mirror of
https://github.com/QuantumPackage/qp2.git
synced 2024-11-07 05:53:37 +01:00
added source/programmers_guide/new_ks.rst
This commit is contained in:
parent
44b9318bae
commit
6ffdccb26d
@ -2,6 +2,138 @@
|
|||||||
Developping new functionals for KS or RS-DFT
|
Developping new functionals for KS or RS-DFT
|
||||||
============================================
|
============================================
|
||||||
|
|
||||||
The |QP| contains all the basic quantities to run |DFT| or |RS-DFT| calculations:
|
The very basics
|
||||||
|
===============
|
||||||
|
|
||||||
|
To develop new functionals for |DFT| (or |RSDFT|) to be used in the SCF programs (:ref:`ks_scf` or :ref:`rs_ks_scf` ) or in multi-configurational |RSDFT| (see `<https://gitlab.com/eginer/qp_plugins_eginer>`_), you need to specify, at the end of the day, two things:
|
||||||
|
|
||||||
|
* *exchange/correlation* **energy functionals** used to compute the **energy**
|
||||||
|
|
||||||
|
* *exchange/correlation* **potentials** for (alpha/beta electrons) used to **optimize the wave function**
|
||||||
|
|
||||||
|
|
||||||
|
The providers to define such quantities, and then used by all the all the |DFT| programs, are (see :ref:`module_dft_one_e`):
|
||||||
|
|
||||||
|
* :c:data:`energy_x` and :c:data:`energy_c` : the *exchange* and *correlation* energy functionals (see :file:`e_xc_genera l.irp.f`)
|
||||||
|
|
||||||
|
* :c:data:`potential_x_alpha_ao` and :c:data:`potential_x_beta_ao` : the exchange potential for alpha/beta electrons on the |AO| basis (se e :file:`pot_general.irp.f`)
|
||||||
|
|
||||||
|
* :c:data:`potential_c_alpha_ao` and :c:data:`potential_c_beta_ao` : the correlation potential for alpha/beta electrons on the |AO| basis ( see :file:`pot_general.irp.f`)
|
||||||
|
|
||||||
|
From the |AO| basis, the providers for the exchange/correlation potentials of alpha/beta electrons on the |MO| basis are automatically obtained by a |AO| --> |MO| transformation.
|
||||||
|
|
||||||
|
So, at the end of the day, adding a new functional consists only in **setting a new value to these providers**.
|
||||||
|
|
||||||
|
|
||||||
|
The general philosphy
|
||||||
|
---------------------
|
||||||
|
|
||||||
|
We have created a quite easy way to develop new functionals that is based on
|
||||||
|
|
||||||
|
* the used of **your own external plugins**
|
||||||
|
|
||||||
|
* the module :ref:`module_new_functionals` acting as **hub**
|
||||||
|
|
||||||
|
A pictorial representation of the main dependencies can be seen here:
|
||||||
|
|
||||||
|
.. image:: /_static/dependencies_func.pdf
|
||||||
|
:align: center
|
||||||
|
:width: 200px
|
||||||
|
:alt: Summary of dependencies
|
||||||
|
|
||||||
|
|
||||||
|
The main idea is the following:
|
||||||
|
|
||||||
|
1. Develop *new providers* for the new functionals (energy and potential for exchange and correlation) in some **external plugin**.
|
||||||
|
|
||||||
|
* Example:
|
||||||
|
|
||||||
|
* In an **external plugin** named **fancy_functionals**, you create *e_c_new_fancy_func* for the energy and *pot_ao_alpha_new_func* for the alpha potential
|
||||||
|
|
||||||
|
* If you want to be able to use the |DFT| programs already available in the |QP|, these *providers* must use the providers for the density defined in :ref:`module_density_for_dft` and :ref:`module_dft_utils_in_r` (see below).
|
||||||
|
|
||||||
|
|
||||||
|
2. Add the name of your **external plugin** to the :file:`NEED` in order to link your new providers to **new_functionals**
|
||||||
|
|
||||||
|
* Example:
|
||||||
|
|
||||||
|
* add **fancy_functionals** to the NEED file of **new_functionals**
|
||||||
|
|
||||||
|
3. Change the file :file:`e_xc_new_func.irp.f` and :file:`pot_xc_new_func.irp.f` to set the value of your new providers to the providers defined in **new_functionals**
|
||||||
|
|
||||||
|
* Example:
|
||||||
|
|
||||||
|
* for the exchange/correlation energy
|
||||||
|
|
||||||
|
|
||||||
|
.. code:: fortran
|
||||||
|
|
||||||
|
BEGIN_PROVIDER[double precision, energy_x_new_functional, (N_states) ]
|
||||||
|
&BEGIN_PROVIDER[double precision, energy_c_new_functional, (N_states) ]
|
||||||
|
implicit none
|
||||||
|
BEGIN_DOC
|
||||||
|
! energy_x_new_functional = define here your functional
|
||||||
|
! energy_c_new_functional = define here your functional
|
||||||
|
END_DOC
|
||||||
|
energy_c_new_functional = e_c_new_fancy_func
|
||||||
|
energy_x_new_functional = e_x_new_fancy_func
|
||||||
|
|
||||||
|
END_PROVIDER
|
||||||
|
|
||||||
|
|
||||||
|
4. Compile at the root of the |QP|
|
||||||
|
|
||||||
|
* Example:
|
||||||
|
|
||||||
|
|
||||||
|
.. code:: bash
|
||||||
|
|
||||||
|
cd ${QP_ROOT}
|
||||||
|
ninja
|
||||||
|
|
||||||
|
|
||||||
|
5. When you want to execute a program with your new functional, just set the options :option:`dft_keywords exchange_functional` and :option:`dft_keywords correlation_functional` to "my_functional".
|
||||||
|
|
||||||
|
|
||||||
|
Using the density for DFT calculations in the |QP|
|
||||||
|
==================================================
|
||||||
|
|
||||||
|
Different ways of defining the density for the DFT
|
||||||
|
--------------------------------------------------
|
||||||
|
|
||||||
|
There are many ways of defining a density, and the keyword to define it is :option:`density_for_dft density_for_dft`.
|
||||||
|
Here are the following options for that keyword:
|
||||||
|
|
||||||
|
* "KS" : density is obtained from **a single Slater determinant**
|
||||||
|
|
||||||
|
* "WFT" : density is obtained from **the wave function** which is stored in the |EZFIO| data base
|
||||||
|
|
||||||
|
* "input_density" : a one-body density matrix on the |MO| basis is read from the |EZFIO| data base, and the density is built from there (see :c:data:`data_one_e_dm_alpha_mo`)
|
||||||
|
|
||||||
|
* "damping_rs_dft" : damped density between "WFT" and "input_density" with the damping factor :option:`density_for_dft damping_for_rs_dft`.
|
||||||
|
|
||||||
|
.. note::
|
||||||
|
If an |MO| basis is already defined in the |EZFIO| data base, the one-body density matrices will be defined
|
||||||
|
according to this |MO| basis. For instance, if "KS", the density constructed will be density of a single Slater
|
||||||
|
determinant built with the current |MO| basis stored in the |EZFIO| data base.
|
||||||
|
|
||||||
|
Once that you have defined how to define the density, you can easily access to the providers associated to it.
|
||||||
|
|
||||||
|
|
||||||
|
Value of the density and its gradients in real space
|
||||||
|
----------------------------------------------------
|
||||||
|
|
||||||
|
The density and its gradients evaluated on all grid points are (see :ref:`module_dft_utils_in_r`):
|
||||||
|
|
||||||
|
* :c:data:`one_e_dm_alpha_at_r` and :c:data:`one_e_dm_beta_at_r` : alpha/beta density at grid points
|
||||||
|
|
||||||
|
* :c:data:`one_e_dm_and_grad_alpha_in_r`, :c:data:`one_e_dm_and_grad_beta_in_r`: alpha/beta gradients (and densities)
|
||||||
|
|
||||||
|
If you want to evaluate the density and its gradients at a given point in space, please refer to:
|
||||||
|
|
||||||
|
* :c:func:`density_and_grad_alpha_beta_and_all_aos_and_grad_aos_at_r`
|
||||||
|
|
||||||
|
If you use these *providers* and subroutines, the density computed will be coherent with the choice of density that you specified
|
||||||
|
with :option:`density_for_dft density_for_dft`, and it will impact automatically the general providers of :ref:`module_dft_one_e`.
|
||||||
|
|
||||||
|
|
||||||
* Handling
|
|
||||||
|
Loading…
Reference in New Issue
Block a user