We have discussed the importance of comparison between simulations and experiments in a previous blog post as well as in Chapter 18, Comparison to Experiment. But we do have to be extremely careful to ensure that the simulations are high quality, directly comparable to the experiment, and do not introduce effects because of the simulation protocols. I want to discuss some of these questions while looking at papers which model the interaction between water and carbon nanotubes. (I should make it clear that I’m using the papers as a springboard, rather than criticizing these papers; nevertheless, it is an important part of science to read papers critically, and ask what might be wrong.)

In a recent Phys. Rev. Lett. [1], Homma et al. measured the photoluminescence (PL) of carbon nanotubes in vacuum and in water vapour, and found a strong shift of the PL emission and excitation; they also used Raman spectroscopy to measure the frequency of the radial breathing mode (RBM) of the nanotube. The phenomena of changing PL and RBM on exposure to the atmosphere are known, though the detailed mechanism is not. To help understand the changes, they also performed MD simulations. Similar simulations have been performed before [2] and the present paper refers to this work. The simulations show a similar shift in the RBM to the experiments, and show strong structure in the water molecules around the nanotube.

Methodological Questions

When I was reading the simulation part of the paper by Homma[1], I was left with various questions. The implication is that they used the same simulation protocols as Longhurst and Quirke[2], but this is not made clear. The water model used (SPC/E) is not the most reliable, and the effect of the water-nanotube interaction on the results is not tested. In fact, there is very little testing of the simulation parameters. How does the density of vapour used compare to the experimental vapour pressure ? Do the results depend strongly on unit cell size ? These questions are important to answer, and I would have liked more detail on these (this type of information is normally found in Supplementary Information).

Scientific Questions

Beside the questions about the simulation protocols, there are interesting scientific questions. How do we understand a macroscopic phenomenon like hydrophobicity in terms of atomistic mechanisms ? Certainly the results from this paper indicate that there is some attraction between the water and the nanotube (the water molecules form a dense layer about 0.8nm away from the nanotube surface, with a weaker layer 1.1nm away). This could be investigated with graphene and either a single water molecule or a sheet of molecules. Indeed, the hydrophobicity of graphene is where the paper starts, but graphene has no PL response, nor indeed a radial breathing mode (the frequency of the RBM scales as the inverse of the nanotube diameter). This makes the experiments impossible but the simulations could be done, and would be illuminating.

It would be extremely hard to model the PL response: exciton recombination is the main source of the PL, which requires the Bethe-Salpeter equation or similar approaches. These simulations are extremely demanding, and the best calculations so far were semi-empirical, with a continuum dielectric. Such an apparently simple system requires care in protocols and, potentially, sophisticated modelling to understand. A large part of the simulator’s job is deciding how to make a sensible model of the problem, and how to ensure that the results are sensible. These ideas are discussed in Chapter 10, Planning a Project.

[1] Phys. Rev. Lett. 110, 157402 (2013) DOI:10.1103/PhysRevLett.110.157402

[2] J. Chem. Phys. 124, 234708 (2006) DOI:10.1063/1.2205852