Structure-Property-Visualizer

Use this app to generate interactive visualizations like these for atomic structures and their properties.

Discover's header

In this study > 66,000 hypothetical Covalent Organic Frameworks (COFs) are screened for their potential application in carbon capture and storage. A subset of ca. 800 COFs with the highest Henry coefficient for CO2 is further analyzed in detail, computing the full CO2 and N2 isotherm and the parasitic energy for the process. This figure presents all the results: note that only the values marked as "high-throughput" (HT) are available for the whole dataset, while the others are computed only for the smallest dataset, selected because of the CO2 Henry coefficient > 0.0001 mol/kg/Pa.

Inner working

There are two kind of groups:

• group_ht_{COF-name} for ca. 66,800 COFs computed with HT approach
• group_pe_{COF-name} for ca. 800 COFs for which PE is computed The nodes of these group have a tag which is stored as the extra with key 'group_tag'. The first group, group_ht_{COF-name}, contains the tags:

'orig_cif'    # CIF input of the HT work chain
'ht_wc'       # High-throughput work chain
'ht_geom_out' # Zeo++'s output Dict
'ht_kh_out1'  # Raspa's output Dict with global info
'ht_kh_out2'  # Raspa's output Dict with component (CO2) info

While the seconds, group_pe_{COF-name}, contains the tags:

'orig_cif'     # CIF input of both CO2 and N2 VolpoKhIsotherm work chain
'isot_co2_wc'  # VolpoKhIsotherm work chain for CO2
'isot_n2_wc'   # VolpoKhIsotherm work chain for N2
'isot_co2_out' # VolpoKhIsotherm's output Dict for CO2
'isot_n2_out'  # VolpoKhIsotherm's output Dict for N2
'pe_out'       # output Dict from calc_PE
'ht_wc'        # (same as the ht group)
'ht_geom_out'  # (same as the ht group)
'ht_kh_out1'   # (same as the ht group)
'ht_kh_out2'   # (same as the ht group)

Re-implementation based on Panel

Use as jupyter notebook:

jupyter notebook
# open figure/main.ipynb

Use with panel:

panel serve detail/ figure/

Features

• interactive scatter plots via bokeh server
• interactive structure visualization via jsmol
• simple input: provide CIF/XYZ files with structures and CSV file with their properties
• simple deployment on materialscloud.org through Docker containers
• driven by database backend:
  1. sqlite database (default)
  2. AiiDA database backend (less tested)

Getting started

Prerequisites

• git
• python >= 2.7
• nodejs >= 6.10

Installation

git clone https://github.com/materialscloud-org/structure-property-visualizer.git
cd structure-property-visualizer
pip install -e .     # install python dependencies
./prepare.sh         # download test data (run only once)

Running the app

bokeh serve --show figure # run app

Customizing the app

Input data

• a set of structures in data/structures
  • Allowed file extensions: cif, xyz
• a CSV file data/properties.csv with their properties
  • has a column name whose value <name> links each row to a file in structures/<name>.<extension>.
• adapt import_db.py accordingly and run it to create the database

Plots

The plots can be configured using a few YAML files in figure/static:

• columns.yml: defines metadata for columns of CSV file
• filters.yml: defines filters available in plot
• presets.yml: defines presets for axis + filter settings

Docker deployment

pip install -e .
./prepare.sh
docker-compose build
docker-compose up
# open http://localhost:3245/cofs/select-figure