Flaviu Cipcigan, Vlad Sokhan, Jason Crain, Glenn Martyna

Transcription

Flaviu Cipcigan, Vlad Sokhan, Jason Crain, Glenn Martyna
Electronically coarse-grained methods for
accurate and efficient molecular modelling
Flaviu Cipcigan, Vlad Sokhan, Jason Crain, Glenn Martyna
flaviu.cipcigan@ed.ac.uk
Introduction
Electronically coarse grained water model
The ability to design materials with desired properties is a major
challenge with both societal and economic impact. The way computer aided design revolutionised manufacturing, atomic and molecular simulation will revolutionise the creation of advanced materials. Computers can explore structures faster than experiment,
study materials in extreme conditions and guide synthesis.
The properties of the model (whose features
are illustrated fig 1) are defined entirely in
terms of isolated molecules, unbiased towards
any thermodynamic state. Yet, it predicts the
following to within 1–2% of experiment:
However, current tools are not always adequate for this job. Classical force-fields, based on fitting model parameters to condensed
phase data transfer poorly to states outside the parametrisation
regime. On the other hand, accurate ab initio methods scale unfavourably with system size.
In a collaboration between University of Edinburgh, National
Physical Laboratory and IBM T.J. Watson Research Center, we are
developing a novel strategy for atomistic modelling, where the
electronic degrees of freedom are coarse grained to a single
quantum harmonic oscillator (called a Quantum Drude Oscillator,
or QDO). The dynamics of interacting QDOs are simulated efficiently via two–temperature path integral molecular dynamics.
In this approach, all many-body forces are on equal footing, dramatically reducing required empirical input. Furthermore, parametrisation needs no condensed phase data, resulting in models
that are extremely transferable between environments.
The ultimate goal is to create a framework to support the design
of technologically important materials, such as synthetic biomolecules and electrolytes. A molecular model for water is the first
step towards this goal.
The model has also revealed new insights into
the physics of water.
800
700
T (K)
600
320
500
280
400
240
0.96 0.98 1.0
300
0
0.2
0.4
0.6
0.8
1.0
density (kg m-3)
Figure 2: The liquid–vapour coexistence curve
of the model matches experiment. The molecular dipole moment reveals an remnant of the
liquid–vapour transition in supercritical water.
The supercritical fluid is thought to be a single
phase. However, we found that abrupt changes
in the dipole moment coincide with the place
of strongest binding (fig 2), separating between
dissociated (gas–like) and associated (liquid–like) regimes in the supercritical fluid.
The hydrogen bonds are thought to be symmetric from the points of view of the oxygen
and hydrogen atoms. We identified an intrinsic
asymmetry (fig 3) and demonstrated it is the
mechanism responsible for the molecular orientation at the surface of liquid water.
positive partial charge
negative partial charge
negative QDO beads
• liquid and vapour densities (fig 2)
• temperature of maximum density (fig 2)
• surface tension
• enthalpy of vaporisation
• dielectric constant
• lattice constants of ice II
• radial distribution functions of supercooled,
ambient and supercritical liquid
locus of heat
nsity
e
capacity maximum
d
s
v
ment
o
m
e
l
dipo
Figure 3: We identified molecules having five
hydrogen bonds, with three neighbours
bonded to the oxygen. This asymmetry is related to the mechanism for the molecular
orientation at the surface of liquid water.
isosurfaces of oxygen density
electronic depletion
electronic
enhancement
polarisation density
T = 203 K
Figure 1: Key features of electronically coarse
grained water for a liquid phase cluster.
T = 372 K
Figure 4: Local structure of liquid water showing
two motifs in the second coordination shell.
Conclusion
We presented a first application of Quantum Drude Oscillators to a complex
molecule, namely water. The method’s success in describing such a challenging
system indicates it can be applied more generally to problems in materials science, such as understanding the movement of ions in electrolytes and modelling
methane, as the first application to a hydrocarbon.
This method is also part of IBM’s offering of HPC solutions, with the plan for
integration with the Watson cognitive platform. The aim is to combine Watson’s
ability to understand complex scientific literature with accurate computer
models to create a materials discovery platform. The envisaged applications are
the search for antimicrobial peptides and new porous materials for batteries.
References
Acknowledgements
Sokhan V, Jones A, Cipcigan F, Crain J, Martyna G (2015) Molecular-Scale Remnants of the Liquid-Gas Transition in Supercritical Polar
Fluids. Phys Rev Lett 115.
Cipcigan F, Sokhan V, Jones A, Crain J, Martyna G (2015)
Hydrogen bonding and molecular orientation at the liquid–vapour interface of water. Phys Chem Chem Phys 17:8660.
Sokhan V, Jones A, Cipcigan F, Crain J, Martyna G (2015) Signature
properties of water:Their molecular electronic origins.
Proc Natl Acad Sci 112:6341.
Jones A, Cipcigan F, Sokhan V, Crain J, Martyna G (2013)
Electronically Coarse-Grained Model for Water.
Phys Rev Lett 110.