Since their discovery in 1985,(1) fullerenes have attracted much attention for their unique properties. For example, their remarkable electron acceptor ability makes them suitable components of donor–acceptor complexes for artificial photosynthesis.(2) Numerous applications have been described for fullerene in several areas of science and technology, and particularly in the field of photovoltaic devices.(3) Applications in biomedical sciences have also been proposed, mostly related to fullerene high permeability through biological barriers,(4) its capability to entrap metals and small molecules,(5) and its antioxidant properties.(6) Endohedral metallofullerenes, in which metal ions are trapped inside the fullerene cage, have been proposed as powerful contrast agents for magnetic resonance imaging.(5)
Simulations of fullerenes and carbon nanotubes at different levels have been carried out since the early 1990s. Quantum mechanics calculations have been used to predict, among others, electronic and structural properties.(7, 8) Classical simulations have been used to study fullerene interactions with water and other solvents (reviewed in ref 9). The results of classical simulations depend critically on the potential energy functions used to represent the molecule, and on the properties used as a target in the force field parametrization. In the specific case of fullerene, a large number of atomistic simulations has been published in the literature. However, little work has been done to validate the potentials used, and few papers have been published with comparisons between results from atomistic simulations and experimental data.(9) Rivelino and co-workers compared the free energy of transfer of fullerene between different solvents for six different fullerene parametrizations, showing significant differences among the models.(10) While free energies of transfer are of fundamental importance to assessing the interaction of fullerene with different solvents, they do not contain information on the interaction of fullerene with other fullerene molecules—in other words, on the cohesive forces that are present in the solid phase or in solid-like aggregates (e.g., clusters in solution).
Large-scale behavior of fullerene has been explored using coarse-grained (CG) simulations.(11-14) Again, little work has been done to validate coarse-grained potentials. CG studies on the aggregation behavior of fullerene yielded different and sometimes contrasting results.(11, 13) Those results clearly depend on the force field employed and will reproduce the experimental behavior only if both fullerene–solvent and fullerene–fullerene interactions are carefully calibrated.
In the present study, we compare and validate both atomistic and coarse-grained models for fullerene. The manuscript is divided into two parts: first, we present an assessment of the quality of four atomistic force fields currently in use, with particular emphasis on two models; second, we optimize the MARTINI coarse-grained (CG) force field for fullerene. The performance of atomistic force fields is assessed by comparing (a) the partitioning behavior, as a stringent test for fullerene–solvent interactions, and (b) the enthalpy of sublimation, a measure of the cohesive forces in the solid. Refinement of the MARTINI CG model for fullerene is based on the simultaneous optimization of several quantities, including experimental partitioning coefficients, sublimation energies, and fullerene–fullerene potentials of mean force in water and alkanes.