Upload
lamdang
View
216
Download
1
Embed Size (px)
Citation preview
New XC free energy approximations from TDDFTAurora Pribram-Jones
Lawrence Livermore National LaboratoryUC Berkeley Department of Chemistry
Part of this work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344.
IPAM DFT WorkshopAugust 23, 2016
LLNL-PRES-683261
http://newscenter.lbl.gov/2011/10/19/part-i-energy-stars-earth/
August 23, 2016 1
Inertial Confinement Fusion
IPAM DFT
Promotional Materials, SLAC, Stanford University (2015)
August 23, 2016 2
“Warm” Dense Planetary Cores
IPAM DFT
• Magnetic pressure of half a billion refrigerator magnets
• 1 million times atmospheric pressure
• Power >2000 lightning bolts
• 2.5 sticks of dynamite § in 1 cc § over a few ns
Real world equivalents
August 23, 2016 IPAM DFT 4
Basic Research Needs for HEDLP: Report of the Workshop on HEDLP Research, DOE (2009)August 23, 2016 5
The Malfunction Junction
IPAM DFT
R.F. Smith et al., Nature 511 (2014) 330-333August 23, 2016 6
Probing Planetary Conditions
IPAM DFT
R.F. Smith et al., Nature 511 (2014) 330-333August 23, 2016 7
Probing Planetary Conditions
IPAM DFT
Simulations are difficult!
• Quantum effects, strong correlation, partial ionization…
• Approximations affect calculated material properties
Simulations are important!
• Data used in core structure modeling, experimental design
• Experiments hard, expensive, limited
August 23, 2016 8
Quantum and Classical
IPAM DFT
Popular, but...
•no explicit temperature dependence in electrons
•computationally expensive
•need TD electrons for response
•no energy transfer between electrons and ions
IPAM, UCLA
August 23, 2016 9
Quantum Molecular Dynamics
IPAM DFT
…some progress:
•building “math toolbox” for approximations
•novel orbital-free method
•derived new linear response proof for thermal ensembles
•no energy transfer between electrons and ions IPAM, UCLA
August 23, 2016 10
Quantum Molecular Dynamics
IPAM DFT
August 23, 2016 IPAM DFT 11
Heating Things Up
a highly accurate density approximation in one dimension to illustrate our general result,
and (iv) perform (orbital-free) PFT calculations in the WDM regime. Our method gen-
erates highly accurate density and kentropy approximations, skirts the need for separate
kentropy approximations, provides a roadmap for systematically improved approximations,
and converges more quickly as temperatures increase while maintaining accuracy at low
temperatures. At the same time, it bridges low and high temperature methods, and so is
uniquely suited to WDM.
7.2 Theory
At non-zero temperature, the energy is replaced by the grand canonical potential as the
quantity of interest[56, 169]. The grand canonical Hamiltonian is written
⌦ = H � ⌧ S � µN, (7.1)
where H, S, and N are the Hamiltonian, entropy, and particle-number operators. In elec-
tronic structure theory, we typically deal with non-relativistic electrons, most commonly
within the Born-Oppenheimer approximation. The electronic Hamiltonian (in atomic units
here and thereafter) reads
H = T + Vee + V , (7.2)
where T denotes the kinetic energy operator, Vee the interelectronic repulsion, and v(r) the
static external potential in which the electrons move. (We suppress spin for simplicity of
notation.) The grand canonical potential can be written in terms of potential functionals
165
a highly accurate density approximation in one dimension to illustrate our general result,
and (iv) perform (orbital-free) PFT calculations in the WDM regime. Our method gen-
erates highly accurate density and kentropy approximations, skirts the need for separate
kentropy approximations, provides a roadmap for systematically improved approximations,
and converges more quickly as temperatures increase while maintaining accuracy at low
temperatures. At the same time, it bridges low and high temperature methods, and so is
uniquely suited to WDM.
7.2 Theory
At non-zero temperature, the energy is replaced by the grand canonical potential as the
quantity of interest[56, 169]. The grand canonical Hamiltonian is written
⌦ = H � ⌧ S � µN, (7.1)
where H, S, and N are the Hamiltonian, entropy, and particle-number operators. In elec-
tronic structure theory, we typically deal with non-relativistic electrons, most commonly
within the Born-Oppenheimer approximation. The electronic Hamiltonian (in atomic units
here and thereafter) reads
H = T + Vee + V , (7.2)
where T denotes the kinetic energy operator, Vee the interelectronic repulsion, and v(r) the
static external potential in which the electrons move. (We suppress spin for simplicity of
notation.) The grand canonical potential can be written in terms of potential functionals
165
Mermin, N.D. Phys. Rev. A, 137: 1441 (1965).Pittalis, S. et al. Phys. Rev. Lett., 107: 163001 (2011).
Grand canonical potential operator
Electronic Hamiltonian
August 23, 2016 IPAM DFT 12
Entropy and Statistics
ground-state problem. The former is written
⌦ = H � ⌧ S � µN, (3.83)
where H, S, and N are the Hamiltonian, entropy, and particle-number operators. The crucial
quantity by which the Hamiltonian di↵ers from its grand-canonical version is the entropy
operator:5
S = � kBln� , (3.84)
where
� =X
N,i
wN,i| N,iih N,i| . (3.85)
| N,ii are orthonormal N -particle states (that are not necessarily eigenstates in general) and
wN,i are normalized statistical weights satisfyingP
N,i wN,i = 1. � allows us to describe the
thermal ensembles of interest.
Observables are obtained from the statistical average of Hermitian operators
O[�] = Tr {�O} =X
N
X
i
wN,ih N,i|O| N,ii . (3.86)
These expressions are similar to Eq. (3.53), but here the trace is not restricted to the ground-
state manifold.
In particular, consider the average of the ⌦, ⌦[�], and search for its minimum at a given
temperature, ⌧ , and chemical potential, µ. The quantum version of the Gibbs Principle en-
sures that the minimum exists and is unique (we shall not discuss the possible complications
5Note that, we eventually choose to work in a system of units such that the Boltzmann constant is kB = 1,that is, temperature is measured in energy units.
72
ground-state problem. The former is written
⌦ = H � ⌧ S � µN, (3.83)
where H, S, and N are the Hamiltonian, entropy, and particle-number operators. The crucial
quantity by which the Hamiltonian di↵ers from its grand-canonical version is the entropy
operator:5
S = � kBln� , (3.84)
where
� =X
N,i
wN,i| N,iih N,i| . (3.85)
| N,ii are orthonormal N -particle states (that are not necessarily eigenstates in general) and
wN,i are normalized statistical weights satisfyingP
N,i wN,i = 1. � allows us to describe the
thermal ensembles of interest.
Observables are obtained from the statistical average of Hermitian operators
O[�] = Tr {�O} =X
N
X
i
wN,ih N,i|O| N,ii . (3.86)
These expressions are similar to Eq. (3.53), but here the trace is not restricted to the ground-
state manifold.
In particular, consider the average of the ⌦, ⌦[�], and search for its minimum at a given
temperature, ⌧ , and chemical potential, µ. The quantum version of the Gibbs Principle en-
sures that the minimum exists and is unique (we shall not discuss the possible complications
5Note that, we eventually choose to work in a system of units such that the Boltzmann constant is kB = 1,that is, temperature is measured in energy units.
72
ground-state problem. The former is written
⌦ = H � ⌧ S � µN, (3.83)
where H, S, and N are the Hamiltonian, entropy, and particle-number operators. The crucial
quantity by which the Hamiltonian di↵ers from its grand-canonical version is the entropy
operator:5
S = � kBln� , (3.84)
where
� =X
N,i
wN,i| N,iih N,i| . (3.85)
| N,ii are orthonormal N -particle states (that are not necessarily eigenstates in general) and
wN,i are normalized statistical weights satisfyingP
N,i wN,i = 1. � allows us to describe the
thermal ensembles of interest.
Observables are obtained from the statistical average of Hermitian operators
O[�] = Tr {�O} =X
N
X
i
wN,ih N,i|O| N,ii . (3.86)
These expressions are similar to Eq. (3.53), but here the trace is not restricted to the ground-
state manifold.
In particular, consider the average of the ⌦, ⌦[�], and search for its minimum at a given
temperature, ⌧ , and chemical potential, µ. The quantum version of the Gibbs Principle en-
sures that the minimum exists and is unique (we shall not discuss the possible complications
5Note that, we eventually choose to work in a system of units such that the Boltzmann constant is kB = 1,that is, temperature is measured in energy units.
72
Pittalis, S. et al. Phys. Rev. Lett., 107: 163001 (2011). Pribram-‐Jones et al., “Thermal DFT in Context,” Frontiers and Challenges in Warm Dense Matter, Springer Publishing (2014), p 25-‐60.
Entropy operator:
Statistical operator:
Observables:
August 23, 2016 13
Finite-‐temperature Kohn-‐ShamMap interacting system to non-interacting system with same density.
Kohn and Sham, 1965.
IPAM DFT
Free energies: Helmholtz and XC
Pittalis, S. et al. Phys. Rev. Lett., 107: 163001 (2011). Pribram-‐Jones et al., “Thermal DFT in Context,” Frontiers and Challenges in Warm Dense Matter, Springer Publishing (2014), p 25-‐60.
Temperature-dependent free energy:
August 23, 2016 IPAM DFT 14
Kinetic, potential, entropic exchange-correlation:
Correlation RelationsCorrelation free energy: kentropic, potential, kinetic, entropic
August 23, 2016 IPAM DFT 15
APJ and K. Burke, Phys. Rev. B 93, 205140 (2016)
Correlation free energy: kentropic, potential, kinetic, entropic
Combine with ACF to get a set of relations, such as:
ACF via scaled density/temperature
August 23, 2016 IPAM DFT 16
Change of variables yields thermal connection formula:
Combine finite-temperature ACF (Pittalis, et al., 2011)
with coupling constant-coordinate-temperature scaling (Pittalis, et al., 2011)
APJ and K. Burke, Phys. Rev. B 93, 205140 (2016)
Thermal Connection Formula
August 23, 2016
• Relates exact XC free energy to high temperature, high density limit• Need knowledge of XC potential energy at scaled densities, not at scaled interaction strengths
• Reduces to plasma physics coupling-constant relation for uniform systems
• Generalization of plasma physics formula to density functionals and inhomogeneous systems
IPAM DFT 17
APJ and K. Burke, Phys. Rev. B 93, 205140 (2016)
An Application
Useful for computation and theory:
•Generates new XC approximations for FT DFT•Provides link between finite-temperature and infinite-temperature limit
August 23, 2016 IPAM DFT 18
Using finite-temperature fluctuation-dissipation theorem for the correlation free energy in terms of the thermal density-density response function:
APJ, P.E. Grabowski, and K. Burke, Phys. Rev. Lett. 116, 233001 (2016)
Approximations to thermal XC kernel:
• = 0 à thermal RPA
•
à Approximate
XC ApproximationsExact expression, as long as exact thermal kernel is used:
August 23, 2016 IPAM DFT 19
APJ, P.E. Grabowski, and K. Burke, Phys. Rev. Lett. 116, 233001 (2016)
AcknowledgmentsMentors and Collaborators
• F. Graziani (LLNL), E. Schwegler (LLNL), M. Head-Gordon (UCB), K. Burke (UCI), A. Cangi (MPI), Z.-Y. Yang (Temple), C.A. Ullrich (UMC), J.C. Smith (UCI), D. Feinblum (UCI), S. Pittalis (CNR-NANO), E.K.U. Gross (MPI), P.E. Grabowski (UCI/LLNL), R.J. Needs (Cambridge), J.R. Trail (Cambridge), D.A. Gross (UCI), M.P. Desjarlais (SNL), K.R. Cochrane (SNL), M.D. Knudson (SNL)
Funding Sources• Lawrence Fellowship, Lawrence Livermore National Laboratory•UC President’s Postdoctoral Fellowship, UC Berkeley Chemistry•Quantum Simulation Group and WCI, LLNL•Krell Institute and the DOE CSGF (DE-FG02-97ER25308)
August 23, 2016 IPAM DFT 20
Summary
August 23, 2016
For further information• Linear response of thermal ensembles: A. Pribram-‐Jones, P.E.
Grabowski, and K. Burke, Phys. Rev. Lett. 116, 233001 (2016).• Thermal connection formula and exact conditions: A. Pribram-‐Jones
and K. Burke, Phys. Rev. B 93, 205140 (2016).• Thermal DFT: A. Pribram-‐Jones, S. Pittalis, E.K.U. Gross, K. Burke,
Frontiers and Challenges in Warm Dense Matter, Springer Publishing (2014), p 25-‐60.
• Quirky overview of DFT: A. Pribram-‐Jones, D.A. Gross, K. Burke, Ann. Rev. Phys. Chem 66 (2015).
IPAM DFT 21
• Exact conditions relating components of correlation free energy• Thermal connection formula: adiabatic connection via
temperature integral• Can generate XC free energy approximations from thermal kernel