Rationalizing the interphase stability of Li|doped-Li7La3Zr2O12via automated reaction screening and machine learning. Issue 34 (16th August 2019)

• Record Type: Journal Article
• Authors: Liu, Bo; Yang, Jiong; Yang, Hongliang; Ye, Caichao; Mao, Yuanqing; Wang, Jiping; Shi, Siqi; Yang, Jihui; Zhang, Wenqing
• Is Part Of: Journal of materials chemistry. Volume 7:Issue 34(2019)
• Page Start: 19961
• Page End: 19969
• Publication Date: 2019-08-16
• Subjects: Materials -- Research -- Periodicals; Chemistry, Analytic -- Periodicals; Environmental sciences -- Research -- Periodicals; 543.0284
• DOI: 10.1039/c9ta06748e
• Languages: English
• ISSNs: 2050-7488

Abstract

Abstract : Lithium metal batteries are a promising candidate for future high-energy-density energy storage. Abstract : Lithium metal batteries are a promising candidate for future high-energy-density energy storage. However, dendrite growth and the high reactivity of the Li metal anode result in low cycling efficiency and severe safety concerns. Here, we present a strategy to stabilize the lithium metal anode through cation doping in Li7 La3 Zr2 O12 (LLZOM, M = dopant). High-throughput automated reaction screening together with a machine learning approach are developed to evaluate possible reactions and the thermodynamic stability of the Li|LLZOM interfaces under various chemical conditions. It is discovered that some dopants, such as M = Sc 3+ (doping on Zr site), Ce 3+ (La or Zr), Ac 3+ (La), Y 3+ (La or Zr), Tm 3+ (La or Zr), Er 3+ (La or Zr), Ho 3+ (La or Zr), Dy 3+ (La or Zr), Nd 3+ (La or Zr), Tb 3+ (La or Zr), Pr 3+ (La), Pm 3+ (La or Zr), Sm 3+ (La or Zr), Gd 3+ (La or Zr), Lu 3+ (La), Ce 4+ (Zr), Th 4+ (Zr), and Pa 5+ (Zr), exhibit thermodynamic stability against Li; while others, M = Ca 2+ (La or Zr), Yb 3+ (La), Br 3+ (Li), Te 4+ (Zr), Se 4+ (Zr), S 4+ (Zr), Hf 4+ (Zr), Cl 5+ (Zr), and I 5+ (Zr), may lead to the spontaneous formation of a stable, passivating solid electrolyte interphase (SEI) layer on the Li metal, and alleviate dendritic lithium growth. From the machine learning approach, the formation energy of oxides M x O y emerges as the most crucial feature dominating the route of interface reactions, implying that the M–O bond strength governs the interface stability of the cation-doped LLZOM. The machine learning model then predicts 18 unexplored LLZOM systems, which are all validated in subsequent calculations. Our work offers practical insights for experimentalists into the selection of appropriate dopants in LLZO to stabilize Li metal anodes in solid-state batteries.