(546e) Development of a Multiscale Simulation Approach for Modeling Nanostructured Si-Based Anodes in Lithium Ion Batteries

Recently, significant interest has been focused on Lithium ion batteries containing silicon based anodes for their ability to lead to increased storage capacity due to the significant gain in Lithium uptake evidenced by the silicon anode in comparison to more traditional graphitic or carbon based anodes.  However, the increased capacity seen in lithium insertion into amorphous silicon (a-Si) has a substantial drawback:  The full insertion of lithium into the anode is accompanied by a roughly 300% volume increase of the anode material upon full lithiation, coupled with a substantial reduction in the mechanical properties of the lithiated silicon.  As such, while the capacity of a-Si to accept lithium is prodigious, the stabilitiy of the silicon anode on repeated cycling remains problematic due to the evolution of high internal stresses from expansion coupled with the decrease in mechanical stability as the silicon bonding network is disrupted by lithium insertion and bonding.

Since the chemical process of lithium insertion is fundamental to the destabilization of the silicon anode mechanical properties, avenues apart from chemical modification of the anode material have been investigated in order to improve the cyclability and performance of the silicon anode.  Some success has been seen by reinforcing the silicon phase with external coatings, or changing the physical size and shape of the silicon composing the anode phase to allow for more even lithium penetration.  While such techniques are evocative of potential avenues for increasing the mechanical stability while preserving the viability of lithium storage in amorphous silicon anodes, they are not prescriptive.  In order to try to enable rational design of electrodes, we have therefore embarked on an effort to develop a robust computational multi-scale model of the Li/a-Si composite system which will allow for a phenomenological understanding of the stress evolution and failure of these systems, and which may therefore serve as a basis for establishing design parameters to promote more reliable future anode designs.

This talk will focus on our efforts to build a dynamically launched, computational framework to support the necessary elements of multi-scale modeling of the Li/a-Si anode evolution upon charge/discharge cycles.  While the context is the specific sets of models and equations which allow for the calculation of this particular system, we have developed a generic framework for such multi-scale approaches which allows us to couple microscopic sensitivity (in the guise of the changing mobility of the Lithium within the partially lithiated a-Si at arbitrary stress state) with macroscopic response (via a chemo-reactive material-point method (MPM) simulation) by enacting dynamic and on-demand calculation of the key elements of the MPM constitutive equations via atomistic molecular dynamics and ab initio simulations.  We will discuss the construction of this framework, the necessary assumptions and relations that must hold true in order to allow for this coupling, and the statistical techniques (based upon uncertainty quantification) used to guide the launch of the on-demand sub-grid calculations, which modify the MPM material response.  By providing a concrete demonstration of our multi-scale approach, we will also demonstrate the generic sets of constraints under which it is possible to enact clean transition between differing simulation scales, in order to help to begin to establish the capability for consistently tying dynamic multiple scale simulations into a cohesive overall material model which may span multiple time/length scales and dynamic processes seamlessly.