How to select the right calculator¶
A calculator is the engine that performs the atomic simulations, defining how we describe the interactions of the electronic system. It both represents a level of theory (full quantum mechanical approach, semi empirical approach, or classical approach) and algorithmic settings.
In this tutorial, you will receive a brief overview of the different calculators and learn to select the calculator that best fits your needs in QuantumATK NanoLab.
Table of Contents
The Calculator types¶
When setting up a calculation in the Workflows tool (
) you
must select a calculator to perform the atomic scale modelling.
Quantum Mechanical (DFT) Calculators |
|
The two quantum mechanical calculators both model the electronic properties within the framework of density functional theory (DFT). These DFT calculators provide the most accurate results, but require the largest computational resources. It is recommended to use these when the accuracy of the results is of foremost concern. |
|
Semi Empirical Calculator |
|
The semi empirical calculator is also based on DFT, but it approximates the electron-electron interactions, using experimental data. Due to these approximations the semi empirical calculator is orders of magnitudes faster than the quantum mechanical calculators, but the accuracy is correspondingly reduced. The semi empirical calculator is often used when qualitative results will suffice, or when a DFT calculator cannot be used. |
|
The Classical Calculator |
|
The force field calculator, is based on classical mechanics and estimate the forces between atoms using a range of element-pair specific potentials. The force field calculator is the fastest calculator, orders of magnitudes faster than the semi empirical calculator, and is often the calculator of choice when dealing with huge systems (many thousands of atoms) or when modelling the full dynamic evolution of the system on longer time frames (molecular dynamics). |
The DFT Calculators (LCAO and Plane Wave)¶
Both the plane wave calculator and the LCAO calculator solve the time independent Kohn–Sham equations. The main difference is that the LCAO calculator employs numerical LCAO (Linear Combination of Atomic Orbitals) basis sets, while the plane wave calculator employs a PW (Plane Wave) basis set.
Neither of the two DFT calculators require any system specific input, and can in general be expected to produce results of similar quality, independent of the system. A basis set is required for each element species in the system under investigation (for plane wave, the basis set is only used for initialization and projections). Nearly all basis set provide support for all elements, with only a few exceptions.
The Plane Wave Calculator¶
The plane wave bases incorporate the periodicity of the bulk system into the basis set, which allows a higher accuracy to be obtained. The plane wave calculator is the state of the art calculator, though it is often slow and can only be applied to bulk systems. Introducing the QuantumATK plane-wave DFT calculator
See DFT: Plane Wave for more details.
The plane wave calculator is suitable for treating system sizes on the order of hundreds of atoms.
The LCAO Calculator¶
In general, it is recommended to use the LCAO calculator, which can work with both molecules and bulk systems. In most cases, an accuracy close to the plane wave calculator can be achieved at significantly reduced computational effort. The accuracy of the LCAO calculator can be improved even more by using a larger basis set, at the cost of longer computation times. See DFT: LCAO for more details.
The LCAO calculator is suitable for treating system sizes on the order of thousands of atoms.
Note
Computational effort for the DFT calculators depends heavily on the exchange correlation functional and calculator settings.
The Semi Empirical Calculator¶
The semi empirical calculator requires a parameter set.
QuantumATK NanoLab includes a selection of parameter sets designed for a range of purposes. However, each parameter set is only constructed for a subset of elements on the periodic table. Thus, it may be that there is no parameter set which provides support for all elements in the system under investigation.
The semi empirical calculator is suitable for treating system sizes on the order of thousands of atoms.
The is the calculator of choice when performing large scale NEGF calculations (NEGF: Device Calculators, see Transport calculations with QuantumATK)
See Semi Empirical for more details.
The Force Field Calculator¶
The force field calculator requires a potential set, comprised of potential components, which defines the types of atomic interactions to be modeled between different atomic element pairs. Consequentially, the potential set must include support for all elements in the system under investigation.
QuantumATK NanoLab includes a selection of potential sets designed for a range of purposes (See Force Field for more details.). However, each potential set is only constructed for a subset of elements on the periodic table. Thus, it may be that there is no potential set which provides support for all elements in the system under investigation. In such a case, there is an advance feature to create a custom potential set, which must be manually configured to handle all relevant atomic interactions between all relevant atomic element pairs.
The force field calculator is suitable for treating system sizes on the order of tens of thousands of atoms.
See Force Field for more details.
Bonding and non-bonding potential sets¶
Roughly speaking, potential sets come in two flavors: rigid and flexible.
Rigid potential sets do not handle molecular bonding, and are thus geared towards modelling non-reactive interactions such as angle-strain (represented by harmonic potentials) or dispersion (represented by van der Waals potentials).
To model chemical reactions, a flexible potential set should be used instead. Flexible potential sets can model the formation and breaking of bonds, allowing the calculator to reproduce reaction chemistry without explicit quantum mechanical considerations.
Machine Learned Potentials¶
MTPs (Moment Tensor Potentials) are machine-learned potentials that allow dynamics modelling of arbitrary materials systems. The accuracy of the results, as well as what scenarios can be modeled, are dependent on the training data used to produce the MTP. MTPs are suitable to use when large systems (e.g., amorphous materials or multi-layer stacks) need to modeled with near-ab initio accuracy.