This extension contains an app able to analyze images of transition metal dichalcogenide (TMD) monolayers and build 3D structures from them, which can then be used for further analysis, export, etc.

Setup

Configure the analysis in the first part of the interface (1 - Setup).

1. Click the 'Open image...' button to import an experimental image. Supported image types are bmp, gif (the first image), jpg, jpeg, png, pbm, pgm, ppm, xbm, xpm.

2. If preferable, invert the image using the 'Invert image' button.

3. Optionally, reduce noise in the image using the 'Smooth image' button.

4. Choose the transition metal type (Molybdenum or Tungsten)

5. Choose the chalcogen type (Selenium or Sulfur)

Training

Then, you must provide a few examples of atomic columns to the app.

1. Choose the column type in the combo box. There are three possibilities: transition metal, chalcogen (two atoms) and chalcogen (one atom).

2. Press, hold and release the left mouse button to draw individual columns:

   • Press the left mouse button to select the column center
   • Move the mouse to choose the atom radius
   • Release the left mouse button to create the column

You should provide at least one example of typical patterns found in the image. For example, if the image shows two main regions (e.g. two different surface orientations), you should provide at least two examples of columns per region (one for the transition metal, and one for the two-atom chalcogen column). Providing examples of columns at the interface of regions will help the detection algorithm too.

To delete columns, click them. To change a column type, press Alt and left click an existing column.

To navigate the image:

• Press Ctrl / Cmd and use the left mouse to pan the image.
• Press Ctrl / Cmd and use the mouse wheel to zoom and unzoom.

Building

Once a few training examples have been provided (you need at least one transition metal example and one double chalcogen example), you can start using the detection algorithm to automatically find the other columns.

1. Choose a detection threshold. It's preferable to start high enough and progressively lower the threshold to avoid false positives.

2. Choose whether you want to force alternating neighbor types (this is checked by default and should be checked in most cases).

3. Click the 'Find columns' button to automatically detect other columns in the image. The detection algorithm will use the example columns provided during the training phase.

4. If necessary, add more training examples to help the detection algorithm and change the detection threshold (the more examples you provide, the higher the correlation threshold can be), and go back to Step 3.

5. Once you're happy with the detection, click the 'Generate surface' button to create a 3D model.

About

This extension was developed through a collaboration between OneAngstrom and the Laboratory of Electron Microscopy and Material Advances (LEMMA) group at CEA headed by Dr. Hanako Okuno.