CAES/INL C3 Summer Boot Camp

Sponsored by Center For Advanced Energy Studies

Workshop Intro & Ab Initio Modeling

Lan Li
Associate Professor, Micron School of Materials Science & Engineering,
Boise State University
CAES Fellow, Center for Advanced Energy Studies

A C O L L A B O R AT I O N B E T W E E N
CAES/INL C3 Summer Boot Camp
Sponsored by Center For Advanced Energy Studies
• Schedule for next three days
• Create a nanoHUB account here: The computational
modeling and data science tools the workshops will
demonstrate are available on the nanoHUB.
• Useful info
• Post survey, please fill out after you attend all the workshops
you have registered.
Check on Summer Boot Camp Website for Details!
https://caesenergy.org/remote-summer-boot-camp/

A C O L L A B O R AT I O N B E T W E E N
Different Computational Modeling Methods
• Quantum Mechanics (QM)
• Ab-initio / first-principles simulation
• Electronic degrees of freedom, interactions,
bonding, energy estimation
• Schrödinger equation, wave functions
• Molecular Dynamics (MD)
• Atomistic modeling
• F​ orces between the particles and potential energy
defined by molecular mechanics force fields.
• Mesoscale to Continuum
• Phase field simulation: Thermodynamics-based
approach to model phase changes and evolving
microstructures in materials
• Finite element approach: Partial differential equation
solver for use to simulate structural responses.
• Most of physical properties are predictable based on the quantum
mechanics laws.
• Basic ingredients: nuclei and electrons.
• Schrödinger equation:
^
𝐻  ( 𝑟 𝑖 , 𝑟 𝐼 , 𝑡 ) =𝐸  ( 𝑟 𝑖 ,𝑟 𝐼 ,𝑡 )

2 2 Ζ e2 1 e 2 1 Ζ Ζ e2
Η     Ι     Ι J
2m ι r R 2ι  j r r 2I  J Ρ Ρ
 ι  e   ι,Ι
        ι   j      I  J
ι Ι
Τ el V Vion ion
el  ion V
el  el

Tel: Kinetic energy of electrons

Vel-ion: Electron-ion interactions
Vel-el: Electron-electron interactions
Vion-ion: Ion-ion interactions
• The price to solve total Hamiltonian for the interacting electron-ion system
is a high computational cost.
 Modeling Solid-State Materials  Density Functional Theory (DFT)

• Solving the Schrӧdinger wave equation only

possible for H or He (up to 2 atoms)
– Cannot treat all “bodies” interacting at same time

Density Functional Theory (DFT)

• Turn “many-body” system into non-
interacting system under external potential
• Determine ground state ion and electron
configuration
• All energy terms written as a function of
electron density

Schrӧdinger’s Equation
^
𝐻 𝐾𝑆 ( 𝒓 ) Ψ ( 𝒓 ) =𝐸 ( 𝒓 ) Ψ ( 𝒓 ) Ion Potential

^ 𝑛𝑜𝑛
𝐻 𝐾𝑆=𝐸 𝑘𝑖𝑛 +𝑈 𝑒𝑥𝑡 +𝑈 𝐻 +𝑈 𝑥𝑐 Exchange
Correlation
Potential
*June Gunn Lee, Computational Materials Science Kinetic
An Introduction (2nd Edition), CRC Press Taylor & Hartree (Mean-Field)
Francis Group, Boca Raton, FL, USA 2017 Energy
Potential
 Exchange-Correlational Functional
• Local density approximation (LDA) – Use only local density to define the
approximate exchange-correlation functional
• Generalized gradient approximation (GGA) – Use local electron density and
local gradient in the electron density

*June Gunn Lee, Computational Materials Science An Introduction (2 nd Edition), CRC Press Taylor & Francis Group, Boca Raton,
FL, USA 2017
6
Key DFT Parameters To Be Determined
• Atomic coordinates of the structure you would want to simulate
• Exchange-correlation (XC) functional - a correction term for electron-
electron interaction, LDA or GGA
• Basis set – a set of basis functions (e.g. Gaussian basis functions,
plane wave expansion) to represent electronic wave function
• Mesh cutoff – an energy corresponds to a finite 3D grid for the
calculation of electron density ρ(𝑟) and potentials V(r).

 𝑟 =∑ ¿ 𝑖 𝑟 ∨¿ ¿
( ) ( ) 2

• k-point density sampling in the first Brillouin zone of the material

Ab Initio Software
• 2020 June Bootcamp Website
The computational modeling and data science tools the workshops will
demonstrate are available on the nanoHUB.
• Ab Initio Software available on nanoHUB
• SIESTA (Spanish Initiative for Electronic Structure Simulations with
Thousands of Atoms): total energy calculation, geometry optimization,
band structure calculation, and quantum molecular dynamics
• What we simulate today:
– Solid: Cu crystal structure
– Electronic band structure
– Water dissociation
• References: Check on 2020 June Bootcamp Website
• Use “MIT Atomic Scale Modeling Toolkit” to run SIESTA
Cu Crystal Structure
What to simulate:
• Simple cubic, FCC & BCC crystal structures of Cu

• Which one is more stable / energetically favorable through calculating total

energy/Cu atom?

Lattice Lattice Atomic Run type XC Basis set Mesh k-point Energy/Cu
type constant coordinates functional cutoff density (eV)
Simple 2.39 Ang 000 Geometry GGA/PBE DZP 100 4 -1197.9752
cubic Optimization
FCC 3.60 Ang 000 Geometry GGA/PBE DZP 100 4 -1198.3846
Optimization
BCC 2.89 Ang 000 Geometry GGA/PBE DZP 100 4 -1198.3057
Optimization
Overview of Input Parameters in SIESTA
Parameter Science What we use

XC Functional Simplify n-electron (interacting) GGA/PBE

system to a 1-electron (non-
interacting) system plus an
external energy

Basis Set Determines wavefunction Ψ DZP

SIESTA uses linear combination of
atomic orbital basis sets
Consider radial and angular
flexibility of atomic orbital
Mesh Cutoff The finite 3D grid of electron 100
density and potentials
Larger cutoff ⇒ smaller
spacing ⇒ more dense 3D
mesh
GaAs Band Structure
SIESTA Inputs
• XC Functional: GGA
• Run type: Geometry Optimization
• Basis set: DZP,
As
• Type of System: Solid
• Lattice type: face-centered cubic Ga
• k-point density: 8
• Mesh cutoff: 300 Ry

Questions: Analyze GaAs band structure, and determine

– What is the band gap size?
– Is it a direct or indirect band gap?

11
Key Note for k-Point Sample
• Crystallography and reciprocal space
• Generate a grid of k-points spaced evenly throughout the Brillouin Zone (or along each reciprocal
lattice vector).
• Any crystal structure can be represented in two spaces: real space (primitive cell or unit cell in x, y, z)
and reciprocal space (Brillouin Zone in kx, ky, kz).
• Many physical properties of crystals as well as the geometry of the three-dimensional patterns
resulting from a diffraction event (e.g. X-ray diffraction) are represented using the reciprocal lattice.
• Diffraction occurs in inverse proportion to the spacing between objects causing diffraction.
• Any point in the Brillouin Zone can represent a k-point, there are an infinite number of discrete k-
vectors well qualified to be a wave function.
• Specify high-symmetry point positions in the Irreducible Brillouin zone.
• The wave function and other properties (e.g. band structure) vary smoothly over the Brillouin Zone,
so we can just sample a finite number of k-points (i.e. high symmetry points) that represent each
small region.
• Each k-point contains rich information: Wave length (λ=2𝜋/𝑘), and kinetic energy ( 𝐸=); All the
incoming energies on each k-point, it will form band structure.
• More k points lead to a more accurate simulation, but it costs more computational time.
• Smaller unit cell, high k-point density.

12
Modeling of Water Dissociation
• Predict how difficult to dissociate a H atom from a H2O molecule
• Manually increase the distance between H and OH
• Record total energy step by step and then plot energy as a function of
distance
• Energy difference will result in dissociation energy
• Initial O-H bond length = 0.97 Ang (Literature)
• Angle H-O-H = 104.50 (Literature)

H H H H H
O O O O O
H
H
H

H
Key Input Parameters
• XC Functional: GGA
• Run type: Total Energy
• Basis set: DZP
• Type of System: Molecule (Cartesian Coordinates required, be careful the format)
• Coordinates (Cartesian): Upload files provided on the nanoHUB website
• Mesh cutoff: 100 Ry
• No k point density needs to be setup due to a molecule structure.
• Write down total energy and check on the optimized HO-H bond length.

HO-H 0.75 0.88 0.97 2.00 3.00 5.00 10.00 15.00

Distance
(Ang)
Total
Energy
(eV)