Disentangling the origin of chemical differences using GHOST
qmms_wines.pptx
1. Utilizing the JARVIS Infrastructure to
Discover and Accurately Characterize Next-
generation Quantum Materials
1/31/2023
Daniel Wines
NRC Postdoctoral Associate
NIST, Materials Science and Engineering Division
Joint Automated Repository for Various Integrated
Simulations
https://jarvis.nist.gov/
2. Outline
• Introduction
• JARVIS-DFT
• Bulk Superconductors
• 2D Superconductors
• Topological Materials
• JARVIS-QMC
• Motivation and Background
• 2D CrX3 Magnets
• Conclusions and Outlook
3. Acknowledgement and Collaboration
3
A. Biacchi
(NIST)
D. Wines
(NIST)
R. Gurunathan
(NIST)
B. DeCost
(NIST)
Bobby sumpter
(ORNL)
A. Agarwal
(Northwestern
University)
S. Kalidindi
(GAtech)
A. Reid
(NIST)
Ruth Pachter
(AFRL)
Karen Sauer
(George Mason University)
K. Garrity
(NIST)
David Vanderbilt
(Rutgers University)
Sergei Kalinin
(ORNL)
F. Tavazza
(NIST)
K. Choudhary
(NIST)
4. User-comments:
• “There are many different theoretical levels on which you can
approach the field. JARVIS is unusual in that it spans more
levels than other databases.”
• “A pure gold-mine for the data-quality effort…”
• “You guys are doing something really beneficial…”
• “I find JARVIS-DFT very useful for my research…”
Databases, Tools, Events, Outreach
https://jarvis.nist.gov
Established: January 2017
Published: >40 articles
Users: >10000+ users worldwide
Downloads: >500K
Events:
• Quantum Matters in Materials Science (QMMS)
• Artificial Intelligence for Materials Science (AIMS)
• JARVIS-School
Requires login credentials, free registration
Choudhary et al., npj Computational Materials 6, 173 (2020).
GitHub:
Notebooks:
Docs:
7. JARVIS-DFT
Motivation: Functional and structural materials design using quantum mechanical methods
~70,000 materials, millions of calculated properties, compared with experiments if possible
https://jarvis.nist.gov/jarvisdft/
K. Choudhary, K. Garrity, et al. npj Comp. Mater. 6 173
(2020): https://doi.org/10.1038/s41524-020-00440-1
8. Efficient energy conversion
Superconductivity
Nano Lett. 13, 3664–3670 (2013)
2D Transistors
• LEDs
• Flexible
electronics
https://www.nextplatform.com/2019/09/13/tsmc-thinks-it-can-uphold-moores-law-for-decades/
Nano Lett. 21, 3435 - 3442 (2021)
2D Magnets
• Spintronics
• Magnetic
Storage
J. Chem. Phys. 156, 014707 (2022)
Next Generation Materials
https://phys.org/news/2014-01-quantum-natural-3d-counterpart-graphene.html
9. • Need for High-TC, Ambient condition superconductors +large dataset to choose from
• Experimental datasets (NIMS-SuperCon) contains chemical formula only
• Expensive experiments as well as computation
• Need for High-throughput computation workflow-DFT
• Verify candidates with fast experimental techniques
Superconductors: Materials to conduct electricity without energy loss when they are cooled below a critical temperature, TC
MgB2 (TC = 39 K): Highest TC ambient condition conventional superconductor
https://doi.org/10.1016/j.isci.2021.102541
https://en.wikipedia.org/
Nobel prizes:
1913, 1972,
1973, 1987,
2003
Superconductors
10. JARVIS: Superconductors & E-Ph coupling
Eliashberg
spectral function
Electron-phonon
coupling (EPC)
Effective Coulomb
potential (empirical),
taken as 0.1
McMillan-Allen-Dynes Eq.
• EPC derived from Eliashberg spectral
function
• Obtained from DFPT calculations
• Interpolation method used (broadening
converged)
• PBEsol and GBRV pseudopotentials
Choudhary et al., npj Computational Materials, 8, 244 (2022)
12. JARVIS: Bulk Superconductors
• Benchmarked against several well-known
superconductors (experiment and theory)
• Revealed previously undiscovered
superconductors: h-MoN, h-ZrN, LaN2, and
several others
Choudhary et al., npj Computational Materials, 8, 244 (2022)
13. JARVIS: 2D Superconductors
• 2D superconductivity is emerging field,
very few materials with high Tc
• Computational screening is necessary
precursor to experimental investigation
• Utilized JARVIS framework to screen new
2D superconductors, modified criteria
Wines et al., Nano Letters, 10.1021/acs.nanolett.2c04420 (2023)
14. JARVIS: 2D Superconductors
• Distribution of
EPC data for 2D
superconductors
• Experimental verification of
selected commercially
available materials
(magnetometry)
Wines et al., Nano Letters, 10.1021/acs.nanolett.2c04420 (2023)
Tc, exp = 8.3 K
Tc, DFT = 6.4 K
Tc, exp = 7.1 K
Tc, DFT = 9.3 K
15. Spin-orbit coupling & Topological Materials
New class of materials
(electronic bandgap perspective)
https://phys.org/news/2014-01-quantum-natural-3d-counterpart-graphene.html
https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSzMKD5ICIkR9neJRre3prqIjp_iqLMu6TQp7mXKJqmmh-HqjFB
(2016 Nobel prize)
Metal
Semiconductor
Insulator
16. Spin-orbit Spillage
• Majority of the topological materials driven by spin-orbit coupling (SOC)
• Simple idea: Compare wavefunctions of a material with and without SOC?
• Spillage initially proposed for insulators only, now extended to metals also
• For trivial materials, spillage 0.0, non-trivial materials ≥ 0.25
16
https://www.ctcms.nist.gov/~knc6/jsmol/JVASP-1067
𝜂 𝐤 = 𝑛𝑜𝑐𝑐(𝐤) − Tr 𝑃𝑃 ; 𝑃 𝐤 =
𝑛=1
)
𝑛𝑜𝑐𝑐(𝐤
|𝜓𝑛𝐤 𝜓𝑛𝐤|
Sci. Rep., 9, 8534 (2019)
NPJ Comp. Mat., 6, 49 (2020)
Phys Rev B, 103, 054602 (2021)
17. Spin-orbit coupling & Topological Materials
• A number of high-spillage
materials have been
verified experimentally to
be topological insulators
Nature Materials 21, 1111–1115 (2022)
Choudhary, et al. Phys. Rev. B 103, 155131 (2021)
18. DFT: Success and Limitations
• Results depend directly on which XC
functional is used
• van der Waals interactions (corrections)
• Systems with strongly localized and
correlated electrons (DFT+U)
• Band gaps (underestimated)
Proposed Solutions
• Post DFT methods (many-body
perturbation theory)
• Stochastic methods (Quantum
Monte Carlo)
• Reduces 3N-dimensional problem to 3
• Good balance between computational
efficiency and accuracy
DFT Successes
DFT Shortcomings
19. Computational Metrology: Quantum Monte Carlo
• A class of algorithms that apply MC integration to solve
quantum problems (many-body)
• Variational MC (VMC) and Diffusion MC (DMC) are most
common for studying crystals
• Scales ~Ne
3 (similar to DFT), accuracy beyond DFT
• Current state of the art software: QMCPACK
20. QMC: Variational MC WF Optimization
From DFT For correlation
SLATER JASTROW
x x x
Trial Wavefunction
• Types of Jastrow factors:
• Electron-electron
• Electron-nucleus
• Electron-electron-nucleus
• Slater determinant from DFT and Jastrow
factor has some functional form and
recovers correlation energy
• Parameters of the Jastrow are optimized
with VMC before DMC
• Jastrow optimization decreases error in
DMC
J. Chem. Phys. 146, 244101 (2017)
21. QMC: Diffusion MC
Diffusion Monte Carlo (DMC)
Diffusion of walkers in imaginary time
Imaginary-time Schrödinger Eq.
Fixed-nodal surface
• DMC: Simulate diffusion of walkers
in imaginary-time until you reach
steady state
• Timestep errors
• Finite size errors
Rev. Mod. Phys., 73, 1, (2002) Rev. Mod. Phys., 73, 1, (2002)
23. JARVIS-QMC: 2D CrX3 Magnets
• Case study of 2D correlated magnets
with CrX3 stoichiometry
• QMC added to JARVIS framework
24. JARVIS-QMC: 2D CrX3 Magnets
Wines, et al. J. Phys. Chem. C 127, 2, 1176-1188 (2023)
2D Model Spin Hamiltonian:
J Isotropic Heisenberg Exchange
D Easy Axis Single Ion Anisotropy
λ Anisotropic Symmetric Exchange
*Tc (Curie Temperature)
estimated by method of
Torelli and Olsen
2D Materials, 6, 015028 (2019)
Strong variability
in DFT results
25. JARVIS-QMC: 2D CrX3 Magnets
• Optimal trial WF can be created
by tuning U parameter
• U = 2 eV variationally yields
optimal WF
Wines, et al. J. Phys. Chem. C 127, 2, 1176-1188 (2023)
26. JARVIS-QMC: 2D CrX3 Magnets
• Accurate statistical bound on magnetic
exchange and Curie Temperature
• Maximum Tc: 43.56 K for CrI3 and
20.78 K for CrBr3
• Less dependence on starting functional
and Hubbard (U) parameter
• Same workflow can be applied to other
2D ferromagnets
• Goal: JARVIS-QMC database
Wines, et al. J. Phys. Chem. C 127, 2, 1176-1188 (2023)
27. JARVIS-QMC: 2D CrX3 Magnets
• Can obtain accurate estimates
for spin density and magnetic
moment with DMC
Wines, et al. J. Phys. Chem. C 127, 2, 1176-1188 (2023)
29. Conclusions and Outlook
• JARVIS-DFT framework can be
used to screen exotic next
generation materials:
o Superconductors, topological
insulators, magnets
• When DFT yields inconclusive
results, QMC methods can be used
for higher accuracy
• These open access tools and
datasets are intended to benefit
materials science community
30. Resources
• NIST-JARVIS Infrastructure:
• Databases:
• DFT, Classical Forcefield, Tight-binding, Experimental …
• Coming Soon: QMC database
• Tools:
• ALIGNN, Quantum computation, high-throughput DFT …
• Events! Conferences and JARVIS Schools
Email: daniel.wines@nist.gov , ramya.gurunathan@nist.gov,
kamal.choudhary@nist.gov, francesca.tavazza@nist.gov
Slides: https://www.slideshare.net/
Website:
https://jarvis.nist.gov/
GitHub:
https://github.com/usnistgov/jarvis
https://github.com/usnistgov/alignn
https://github.com/usnistgov/atomvision
https://github.com/usnistgov/chemnlp
https://github.com/usnistgov/atomqc
Artificial Intelligence for Materials Science
Summer, 2023
Invited speakers from academia, industry, and
government + contributed talks
https://jarvis.nist.gov/events/aims
NRC Postdoc Opportunities:
Many project opportunities for recent PhDs
interested in quantum materials, machine
learning, computation, and materials design.
Notas do Editor
Hello everyone, my name is Daniel Wines and I am a postdoc at NIST.
These are the primary contributors and collaborators to JARVIS.
What are 2d materials?
Crystalline materials consisting of a single layer of atoms
These materials exhibit interesting properties, often much different than their bulk counterparts
Of course, Graphene was one of first, started 2d revolution
So why should we care about 2D materials?
-Since they are so different from their bulk counterparts they have interesting properties that we can utilize for applications such as transistors and electronics, energy conversion and H2 generation
-Also in accordance with Moore’s law
-2d materials are the next logical step as these chips and technologies get smaller and smaller
Currently the most popular electronic structure method is DFT
Maps a fully interacting electronic system to a fully noninteracting system using a functional* of the electron density
-Explain error, errors increase when materials are correlated
-obviously DFT has some shortcomings despite its successes: read off slides “most importantly band gap”
-mention corrections
-Some of the proposed solutions for a more accurate electronic structure are many-body perturbation theory and QMC
Solutions (read)
-The first type of QMC we will talk about it VMC. Here a trial WF is created and then the integral in the variational equation are solved using MC integration.
-Read off slide
-It is essential that the trial WF for QMC is good, for accurate results and convergence purposes
-For the most accurate results, it is useful to optimize the WF with VMC. This involves multiplying the single det WF by a Jastrow factor which is a functional expression that adds additional correlation effects to the many-body WF
-Read types of Jastrows
-VMC and WF optimization are usually a precursor to more accurate DMC
In the following step: diffusion monte carlo, the Schro eq is recast into the imaginary time Schro eq….where walkers diffuse in imaginary time until a steady state is reached
The main approximation in DMC is the fixed node approx., which prevents the walkers from changing sign in the simulation, solves fermion sign problem, bound on the energy
In addition there are time step and finite size errors that must be addressed to achieve an accurate DMC result
We can also excite the system to obtain the quasiparticle and optical gaps with DMC
Mention QMCPACK and Nexus to automate DFT->VMC->DMC
**Emphasize the difficulty of QMC over DFT, but talk about payoff in accuracy