Adsorption of CO₂ by amine-functionalized metal–organic frameworks using GCMC and ReaxFF-based metadynamics simulations
Grand canonical Monte Carlo (RASPA) screens CO₂ and H₂O uptake on amine-decorated Zn-MOF candidates; Jaguar DFT (M06-2X-D3/6-311++G**) ranks linker-level physisorption vs chemisorption; a ZIF-oriented ReaxFF is extended with CO₂–amine chemistry and metadynamics (LAMMPS + PLUMED) estimates chemisorption barriers, alongside experimental adsorption tests on synthesized samples.
Summary¶
Twenty primary-amine Zn MOFs (CSD or linker-substituted models) are downselected using Zeo++ pore metrics, RASPA GCMC for CO₂ and H₂O at 100 kPa and 298 K (including dilute CO₂–O₂–N₂ mixtures), and MPNN-predicted charges under charge neutrality. DFT on gas-phase linkers compares physisorbed vs chemisorbed CO₂ and probes water-assisted stabilization (e.g., CH₃NH₂ + CO₂ + H₂O clusters, transition states for carbamate-like paths). A ReaxFF trained for ZIF water stability is augmented with CO₂ physisorption/chemisorption energies and CH₃NH₂–CO₂ reaction energetics; parameters are listed in Supporting Information. Well-tempered metadynamics (Gaussian hills, PLUMED) with distance-based CVs drives CO₂ toward −NH₂ in dry, water-assisted, and (for PEI-like cases) methylamine-assisted setups; NVT at 300 K, Δt = 0.25 fs, 1–4 ns runs, multiple initial velocities for barrier statistics. Experiments measure CO₂ uptake under humid air (2500 ppm CO₂) and related characterization (article §2.6).
Methods¶
- Screening: RASPA GCMC; Generic MOFForce field for physisorption; MPNN charges; Zeo++ voids; multiple gas compositions including 400 ppm and 1% CO₂ scenarios in tables.
- DFT: Jaguar, M06-2X-D3, 6-311++G**; full relaxations; transition-state searches with one imaginary frequency where reported (SI).
- ReaxFF + metadynamics: Extended ReaxFF for amine–CO₂ chemistry; LAMMPS; PLUMED metadynamics (Gaussian height ~10 kJ/mol, σ = 0.1, hill frequency 1000 steps, well-tempered bias factor ~100, lower walls to suppress spurious inter-amine reactions); collective variables primarily interatomic distances among N, C (CO₂), O, and helper atoms (Table 2).
- Experiment: Synthesis and CO₂ uptake protocols for selected MOFs and ZIF-8/PEI blends (§2.6).
1 — MD application (ReaxFF + metadynamics). Engine: LAMMPS + PLUMED; NVT at 300 K; PBC 3D MOF unit cells with PLUMED-wrapped PBC as in the SI; NVT thermostat chain as in the article (confirm details in pdf_path); Δt = 0.25 fs; 1–4 ns restarts for barrier stats; well-tempered metadynamics (Gaussian hills ~10 kJ/mol, σ = 0.1, hill period 1000 steps, well-tempered bias ~100 as summarized; lower walls to block spurious amine–amine chemistry); CVs are interatomic distances among N, C (CO₂), O, and selected helpers (Table 2). GCMC (RASPA) legs for adsorption use 100 kPa / 298 K and dilute air-like mixtures as in Tables (not the same code path as LAMMPS). N/A — NPT metadynamics in the summary here; N/A — external uniform E-field in the stated ReaxFF leg (field-like biases only via PLUMED as written for metadynamics). N/A — replica exchange in the brief text.
2 — Force-field training — ReaxFF extended for ZIF + amine + CO₂ with Jaguar M06-2X-D3/6-311++G** reaction/adsorption targets and SI-listed parameters. 3 — Static QM (Jaguar). Cluster-level physisorption vs chemisorption and TS searches as described in article/SI. 4 — Grand canonical Monte Carlo (RASPA). Screening isotherms; not a ReaxFF MD block but part of the multi-method workflow.
Findings¶
- At high CO₂ fugacity, uptake is dominated by physisorption; at dilute CO₂ (e.g., air-like 2500 ppm), chemisorption contributions become prominent in the analysis.
- Gas-phase DFT shows water can stabilize carbamate-like products and make chemisorption more exothermic; metadynamics reports high intrinsic barriers for dry chemisorption on several frameworks, reduced with explicit water or −NH₂ assistance pathways.
- ZIF-8 mixed with polyethylenimine shows high initial chemisorption but degrades on cycling, while amine-functionalized MOFs show lower amine utilization under humid air in the experiments—contrasting simulation-experiment roles of water and diffusion.
- The paper positions the workflow as an example of combining screening, reactive FF development, and rare-event simulation for MOF CO₂ capture.
Corpus honesty — see ## Limitations for ReaxFF transfer across linkers and sparse experimental statistics.
Limitations¶
ReaxFF accuracy for full periodic MOF reactivity, force-field transferability across linker libraries, and limited experimental statistics on every candidate remain constraints.
Relevance to group¶
Co-authored by van Duin; extends ReaxFF for MOF–amine–CO₂ chemistry with metadynamics and experimental validation.