StatMechGlass 0.1.5

Calculates structure of oxide glasses using statistical mechanics

1. Introduction

This python package uses statistical mechanics-based modeling to accurately predict the short-range order structural distribution in oxide glasses. To make the predictions, the model requires interaction enthalpies obtained by fitting to experimental data (typically obtained with NMR spectroscopy techniques).

The package is designed for building a library of enthalpies by providing relevant data to the package.

For a detailed guide, please refer to "the article" (will be filed when published). Any bugs or questions, do not hesitate to contact [e-mail] or [e-mail]

2. Package format

The basic format of the package follows:

StatMechGlass
.
├── Data
│   ├── SiB
│   │   ├── Na.csv
│   └── SiO2
│   ├── Na.csv
│   ├── Na_Tg.csv
├── Parameters
│   ├── MF
│   │   ├── BAl.csv
│   │   └── SiB.csv
│   └── SiO2
│   ├── K.csv
│   └── Na.csv
├── stat_mech_module
│   ├── __init__.py
│   ├── stat_mech_borate.py
│   ├── stat_mech_phosphate.py
│   └── stat_mech_silicate.py
├── __init__.py
├── stat_mech_glass.py

Here, experimentally obtained data should be placed in the /Data directory. The /parameter directory is where the package will automatically store the enthalpies obtained by fitting the provided data. Some enthalpies are provided already.

3. Usage

When using the package, either type commands in the stat_mech_glass.py file directly or import the package to a python script. It is advised to write code within if __name__ == '__main__'. Alternatively, place the package in the working directory and import it with

import StatMechGlass.stat_mech_glass as smg

or import with pip:

pip install StatMechGlass
from StatMechGlass import stat_mech_glass as smg

The four main functions:

smg.smg_binary_par(former, modifier, it=10, path_in=None)  
smg.smg_ternary_par(formers, modifier, it=10, path_in=None)  
smg.smg_structure(val, Tg, p = None)  
smg.smg_plot(comps, free_comp, Tg, plt_save = False)  

3.1 Fitting enthalpy parameters on binary oxide glasses

When building the enthalpy database, use smg.smg_binary_par:

1. Place data in the Data/"Former" directory, where "Former" corresponds to the network forming species of the glass. Currently supported formers:
   "SiO2", "B2O3", "P2O5"
   The data file should be named appropriately such as "Na.csv" or "K.csv"
   Refer to Section 4 for data file layout.

2. Place T_g data for the same glass system in the same directory with the name:
   "modifier"_Tg.csv
   Here, "modifier" should be the same as in 1. such as "Na_Tg.csv" or "K_Tg.csv" The T_g data does not need to be for the same glass compositions as the structural data

3. Execute the function

   smg.smg_binary_par(former, modifier, it=10)
   

   Example:

   smg.smg_binary_par("Si", "Na", it=500)  
   

   500 or more iterations are advised for accurate enthalpies (refer to the manuscript for more details)

path_in can be used to change the path for the input data (naming convention must still be followed)

3.2 Fitting interaction parameters on ternary oxide glasses

When building the former/former interaction database, use smg.smg_ternary_par:

1. Place data in the Data/"Former""Former" directory, where "Former" corresponds to the network forming or intermediate specie of the glass. Currently supported formers and intermediates:
   SiO₂, B₂O₃, P₂O₅, Al₂O₅
   Here, the folder must be named according to the naming convention (section 4):
   "SiB", "BP", "PSi", "AlB"
   The data file should be named appropriately such as "Na.csv" or "K.csv"

2. T_g data should be provided for each glass in the data file. Refer to section 4 for clarification on data file content

3. Execute the function

   smg.smg_ternary_par(formers, modifier, it=10)  
   

   Example:

   smg.smg_ternary_par(["Si", "B"], "Na", it=100)  
   

   100 or more iterations are advised for accurate parameter (refer to the manuscript for more details)

path_in can be used to change the path for the input data (naming convention must still be followed)

3.3 Predicting the structural distribution in a given composition

The number of compositions to be predicted by the model increases exponentially when building the enthalpy database. This is due to the tranferability of the interaction enthalpies. That is, enthalpies established from Na₂O-SiO₂ and Na₂O-B₂O₃ glasses may also be used to predict any Na₂O-B₂O₃-SiO₂ composition.

When using the model to predict structural distributions, use smg.smg_structure:

1. Define the glass composition using python directory
   Example:

   glass_comp = {"Si": 25, "B": 25, "Na": 25, "K":25}  
   

2. Run the function with a defined T_g Example:

   results = smg.smg_structure(glass_comp, 700)  
   

This way, users may easily build a database of structures from a large set of glass compositions

3.4 Simple 2D plotting

As glass compositions may consist of many different elements, smg.smg_structure can be used to return structures which the user may plot themselves. For simple visualization, the smg.smg_plot function may be used:

1. Define the glass composition using python directory
   Example:

   glass_comp = {"Si": 25, "B": 25, "Na": 0}  
   

   Alternatively, leave out the free component:

   glass_comp = {"Si": 25, "B": 25}  
   

2. Run the function with a defined T_g and free component
   Example:

   smg.smg_plot(glass_comp, "Na", 800, plt_save = True)  
   

   Set plt_save to True for saving the plot as .png file

This will make a plot of 25SiO₂-25B₂O₃ as a function of Na₂O content.

4. Naming convention

For the script to locate files, parameters and functions, a certain naming convention was introduced which must be followed:

4.1 Network formers and modifiers

When calling network formers in any function these are abbreviated to their basic atom:
"Si", "B", "Al", "P"
Example:
25SiO₂-25B₂O₃-50Na₂O should be {"Si": 25, "B": 25, "Na": 50}. Note that the glass contains 2 boron atoms for each silicon atom but in the naming convention they seem to contain the same number of atoms.

When using modifiers in the functions, these should be named according to the data files provided by the user. If the data file is named "Na.csv", "Na" should be used in the functions.

4.2 Datafiles

All data files should consist of numbers only. No letters

4.2.1 Binary oxide glass data

All binary oxide glass data files must contain the modifier concentration in the first column and the structure distributions in the following columns:

modifier_mol%,former_structure1,former_structure2,...former_structureN

Example for Na₂O-SiO₂ data:
Na₂O%,Si⁴%,Si³%,Si²%,Si¹%,Si⁰%
Example with numbers:

20,50,48,2,0,0
25,38,61,1,0,0
28.6,27,68,5,0,0
33.3,11,79,10,0,0

The structures reported depend on the network forming species:

SiO₂:
Si⁴,Si³,Si²,Si¹,Si⁰

B₂O₃:
B⁴

P₂O₅:
P³,P²,P¹,P⁰

Al₂O₃:
Al⁴,Al^5/6*

* Note that Al⁵ and Al⁶ species are combined.

The T_g file should contain:
modifier_mol%,T_g

4.2.2 Ternary oxide glass data

All ternary oxide glass data files must contain "modifier concentration, first former concentration, second former concentration, T_g" in columns 1, 2, 3, and 4, respectively:

modifier_mol%,former1_mol%,former2_mol%,T_g,former1_structures,former2_structures.

The structures provided should follow the convention explained in Section 4.2.1 and be in order of network forming species.
Example data file for smg.smg_ternary_par(["Si", "B"], "Na", it=100):
Na₂O%,SiO₂%,B₂O₃%,T_g,Si⁴,Si³,Si²,Si¹,Si⁰,B⁴

Example for smg.smg_ternary_par(["B", "Si"], "Na", it=100):
Na₂O%,B₂O₃%,SiO₂%,T_g,B⁴,Si⁴,Si³,Si²,Si¹,Si⁰

5. Final remarks

Thank you for using the software!
Upon building on the software or enthalpy database, you are encouraged to make a pull request by first making a seperate branch. Make sure to describe the change in detain and how you tested the change. Possible changes could include: adding new data or parameters to the database, building upon the modules to include more glass families, or general bug fixes/improvements.

Thanks,
Mikkel Bødker, [e-mail]
Oxide Glass Chemistry Group, Department of Chemistry and Bioscience, Aalborg University, Denmark