Embed Size (px)
Citation preview
Artem R. Oganov (ARO)
Moscow Institute of Physics and Technology, Dolgoprudny, Russia
State University of New York, Stony Brook, USA
Northwestern Polytechnical University, Xi’an, China
Computational Materials Discovery: NewMethods and Results
(from http://nobelprize.org)
Structure is the basis for understanding materials andtheir properties
Zincblende ZnS.One of the firststructures solvedby Braggs in 1913.
Structure Diffraction
With time, incredibly complex structures werediscovered
Host-guest elements(Rb-IV phase, U.Schwarz’99)
Quasicrystals
New state of matter discovered in labin 1984. In nature found only in 2009!
Proteins
Need to find GLOBAL energy minimum.
Trying all structures is impossible: Natoms Variants CPU time
1 1 1 sec.
10 1011 103 yrs.
20 1025 1017 yrs.
30 1039 1031 yrs.
J. Maddox(Nature, 1988)
The USPEX (Universal Structure Prediction: EvolutionaryXtallography) project
http://uspex.stonybrook.edu
•The most popular code for computational materials design in the world(>2000 users).•The largest , the most versatile, the fastest and the most reliable code in thisfield. Many of its capabilities are unique. 3D-, 2D-, 1D-, 0D- systems can betreated. Prediction of crystal structures, stable compounds, phase transitionmechanisms.•THE CODE IS FREE.•Effort of ~50 man-years.•>200 publications, 2 US patents.
New developments in crystal structure predictionextend the range of problems that can be solved
2. Predicting new chemistry
1. Predicting crystalstructures by evolution
Can Periodic System change at extreme conditions?What is the chemical formula of sodium chloride?What is the most inert element?What is the cleanest fuel material?Why are surfaces anomalous?
1. Predicting structures by evolution
ARO, Lyakhov, Valle (2011).How evolutionary crystal structure prediction works - and why.Acc. Chem. Res. 44, 227-237.
Evolutionary simulations learn & explore themost promising regions of search space
Evolutionary simulations learn & explore themost promising regions of search space
Evolutionary simulations learn & explore themost promising regions of search space
Evolutionary simulations learn & explore themost promising regions of search space
Evolutionary simulations learn & explore themost promising regions of search space
USPEX(Universal Structure Predictor: Evolutionary Xtallography)
• (Random) initial population• Evaluate structures by relaxed (free) energy• Select lowest-energy structures as parents for new
generation• Standard variation operators:
(1) Heredity (crossover)
(2) Lattice mutation (3) Permutation
Test: High-pressure phases of carbon
2000 GPa: bc8 phase, potentially
important in astrophysics
100 GPa: diamond is stable
Metastable bc8 form of SiIs known (Kasper, 1964)
[ARO & Glass, J.Chem.Phys. (2006)]
+found metastable formthat matches„superhard graphite“ of W.Mao(Li, ARO, et al., PRL 2009)
Increases diversity and fitness of the population
Symmetric initialization
Crystals: 230 space groupsNanoparticles: point groups
[Lyakhov, ARO, et al. (2013)]
Atoms are moved along the eigenvector of the softest mode
(both positive and negative directions need to be tried)
Requires the calculation of the dynamical matrix
Soft-mode mutation
[Lyakhov, ARO, et al. (2013)]
The metastable state is found first and the ground state is found shortly after
‘Ageing’ technique (antiseeds)
[Lyakhov, ARO, et al. (2013)]
Crystal structure prediction methods
1. Random sampling (Freeman & Catlow, 1992)
2. Simulated annealing (Pannetier, 1990)
3. Molecular dynamics and metadynamics (Martonak, 2003)
4. Data mining (Curtarolo, 2003)
5. Minima hopping (Goedecker, 2004)
6. Evolutionary algorithms / PSO
Overview of USPEX(ARO & Glass,J.Chem.Phys. 2006)
Why does USPEX work so well? A few tips and tricks
I. Reduction of dimensionality throughunbiased symmetric initialization.
II. Reduction of effective dimensionalityof problem by structure relaxation (alsoreduces “noise” and transforms energylandscape to a convenient shape).
III. Variation operators are defined insubspaces of reduced dimensionality andinvolve cooperative transformations.
New developments of USPEX
A. Improved efficiency
B. Molecular crystals
C. Variable composition
D. Low-dimensional systems
-0D (nanoparticles)
-1D (polymers)
-2D (surfaces and 2D-crystals)
E. Evolutionary metadynamics
F. vc-NEB for transition pathways
G. Transition path sampling
MnB3 was predicted and then synthesized(Niu, ARO, et al., PCCP 2014)
1. New compound discovered – MnB3.2. For MnB4, reported experimental structure was wrong – new experiments
confirm our structure.
Predicting new compounds:example of “well-known” system Mn-B
How to make it?Hierarchical approach & new capabilities
1. Evolutionary metadynamics:Identify possible mechanisms(Zhu & ARO, 2012)
2. Variable-cell NEB: refinemechanisms in the staticmean-field approximation(Qian & ARO, 2013)
3. Transition path sampling:Include nucleation andtemperature (Dellago, 1998;Boulfelfel & ARO, 2012)
2. Predicting new chemistry
Applications of this method proved its great utlilty:
Units: 100 GPa = 1 Mbar =
200x
Matter under pressure: new phenomenaand ubiquity in nature
P.W. Bridgman1946 Nobel Prize for Physics
BDiscovery of -B28, one of the hardest knownmaterials
Overview of hard materials (Chen, 2011)-B28 has hardness of 50 GPa (Solozhenko, ARO, 2008)
NaCl-like structure of -B28
(ARO et al., Nature 2009)First phase diagram of boron(ARO et al., Nature 2009)
Sodium, a typical metal, under pressurebecomes transparent
• Surprise: core electrons become essential and cause demetallization ofNa. Na becomes transparent at 200 GPa (Ma, Eremets, ARO, Nature 2009).Not alkali METAL anymore!
Localized interestitial electron pairs make Na insulating.
corecore
corecore
e
e
e
e
e
core
Explanation: Models of Ashcroft(1999, 2008)
Pressure
core
corecore
Na
[Ma, Eremets, ARO, Nature 2009]
Transparent sodium Na
[Ma, Eremets, ARO, Nature 2009]
Salt as we know it
Na-Cl
Crazy salt
Weiwei Zhang
Na-Cl
Salt as we know it
Na-Cl
What classical chemistry tells us
• Large electronegativity difference (2.2) ionic bonding.
• Na and Cl have valence 1 Na+ and Cl-.
• The only possible compound is NaCl.
Na-Cl
Na Cl
Crystal structure of NaCl(rocksalt)
“The only possible compound is NaCl”
All other possibilities – forbidden or unlikely.
Na-Cl
-
+
Correct structure Incorrect structure
Defects, coming from tinyerrors in the Na:Cl ratio, leadto rare colors of NaCl
-
-
--
+
+
+
-
+
-
--
+
+
+
+
Unusual chemistry of the most usual compound
• Na-Cl system: compounds Na3Cl, Na2Cl, Na3Cl2, NaCl, NaCl3, NaCl7 are allstable under pressure (Zhang, ARO, et al. Science, 2013). Confirmed byexperiment!
Phase stability in the Na-Cl systemAtomic and electronicstructure of cubic NaCl3
Na-Cl
[Zhang, ARO, et al., Science (2013)]
Crazy sodium chlorides
Peculiar Na-Cl compounds:
• NaCl7: some Cl atoms have POSITIVE Bader charge (+0.07).
• Na3Cl: 2D-metal
NaCl7 NaCl3 Na3Cl
Na-Cl
[Zhang, ARO, et al., Science (2013)]
Experimental confirmation of NaCl3 and Na3Cl
Pm3n NaCl3a= 4.6110(3) Å - experimenta=4.602 Å - theory
[Zhang, ARO, et al., Science (2013)]
Na-Cl
Equations of state of NaCl3 (left) andNa3Cl (right): theory vs experiment
K-Cl: extreme richness of the phase diagram
(Zhang, ARO, Goncharov, 2014). Predictions confirmed by experiment!
P-x phase diagram of the K-Cl system
Electronic structure of K3Cl5
Electronic DOS of K-Cl compounds
Experimental X-ray
diffraction of KCl3
New class of compounds &new chapter of chemistry
Mg-O system: compounds Mg3O2, MgO, MgO2 are all stable under pressure(Zhu, ARO, Phys. Chem. Chem. Phys. 2013).
Phase stability in the Mg-O system
Crystal structure and electronlocalization function of MgO2 andMg3O2
Mg-O
Qiang Zhu
What is the chemically most inertelement?
• Helium or neon.
• Helium is the 2nd most abundant element in the Universe (24 wt.%).
• Helium:
ionization potential = 24.39 eV (record!)
electron affinity = 0.08 eV
• No stable compounds are known at normal conditions.
• Van der Waals compound NeHe2 exists at high pressure (Loubeyreet al., 1993).
Helium’s unexpected chemistry:stable compound Na2He (Dong, ARO, et al., 2014)
Na-He
1. Na2He is stable at >120 GPa, at least up to 1000 GPa.2. It is an insulating electride with direct gap.3. Its band gap anomalously increases with pressure.
[Dong, ARO, et al., submitted (2014)]
Helium’s unexpected chemistry:stable compound Na2He (Dong, ARO, et al., 2014)
Na-He
Experimental X-ray diffraction pattern of Na2He at 130 GPa.
[Dong, ARO, et al., submitted (2014)]
Helium’s unexpected chemistry:stable compound Na2He (Dong, ARO, et al., 2014)
Na-He
1. Put acceptor of an electron pair onthe “2e” site.
2. Na2HeO – stable already at 14 GPa.
[Dong, ARO, et al., submitted (2014)]
Is ice (H2O) the only stable compound of H and O?H-O
Phase diagram of the H2O-H2 system
[Qian, ARO, et al., Scientific Reports (2014)]
Structure and Raman shift ofnew C3 phase
1. H2O*2H2 (=H6O!) has 18wt% of easily removable hydrogen!2. Cleanest fuel, but stable only at >38 GPa.3. Explains experimental data of Machida (2011).
Сs-F
[Zhu & ARO, submitted (2014)]
“Forbidden” compounds can be practically useful:
Phase diagram of the Cs-F - Miao (2013) and corrected (Zhu & ARO, 2014)
Structure of CsF5,at 1 atm
• СsF2, CsF3, CsF5 are stable at 1 atm and can be used forfluorine storage.• Decomposition temperatures ~250-400 K.• Analogous Cl-rich phases predicted.• US patent (2013).
Some “forbidden”, yet stable compounds
Mg3O2, MgO2 - [MgO]SiO, SiO3 - [SiO2]Al4O7, AlO2 - [Al2O3]
N-H: Unique diversity of chemistry. Nitrogen life?
Cr-H: Unexpected compounds –common phenomenon
K-Cl: Huge number of stablecompounds
• Uranus, Neptune =H2O:CH4:NH3 = 59:33:8
• High pressures andtemperatures.
• Life would have been likelyat low temperatures
Nitrogen life?
Polymerization of N3H at 10 GPa fromvc-NEB calculation.Could be a mechanism for storing/releasing energy(similar to ATP=ADP+PO4)
Molecular – greenIonic – purplePolymeric – blue2D-polymeric - red
Extreme chemistry is everywhere
Boron surface: metallic or semiconducting?
2013: Amsler et al. (PRL) predicted metallic (111) surface of alpha-boron.Metallic surfaces of semiconducting materials are known for topological insulators.
2014: Zhou & ARO found a much more stable semiconducting surface
Amsler (2013): complex metallic surface.Surface energy 171 meV/A2
Zhou+ARO (2014): simple semiconductingsurface. Surface energy 128 meV/A2
2D-crystals: example of 2D-boron
2007: Tang et al. (PRL) construct alpha-sheet.2011: using CALYPSO, Luo et al. (JACS) confirm alpha-sheet.2013: using USPEX, Zhou, Zhu, ARO find two MUCH more stable
2D-structures.
alpha-sheet Structure 1,60 meV/atom better
Structure 2,80 meV/atom better
Structure 1 is a semimetal withmassless Dirac fermions andanisotropic Dirac cone[Zhou, ARO et al., 2013]
2D: Surfaces of crystals, new physics and chemistry…and medicine?
Structure and composition of surface phases of GaN (10-11) Phase diagram inpresence of oxygen
Structure of (111) surface of MgO.One can see peroxide-groups, О3-groups
Note peroxide (O22-) groups at the surface
Surfaces of crystals: new physics, chemistry,… andmedicine?
(111) surface of MgO
Peroxide (O22-) and other reactive oxygen forms on the surface.
Explanation of carcinogenic effect of oxide dust?
(011) Surface of quartz (SiO2)
(110) surface of rutile (TiO2)Wang & ARO, PRL 2014
Mineral nanoparticles (quartz,asbestos) cause lung cancer.Cause – oxidative stress. Peroxideions can form on surfaces of evennormally inert minerals.
New developments in crystal structure predictionextend the range of problems that can be solved
Predicted new materialsand phenomena
Powerful methods forcrystal structure prediction
Can Periodic System change at extreme conditions?What is the chemical formula of sodium chloride?What is the most inert element?What is the cleanest fuel material?Why are surfaces anomalous?
Q. Zhu
Acknowledgments:
W. Zhang X. DongG.R. Qian
US laboratory:Qiang ZhuAndriy LyakhovSalah Eddine BoulfelfelXiang-Feng ZhouGuangrui QianHuafeng DongQianku HuDongxu LiYue LiuChaohao HuXiao DongMaksim RakitinShengnan WangMahdi DavariJin Zhang
