Revealing Short- and Long-range Li-ion diffusion in Li$_2$MnO$_3$ from finite-temperature dynamical mean field theory

By combining finite-temperature dynamical mean field theory with DFT+$U$ and nudged-elastic-band calculations, this study reveals that dynamical correlations significantly lower Li-ion migration barriers in paramagnetic Li$_2$MnO$_3$, providing a unified explanation for both short-range and long-range transport activation energies without requiring extrinsic disorder or clustered vacancies.

The Gist

Imagine a lithium-ion battery as a busy city where tiny lithium ions are the commuters trying to get from one side of town to the other to power your phone or car. The "roads" they travel on are inside a material called Li₂MnO₃.

For a long time, scientists were confused about how fast these commuters could move. Some experiments (looking at very short distances) said the roads were super smooth and fast. Other experiments (looking at long distances) said the roads were full of traffic jams and very slow. It was like saying, "You can run a sprint in 10 seconds!" but also, "You can't run a marathon because the track is broken."

This paper solves that mystery by using a super-advanced computer simulation to look at the "traffic" in a new way.

The Old Map vs. The New Map

Previously, scientists used a standard computer model (called DFT+U) to map out the roads. This model was like a basic GPS: it saw the lithium ions trying to jump over walls, but it calculated the walls to be very high (about 0.6 to 0.9 eV). This suggested the ions would move very slowly, which didn't match the fast "sprint" data from the short-distance experiments.

The authors realized that the old model was missing a crucial ingredient: heat and chaos. In the real world, the atoms in the battery aren't frozen in place; they are jiggling and vibrating because of the heat (temperature). The manganese atoms in the material also have tiny magnetic spins that are flipping around randomly. The old model treated these spins as if they were frozen in a perfect line, which isn't true for a working battery.

The "Dynamic" Simulation

To fix this, the authors used a more powerful tool called DFT+DMFT. Think of this as upgrading from a static 2D map to a 3D, real-time simulation that accounts for the heat and the random flipping of magnetic spins.

They simulated a single "empty seat" (a vacancy) in the lithium city. The lithium ions need to jump into this empty seat to move forward.

The Two Speeds of Travel

When they ran their new, "hot and chaotic" simulation, they found something amazing. The energy barriers (the walls the ions have to climb) dropped significantly, but only for specific types of jumps.

1. The Short Hop (The Sprint):
   For the very shortest jump between two neighboring spots, the new simulation showed the wall was only 0.18 eV high.

   • The Result: This perfectly matches the "fast sprint" data from the short-distance experiments.
   • The Analogy: Imagine a commuter stepping over a small curb. It's easy and fast. The old model thought the curb was a 10-foot fence; the new model realized it was just a small step.

2. The Long Haul (The Marathon):
   However, to travel a long distance across the whole city, the commuter can't just take the easy steps forever. They eventually have to take a slightly harder step. The simulation found a second, slightly higher wall at 0.50 eV.

   • The Result: This matches the "slow marathon" data from the long-distance experiments.
   • The Analogy: To get across town, you have to take many easy steps, but occasionally you hit a hill. Even if most steps are easy, your overall speed is limited by that one hill.

Why This Matters

The big discovery is that you don't need to invent complicated explanations to fix the speed problem. You don't need to assume the battery is full of "clumps" of empty seats or that the material is broken.

The paper shows that Li₂MnO₃ is actually a very good material (nearly perfect, or "stoichiometric"). The reason we see different speeds in different experiments is simply because:

• Short-range experiments only see the easy, low hills (0.18 eV).
• Long-range experiments see the whole journey, which is slowed down by the occasional higher hill (0.50 eV).

The Takeaway

By accounting for the heat and the magnetic "jitter" of the atoms, the authors created a single, unified story. They proved that the lithium ions can zip around easily on a local scale, but their overall journey is controlled by a few slightly tougher steps. This explains why the battery behaves differently depending on how you measure it, without needing to blame defects or impurities in the material.

In short: The battery isn't broken; we just needed a better map that accounted for the heat and the magnetic dance of the atoms to understand how the lithium ions really move.

Technical Summary

Technical Summary: Revealing Short- and Long-range Li-ion diffusion in Li2MnO3 from finite-temperature dynamical mean field theory

Problem Statement
Li2MnO3 is a critical component of Li-excess layered cathodes, yet its role in limiting Li-ion transport remains debated. Experimental estimates of Li-ion migration in Li2MnO3 exhibit a significant discrepancy depending on the probe length scale. Microscopic $\mu^+$SR measurements on nearly stoichiometric powders report a low short-range activation energy ($E_a \approx 0.156$ eV) in the paramagnetic regime. In contrast, macroscopic ac-impedance measurements yield a substantially higher apparent barrier ($E_a \approx 0.46$ eV) for long-range transport.

Previous first-principles studies based on Density Functional Theory with Hubbard U corrections (DFT+U) have struggled to reconcile these values. Standard DFT+U calculations for a single Li vacancy typically predict barriers of 0.5–0.8 eV, which are too high to explain the short-range experimental data. To match the low experimental barrier, previous models often invoked clustered vacancy configurations (divacancies or trivacancies) or assumed specific magnetic orders. However, such vacancy clusters are unlikely in nearly stoichiometric samples, and Li2MnO3 is paramagnetic under battery operating conditions, whereas many DFT+U studies assumed ferromagnetic or nonmagnetic states that suppress local Mn moments.

Methodology
This work employs a multi-scale computational approach to study Li+ migration in paramagnetic Li2MnO3 with a single Li vacancy at room temperature (300 K). The methodology integrates three components:

1. DFT+U: Used to optimize migration pathways and saddle-point geometries via the Nudged-Elastic-Band (NEB) method. Spin-polarized DFT+U calculations were performed using the PBEsol functional and a 1×2×2 supercell.
2. DFT+DMFT: Finite-temperature Dynamical Mean Field Theory (DMFT) with a continuous-time quantum Monte Carlo (CTQMC) impurity solver was applied to evaluate total energies along the fixed NEB geometries. This framework captures dynamical electronic correlations in the paramagnetic phase.
3. Electronic Structure Handling: To address the low crystallographic symmetry (C2/m) of Li2MnO3, which causes large off-diagonal matrix elements in standard Wannier projections, the authors diagonalized the Mn-d block. This created an effective low-energy "d-only" model where the Wannier basis implicitly encodes Mn-d–O-p hybridization, allowing for accurate reproduction of the insulating gap.

The study evaluates six symmetry-inequivalent migration pathways (four intralayer and two interlayer) by calculating DMFT total energies for the initial, saddle-point, and final images, allowing the correlated electronic state to relax at each point.

Key Results
The application of finite-temperature dynamical correlations significantly renormalizes the activation energies for specific migration pathways, resolving the discrepancy between microscopic and macroscopic experimental data without invoking clustered vacancies:

• Short-Range Diffusion: For the lowest-barrier intralayer hop (Path ii: $4h \to T_{2b}^{Li} \to 4h$), the DFT+DMFT activation energy is reduced to 0.18 eV. This value quantitatively reproduces the short-range activation energy ($0.156$ eV) observed in $\mu^+$SR measurements. The reduction is driven primarily by the eigenvalue-correction term in the DMFT total energy, which stabilizes the saddle point configuration.
• Long-Range Diffusion: The next-lowest barrier pathway (Path iv: $4h \to T_{2b}^{Li} \to 2c$) has an activation energy of 0.50 eV under DMFT. This barrier controls the long-range percolation of Li ions. It is consistent with the macroscopic activation energy ($\approx 0.46$ eV) extracted from ac-impedance measurements.
• Path Dependence: The barrier reduction is highly path-dependent. While Path (ii) sees a dramatic drop from the DFT+U value (~0.68 eV) to 0.18 eV, other paths (e.g., Path iii) show minimal change. The authors correlate this with the local registry of the migrating Li ion relative to the MnO6 network; Path (ii) passes through a more open region with longer Li-Mn distances, which is more effectively stabilized by dynamical correlations.
• Electronic Structure: Unlike DFT+U, which shows a weakly site-dependent change in Mn-d occupation upon vacancy creation, DMFT reveals a strong site-dependent redistribution of spectral weight, with significant depletion of occupied Mn-d weight on Mn sites further from the vacancy.

Significance and Claims
The paper claims to provide a unified microscopic interpretation of both local and macroscopic Li-ion transport in Li2MnO3. By utilizing a single-vacancy, paramagnetic DMFT description, the authors demonstrate that the hierarchy of activation energies (low for short-range, higher for long-range) arises naturally from the interplay of specific migration pathways and finite-temperature dynamical correlations.

The study concludes that:

1. The experimentally observed short-range activation energy can be rationalized by single-vacancy migration in a paramagnetic host, eliminating the need to invoke extrinsic disorder or clustered vacancy configurations to explain the data.
2. Long-range transport is controlled by a higher-barrier step (0.50 eV) within a network of hops, consistent with macroscopic impedance data.
3. Finite-temperature dynamical correlations are an essential ingredient for predicting ionic migration energetics in correlated oxide electrodes, as static DFT+U approaches fail to capture the necessary renormalization of barrier heights.

The authors note that these findings highlight the necessity of beyond-DFT treatments for ionic diffusion in correlated oxides and suggest that future work should extend this framework to higher vacancy concentrations and mixed-cation Li-rich layered cathodes, particularly where oxygen redox becomes active.