$ crystal structure prediction for next-generation battery anodes
Matthew Evans
16th November 2017
Supervisor: Dr Andrew Morris
 |
 |
Outline
- Introduction
- Crystal structure prediction
- Application: K-Sn-P
- Summary
Introduction
Why batteries?
- Electrification of transport
- both personal and public
- Computation
- Mobile devices
- Internet of Things etc.
- Grid-level storage
- load smoothing
- reduce cost of renewables
Introduction
3 materials problems:
https://www.iam.kit.edu/wpt/
Introduction
>> 3 materials problems:
https://www.iam.kit.edu/wpt/
Introduction
Why anodes? What's wrong with graphite?
Introduction
Why anodes? What's wrong with graphite? ...limited capacity, rate capability.
Introduction
Why beyond-Li? ...cheaper, more sustainable, safer!
Introduction
Why beyond-Li? ...cheaper, more sustainable, safer!
Introduction
Why beyond-Li? ...cheaper, more sustainable, safer!
Introduction
Why K-Sn-P?
Introduction
Why computation?
- Density functional theory
- efficient, reliable method for energies and forces
- Huge materials space
- too large and expensive to be spanned with experiment alone
- Computational resouces
- iPhone more power than moon computer blah blah
- Software; algorithms faster, more robust
“It is a understatement to say that materials properties depend critically on where the atoms are.”
Richard Needs
Where are the atoms?
Crystal structure prediction
- Ab initio random structure searching (AIRSS)
- Data mining
- Genetic algorithms (GA)
I. Ab initio random structure searching (AIRSS)
![airss svg]()
I. Ab initio random structure searching (AIRSS)
![nature]()
![nature]()
![nature]()
![nature]()
- High pressure phases (Pickard and Needs, 2006)
- SiH4, AlH3, metallic H, N2, H2O NH3, Al etc...
- Superconducting H3S ($T_c$ = 203 K at 155 GPa).
- Defects
- {H, N, O, Li} complexes in Si
- Intrinsic defects in CaZrTi2O7
- Battery materials
- Li-S, Li-Si, Na-Sb, Li-MoS2, Li-FeS
- Li-Ge, Li-P, Na-P
- Na-Sn, Li-Sn, Li-Sb
- Encapsulated nanowires
II. Data mining
Database of ~600k relaxed structures, or 1M+ when combined with OQMD.
![matador]()
III. Genetic algorithms (GA)
![airss]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
Phase stability
Formation Energy ↓
Composition →
Analysis
Electrochemistry:
![ksnp]()
![ksnp]()
![ksnp]()
![ksnp]()
K-Sn-P
- 3 abundant, cheap materials.
- K has potential to outperform Na batteries in terms of energy density.
- Recent experimental studies show reasonable performance of Sn4P3.
- W. Zhang et al, JACS 139(9), 2017
- Phase chemistry of even the binaries is not well known: K3P forms?
- Builds on previous work in our group on phosphides and stannides: ~50k structures in our database.
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
![ksnp]()
K-Sn-P
4 SnP3 + 25 K → 3 K8SnP4 + KSn
4 Sn4P3 + 37 K → 3 K8SnP4 + 13 KSn
- 31% and 19% increase in predicted gravimetric capacity
- High insertion voltages of 1.55 V and 1.15 V improve safety
- ...at the expense of energy density
- Volume expansion of 311% and 177%, compared to 600% predicted for K3P formation!
![ksnp]()
Summary
- We are performing crystal structure prediction on metal phosphides, with the aim of finding the metal which maximises capacity with reasonable volume expansion.
- Sampling a ternary composition space is tricky!
- We have developed two main software packages, MATADOR and ilustrado to do so.
- So far, this has been applied to the K-Sn-P system, with some success.
- 3 novel ternary phases have been found, K8SnP4, KSnP and KSn2P2.
- These phases extend predicted theoretical capacity by ~30% and ~20%.
- What next?
- Experimental verification!
- More (all?) systems!
- Extend beyond 0 K; use wealth of metastable phases sampled.
Acknowledgements
![matador]()
K-Sn-P: Andrew Morris, Kent Griffith
![matador]()
![matador]()
![matador]()
AIRSS review (C. Pickard and R. Needs, JPCM 23(5), 2011) 10.1088/0953-8984/23/5/053201
Computational phosphide work (M. Mayo, K. Griffith et al., Chem. Mater. 28(7), 2016) 10.1021/acs.chemmater.5b04208
Computational stannide work (M. Mayo and A. Morris, Chem. Mater. 29(14), 2017) 10.1021/acs.chemmater.6b04914
K-Sn-P experimental paper (W. Zhang et al., JACS 139(9), 2017) 10.1021/jacs.6b12185
- High pressure phases (Pickard and Needs, 2006)
- SiH4, AlH3, metallic H, N2, H2O NH3, Al etc...
- Superconducting H3S ($T_c$ = 203 K at 155 GPa).
- Defects
- {H, N, O, Li} complexes in Si
- Intrinsic defects in CaZrTi2O7
- Battery materials
- Li-S, Li-Si, Na-Sb, Li-MoS2, Li-FeS
- Li-Ge, Li-P, Na-P
- Na-Sn, Li-Sn, Li-Sb
- Encapsulated nanowires
- 3 abundant, cheap materials.
- K has potential to outperform Na batteries in terms of energy density.
- Recent experimental studies show reasonable performance of Sn4P3.
- W. Zhang et al, JACS 139(9), 2017
- Phase chemistry of even the binaries is not well known: K3P forms?
- Builds on previous work in our group on phosphides and stannides: ~50k structures in our database.
- 31% and 19% increase in predicted gravimetric capacity
- High insertion voltages of 1.55 V and 1.15 V improve safety
- ...at the expense of energy density
- Volume expansion of 311% and 177%, compared to 600% predicted for K3P formation!
- We are performing crystal structure prediction on metal phosphides, with the aim of finding the metal which maximises capacity with reasonable volume expansion.
- Sampling a ternary composition space is tricky!
- We have developed two main software packages, MATADOR and ilustrado to do so.
- So far, this has been applied to the K-Sn-P system, with some success.
- 3 novel ternary phases have been found, K8SnP4, KSnP and KSn2P2.
- These phases extend predicted theoretical capacity by ~30% and ~20%.
- What next?
- Experimental verification!
- More (all?) systems!
- Extend beyond 0 K; use wealth of metastable phases sampled.
II. Data mining
Database of ~600k relaxed structures, or 1M+ when combined with OQMD.
![matador]()
Database of ~600k relaxed structures, or 1M+ when combined with OQMD.
III. Genetic algorithms (GA)
![airss]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
III. Genetic algorithms (GA)
![]()
Phase stability
Formation Energy ↓
Composition →
Analysis
Electrochemistry:
![ksnp]()
Electronic and optical spectroscopies:
![ksnp]()
NMR:
![ksnp]()
PDF:
![ksnp]()
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
K-Sn-P
4 SnP3 + 25 K → 3 K8SnP4 + KSn
4 Sn4P3 + 37 K → 3 K8SnP4 + 13 KSn
Summary
Acknowledgements
K-Sn-P: Andrew Morris, Kent Griffith
AIRSS review (C. Pickard and R. Needs, JPCM 23(5), 2011) 10.1088/0953-8984/23/5/053201 Computational phosphide work (M. Mayo, K. Griffith et al., Chem. Mater. 28(7), 2016) 10.1021/acs.chemmater.5b04208 Computational stannide work (M. Mayo and A. Morris, Chem. Mater. 29(14), 2017) 10.1021/acs.chemmater.6b04914 K-Sn-P experimental paper (W. Zhang et al., JACS 139(9), 2017) 10.1021/jacs.6b12185