Mapping uncharted territory in ice from zeolite networks to ice structures

Edgar A. Engel, Andrea Anelli, Michele Ceriotti, Chris J. Pickard, R. J. NeedsNature Communications 2018 · DOI

TL;DR

This paper presents a comprehensive computational exploration of water ice's configuration space, using a novel "generalized convex hull" approach to identify potentially stable ice structures. They've mapped out the landscape of ice structures, rediscovered known phases, and proposed 34 new candidate ice phases (XVIII-LI) that could be experimentally synthesized.

What Problem Does This Paper Address?

Water ice, despite being a ubiquitous and extensively studied substance, still holds many mysteries. While 18 crystalline phases are known, several are metastable, and new ones continue to be discovered. The challenge lies in the sheer complexity of ice's configuration space – the vast number of possible arrangements of water molecules and their hydrogen bonds. Traditional methods struggle to explore this space comprehensively and to filter the resulting theoretical structures down to those that are experimentally relevant and likely to form. This paper aims to systematically map this uncharted territory, identify new stable ice phases, and provide a framework for understanding their formation.

Methods

The researchers employed a multi-pronged computational strategy:

  • Leveraging Zeolite Databases: They exploited the known isomorphism between tetrahedral networks in zeolites and ice structures. By drawing from extensive databases of theoretical zeolite networks (Treacy, Deem, IZA), they generated a large pool of potential ice structures.
  • Density-Functional Theory (DFT): First-principles DFT calculations (specifically PBE-DFT) were used for geometry optimization and energy calculations of these candidate structures. They found that ReaxFF force fields were sufficient for initial screening, followed by more accurate DFT refinement.
  • "Generalised Convex Hull" (GCH): This is the paper's core novel contribution. Instead of relying solely on energy-density relationships (which primarily identify pressure-stabilized phases), the GCH uses a kernel principal component analysis (KPCA) to extract abstract structural features. This allows for the identification of structures stable under a broader range of thermodynamic and kinetic constraints, not just pressure.
  • "Sketch-Map" Dimensionality Reduction: To visualize and navigate the complex configuration space, they used the sketch-map algorithm. This creates a 2D map where similar structures are clustered together, providing an intuitive overview of the relationships between different ice polymorphs and highlighting novel candidates.

Datasets

The primary "dataset" for generating candidate structures was derived from:

  • Treacy Database: A database of theoretical zeolite networks.
  • Deem Database: Another database of theoretical zeolite networks.
  • International Zeolite Association (IZA) Database: A database of experimentally synthesized zeolite networks.

From these, they generated and optimized 74,963 initial structures, which were then refined to 15,869 distinct structures for analysis.

Key Insights and Ideas

  • Zeolite-Ice Isomorphism is Powerful: The strong structural similarity between water ice and silica networks (zeolites) provides a robust starting point for exploring ice's configuration space.
  • Beyond Pressure: The GCH is Key: The traditional energy-density convex hull is limited to pressure-stabilized phases. The GCH, by considering abstract structural features, unlocks the discovery of phases stabilized by other factors, making it a more general tool for identifying synthesisable structures.
  • A Navigable Map of Ice: The sketch-map visualization provides an unprecedented global view of ice's structural landscape, revealing relationships between known and novel phases and offering insights into potential formation pathways.
  • Discovery of New Ice Phases: The approach successfully identified all known ice phases (except the metastable ice IV) and, crucially, proposed 34 new candidate phases (ices XVIII-LI) that are predicted to be stable and potentially synthesizable.

Key Results

  • 74,963 structures geometry optimized: This represents a massive computational effort to explore the configuration space.
  • Rediscovery of known phases: The GCH approach successfully identified all known ice phases (I-XVII, i, O, quartz) except ice IV.
  • Identification of 34 new candidate ice phases: These are proposed as ices XVIII through LI and are highlighted as promising for experimental synthesis.
  • Sketch-map visualization: A navigable map of the configuration space was created, showing structural similarities and clustering of known phases.

Limitations and Future Work

  • Preselection Cutoff: The current study is limited by a preselection cutoff on system size, meaning that very large unit cell structures were not exhaustively explored. Relaxing this cutoff is suggested as a next step for a more complete search.
  • Experimental Validation Needed: The proposed 34 new ice phases are theoretical predictions and require experimental verification.
  • Formation Pathways: While the sketch-map and GCH provide a starting point, further work using advanced simulation techniques (e.g., transition state sampling, umbrella sampling) is needed to fully elucidate formation pathways.
  • Other Stabilizing Factors: While the GCH accounts for constraints beyond pressure, the paper notes that the influence of factors like temperature, electric fields, and guest molecule concentrations on stability could be further investigated.

Summary updated

Abstract

Ice is one of the most extensively studied condensed matter systems. Yet, both experimentally and theoretically several new phases have been discovered over the last years. Here we report a large-scale density-functional-theory study of the configuration space of water ice. We geometry optimise 74,963 ice structures, which are selected and constructed from over five million tetrahedral networks listed in the databases of Treacy, Deem, and the International Zeolite Association. All prior knowledge of ice is set aside and we introduce "generalised convex hulls" to identify configurations stabilised by appropriate thermodynamic constraints. We thereby rediscover all known phases (I-XVII, i, 0 and the quartz phase) except the metastable ice IV. Crucially, we also find promising candidates for ices XVIII through LI. Using the "sketch-map" dimensionality-reduction algorithm we construct an a priori, navigable map of configuration space, which reproduces similarity relations between structures and highlights the novel candidates. By relating the known phases to the tractably small, yet structurally diverse set of synthesisable candidate structures, we provide an excellent starting point for identifying formation pathways.
BibTeX
@inproceedings{engel2018mapping,
  author = {Edgar A. Engel and Andrea Anelli and Michele Ceriotti and Chris J. Pickard and R. J. Needs},
  title = {{Mapping uncharted territory in ice from zeolite networks to ice structures}},
  booktitle = {Nature Communications},
  year = {2018},
  doi = {10.1038/s41467-018-04618-6},
}
Figure