Grain Boundary Dendrite Growth in Solid-State Electrolytes

grain boundaries and dendrites in solid state

Lithium dendrites growing through a solid electrolyte do not behave like the diffuse, roughly uniform plating seen at a liquid electrolyte interface.

In a polycrystalline ceramic or a cold-pressed sulfide pellet, the microstructure itself dictates where lithium goes: dendrites overwhelmingly nucleate and propagate along grain boundaries rather than through the bulk of individual grains, even in materials whose bulk mechanical properties should, on paper, resist them. That gap between theoretical and real-world dendrite resistance is the central puzzle in solid-state electrolyte design, and it plays out differently in oxide/garnet and sulfide chemistries.

The Monroe-Newman Paradox

The starting point for most solid electrolyte dendrite discussion is the Monroe-Newman criterion: a widely cited 2005 analysis concluding that an electrolyte with a shear modulus roughly twice that of lithium metal (lithium’s shear modulus is only a few GPa) should be mechanically stiff enough to suppress the growth of a surface protrusion before it can propagate into a short circuit [1]. Garnet-type oxide electrolytes such as Li₇La₃Zr₂O₁₂ (LLZO) comfortably clear that bar, with a measured shear modulus around 56-61 GPa and a Young’s modulus around 130-150 GPa depending on dopant and porosity [2] — roughly an order of magnitude above the Monroe-Newman threshold. By that logic, garnets should be effectively dendrite-proof.

In practice they are not. Garnet cells still short well below their theoretical current-carrying capacity, typically failing somewhere in the region of 1 mA/cm² rather than at the much higher values the bulk modulus alone would predict [3]. The reason is that the Monroe-Newman criterion assumes a homogeneous, defect-free electrolyte, and no sintered polycrystalline ceramic is one. Real garnet pellets are full of grain boundaries, and it is along those boundaries — not through the grains — that dendrites actually travel.

Why Grain Boundaries Are the Preferred Path

Two separate mechanisms converge at a grain boundary, and both favour it as a growth path over the bulk grain interior:

Elevated electronic conductivity. Grain boundaries in garnet electrolytes carry higher electronic (as opposed to ionic) conductivity than the bulk grains. Since dendrite nucleation inside a solid electrolyte requires local electron transfer to reduce Li⁺ to Li⁰, a boundary that leaks electrons more readily is a preferential nucleation site before any crack has even formed [4].
Mechanically driven fracture. Once a small amount of lithium metal has nucleated in a subsurface void or flaw, continued plating builds up internal pressure inside that pocket of lithium. Recent operando work shows this pressure buildup is what mechanically fractures the surrounding ceramic, and the crack preferentially follows grain boundaries because fracture toughness is lowest there — the dendrite is less “growing through” the ceramic than levering it apart from the inside [5]. Engineered void arrays that give the pressure somewhere else to go have been shown to redirect crack propagation away from the direct electrode-to-electrode path, which is a direct demonstration that the failure is mechanical rather than purely electrochemical.

Oxide and Garnet Electrolytes

For garnets, the two mechanisms above compound each other over cycling. Grain-boundary conduction seeds lithium metal inside the ceramic; the resulting internal pressure fractures the (locally weaker) grain-boundary network; and the fresh crack surface provides yet more high-conductivity boundary area for the next cycle. Mitigation strategies now target this loop directly rather than trying to raise bulk modulus further, since bulk modulus was never the limiting factor. Selective grain boundary amorphization — converting the crystalline boundary phase to an amorphous one — has been shown to suppress dendrite growth by removing the fast electronic pathway the crystalline boundary provided [6]. Doping with aluminium or gadolinium reduces grain boundary resistance by generating Li⁺ vacancies and improving sintered density, which shrinks the population of residual pores that would otherwise act as nucleation sites [7].

Sulfide Electrolytes

Sulfide electrolytes such as Li₆PS₅Cl (LPSCl) sit at the opposite end of the stiffness scale from garnets — an elastic modulus around 30 GPa, roughly a fifth of LLZO’s — and are soft enough to be densified by cold pressing alone rather than requiring the high-temperature sintering garnets need. That plasticity was originally expected to help: a softer, more deformable electrolyte should flow to accommodate an incipient protrusion rather than cracking around it. In practice it only partly delivers. Even under cold-pressing loads of 370-530 MPa, sulfide pellets retain residual porosity of roughly 12% or more, and plastic flow closes some but not all of the void network between particles [8]. Those residual voids behave much like the pores in a garnet ceramic: they are where lithium metal preferentially deposits during charging, and imaging of lithium stripping from LPSCl has shown small voids nucleating and coalescing into larger cracks that drive interfacial delamination [9]. Simply pressing sulfide pellets harder to reduce grain boundary area has not been sufficient to eliminate dendrite formation on its own, which points to the same underlying lesson as the garnet case: densification helps, but the residual defect population, not the bulk modulus, sets the practical limit.

What This Means for Cell Design

The practical takeaway for pack and cell designers is that a solid electrolyte’s headline modulus number is a poor proxy for its real dendrite resistance. Garnets and sulfides fail by related but distinct routes — grain-boundary conduction and mechanical fracture in garnets, residual void nucleation and interfacial delamination in sulfides — and in both cases the failure mode is governed by microstructural defects introduced during processing rather than by the electrolyte’s intrinsic bulk properties. That is why solid-state cell development puts so much emphasis on processing conditions (sintering profile and dopant chemistry for garnets, pressing pressure and particle size distribution for sulfides) alongside stack pressure and current density limits during operation: the defect population set during manufacture is what current density limits and stack pressure are ultimately trying to manage around.

References

The Impact of Elastic Deformation on Deposition Kinetics at Lithium/Polymer Interfaces (Monroe & Newman), UC Berkeley
Elastic Properties of the Solid Electrolyte Li7La3Zr2O12 (LLZO), OSTI
Grain Boundaries in Ceramic Solid-State Lithium Metal Batteries: A Review, arXiv
Understanding the evolution of lithium dendrites at Li6.25Al0.25La3Zr2O12 grain boundaries via operando microscopy techniques, PMC
Mechanically driven Li dendrite penetration in garnet solid electrolyte, Nature
Grain boundary amorphization as a strategy to mitigate lithium dendrite growth in solid-state batteries, Nature Communications
Grain boundary modification in garnet electrolyte to suppress lithium dendrite growth, Chemical Engineering Journal
Densifying Solid Electrolytes through Surface Lubrication to Mitigate Shorting of Solid-State Batteries, UC San Diego
Recent Advances in Dendrite Suppression for Sulfide-Based All-Solid-State Batteries, Preprints.org

Disclaimer: This page summarises an active area of research; specific numerical values (modulus, porosity, critical current density) vary significantly between material formulations, dopants, and processing routes reported in the literature, and should be checked against the primary source for any specific electrolyte composition.