This poster elucidated the governing factors for dendrite growth in lithium metal batteries (LMBs).
Mechanistic insights are obtained by modeling dendrite growth on a coarse–grained (CG) and atomistic scale.
Governing trends derived from the model have been compared to experimental observation.

Driven by the increasing demand for portable high–capacity energy storage, there is a renewed interest in Lithium metal batteries.
In LMBs one electrode is replaced by lithium metal anchored on a current collector.
Lithium metal is electrochemically plated on the anode side when charging the battery.
Depositing the metal uniformly is a hugely challenging undertaking.
The intrinsically high reactivity of lithium metal, electric field effects, and spatial variations of the local composition in the electrolyte all drive dendritic deposition.
This dendrititc growth can creates a high surface area (high surface area Lithium, HSAL) and short circuits.
A high surface area fosters the occurrence of parasitic side reactions which consume Lithium and electrolyte components.
Furthermore, the dendritic Lithium can become detached from the surface and creates dead Lithium.
This too hinders the ion transport and contributes to the loss of active material.

To understand the influence of various electrochemical conditions on the dendrite morphology on a microscopic level a versatile coarse–grained model is employed.
Besides the investigations of fundamental driving forces like the electric field, cation concentration, and cation mobility, the evaluation of more intricate procedures e.g.~pulse charging is also possible.
Local passivation and solid electrolyte interface (SEI) effects can be mimicked by the implementation of different particle types with varying reactivity.
This CG model simulates the movement of cations only.
Solvent and anions are treated implicitly.
It is assumed that the implicit anions effectively screen the cations which causes no interactions between the cations.
The overall cation movement can be modeled as sum of two contributions:
First, the cations perform a random walk through the electrolyte due to diffusion.
Second, the applied electrode voltage for charging may create an electric field in the electrolyte close to the interface.
The electric field is accounted for by a migratory term in the cation movement.
Also, when the cations come into contact with the electrode, a reaction may occur.
The elementary reaction rate is simply expressed as a reaction probability with which a reaction may occur.
If a reaction is accepted, the cation is reduced and forms a metal particle at the surface.
Else, the cation steps back into the electrolyte and continues the its movement according to diffusion and migration.
These assumptions hold true in the diffusion–limited case as otherwise electrostatic interactions and layering close to the electrode become dominant.
The importance of the presence of the electric field is discussed in the second part concerning the atomistic simulation of a nano–scale battery.

When starting from an equilibrium distribution in the CG model, the cations are homogeneously distributed through the electrolyte.
After applying the electric field and allowing for reactions, cations close to the interface deposit as metallic Lithium.
This step is reaction limited as there are enough cations in close proximity to the electrode.
The consumption of cations causes a depletion layer to form and a transition to diffusion–limited growth is observed.
Now the growth happens predominantly at the dendrite tips where the electric field causes a steady supply of cations due to spherical diffusion along the field lines.

Pulse charging can be simulated by repeatedly applying the electric field and allowing for reactions during an on–time which is followed by an off-time without field and reactions.
The ratio between the two pulses can be varied as well as the length of the on–time.
Here, a fixed duty cycle is chosen and different pulse lengths are compared.
In these simulations, the growth velocity of the dendrites is observed for different reaction probabilities.
As expected, a higher reaction probability leads to faster dendrite growth.
If the on–time is short, dendrites grow fast.
This is because during the following resting period, which is short as well, the cations have insufficient time to diffuse between the dendrites.
Reactions happen mostly at the dendrite tips which drives the growth of dendrites.
For medium–length pulses, cations can diffuse between the dendrites during the resting period.
In the following on–pulse, the cations are consequently reduced within the porous structure and do not contribute to the increase in dendrite height.
The dendrites grow slower.
If the charging pulse becomes too long, all cations between the dendrites are consumed and tip–induced growth continues.
The dendrite–growth velocity increases again.
Therefore, the CG simulation facilitates the optimization of pulse sequences for a given diffusivity and electric field.

Solid electrolyte interface (SEI) effects can be modeled by an implicit SEI in the CG model.
In this approach, the cationic diffusion coefficient is reduced within a specific distance to the electrode interface.
Systematically increasing the diffusion coefficient in the SEI towards the value in the electrolyte shows expectedly convergent growth speed and structure.
Interestingly, increasing the thickness of the SEI layer shows a more complex dependence.
Even a thin SEI layer reduced the growth speed of the dendrites quickly due to higher interfacial resistance.
The slow–down effect becomes less for thicker SEI layers.
The local dendrite structure captured by the metal–metal coordination number (CN) shows an instantaneous switch from the SEI–free system with denser dendrites (higher CN) to the SEI system with lower local density.
Apparently, even a thin SEI layer has a significant influence on the dendrite growth.
Alternatively, the SEI can be modeled by explicit particles on the surface.
The mechanical properties of the SEI layer can be tuned by changing the stiffness of the interconnecting bonds of the layer beads.
If the layer is too soft, dendrites can rupture the layer.
If the layer is too stiff, metallic deposits can spike the layer.
The simulation allows finding the optimal value for the mechanical layer properties depending on cation mobility and field conditions.

Lastly, in order to estimate the strength of the electric field, reactive atomistic simulations have been conducted.
Between two oppositely charged graphite electrodes a glyme–based electrolyte is placed.
In a reactive step, cations are removed from one electrode an placed on the other electrode so that a target rate is reached.
These simulations show, that without reactions at the interface the surface charge is effectively screened by an ion layer.
However, under reactive conditions this layer is significantly reduced causing a well pronounced an far–reaching electric field.