This builds upon #21, using the file containing the melting points as the input for normalising the results.

All the figures should be assesed, in particular with a couple of points.

1. The naming of the figures should be consistent
2. The data in the figures should be accurate and up to date.

The first point is important because there are currently a couple of figures in my thesis with the incorrect name. Figures which contain the summary of the relaxation should all include the word _summary in the filename or something similar.

Update to the latest versions of

• sdanalysis
• sdrun
• experi

The required version of the cudatoolkit package, cudatoolkit==8.0 is no longer available as a package on anaconda cloud. This version is required by the version of hoomd I am using so to keep working an updated version of hoomd is required.

Unfortunately an updated versions of hoomd on conda are no longer built with MPI support, so using on clusters will require compilation from source.

I should be calculating the scaling of a variety of quantities, how they change with temperature, however this should be compared with a variety of theoretical values, including Vogel--Tamman--Fulcher and other similar relationships.

This will provide a single value for how unusual the dynamics of these systems actually are.

I have had some issues with results being incorrectly calculated due to trying to get the results quickly. I am not entirely confident with the results I have, and whether they are up to date, so I should go through and confirm the correctness of the results I have collected, taking note of any peculiarities.

Rather than using the short time relaxations, I now have the ability to calculate the long timescale rotational diffusion.

Calculating the harmonic mean provides an alternate method of aggregating the molecular relaxation times. Which can also be used as a measure of dynamic heterogeneity.

Using the fit of the Vogel-Fulcher-Tamman function, it is possible to get an estimate on when the dynamics reach the point of slowing down enough. This would allow a comparison with other values using the glass transition temperature.

Create a clearly defined process for taking raw simulation data and having it end up somewhere as
computed dynamics quantities. This is particularly important as I develop notebooks that use these
quantities.

As an additional test to the actual tests, which are very lacking. I should run some static analysis tools on the source code.

The readme should include instructions for getting started with replicating the results which I have produced. Included in these instructions should be the practices I am using for developing, including things like structure and commit messages.

All the quantities presented should also include an indication of the estimated error of the result. This can be calculated from the step which averages over all the individual results.

I have this idea of reversible relaxations, but I don't know how I want to include them or what types of analysis are useful for this.

The idea is that structural relaxation is reversible, way beyond what would be expected from random motion. There are larger fluctuations than the structural relaxation distance.

This discussion is going to be related to #28 which will paint part of the picture. Also inclusion of first passage times, last passage times and diffusion constants.

I have jupyter notebook with analysis of various dynamic properties, these need to be added to this repository.

I need to be able to easily take the raw simulation data and generate the dynamics quantities for use in the various figures that I have. I would like for all this data to be stored in a single hdf5 file with datasets for each temperature. Although I am not entirely sure if that is technically feasible, since I may need support for multiple open file targets.

I will need to store;

• Dynamics values (msd, rotational relaxation ...)
• Relaxation values (diffusion constant ...)
• molecular relaxation values (tau_1, tau_0.4 ...)

Additionally I will need a data dictionary explaining all these values. Ideally a yaml file or similar which includes the latex and unicode symbols so I can just import the dictionary into any code.

The relationship T_m/T is an important one, allowing the comparison of dynamics from a number of different materials and conditions.

This depends on malramsay64/Crystal_Melting#21 for the melting points.

There is something really wrong with all the results from the dynamics, and I am
currently not sure what it is. So I need to look into why so many of the points appear
to be missing.

This plot should include the slopes indicating the diffusive and ballisitc regimes. Would be nice to also annotate the slopes.

Ref malramsay64/phd-thesis#241

There is some issue with the detection of the datatype of the normalised temperature. This is required to ensure that the temperature is converted from a string to a float. However it appears there are issues with false positives.

The lowest temperature simulation appears to not be equilibrated, displaying anomalous results.

The points at the end have a large spread, so increasing the minimum number of points in the cleanup step should help with this.

The figures for my thesis need to be created, which should be high quality pdfs.

A possible idea is to use latex text rendering for the figures allowing for consistency throughout the document.

https://matplotlib.org/users/usetex.html

In the calculation of rotational diffusion, I need to keep track of the periodicity of the rotations so I can describe rotations which are larger than pi. The main reason for this is the calculation of the quantity gamma, which is a combination of the rotational and translational quantities.

Additionally this is required for finding the rotational heterogeneities.

There is an issue with the calculation of the molecular relaxation. There seems to be an issue with not removing the missing values -> i.e. 2**32-1

A key feature of this repository is that all the data should be reproducible
from the components checked into the repository, even if the data isn't. To
this end being able to reproduce the data I already have available is
important.

I need a juptyer notebook I can use as a reference for the timescale of all the dynamics quantities I
am interested in. This is to make it simple to just look up a value.

All the notebooks should be numbered, to provide some direction for the analysis narrative. Additionally all the notebooks should include descriptions of what is going on and the results which are obtained.

This should include the melting points in a simple file format so that it is possible to easily change or add to these values which can then be propagated across all analysis.

The dynamics of single molecules introduces a quantity, tau_DL04 which is defined as the
timescale of the last passage of a molecule, where the last passage is the final time
before the molecule moves a distance 1. This quantity is then compared with the tau_D1,
for which there is a high correlation. Is this purely a result of the construction of
the quantity? Both quantities do include the distance of 1. Is there a better method of
measuring this?

I need some analysis which indicates the presence of rotational heterogeneities, matching the idea of the non-gaussian parameter for translational motion.

The idea is to bring the ideas of heterogeneous dynamics together.

Figures which show the heterogeneous dynamics in the molecular systems.

The implementation of the Brownian Motion should be in the src directory and imported
into the notebook. This puts the emphasis of the notebook on the results rather than the
code to generate them.

Give the project a DOI, including the computational environment used to run it.

It appears that large jumps are an important characteristic of the dynamics in this system. Both for translational and rotational motion. This is going to need some narrative, or cohesive idea to tie the results I have together.

The general result is taking the different means of the relaxation times (#27) however this has to fit into the wider discussion on breakdown of Stokes--Einstein. Comparison with Diffusion constant and rotational relaxation.

This is a picture of the translational vs rotational motion.

The idea is to show the types of motion which a particle can undergo. It would be nice to also plot the different thresholds for relaxation, particularly to see how the motions fit with them.

This would likely need a few different trajectories, something like;

• the fastest,
• the slowest, and
• the middle.

Rather than using the timesteps, I should be using the base unit time since this allows for comparison with existing results.

Create a setup.py which will install the src files as a python package. The installation can be managed by the environment file.

In having a setup.py file, I no longer need the sys path append nonsense at the start of every notebook and it will just generally make things easier.

Fit the values I have to the curve

As an extra I should use these quantities to estimate the glass transition.