blaugroup / nanoparticletools Goto Github PK

In the analysis, the occurrences_per_atom is incorrectly calculated. It should be calculated without division by simulation time.

In the meantime, this value can be easily obtained by multiplying the dndt_per_atom by simulation_time

Some dopants (such as Er) have many more energy levels than others. It is unlikely that we get much additional accuracy from including these extra energy levels into the simulation.

Proposed changes:

1. Add the ability to limit the energy levels by energy value.
2. Add the ability to limit the max energy of dopant states in the SpectralKinetics module.

Simulation setup may fail in some instances:

1. If there are no dopants specified in any part of the nanoparticle or all are zero
2. If the total concentration within a layer is (nearly) 100%, floating point arithmetic errors will cause an error. While erroring when greater than full doping is intended, there should be a small tolerance, where the concentrations can be made to sum to 1.
   • For example, if a core only particle is specified with Yb concentration 1/3 and Er concentration 2/3, the sum may end up being slightly greater than 1

There is an issue with Tb, where entries in the magnetic dipole matrix become NaN. To the best of my knowledge, this comes from Tb's slj values.

Code to reproduce:

from NanoParticleTools.inputs.spectral_kinetics import SpectralKinetics
from NanoParticleTools.species_data import Dopant
sk = SpectralKinetics([Dopant('Tb', 0.1, 3)])
sk.magnetic_dipole_rate_matrix

Output:

array([[0.        , 0.        , 0.        ],
       [       nan, 0.        , 0.        ],
       [0.        , 1.20341175, 0.        ]])

There are 2 possible implementations for this

1. Use an exponential moving average. This is the cheap way to do things, but it would be less accurate.
2. Compute the time averaged population

Examples should cover:

1. Solving rate equations to get spectra
2. Setting up HT campaign
3. Training ML model

The Igor code included differential kinetics modules which solve for dNdt through solving differential equations rather than running a Monte Carlo Simulation.

Some starter code has already been added to the SpectralKinetics class.

Currently, there is no easy way to track or filter documents by batch or metadata. Therefore, it is difficult to query for documents based on a specific experiment or project you may be working on.