Creating custom order parameters¶
Often, you will find the need to create custom order parameters
for your path sampling simulation(s).
In PyRETIS, you can create new order parameters by making use
of the generic OrderParameter
class defined in the library.
Technically speaking, you will have to sub-class OrderParameter
and implement a method (like OrderParameter.calculate()
)
which actually calculates the order parameter.
In addition, you can define additional collective variables to
be calculated.
The short version of what we need to do in order to create a new order parameter is:
- Create a new Python class which is a sub-class of
OrderParameter
- Write a method to initialise this class, that is the
__init__
method. This method will be called when PyRETIS is setting up a new simulation and it will be fed variables from the PyRETIS input file. - Write a method to calculate the order parameter. This should take in a
System
object as its argument. - Setting up the input file to use the new order parameter.
Example: Creating a new order parameter¶
Let us see how this can be done in practice. To be concrete, we will consider an order parameter defined as follows: The distance between a particular particle and a plane positioned somewhere along the x-axis.
Step 1 and 2: Sub-classing OrderParameter¶
In order to define a new class to use with PyRETIS, we import
numpy and the OrderParameter
class:
import numpy as np
from pyretis.orderparameter import OrderParameter
And we define a class, representing our new order parameter:
class PlaneDistanceX(OrderParameter):
"""A positional order parameter.
This class defines a very simple order parameter which is
the distance from a plane for a given particle.
"""
def __init__(self, index, plane_position):
"""Initialise the order parameter.
Parameters
----------
index : integer
Selects the particle to use.
plane_position : float
The location of the plane, along the x-axis.
"""
txt = 'Distance from particle {} to the plane at {}'.format(
index,
plane_position)
super().__init__(description=txt)
self.index = index
self.plane_position = plane_position
Here, we are initialising the class by storing two variables index
and plane_position
which identifies the particle we will consider
and the location of the plane.
In addition, we add some more information when we initialise the parent class (the line
calling super().__init__
). This is simply following the convention
defined by OrderParameter
.
Warning
If the order parameter you are defining depends explicitly on the velocity
(i.e. the direction so that it is not symmetric under time reversal) you
should mark the order parameter as being velocity dependent.
This can be done by setting the velocity
argument to True
while initiating the order parameter:
super().__init__(description='Description', velocity=True)
Step 3: Creating a method for calculating the order parameter¶
Next, we will write a method for actually calculating the
order parameter. In order to do this, we need to know how we
can interact with the System
object, and this
might be a good point to read
the introduction to the API.
In particular, we will mostly be interacting with the
Particles class which we
can access using System.particles
.
Typically, our calculation of the order parameter will correspond to one of the two following cases:
We want to access the positions of the particles directly and calculate the order parameter. Here, just a new method to the
PlaneDistanceX
class we are creating as follows:def calculate(self, system): """Calculate the order parameter.""" pos = system.particles.pos[self.index] dist = -np.abs(pos[0] - self.plane_position) return [dist]
Here, we are making use of the index and position of the plane which we stored previously. Further, we are using the
System.particles
object in order to access the positions. Finally, we are returning the order parameter. In this particular case, we return the negative of the distance, so that the order parameter will increase when we are approaching the place.We are using an external engine (e.g. GROMACS) and we just want to reference the file containing the current configuration, and pass this file reference to another external library or program.
Here, we will make use of some external tools to obtain the order parameter, for simplicity, let us here make use of the mdtraj library. First, at the top of your file, add the import of this library:
Note: mdtraj needs to be installed on top of the regular PyRETIS installation.
In order to access the file containing the configuration, we make use of the
ParticlesExt.config
attribute. Add the following method to thePlaneDistanceX
class:""" txt = f'Distance from particle {index} to a plane at {plane_position}' super().__init__(description=txt) self.index = index self.plane_position = plane_position
Note that the method
some_function_to_obtain_order
has not been defined here and you will have to define his method yourself in order to complete the example.
Step 4: Making use of the new order parameter¶
We make use of the new order parameter by adding the Orderparameter section to the input file:
Orderparameter
--------------
class = PlaneDistanceX
module = orderparameter1.py
index = 0
plane_position = 1.0
As you can see, we specify the following:
- The name of the order parameter class.
- The name of the file where we have stored the new class.
- The value for the
index
parameter of the order parameter. - The value for the
plane_position
parameter of the order parameter.