Magnetic Materials Discovery Demo

The Magnetic Materials Discovery App enables users to automatically view the magnetic phase diagram for a series of materials and make materials predictions.

In this example we will attempt plot the Néel phase of the Lanthanide-Chromite series LNCrO3.

The Home Page

On the home page, use the interactive periodic table to select elements of interest.

Next, specify the element compositions and how to combine the elements into a series. The AND rules specify that an element or series must be present, whereas the OR rules will be combined to form new series.

The Plot Interface

The plotting system makes use of the Plot.ly library. Once loaded, the plot will automatically show the Néel phase for the selected compounds versus their Hill formula. Use the plot settings tab to manipulate the graph and plot the phases against different elemental and materials properties. Here, we plot the Néel phase against the Pauling Electronegativity of the LN ion. We can also hover over the data points to see the sources from which the data were derived.

Property Prediction

We notice that the Ce-based compound, CeCrO3, has no reported Néel transition in the database. Therefore, we will try and predict the transition temperature using the rest of the data

To do this, we access the property prediction tab and enter the compound in its Hill formula, and the desired phases to predict.

Prediction can be performed using four methods:

  • Ridge Regression
  • Support Vector Regression (SVR)
  • Automatic Relevance Determination (ARD)
  • Random Forest Regression

The prediction methods are as implemented in SciKit-learn. The user can also specify a feature-selection method from:

  • Manual
  • K-Best
  • Recursive Feature Elimination

Once submitted, the prediction results will appear in the table, and they can then be displayed on the chart.

Using ARD, with Recursive Feature Elimination yields a predicted Néel temperature of ~260 K.

We can then display this result on the chart and hover to see which features were chosen as best predictors of Néel temperature.

Data Tables

If you wish to see the data that are being plotted, the data tables tab provides tables of the elemental properties, ionic radii and bulk compound Properties, along with their associated source. The Structure data table also provides a reference as to where to find crystallographic information files for the associated compounds from the Crystallography Open Database. All these data can also be downloaded from the Downloads Page.

Associated Data

Once you are happy with your predictions, you can use the DataCite MetaData Search Tool to explore experimental and computational data associated with your chosen compound. This is useful for backward validating your predictions and getting in touch with other researchers that have synthesised or investigated the compound.