Installation

Packages

The following packages are required:
For atomic form factor calculation in GPAW
For atomic form factor calculation in Quantum Espresso (experimental)

  • qe ≥ 7.0

For difference electron density calculation

Windows

As the library relies on GPAW and JAX there is no direct straightforward way to install on Microsoft Windows Python installations. However, the library has been tested on the Windows Subsystem For Linux. This provides a full linux shell under windows and can be installed to run linux programs on Microsoft Windows.

Note

The very short instructions given here only work on a current version of Windows 10/11, which means newer than build 2004. For older versions refer to Microsofts Manual installations steps

Open a Windows PowerShell in Administrator mode. You can do this by typing PowerShell into your Windows Menu, right clicking on the entry and select “Run as Administrator”

In this console you can now type:

wsl --install

Afterwards, you probably need to restart your computer. Now open a new PowerShell and type wsl. If you now now have a linux console, everything works as intended and you can go to the Linux section.

Sometimes it has happened that the default ubuntu installation is not present and wsl does not work. Only in that case you need to type:

wsl --install -d Ubuntu

To start the installation of Ubuntu. Afterwards, start the wsl by typing the wsl command. You can now follow the Linux instructions from that shell.

You can go to you harddrive c: by typing.

cd /mnt/c

Linux

The easiest way to obtain all the necessary package is within a conda environment. So first you need to install Anaconda.

The fastest way from console is installation with the following commands:

wget https://repo.anaconda.com/archive/Anaconda3-2021.11-Linux-x86_64.sh
bash Anaconda3-2021.11-Linux-x86_64.sh
# Accept licence Agreement and run conda init
source ~/.bashrc
rm Anaconda3-2021.11-Linux-x86_64.sh

Note that we are playing it save for the instructions. Installation without the version numbers (i.e. using the newest version) should also work, at least for some time.

In Anaconda create and activate a new environment using

conda create -n xharpy
conda activate xharpy

If you want to get everything as quickly as possible just type:

conda install python=3.11 numpy scipy pandas jax cctbx ase gpaw qe

This will install the somewhat new versions which should work at the time of writing (16. August 2023). If it does not please raise an issue on Github and try the older explicit versions and instructions given below.

You can download the XHARPy library from: https://github.com/Niolon/XHARPy. You need to add the folder where your xharpy directory is located to the $PYTHONPATH. If you are unsure the directory should be one level above the one where the __init__.py is located, so if you have downloaded the complete repository it is the folder containing the docs, examples and xharpy folder. As long as there is no installation routine you can do this by:

conda develop /path/to/xharpy

Detailed installation with tested version numbers

You can now start by installing python in that environment

conda install -c conda-forge python=3.8

Followed by installing the necessary packages

conda install -c conda-forge jax=0.2.26 numpy=1.20.3 scipy=1.7.3 pandas=1.3.5

This will create a .pth file in the site-packages of your conda environment, which acts as the necessary link for conda/python.

Currently XHARPy has no working possibility for calculating difference electron densities on its own. For this purpose we need the cctbx library. The library will however run without cctbx, with some features unavailable.

conda install -c conda-forge cctbx=2021.11

If you want to use gpaw as source for the atomic form factors (recommended)

conda install -c conda-forge ase=3.22.1 gpaw=21.6.0

You can download the XHARPy library from: https://github.com/Niolon/XHARPy. You need to add the folder where your xharpy directory is located to the $PYTHONPATH. If you are unsure the directory should be one level above the one where the __init__.py is located, so if you have downloaded the complete repository it is the folder containing the docs, examples and xharpy folder. As long as there is no installation routine you can do this by: If you also want to try out the atomic form factor calculation in Quantum Espresso, you need to install the program. You can do this with conda. You can also install this separately of course, as long as pw.x and pp.x are directly callable.

conda install -c conda-forge qe=7.0

Finally, the examples are written as jupyter notebooks. It also has proven to be good practice to write the refinements in jupyter, as it is easy to further analyse the results. This is however not necessary.

conda install -c conda-forge jupyter jupyterlab

Using other functionals in GPAW

If installed via conda GPAW does bring the PAW setups for some functionals. If you wand to use metaGGA functionals further action is also not necessary, as these will use the PBE setups. For usage of GGA or LDA functionals not included follow these steps.

  1. Go to your GPAW path by typing into the console:

cd $GPAW_SETUP_PATH

  1. For all the main and transition group elements with the functional type xc in:

gpaw-setup -f *xc* H He Li Be B C N O F Ne Na Mg Al Si P S Cl Ar K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr Rb Sr Y Zr Nb Mo Ru Rh Pd Ag Cd In Sn Sb Te I Xe Cs Ba La Ce Hf Ta W Re Os Ir Pt Au Hg Tl Pb Bi

For more details consult the gpaw documentation