Result Dissemination

Mike Tonks, Damien Tourret

How you present and disseminate your results is also a critical aspect of applying the phase-field method. Here, we provide recommended practices on disseminating your results.

Recommended Practices for Postprocessing and Plotting:

Present dimensional results whenever possible

The phase-field method is inherently dimensional, meaning that each of its model parameters have specific dimensions in terms of length, time, energy, etc. These values vary for different materials and systems, based on material properties such as free energies, diffusivities, interfacial energies, and so on. However, it is a common practice to nondimensionalize a model to simplify the numerical solution. Nondimensionalization has also been used to avoid the need to have accurate material properties for a specific material system. In those cases, results are qualitative at best and are not material specific.

It is recommended to present material-specific results using dimensional properties. Therefore, if you nondimensionalize your model, always dimensionalize your results before plotting if at all possible. Quantitative, material specific results are always higher impact than qualitative results.

Automate postprocessing and plot generation

When preparing results for dissemination, whether for a presentation, report, or paper, you will often find things that need to be changed and therefore have to regenerate figures and plots. This is a normal part of research. However, the work required to regenerate figures and plots can be significantly reduced using automated scripts or a well-defined and documented reproducible protocol. The goal is to reduce the amount of "manual" operations to a minimum.

For plots, data is often outputted from research codes and then plotted using software such as Excel, Python, Matlab, or GNUPlot. With Python, Matlab, and GNUPlot, plots can be generated and then edited manually or scripts can be used that define all aspects of the plot. The use of scripts may make the time it takes to create the plot the first time slightly longer but will reduce the time to regenerate the plots to almost zero. Scripts also ensure that your plots have a consistent look when regenerated. We recommend that you use tools that allow scripts; such tools include:

• Python, using matplotlib for plotting and pandas for csv reading
• MATLAB
• GNUPlot

For images, visualization applications such Paraview often have ways of saving a given configuration that can then be reloaded to quickly generate a similar image. In Paraview, this is done by saving and loading the state.

Recommended Practices for Papers

Provide all of the information needed for someone else to reproduce your results

A critical aspect of scientific publication is the reproduction of results by other researchers. For papers using the phase-field method, that means you need to provide enough information that the reader could reproduce your results. This information does not have to be provided in the main body of the paper, but can be in appendices or supplemental information. The provided information should include:

• The value used for every parameter in the model. Even if the values are given in one of your references, you should still provide the values in your paper.
• The domain dimension and size
• Initial conditions and boundary conditions
• Other conditions used for each simulation, including temperature and length of time.
• The discretization scheme, including types of spatial and temporal discretization and spatial discretization spacing and time step size.

Clearly state all assumptions

All models are generated by making assumptions and approximations. You should state every assumption and approximation made in the formulation of the model and why they were made. This includes any assumptions involved in equations you have taken from the literature. It is impossible for a reader to understand the strengths and weaknesses of your model without knowing the assumptions and approximations.

Prove that your new models solve correctly using simple problems

It is natural to present large simulation results that capture complex behaviors that occur in nature. However, it can often be difficult to assess how well a model is performing in such large simulation results since the behavior is so complex. Therefore, when presenting a new phase-field model, it is important to first show more simple problems that more clearly demonstrate that your model reproduces accurate behavior. Such simple problems may include model verification, where you compare against known analytical solutions (see the section on Verification here: Software Development).

Share your input file, output data, and/or codes

The best practice is to share key output data and input files where it is findable and accessible in a manner that the data is interoperable and reusable (FAIR); see Data Generation and Curation. For projects funded by federal agencies, your data management plan may also require you to publish your data associated with your publication. There are several public repositories to which your data/files can be uploaded and shared, which are described in Data Generation and Curation. The data should be accompanied with key metadata, including the information on the software used to generate the output and the associated publication. The publication should also be associated with the shared data via DOI. If the codes are not publicly available, it is also encouraged that they are shared as well.