Xiliang Lian
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xiliang Lian has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Mathieu Salanne
Sorbonne Université, CNRS
9 shared publications - Maria R. Lukatskaya
ETH Zurich
4 shared publications - Akhil Tayal
National Synchrotron Light Source II Brookhaven National Laboratory, Upton New York USA
2 shared publications Alfredo Pasquarello
2 shared publicationsDario Gomez Vazquez
Electrochemical Energy Systems Laboratory, Department of Mechanical and Process Engineering
2 shared publicationsShianlin Wee
Electrochemical Energy Systems Laboratory, Department of Mechanical and Process Engineering
2 shared publications
Latest Publications
- Open Access
- 21 August 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (17) , 8730-8739
Data-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations by combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3 times less training data than DeePMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ∼10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- 21 August 2026
- Published by American Chemical Society (ACS) in Journal of Chemical Theory and Computation
- Vol. 22 (17) , 8730-8739
Open AccessData-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations by combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3 times less training data than DeePMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ∼10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- Open Access
- 10 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Data-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations, combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3 times less training data than DeePMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ~10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- 10 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessData-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations, combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3 times less training data than DeePMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ~10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- Open Access
- 11 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Data-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3.5 times less training data than DeepMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ∼10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- 11 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessData-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
We develop data-efficient machine learning interatomic potentials (MLIPs) for fast molecular dynamics simulations combining DeePMD and MACE models within an active learning and knowledge distillation framework. Using liquid water as a case study, we first independently train DeePMD and MACE models from scratch through active learning. We find that MACE requires around 3.5 times less training data than DeepMD, but its inference speed is 10 times lower. We also show that starting from a pretrained foundation model based on the MACE architecture further reduces the training data by a factor of 7, resulting in a fine-tuned foundation model with a 25 times data reduction compared to DeePMD. To overcome the limitation associated with the lower inference speed of MACE potentials, we next develop a knowledge distillation scheme to train a DeePMD potential from the fine-tuned foundation model through an inexpensive active learning workflow. The distilled model is generated with ∼10 times less computer time than the DeePMD model trained from scratch, while showing the same fast inference speed. Comparison with ab initio calculations shows that all the models reach the same level of accuracy in reproducing structural, vibrational, and diffusive properties of liquid water. Our approach enables practical, data-efficient training of customized MLIPs with high speed and accuracy.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Machine learning potentials for the simulation of fluoride ion batteries electrolytes
Potentiels d'apprentissage automatique pour la simulation des électrolytes de batteries à ions fluorure: Les batteries à ions fluorure (FIB) représentent une technologie prometteuse post-Li-ion, mais l'identification d'électrolytes avec une haute mobilité des ions fluorure à température ambiante demeure un défi majeur. En effet, la recherche dans ce domaine est freinée par une compréhension limitée du mécanisme de conduction des ions fluorure. Dans ce manuscrit de thèse, je présenterai mon travail sur l'utilisation des dynamiques moléculaires basées sur un potentiel d'apprentissage automatique (MLP) pour comprendre le mécanisme de diffusion ionique dans les électrolytes solides et liquides. En particulier, nous utiliserons le BaSnF4 comme électrolyte solide de référence pour comprendre comment les ions Sn2+ déterminent de manière critique le mouvement des ions F-. À la suite de travaux antérieurs, nous avons conçu un électrolyte à haute entropie, où le Np1F/ACN et le CsF/H2O sont mélangés, pour explorer des électrolytes liquides potentiels avec une haute mobilité des ions F-.Dans le chapitre 3, nous nous concentrerons sur la précision et l'efficacité de la méthode MLP comparée aux potentiels d'interaction classiques et aux dynamiques moléculaires ab initio. Le MLP a émergé comme une méthode prometteuse pour approcher la surface d'énergie potentielle avec une grande précision et efficacité. Cependant, la polyvalence et l'efficacité d'une telle méthode par rapport aux potentiels d'interaction classiques sont rarement étudiées. En utilisant le BaSnF4 et un système auxiliaire plus simple, NaF, qui possède une structure de type sel gemme, nous montrons comment un MLP peut capturer les interactions subtiles des paires solitaires de Sn, tandis qu'un champ de force polarisable échoue. La précision de notre MLP est validée en comparant les propriétés vibrationnelles telles que la dispersion des phonons et les équations d'état avec les résultats obtenus par la théorie de la fonctionnelle de la densité, et elle montre un excellent accord avec cette théorie. Le MLP présente également une efficacité computationnelle nettement améliorée par rapport à la méthode de dynamiques moléculaires ab initio de référence.Dans le chapitre 4, nous fournirons une investigation complète du comportement de diffusion des ions F⁻ pour le BaSnF4, qui présente une haute conductivité ionique de 1×10-4 S/cm à température ambiante, mais dont le mécanisme de conduction ionique sous-jacent et les implications des paires solitaires (LP) de Sn2+ dans ce processus restent obscurs. En utilisant les dynamiques moléculaires basées sur l'apprentissage automatique, nous avons étudié comment le mouvement des différentes espèces affecte le paysage énergétique de diffusion. Nous avons démontré que les ions Sn2+ jouent un rôle fondamental. Leur mouvement translationnel conduit à un canal de conduction ionique avec un rayon plus large, augmentant ainsi la conductivité ionique du BaSnF4. Cela modifie également la dynamique d'orientation des LP. En outre, nous avons examiné de manière exhaustive le comportement de diffusion des ions fluorure dans le BaSnF4 et confirmé une diffusion bidimensionnelle au sein des couches Ba-Sn, tandis que dans l'autre direction, les ions F- traversent rapidement la couche Sn-Sn, créant des défauts de Frenkel dans la structure. Notre travail élucide le mécanisme atomistique régissant la conduction des ions fluorure dans les conducteurs ioniques rapides et contribuera à la conception rationnelle des électrolytes solides pour les FIB.Dans le chapitre 5, nous présenterons nos premières tentatives sur les électrolytes liquides mixtes. Les propriétés prometteuses des solutions isolées de Np1F/ACN et CsF/H2O ainsi que les avantages tirés du système à haute entropie rendent le mélange intéressant. Dans ce travail, nous avons d'abord entraîné le MLP pour le mélange et validé notre MLP par rapport aux résultats calculés par la DFT. Nous avons ensuite réalisé des dynamiques moléculaires à grande échelle et analysé le comportement des ions F-.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Machine learning potentials for the simulation of fluoride ion batteries electrolytes
Potentiels d'apprentissage automatique pour la simulation des électrolytes de batteries à ions fluorure: Les batteries à ions fluorure (FIB) représentent une technologie prometteuse post-Li-ion, mais l'identification d'électrolytes avec une haute mobilité des ions fluorure à température ambiante demeure un défi majeur. En effet, la recherche dans ce domaine est freinée par une compréhension limitée du mécanisme de conduction des ions fluorure. Dans ce manuscrit de thèse, je présenterai mon travail sur l'utilisation des dynamiques moléculaires basées sur un potentiel d'apprentissage automatique (MLP) pour comprendre le mécanisme de diffusion ionique dans les électrolytes solides et liquides. En particulier, nous utiliserons le BaSnF4 comme électrolyte solide de référence pour comprendre comment les ions Sn2+ déterminent de manière critique le mouvement des ions F-. À la suite de travaux antérieurs, nous avons conçu un électrolyte à haute entropie, où le Np1F/ACN et le CsF/H2O sont mélangés, pour explorer des électrolytes liquides potentiels avec une haute mobilité des ions F-.Dans le chapitre 3, nous nous concentrerons sur la précision et l'efficacité de la méthode MLP comparée aux potentiels d'interaction classiques et aux dynamiques moléculaires ab initio. Le MLP a émergé comme une méthode prometteuse pour approcher la surface d'énergie potentielle avec une grande précision et efficacité. Cependant, la polyvalence et l'efficacité d'une telle méthode par rapport aux potentiels d'interaction classiques sont rarement étudiées. En utilisant le BaSnF4 et un système auxiliaire plus simple, NaF, qui possède une structure de type sel gemme, nous montrons comment un MLP peut capturer les interactions subtiles des paires solitaires de Sn, tandis qu'un champ de force polarisable échoue. La précision de notre MLP est validée en comparant les propriétés vibrationnelles telles que la dispersion des phonons et les équations d'état avec les résultats obtenus par la théorie de la fonctionnelle de la densité, et elle montre un excellent accord avec cette théorie. Le MLP présente également une efficacité computationnelle nettement améliorée par rapport à la méthode de dynamiques moléculaires ab initio de référence.Dans le chapitre 4, nous fournirons une investigation complète du comportement de diffusion des ions F⁻ pour le BaSnF4, qui présente une haute conductivité ionique de 1×10-4 S/cm à température ambiante, mais dont le mécanisme de conduction ionique sous-jacent et les implications des paires solitaires (LP) de Sn2+ dans ce processus restent obscurs. En utilisant les dynamiques moléculaires basées sur l'apprentissage automatique, nous avons étudié comment le mouvement des différentes espèces affecte le paysage énergétique de diffusion. Nous avons démontré que les ions Sn2+ jouent un rôle fondamental. Leur mouvement translationnel conduit à un canal de conduction ionique avec un rayon plus large, augmentant ainsi la conductivité ionique du BaSnF4. Cela modifie également la dynamique d'orientation des LP. En outre, nous avons examiné de manière exhaustive le comportement de diffusion des ions fluorure dans le BaSnF4 et confirmé une diffusion bidimensionnelle au sein des couches Ba-Sn, tandis que dans l'autre direction, les ions F- traversent rapidement la couche Sn-Sn, créant des défauts de Frenkel dans la structure. Notre travail élucide le mécanisme atomistique régissant la conduction des ions fluorure dans les conducteurs ioniques rapides et contribuera à la conception rationnelle des électrolytes solides pour les FIB.Dans le chapitre 5, nous présenterons nos premières tentatives sur les électrolytes liquides mixtes. Les propriétés prometteuses des solutions isolées de Np1F/ACN et CsF/H2O ainsi que les avantages tirés du système à haute entropie rendent le mélange intéressant. Dans ce travail, nous avons d'abord entraîné le MLP pour le mélange et validé notre MLP par rapport aux résultats calculés par la DFT. Nous avons ensuite réalisé des dynamiques moléculaires à grande échelle et analysé le comportement des ions F-.
- Open Access
- 5 September 2025
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 19 (36) , 32183-32191
Intercalation of Transition Metals into MXenes: Impact on Electronic and Pseudocapacitive Properties
MXenes are two-dimensional transition metal carbides and nitrides characterized by versatile electronic and electrochemical properties. Herein, we investigate the electronic interactions between various redox-active transition metals (Ni, Co, Mn, and Zn) intercalated into the conductive Ti3C2Tx MXene host. Employing X-ray absorption spectroscopy (XAS) and Bader charge analysis, we reveal that the oxidation states of the intercalated ions remain unchanged upon insertion, whereas Ti atoms within the MXene layers become progressively oxidized with increasing intercalant concentration. Consequently, the electrical resistivity of the intercalated MXenes increases. Ab initio molecular dynamics (AIMD) and density functional theory (DFT) demonstrate distinct spatial arrangements and coordination environments of the intercalated cations, significantly influencing their electronic density of states and interactions with MXene surfaces. Pseudocapacitance measurements in 0.1 M NaOH show distinct behaviors: Co exhibits significant redox activity with less participation from Ti of MXene, while Ni ions show negligible oxidation state changes with predominant Ti redox involvement. Our findings reveal the complex electronic and redox behavior of transition metal-intercalated MXenes, guiding the targeted modification of 2D material properties through careful selection of intercalant species.
- 5 September 2025
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 19 (36) , 32183-32191
Open AccessIntercalation of Transition Metals into MXenes: Impact on Electronic and Pseudocapacitive Properties
MXenes are two-dimensional transition metal carbides and nitrides characterized by versatile electronic and electrochemical properties. Herein, we investigate the electronic interactions between various redox-active transition metals (Ni, Co, Mn, and Zn) intercalated into the conductive Ti3C2Tx MXene host. Employing X-ray absorption spectroscopy (XAS) and Bader charge analysis, we reveal that the oxidation states of the intercalated ions remain unchanged upon insertion, whereas Ti atoms within the MXene layers become progressively oxidized with increasing intercalant concentration. Consequently, the electrical resistivity of the intercalated MXenes increases. Ab initio molecular dynamics (AIMD) and density functional theory (DFT) demonstrate distinct spatial arrangements and coordination environments of the intercalated cations, significantly influencing their electronic density of states and interactions with MXene surfaces. Pseudocapacitance measurements in 0.1 M NaOH show distinct behaviors: Co exhibits significant redox activity with less participation from Ti of MXene, while Ni ions show negligible oxidation state changes with predominant Ti redox involvement. Our findings reveal the complex electronic and redox behavior of transition metal-intercalated MXenes, guiding the targeted modification of 2D material properties through careful selection of intercalant species.
- Open Access
- 10 January 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Elucidating intercalation of transition metal guests to tailorproperties of layered MXenes
Transition metal (TM) cations exhibit distinct electronic properties, compared to alkali and alkaline earth metal cations. In this work, we aim to explore the electronic coupling between different redox-active TMs guest and conductive MXene host. Specifically, within the confined environment of MXene layers, different inserted TM cations demonstrate varying electronic structures and exert distinct influences on the MXene. Unlike Cu which undergoes unusual reduction upon insertion within MXenes, our Bader charge analysis and X-ray absorption spectroscopy (XAS) results indicate that the oxidation states of inserted Ni, Co, Mn, and Zn ions remain unchanged. Despite the higher degree of Ti oxidation led to higher resistivity of intercalated MXenes, the density of states analysis of Ni and Co highlighted the important contributions of these TM cation guests on the overall electrical conductivity of intercalated-MXenes. Moreover, the TM-intercalated MXenes displayed varying redox contributions to the charge storage, where the redox properties of transition metal can be suppressed. This study highlights the unique influence of each transition metal cation under MXene confinement and lays the groundwork for utilizing transition metal guests in MXenes for various electrochemical applications.
- 10 January 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessElucidating intercalation of transition metal guests to tailorproperties of layered MXenes
Transition metal (TM) cations exhibit distinct electronic properties, compared to alkali and alkaline earth metal cations. In this work, we aim to explore the electronic coupling between different redox-active TMs guest and conductive MXene host. Specifically, within the confined environment of MXene layers, different inserted TM cations demonstrate varying electronic structures and exert distinct influences on the MXene. Unlike Cu which undergoes unusual reduction upon insertion within MXenes, our Bader charge analysis and X-ray absorption spectroscopy (XAS) results indicate that the oxidation states of inserted Ni, Co, Mn, and Zn ions remain unchanged. Despite the higher degree of Ti oxidation led to higher resistivity of intercalated MXenes, the density of states analysis of Ni and Co highlighted the important contributions of these TM cation guests on the overall electrical conductivity of intercalated-MXenes. Moreover, the TM-intercalated MXenes displayed varying redox contributions to the charge storage, where the redox properties of transition metal can be suppressed. This study highlights the unique influence of each transition metal cation under MXene confinement and lays the groundwork for utilizing transition metal guests in MXenes for various electrochemical applications.
- 16 December 2024
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 37 (1) , 378-386
Stereoactive Electron Lone Pairs Facilitate Fluoride Ion Diffusion in Tetragonal BaSnF4
- 16 December 2024
- Published by American Chemical Society (ACS) in Chemistry of Materials
- Vol. 37 (1) , 378-386
Stereoactive Electron Lone Pairs Facilitate Fluoride Ion Diffusion in Tetragonal BaSnF4
- Open Access
- 27 September 2024
- Published by American Chemical Society (ACS) in ChemRxiv
Stereoactive Electron Lone Pairs Facilitate Fluoride Ion Diffusion in Tetragonal-BaSnF4
Solid state ionic conductors are of primary importance for the design of tomorrow's batteries. In lithium ion or sodium ion-based materials, the alkali cations diffuse through three-dimensional channels consisting of interconnected tetrahedral or octahedral sites with low free energy barriers between them. Fluoride ion conductors stand out in this landscape since the materials with the highest conductivities belong to the MSnF4 family (in which M2+ is a divalent cation), which structure is layered and characterized by double-layers of Sn2+ and M2+ cations along a given direction. Importantly, these materials display stereoactive electron lone pairs (LPs) that seemingly play an important role not only in stabilizing of the Sn-Sn layer but also in modulating the fluoride ion diffusive behavior. However, despite previous experimental and simulation studies, the involvement of the LPs in the fluoride ions conduction mechanism remains to be quantitatively understood. In this work, we simulate the BaSnF4 tetragonal structure using machine learning-based molecular dynamics, in which the interaction potential is trained on density functional theory data. We investigate the role of the Sn-LP-Sn layer in lowering the diffusion energy landscape. In particular, we show how the F- ions jump across this layer occur much more frequently than in the Ba-F-Ba one, resulting in the formation of vacancies in the Ba-Sn layers. Concurrently, the LPs stereochemical activity fluctuates to accommodate the F ions jumping. In addition, the presence of the LP layer enhances the flexibility of the Sn ions, which leads to an increase of the two-dimensional diffusion by several orders of magnitude. These results contribute to the understanding of the interplay between LPs and ionic diffusion, helping to explain the good performance of the material in fluoride-ion batteries.
- 27 September 2024
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessStereoactive Electron Lone Pairs Facilitate Fluoride Ion Diffusion in Tetragonal-BaSnF4
Solid state ionic conductors are of primary importance for the design of tomorrow's batteries. In lithium ion or sodium ion-based materials, the alkali cations diffuse through three-dimensional channels consisting of interconnected tetrahedral or octahedral sites with low free energy barriers between them. Fluoride ion conductors stand out in this landscape since the materials with the highest conductivities belong to the MSnF4 family (in which M2+ is a divalent cation), which structure is layered and characterized by double-layers of Sn2+ and M2+ cations along a given direction. Importantly, these materials display stereoactive electron lone pairs (LPs) that seemingly play an important role not only in stabilizing of the Sn-Sn layer but also in modulating the fluoride ion diffusive behavior. However, despite previous experimental and simulation studies, the involvement of the LPs in the fluoride ions conduction mechanism remains to be quantitatively understood. In this work, we simulate the BaSnF4 tetragonal structure using machine learning-based molecular dynamics, in which the interaction potential is trained on density functional theory data. We investigate the role of the Sn-LP-Sn layer in lowering the diffusion energy landscape. In particular, we show how the F- ions jump across this layer occur much more frequently than in the Ba-F-Ba one, resulting in the formation of vacancies in the Ba-Sn layers. Concurrently, the LPs stereochemical activity fluctuates to accommodate the F ions jumping. In addition, the presence of the LP layer enhances the flexibility of the Sn ions, which leads to an increase of the two-dimensional diffusion by several orders of magnitude. These results contribute to the understanding of the interplay between LPs and ionic diffusion, helping to explain the good performance of the material in fluoride-ion batteries.
- Open Access
- 25 July 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (31) , 19931-19949
Modeling of Nanomaterials for Supercapacitors: Beyond Carbon Electrodes
Capacitive storage devices allow for fast charge and discharge cycles, making them the perfect complements to batteries for high power applications. Many materials display interesting capacitive properties when they are put in contact with ionic solutions despite their very different structures and (surface) reactivity. Among them, nanocarbons are the most important for practical applications, but many nanomaterials have recently emerged, such as conductive metal-organic frameworks, 2D materials, and a wide variety of metal oxides. These heterogeneous and complex electrode materials are difficult to model with conventional approaches. However, the development of computational methods, the incorporation of machine learning techniques, and the increasing power in high performance computing now allow us to tackle these types of systems. In this Review, we summarize the current efforts in this direction. We show that depending on the nature of the materials and of the charging mechanisms, different methods, or combinations of them, can provide desirable atomic-scale insight on the interactions at play. We mainly focus on two important aspects: (i) the study of ion adsorption in complex nanoporous materials, which require the extension of constant potential molecular dynamics to multicomponent systems, and (ii) the characterization of Faradaic processes in pseudocapacitors, that involves the use of electronic structure-based methods. We also discuss how recently developed simulation methods will allow bridges to be made between double-layer capacitors and pseudocapacitors for future high power electricity storage devices.
- 25 July 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (31) , 19931-19949
Open AccessModeling of Nanomaterials for Supercapacitors: Beyond Carbon Electrodes
Capacitive storage devices allow for fast charge and discharge cycles, making them the perfect complements to batteries for high power applications. Many materials display interesting capacitive properties when they are put in contact with ionic solutions despite their very different structures and (surface) reactivity. Among them, nanocarbons are the most important for practical applications, but many nanomaterials have recently emerged, such as conductive metal-organic frameworks, 2D materials, and a wide variety of metal oxides. These heterogeneous and complex electrode materials are difficult to model with conventional approaches. However, the development of computational methods, the incorporation of machine learning techniques, and the increasing power in high performance computing now allow us to tackle these types of systems. In this Review, we summarize the current efforts in this direction. We show that depending on the nature of the materials and of the charging mechanisms, different methods, or combinations of them, can provide desirable atomic-scale insight on the interactions at play. We mainly focus on two important aspects: (i) the study of ion adsorption in complex nanoporous materials, which require the extension of constant potential molecular dynamics to multicomponent systems, and (ii) the characterization of Faradaic processes in pseudocapacitors, that involves the use of electronic structure-based methods. We also discuss how recently developed simulation methods will allow bridges to be made between double-layer capacitors and pseudocapacitors for future high power electricity storage devices.
- Open Access
- 21 March 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (14) , 10124-10132
Tuning MXene Properties through Cu Intercalation: Coupled Guest/Host Redox and Pseudocapacitance
MXenes are 2D transition metal carbides, nitrides, and/or carbonitrides that can be intercalated with cations through chemical or electrochemical pathways. While the insertion of alkali and alkaline earth cations into Ti3C2Tx MXenes is well studied, understanding of the intercalation of redox-active transition metal ions into MXenes and its impact on their electronic and electrochemical properties is lacking. In this work, we investigate the intercalation of Cu ions into Ti3C2Tx MXene and its effect on its electronic and electrochemical properties. Using X-ray absorption spectroscopy (XAS) and ab initio molecular dynamics (AIMD), we observe an unusual phenomenon whereby Cu2+ ions undergo partial reduction upon intercalation from the solution into the MXene. Furthermore, using in situ XAS, we reveal changes in the oxidation states of intercalated Cu ions and Ti atoms during charging. We show that the pseudocapacitive response of Cu-MXene originates from the redox of both the Cu intercalant and Ti3C2Tx host. Despite highly reducing potentials, Cu ions inside the MXene show an excellent stability against full reduction upon charging. Our findings demonstrate how electronic coupling between Cu ions and Ti3C2Tx modifies electrochemical and electronic properties of the latter, providing the framework for the rational design and utilization of transition metal intercalants for tuning the properties of MXenes for various electrochemical systems.
- 21 March 2024
- Published by American Chemical Society (ACS) in ACS Nano
- Vol. 18 (14) , 10124-10132
Open AccessTuning MXene Properties through Cu Intercalation: Coupled Guest/Host Redox and Pseudocapacitance
MXenes are 2D transition metal carbides, nitrides, and/or carbonitrides that can be intercalated with cations through chemical or electrochemical pathways. While the insertion of alkali and alkaline earth cations into Ti3C2Tx MXenes is well studied, understanding of the intercalation of redox-active transition metal ions into MXenes and its impact on their electronic and electrochemical properties is lacking. In this work, we investigate the intercalation of Cu ions into Ti3C2Tx MXene and its effect on its electronic and electrochemical properties. Using X-ray absorption spectroscopy (XAS) and ab initio molecular dynamics (AIMD), we observe an unusual phenomenon whereby Cu2+ ions undergo partial reduction upon intercalation from the solution into the MXene. Furthermore, using in situ XAS, we reveal changes in the oxidation states of intercalated Cu ions and Ti atoms during charging. We show that the pseudocapacitive response of Cu-MXene originates from the redox of both the Cu intercalant and Ti3C2Tx host. Despite highly reducing potentials, Cu ions inside the MXene show an excellent stability against full reduction upon charging. Our findings demonstrate how electronic coupling between Cu ions and Ti3C2Tx modifies electrochemical and electronic properties of the latter, providing the framework for the rational design and utilization of transition metal intercalants for tuning the properties of MXenes for various electrochemical systems.