Hi Jodi.
You are correct that it is possible to do MD simulations based on DFT. We do not mean to imply that no mesoscopic simulations can answer questions where experimental data is unclear. Our point is that the simulations which have been employed did not link to DFT. They used constrained dynamics is used to simply constrain each thiol to the face of a sphere, whereas DFT and experimental data suggested the thiols form ditholate structures with gold adatoms. The simulations employed simply treat intermolecular reactions simply as harmonic springs with a larger (but arbitrary) repulsion between atoms on unlike molecules. These simulations were never designed to clear up unclear experimental data, they were done to try to tease out the physics of clear experimental data. The parameters were chosen to match the experiments and then the physics driving the phase separation was extracted. Now the experiment is called into question these results cannot be used to as evidence for the structure they were designed to show. This is our point.

I do not disagree that it may be possible to use different mesoscopic simulations to probe these structures, but this is not what was done.

Dear Jodi,

First of all, thank you very much for your interest in the paper, and thank you very much for your insight. There are still large barriers to commenting on papers, and it is in our best interest that these are overcome to get a discussion.

Worried by your comment that we had overreached or overreacted, I read both the "comment on computer simulations" and the relevant paper again. My apologies in advance: I will probably reiterate some of the points already made by my colleagues in this thread. I just wanted to add my opinion on this:

I suspect that the section header may unfortunately have been a bit misleading. The subsequent section does not purport to criticise computer simulations as such, but largely restricts itself to stating that the DPD simulations (reported by Singh et al.) cannot be considered as proof of the existence of stripy nanoparticles. Unfortunately, Stellacci and proponents have indicated that they do consider the simulations to provide proof of their existence.

From my understanding (having had a little bit of experience with DFT simulations in the past), we always need to be extremely careful using simulations to predict structures not known to exist. Even for DFT, it is almost essential that the chosen method, basis sets and optimization parameters have been demonstrated to give comparable results between calculation and experiment for similar systems.

This is underscored by the drastic approximations necessary for the large-scale DPD simulations of striped morphologies. For example, inter-ligand interaction parameters and a ligand density is chosen (18.75, 33.75, and 0.07 chains/Å^2 but it is not immediately clear how these values were chosen). The nanoparticle is assumed to be spherical, not faceted. Ligands are assumed to attach to each nanoparticle in equal amounts. During optimisation, ligands of the same species are briefly moved together. These choices are probably all valid, and as such would indicate that (under the right conditions) such stripe-like domains can form from entropic considerations alone.

Without further evidence that the set of approximations used for this simulation reliably yields experimental results for similar systems, I would be reluctant to assign predictive properties to the simulation. I would furthermore like to highlight that Fig. 1(f) in the PRL paper is a bit misleading as it appears to show very good agreement between simulated stripe width and experimental stripe width, but through the use of two separate axes obfuscates that the one is consistently about 50% larger. Figure 4(f) furthermore ostensibly shows "stripelike morphology" in STM images, but I am hard pressed to see anything more stripy than pure random phase separation.

However, just to bring this to a close, the evidence that entropic considerations alone can drive the formation of stripes (which is very interesting, nonetheless) is not evidence that they have been obtained using the methodology used by Stellacci et al.. What we need is for a separate group to work on ways for making evidence-supported stripy nanoparticles...

Cheers,


Brian.