TL;DR
This paper introduces a novel computational strategy combining crystal structure prediction (CSP) with "structural invariants" to discover new porous materials. This approach successfully identified a rare, low-density mesoporous hydrogen-bonded organic framework (HOF), 3D-CageHOF-1, by finding structural analogies between predicted crystal structures of different molecules.
What Problem Does This Paper Address?
Designing porous molecular crystals, particularly hydrogen-bonded organic frameworks (HOFs), is challenging. While these materials hold promise for applications like separation and catalysis, they are difficult to synthesize because molecules naturally prefer to pack densely, and porous structures often represent kinetically trapped, higher-energy states. Traditional methods like isomorphous substitution (swapping similar molecules in a known structure) are less reliable for organic crystals due to weaker, less directional intermolecular forces. The authors aim to overcome these limitations by developing a computational approach that can systematically compare and identify analogous crystal structures between chemically different molecules, thereby guiding the discovery of new porous HOFs.
Methods
The core of the methodology involves two key components:
- Crystal Structure Prediction (CSP): The researchers used CSP to generate a vast number of potential crystal structures for the organic cage molecule, Cage-3-NH₂. This process explores the energy landscape of possible crystal packings. Initially, rigid-molecule approximations and empirical force fields were used, followed by re-optimization of promising structures using density functional theory-based tight binding (DFTB) to account for molecular flexibility.
- Structural Invariants (Pointwise Distance Distribution - PDD): To quantitatively compare the predicted structures of Cage-3-NH₂ with known crystal structures of a related molecule (T2), a novel "isometry invariant" called the Pointwise Distance Distribution (PDD) was employed. PDD describes the distribution of distances between a molecule's center of mass and its nearest neighbors within a crystal lattice. This metric is invariant to translations, rotations, and reflections, allowing for robust comparisons of crystal packing even when lattice parameters or molecular orientations differ. This is a significant advancement over traditional methods like RMSD, which fail for dissimilar molecules.
Datasets
- Predicted Crystal Structures of Cage-3-NH₂: Generated via CSP.
- Experimentally Observed Crystal Structures of T2: Five known polymorphs of the molecule T2 were used as reference structures. A new polymorph of T2 was also characterized and included.
Key Insights and Ideas
- Bridging the Gap in Analogy: The PDD structural invariant provides a formal, quantitative method to establish structural analogies between molecules that are geometrically similar but chemically distinct. This moves beyond subjective visual comparisons.
- CSP as a Predictive Tool for Kinetically Trapped Structures: The study demonstrates that CSP can identify kinetically trapped, low-density polymorphs, which are often the most interesting for porous materials.
- Discovery of a Novel Mesoporous HOF: The combined approach successfully led to the experimental discovery of 3D-CageHOF-1, a rare example of a non-interpenetrated, mesoporous 3D HOF with significant porosity.
- The Power of "Cage" Building Blocks: The bulky nature of the organic cage molecules (Cage-3-NH₂) plays a crucial role in preventing interpenetration and enabling the formation of stable, porous structures.
Key Results
- The CSP of Cage-3-NH₂ revealed multiple "spikes" on its energy-density landscape, indicating the potential for kinetically trapped, low-density polymorphs.
- Using PDD invariants, five predicted Cage-3-NH₂ structures were identified as strong structural analogues to the five known T2 polymorphs.
- One of these predicted structures, analogous to the porous T2-γ polymorph, was targeted and successfully synthesized as 3D-CageHOF-1.
- 3D-CageHOF-1 exhibits a low density of 0.54 g cm⁻³ and a high accessible void volume of 72.5%.
- The activated material shows a BET surface area of 1750 m² g⁻¹, with 1D mesoporous channels of 2.2 nm width.
- The experimental structure of 3D-CageHOF-1 showed excellent agreement with the computationally predicted structure (RMSD of 0.267 Å).
Limitations and Future Work
- Accessing Predicted Structures: CSP predicts the existence of structures and their relative energies, but it does not directly provide the experimental conditions (solvents, temperature, etc.) needed to crystallize them. The paper notes that computational methods for predicting solvent effects are expensive and were not used here, relying instead on experimental screening.
- Generalizability of the Approach: While successful for Cage-3-NH₂ and T2, the authors anticipate that this strategy of combining CSP with structural invariants will be broadly applicable for discovering other functional molecular crystals.
- Expanding the Scope of Structural Invariants: The PDD invariant is a powerful tool, but further development and application to a wider range of crystal systems and properties are implied.
- Exploring Energy-Structure-Function Landscapes: The paper suggests that this method could enable the cross-comparison of entire energy-structure landscapes for chemically dissimilar molecules, paving the way for more systematic materials design.
Summary updated