Qifan
Yang
ab,
Jing
Xu
ac,
Xiao
Fu
ab,
Jingchen
Lian
ab,
Liqi
Wang
ac,
Xuhe
Gong
ad,
Ruijuan
Xiao
*abc and
Hong
Li
*abc
aBeijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China
bCenter of Materials Science and Optoelectronics Engineering, University of Chinese Academy of Sciences, Beijing 100049, China
cSchool of Physical Sciences, University of Chinese Academy of Sciences, Beijing 100049, China
dSchool of Materials Science and Engineering, Key Laboratory of Aerospace Materials and Performance (Ministry of Education), Beihang University, Beijing 100191, China. E-mail: rjxiao@iphy.ac.cn; hli@iphy.ac.cn
First published on 27th November 2024
The recent study of viscoelastic amorphous oxychloride electrolytes has opened up a new field of research for solid-state electrolytes. In this work, we chose a Li–Al–O–Cl system containing disordered structures with varying O/Cl ratios and Li+ contents to study their structural characteristics and ion transport mechanism using ab initio molecular dynamics (AIMD) simulation and machine learning interatomic potential based molecular dynamics (MLIP-based MD) simulation. It is found that O-doping results in the presence of a skeleton of Al-chains formed by AlOCl tetrahedra and an increase in glass forming ability, causing Cl atoms' rotation around centered-Al within the tetrahedron thus facilitating the motion of Li+ ions. However, a further increase in the O/Cl ratio decreases the number of rotating Cl atoms, weakening the transport of Li+. So increasing glass forming ability without reducing Cl content or by methods through controlling synthesis conditions, is useful to promote Li+ ion conduction of oxychloride electrolytes.
To deepen the understanding of amorphous solid-state electrolytes, we generate and simulate the disordered Li–Al–O–Cl system in various compositions through ab initio molecular dynamics (AIMD), which is a relatively accurate method commonly used for crystalline systems in the simulation of structural evolution and ionic transport.28,29 However, compared to crystals, computational simulation by AIMD has some unique challenges for amorphous materials. Firstly, the atomic arrangement of amorphous materials exhibits long-range disorder without translational periodicity symmetry, which makes it difficult to establish atomic models for simulation. Typically, the construction of the amorphous model in AIMD simulation is performed in a fictitious cubic box through melting and a rapid quenching process to obtain the disordered arrangement of atoms.30 Due to the large computational cost of this method, the systems that are adopted in simulations are generally smaller than 200 atoms, corresponding to a lattice length of 10–20 Å. The size effect of the model and the accessible time scale for evolution become the main limitations for simulation effectiveness. Secondly, room-temperature ionic conductivity in amorphous configurations cannot be extrapolated from high-temperature results as crystals, as amorphous configurations constantly relax, leading to different structures at different temperatures. Direct simulations at room temperature are only possible for systems with high ion transport capabilities, as weak transport cannot generate enough migration events in a limited simulation time to obtain accurate statistical results. Thus although AIMD can accurately illustrate each conducting event and ion transport mechanism,31–33 as well as provide us with initial configurations of disordered structures and simulate ion migration within a certain range, research on larger-spatial and longer-time scales and obtaining macroscopic overall ionic conductivity requires faster computational methods, such as recently developed machine learning interatomic potential based molecular dynamics (MLIP-based MD).
In this work, five different compositions for amorphous LAOCs (Li0.5AlO0.75Cl2, Li1.0AlO0.75Cl2.5, Li1.5AlO0.75Cl3, Li1.0AlO0.25Cl3.5 and Li1.0AlCl4) were simulated through the AIMD method and MLIP method to analyze the atomic structural and ion conducting characteristics on larger and longer scales, identify the physical origin of the excellent Li+ transport performance of amorphous oxychloride, and present possible strategies to explore amorphous SSEs.
Li1.0AlO0.75Cl2.5 is used to illustrate the structural characteristics of amorphous LAOCs. As shown in Fig. 1a, the structure contains long chains consisting of AlOmCl4−m (m = 1, 2, 3, 4) tetrahedra (AOC), and the Al atom on the chains is labeled as Alchain. Isolated AOC tetrahedra also exist, and the Al atom in the center of isolated AOC tetrahedra is marked as Aliso. In Li1.0AlO0.75Cl2.5, two types of Al atom chains are found according to the number of Al atoms in the chain, which are Al5 and Al6 as illustrated in Fig. 1b and c. As we can see, Al atoms are connected by Cl or O atoms in the structure to form Al chains, and when the two adjacent Alchain are connected by Cl atoms, the distance between them reaches a maximum not exceeding 4.2 Å. Therefore, in the analysis of other compositions, the Al atoms with distances less than 4.2 Å are considered to belong to the same chain. The statistical results on the percentage of Al chains with different chain lengths for each LAOC are presented in Fig. 1d. It shows that Al atoms in Li0.5AlO0.75Cl2 are all Alchain, forming Al16 chains. This chain contains 16 Al atoms, and the AOC tetrahedra are mostly connected in an Al–O–Al form, indicating that the formation of these long chains may be related to the high oxygen content in the structure. As the O/Cl ratio decreases, the number of Alchain gradually decreases and the number of Aliso increases, and when the doping concentration of O decreases to 0, the amorphous Li1.0AlCl4 contains the most Aliso. It is noted that Li1.5AlO0.75Cl3 and Li1.0AlO0.25Cl3.5 have the same proportion of Alchain and Aliso, possibly originating from the limited simulation cells used in AIMD and the larger simulation box may exhibit more abundant structural details, thus better distinguishing these two structures. The structural characteristics also reflect in the radial distribution function (RDF) of the Al–Al pair in Fig. S3 and S4.† The first coordination in the Al–Al RDF curve of amorphous LAOCs originates from two adjacent Alchain, since r (distance between Al–Al) is between 2 and 4 Å which is exactly the range for distances between adjacent Alchain. As the O/Cl ratio decreases, the first coordination number gradually decreases, indicating that the number of Aliso increases and the number of Alchain decreases. The above findings indicate that the length and proportion of Al chains in the LAOC structure can be controlled by adjusting the O/Cl ratio. Finally, as Fig. 1e shows, Alchain and Aliso in Li1.0AlO0.75Cl2.5 are distributed in a random way with no segregation phenomenon, which means the structure of LAOC is more similar to that of polymers formed by chains of various lengths.
We conduct a statistical analysis by calculating the proportions of Li+ transport events and their nearest local environment of LAOCs in a 11–50 ps AIMD simulation. The specific statistical methods and processes of the migration of Li+ ions in LAOC structures are demonstrated in Note S1.† The percentage of three different transport events of each LAOC composition at 300 K is displayed in Fig. 2b. The results revealed that as the O/Cl ratio decreases, the intra-chain transport gradually decreases, while the inter-chain and Aliso-assisted transport increase. This is related to the change in atomic structure, in which the length and number of Al chains are gradually decreasing. The chemical environments for all the Li+ migration events are analyzed and the conditions beneficial to Li+ transport are extracted. Fig. 2c indicates the percentage of various tetrahedra in the initial structure (left column) and the percentage of tetrahedral types related to Li+ transport events (right column). This suggests that the majority of the tetrahedra near the migrating Li+ are AlCl4 and AlOCl3 tetrahedra (over 75%), indicating that the migration of Li+ ions is more likely to occur in localized environments with high chlorine content. Furthermore, according to the analysis of the structural features mentioned earlier, O atoms mainly serve as corner- or edge-sharing atoms to connect tetrahedra, and hence a higher chlorine content implies a shorter Al chain length. Therefore, under the same conditions, the higher the sum of the tetrahedral ratios of AlCl4 and AlOCl3 in structures, the shorter the Al chains, and the higher the ionic conductivity expected to be found. While Li+ migrates in the disordered structures, for Cl atoms, unlike remaining stably unmoved in crystal LiAlCl4 (Fig. S5†), some Cl atoms also move in LAOCs, and as shown in Fig. 3c and S6,† the MSD of Cl is quite large. Different from Li, the Cl atom mostly rotates around the central Al of the AOC tetrahedron, without undergoing long distance migration. The changes in the Al–Cl bond lengths throughout the entire simulation process are shown in Fig. S7,† indicating that the Al–Cl bonds are not broken, demonstrating that Cl moves by rotation in its AOC tetrahedron. The only exception is the migration of one Cl atom connected with an AlOCl3 tetrahedron in Li1.0AlO0.25Cl3.5. Fig. 2d shows the tetrahedra with rotating Cl in different compositions proving that Cl movement on AlCl4 and AlOCl3 is the most intense, indicating that the rotation of Cl on AlCl4 as well as AlOCl3 and Li+ transport can promote each other.
The above analysis indicates that the rotation of Cl within AOC tetrahedra of LAOC structures promotes the transport of Li+ ions. Compared with AlO2Cl2, AlO3Cl, and AlO4, the proportion of stable O atoms on the tetrahedra of AlCl4 and AlOCl3 is smaller, and the inhibitory effect on the rotation of Cl connected to the same Al atom is also smaller. At the same time, the proportion of Cl atoms is higher, so once a Cl rotates, it is likely to drive the rotation of the surrounding Cl atoms. Therefore, the rotation of Cl on AlCl4 and AlOCl3 is much more intense than in others. When the ratio of AlCl4 and AlOCl3 is high, it can greatly promote the movement of Li+. Therefore, in amorphous LAOCs, at the same Li and Al concentrations, the lower the O/Cl ratio (the more Cl atoms), the more intense the Li+ transport, and the higher the expected Li+ conductivity.
To prove the aforementioned conclusion, amorphous Li1.0AlCl4, Li1.0AlO0.25Cl3.5 and Li1.0AlO0.75Cl2.5, with the same concentration of Li and Al, were selected. According to the above analysis, the ion conductivity performance among these structures should be: Li1.0AlCl4 > Li1.0AlO0.25Cl3.5 > Li1.0AlO0.75Cl2.5. To confirm the inference, the AIMD simulations were conducted for these systems at 300 K, from which the statistically obtained MSDs are shown in Fig. 3b. Due to the size effect and a lack in statistical samples at large time intervals,35 the Li+ MSDs in Fig. 3b show plateaus, and here the values of MSDs are adopted to qualitatively compare the ion transport capability among different structures. The ionic conductivity will be estimated according to MLIP-based MD later. The MSD of the three structures is ranked as Li1.0AlCl4 > Li1.0AlO0.25Cl3.5 > Li1.0AlO0·75Cl2.5, which further confirms the correctness of the above conclusions.
Based on the above analysis, it is inferred that the doping effect of O in LAOC on Li+ ion transport is a result of two factors. The addition of element O enhances the system's ability to form amorphous structures, promoting the rotation of Cl and the transport of Li. However, the decrease in Cl content reduces the number of AlCl4 and AlOCl3 tetrahedra, which has a negative effect on the transport performance of Li+ ions. This also indicates that besides O doping, if we can find other ways to improve the amorphization degree of LiAlCl4, it may be possible to achieve higher Li+ ion conductivity at higher Cl content.
To build the amorphous model in larger boxes such as 2 × 2 × 2 and 3 × 3 × 3 supercells expanded from the initial box for Li1.0AlO0.75Cl2.5, two different modelling strategies are tested. Although both models have the same initial structure, which is the supercell expanded from the original disordered cell obtained from AIMD simulations, the first method involves a process of re-melting and quenching and then equilibrating by MLIP-based MD, while the second method directly conducts simulations at 300 K on the expanded structure using MLIP-based MD. Note S3† introduces a detailed description of these two methods. Fig. 4 illustrates the simulations' results for these two models in a 3 × 3 × 3 expanded box. Fig. 4a and b show that two methods yield very similar amorphous structures in Li1.0AlO0.75Cl2.5, with a consistent distribution of the number of Al-chains and the proportion of AOC groups. However, model 1 has less Aliso and fewer numbers of the sum of AlCl4 and AlOCl3 groups, resulting in a slightly lower ionic conductivity statistically evaluated from the MSD than that of model 2. An ionic conductivity of 12 mS cm−1 and 17 mS cm−1 was obtained for model 1 and model 2, respectively, indicating that due to the non-uniqueness of the disordered structure, the ionic conductivity of the system varies with the degree of disorder. Since the disorder of the structure in the simulation is much higher than that of the experimentally synthesized structure, the calculated conductivity is also higher than that obtained experimentally (∼1 mS cm−1 at 300 K (ref. 26)).
Therefore, we further investigated the relationship between Li+ ion conductivity and the degree of disorder. Since only LiAlCl4 has a definite crystal structure in the studied system, we used LiAlCl4 structures at different melting stages to manufacture different degrees of disorder in the non-crystalline LiAlCl4 configurations. The AIMD simulations at 300 K were carried out for these models with various amorphous degrees. The results in Fig. S8† indicate that in LiAlCl4 more disordered structures show larger MSD values, which means higher Li+ conductivity and more rotation events of Cl in AOC groups.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta06234e |
This journal is © The Royal Society of Chemistry 2025 |