoms. Only the first loop is reported and the same conventions as Figure 4 are followed. (a,b) Training workflows using different LLZO volumes, cf. text. (c) Diffusion coefficients computed during a workflow where the "maximum bond variation" check was active.ChemSystemsChem Articles

show abstract

Section: Resultssupporting

confidence: 88%

Section: Introductionmentioning

confidence: 83%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Simulating Diffusion Properties of Solid‐State Electrolytes via a Neural Network Potential: Performance and Training Scheme

et al. 2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…scales. In order to support the statistical relevance of our results we developed a classical core-shell potential [64] which is suitable to describe solid-state electrolyte materials [65]. The interatomic potential energy is given by the sum of the electrostatic energy between ions, a Buckingham term describing the short range repulsion, and van-der-Waals interactions between particles i and j at a distance r i j :…”

Section: Molecular Dynamics Simulationsmentioning

confidence: 97%

The solid-state Li-ion conductor Li7TaO6: A combined computational and experimental study

et al. 2020

Self Cite

View full text Add to dashboard Cite

A B S T R A C TWe study the oxo-hexametallate Li 7 TaO 6 with first-principles and classical molecular dynamics simulations, obtaining a low activation barrier for diffusion of ∼0.29 eV and a high ionic conductivity of 5.7 × 10 −4 S cm −1 at room temperature (300 K). We find evidence for a wide electrochemical stability window from both calculations and experiments, suggesting its viable use as a solid-state electrolyte in next-generation solid-state Li-ion batteries. To assess its applicability in an electrochemical energy storage system, we performed electrochemical impedance spectroscopy measurements on multicrystalline pellets, finding substantial ionic conductivity, if below the values predicted from simulation. We further elucidate the relationship between synthesis conditions and the observed ionic conductivity using X-ray diffraction, inductively coupled plasma optical emission spectrometry, and X-ray photoelectron spectroscopy, and study the effects of Zr and Mo doping. ‡ These authors contributed equally to this work between the electrolyte and electrode [18,19]. Li-superionic conductors (LISICON) comprise another family of compounds explored for high ionic conductivity. One such compound is Li 3+x (P 1−x Si x )O 4 , a solid solution of Li 3 PO 4 and Li 4 SiO 4 [20, 21]; Substitutions of P and Si with B, Al, Zr, Ge, Ti, or As led to the discovery of several fast conductors [22][23][24], and the substitution of sulfur with oxygen resulted in the sub-family of thio-LISICONs [25][26][27][28][29][30]. The increased ionic conductivity of these compounds compared to the respective oxygen-based LISICONs is attributed to a higher polarizability of the anion [9,26]. One of the best ionic conductors, tetragonal Li 10 GeP 2 S 12 (LGPS) [31], with an ionic conductivity of 12 mS cm −1 at room temperature, is found in this family. However, sulfur substitution has a deleterious effects on the electrochemical stability [32] of the SSE, compared to their oxygen counterparts. In addition, oxides have higher bulk and shear moduli than sulfides, and this increased mechanical stability could benefit the suppression of Li-metal dendrite growth [33].Switching to thin-film batteries could also lead to technological breakthroughs due to beneficial mechanical properties (lower susceptibility to volume changes) and low resistance due to reduced dimensions [2,34,35]. Lithium phosphorus oxynitrides (LiPON) Li x PO y N z (x = 2y + 3z − 5) can be grown into amorphous thin films, but its activation barriers at ∼0.55 eV, and ionic conductivity at room temperature of 2.3 × 10 −6 S cm −1 are significantly worse than LGPS [36]. Still, these shortcomings are compensated by reductions in the electrolyte thickness for use in thin- arXiv:1910.11079v1 [cond-mat.mtrl-sci]

show abstract

“…One of the limitations of the study presented here is that we consider dopants as ideal, i.e, they affect the defect chemistry of Li 3 OCl only through their effect on the the Fermi energy. In reality, direct dopant-defect interactions may be significant [7,14,15,91]. For example, supervalent dopants, such as Mg 2+ , are predicted to kinetically trap lithium vacancies [15], which will reduce ionic conductivities relative to the values presented in this work.…”

Section: Summary and Discussionmentioning

confidence: 77%

Aliovalent doping response and impact on ionic conductivity in the antiperovskite solid electrolyte Li3OCl

Squires

Dean

Morgan

2021

Preprint

View full text Add to dashboard Cite

Aliovalent doping of solid electrolytes with the intention of increase the concentration of charge carrying mobile defects is a common strategy for enhancing their ionic conductivities. For the antiperovskite lithium-ion solid electrolyte Li3OCl, both supervalent (donor) and subvalent (acceptor) doping schemes have previously been proposed. The effectiveness of these doping schemes depends on two conditions: first, that aliovalent doping promotes the formation of mobile lithium vacancies or interstitials rather than competing immobile defects; and second, that any increase in lithium defect concentration gives a corresponding increase in ionic conductivity. To evaluate the effectiveness of aliovalent doping in Li3OCl, we have performed a hybrid density-functional theory study of the defect chemistry of Li3OCl and the response to supervalent and subvalent doping. In nominally stoichiometric Li3OCl the dominant native defects are predicted to be VLi, OCl, and VCl. Supervalent doping increases VLi and OCl concentrations, with the preferentially formed defect species dependent on synthesis conditions. Subvalent doping increases the concentration of VCl more than the concentration of Lii under all accessible synthesis conditions. While supervalent doping is predicted to be effective at increasing ionic conductivity, particularly under Li-poor synthesis conditions, subvalent doping is predicted to decrease room-temperature ionic conductivities at low-to-moderate doping levels. This effect is due to a reduction in the number of lithium vacancies formed during synthesis, and increased [VLi + Lii] Frenkel-pair recombination upon cooling to room temperature. The strongly asymmetric doping response of Li3OCl with respect to supervalent versus subvalent doping is explained as a consequence of the low [VLi + VCl] Schottky pair formation energy, suggesting analogous behaviour should be expected in other Schottky-disordered solid electrolytes.

show abstract

Doping in garnet-type electrolytes: Kinetic and thermodynamic effects from molecular dynamics simulations

Cited by 14 publications

References 58 publications

Simulating Diffusion Properties of Solid‐State Electrolytes via a Neural Network Potential: Performance and Training Scheme

Simulating Diffusion Properties of Solid‐State Electrolytes via a Neural Network Potential: Performance and Training Scheme

The solid-state Li-ion conductor Li7TaO6: A combined computational and experimental study

Aliovalent doping response and impact on ionic conductivity in the antiperovskite solid electrolyte Li3OCl

Contact Info

Product

Resources

About