Partially reactive force field for the UiO-66 metal-organic framework

This paper introduces nb-UiO-FF, a novel partially reactive force field that accurately models the structural, mechanical, and defect properties of the UiO-66 metal-organic framework and enables molecular dynamics simulations of its solvothermal synthesis and self-assembly mechanisms.

The Gist

Imagine a metal-organic framework (MOF) like UiO-66 as a giant, microscopic 3D puzzle. It's built from metal "nodes" (like Zirconium clusters) acting as the corners, and organic "linkers" (like benzene rings) acting as the connecting rods. Scientists love these puzzles because they are incredibly sturdy and can be tweaked to trap gases, deliver drugs, or speed up chemical reactions.

However, there's a problem: figuring out exactly how these puzzles snap together in the first place is like trying to watch a movie played at the speed of light. The chemical bonds that hold the metal and the linkers together are tricky to simulate on a computer. Most computer models treat these connections as permanent glue; they can't show the glue being applied, the pieces snapping together, or even what happens if a piece is missing.

The Solution: A "Smart Glue" Force Field
In this paper, the authors introduce a new computer tool called nb-UiO-FF. Think of this as a new set of rules for a simulation game that allows the puzzle pieces to be "partially reactive."

Here is how they made it work, using some simple analogies:

• The "Dummy" Atoms (The Invisible Hands):
  In the real world, the Zirconium metal node has a complex electrical charge that pulls the linkers in specific directions. Standard computer models struggle to mimic this without getting messy. The authors solved this by attaching invisible "dummy" atoms (like little magnetic dummies) to the metal nodes. These dummies act like invisible hands that hold the linkers in the correct shape and orientation, ensuring the puzzle builds the right way without needing complex, heavy calculations.

• The "Morse Potential" (The Stretchy Spring):
  Usually, computer models treat bonds like rigid sticks. If you pull them, they break instantly. The authors replaced these rigid sticks with a Morse potential, which acts more like a stretchy spring. This allows the simulation to show the metal and the linker stretching, wobbling, and even snapping together or falling apart dynamically. This is crucial for watching the "birth" of the material.

What They Tested
The authors didn't just build the tool; they put it through a rigorous stress test to make sure it was reliable:

1. The Perfect Puzzle: They checked if the tool could recreate the exact shape of a perfect UiO-66 crystal. It matched real-world measurements almost perfectly (within a tiny fraction of a percent).
2. The Soaked Puzzle: They tested the tool with the crystal soaked in two different liquids used to make it (DMF and ethanol). The model showed that the crystal stays strong and doesn't fall apart when wet.
3. The Broken Puzzle: Real-world crystals often have missing pieces (defects). The authors intentionally removed linkers or entire nodes in the simulation. The tool successfully showed that the crystal could still hold its shape even with these holes, just like the real material does.
4. The Bouncing Puzzle: They tested how hard you could squeeze the crystal before it deformed. The results matched high-level physics calculations, proving the model understands the material's strength.
5. The Cousin Puzzle: They tried the tool on a slightly larger version of the puzzle (UiO-67) and it worked there too, proving the rules are flexible.

Watching the Magic Happen
The most exciting part of the paper is using this new tool to watch the self-assembly process. Imagine dropping all the puzzle pieces (metal nodes and linkers) and the liquid solvent into a box and hitting "play."

The simulation showed the pieces drifting around and slowly starting to stick together.

• They saw the metal nodes and linkers finding each other and forming the initial building blocks.
• They noticed that sometimes pieces get stuck in "wrong" positions (kinetic traps), like a puzzle piece that fits loosely but isn't quite right.
• They observed that the process is slow; the pieces are heavy and move sluggishly, so the full puzzle doesn't assemble completely in the time they simulated.

The Bottom Line
This paper presents a new, highly accurate computer model that acts like a "smart microscope" for the UiO-66 material. It can simulate the material's structure, its strength, and its ability to handle defects. Most importantly, it is the first tool of its kind that can realistically simulate the dynamic process of the material building itself from scratch, helping scientists understand how these amazing materials are born and how to control their imperfections.

Technical Summary

Technical Summary: Partially Reactive Force Field for the UiO-66 Metal-Organic Framework

Problem Statement
Metal-organic frameworks (MOFs), particularly UiO-66, are widely studied for their structural tunability and stability. However, understanding their synthesis mechanisms and defect formation remains challenging due to a lack of computational models capable of simulating reactive processes. Existing force fields (e.g., UFF4MOF, BTW-FF, MOF-FF) typically rely on pre-defined connectivity, rendering them unable to model bond breaking and formation, which is essential for studying MOF self-assembly and defect engineering. While reactive force fields like Reax-FF and machine learning potentials exist, they face challenges regarding accuracy for MOFs, computational cost, or the difficulty of expanding them to include solvents and modulators. Furthermore, specific models for UiO-66 often struggle to accurately represent the hydroxylated form or require computationally expensive dummy atom configurations (e.g., 48 dummy atoms per node).

Methodology
The authors developed nb-UiO-FF, a partially reactive force field designed to model the UiO-66 framework and its isoreticular analog, UiO-67. The methodology builds upon a previously established approach for ZIF-8, utilizing a combination of non-bonded interactions and cationic dummy atoms (CDA).

• Force Field Construction: The model replaces hard, bonded potentials between the metal node and ligands with a non-bonded Morse potential to allow for dynamic bond breaking and formation.
• Dummy Atoms (CDA): To reproduce the anisotropic charge distribution of the Zr atoms and maintain correct coordination geometry without bonded terms, four dummy atoms (D) are introduced per Zr center. These are connected to the Zr atom via strong harmonic bonds and angles. The +1.968 charge of the Zr atom is fully distributed among these four dummy atoms (+0.492 each), rendering the Zr atom electrically neutral.
• Parameterization: The force field parameters were derived starting from a non-reactive force field by Yang et al., adjusted for the hydroxylated Zr6O4(OH)4 node. Bonded parameters for the metal cluster were taken from reported models, while equilibrium bond lengths and angles were sourced from DFT results by Valenzano et al. The Morse potential parameters ($D_0$ and $\alpha$) for the Zr–Oligand interaction were optimized by systematically exploring the parameter space to balance accuracy in number density and local bond lengths (specifically the Zr–Oligand radial distribution function) against DFT and experimental references.
• Simulation Protocols: All simulations were performed using LAMMPS. Validation involved structural analysis, stability tests in solvents (DMF and ethanol), and mechanical property calculations. Self-assembly studies utilized a system of 16 Zr clusters, 96 BDC ligands, and 2136 DMF molecules, simulated over 6 ns in the NPT ensemble.

Key Results

• Structural Accuracy: The nb-UiO-FF reproduces the experimental lattice parameter of pristine UiO-66 with a deviation of approximately 0.046% (20.71 Å vs. 20.7004 Å). Radial distribution functions (RDFs) for key atomic pairs (Zr–$\mu_3$O, Zr–$\mu_3$OH, Zr–Zr, Zr–Oligand) show reasonable agreement with DFT and experimental data, with deviations generally within 10%.
• Solvent and Defect Stability: The force field maintains framework integrity in the presence of DMF and ethanol solvents, as well as in defective structures containing missing linkers and missing nodes. In defective models, charge neutrality was preserved by capping undercoordinated centers with chloride ions or monodentate BDC fragments.
• Mechanical Properties: Calculated elastic constants ($C_{11}$, $C_{12}$, $C_{44}$) and derived mechanical properties (Bulk modulus: 44 GPa, Young's Modulus: 40 GPa) are in good agreement with DFT benchmarks, though slightly higher, likely due to the rigidity introduced by the strong bonded interactions between Zr and dummy atoms.
• Transferability: The model successfully reproduces the structure of the isoreticular UiO-67 framework, demonstrating transferability across different linker lengths.
• Self-Assembly Dynamics: In dynamic simulations, the force field successfully captured the progressive formation of coordination bonds between metal clusters and organic linkers. The study identified transient structural motifs, including initial secondary building units (SBUs) and non-ideal binding configurations (kinetically trapped states). However, full convergence of the coordination number and the formation of large ionic aggregates were not achieved within the 6 ns simulation window, suggesting the need for enhanced sampling methods for longer timescales.

Significance and Claims
The paper claims that nb-UiO-FF provides a consistent and physically reliable description of UiO-66 and UiO-67, bridging the gap between static structural modeling and reactive dynamics. By utilizing a reduced set of dummy atoms (four per Zr) compared to previous models, the authors achieve a balance between model efficiency and the ability to simulate reactive processes. The force field enables the study of self-assembly mechanisms and defect formation, which are inaccessible to traditional fixed-connectivity force fields. The authors position this work as a foundation for investigating the early stages of MOF synthesis and the engineering of point defects, while acknowledging that further work with enhanced sampling schemes (e.g., metadynamics) is required to fully overcome kinetic traps and model complete self-assembly.