computing free energy: replica exchange
TRANSCRIPT
Extending the scale
Essentials of computational chemistry: theories and models. 2nd edition.C. J. Cramer, JohnWiley and Sons Ltd (West Sussex, 2004).Ab initio atomistic thermodynamics and statistical mechanics of surface properties and functionsK. Reuter, C. Stampfl, and M. Scheffler, in: Handbook of Materials Modeling Vol. 1, (Ed.) S. Yip, Springer (Berlin, 2005). http://www.fhi-berlin.mpg.de/th/paper.html
{Ri}
E
Potential Energy Surface: {Ri}
(3N+1)dimensional
109
106
103
1
1015 109 103 1
Length(m)
Time (s)
Microscopicregime
Mesoscopicregime
Macroscopicregime
few processes
few atoms
many atoms
many processes
continuum
average overall processes
more deta
ils
more proce
sses
Thermodynamics:p, T, V, N
(in this page, Helmholtz free energy, F(N,V,T))
if we can calculate E and write analytically on approximation for S for our system, we use this expression. Example: ab initio atomistic thermodynamics.
Ab initio
or similar derivatives that yield measurable quantities (in a computer simulation): one can estimate the free energy by integrating such relations. This is the class of the so called thermodynamicintegration methods.
Thermodynamics
Thermodynamic Integration
Ab initio
Free energy, one quantity, many definitions
Classical statistics (for nuclei):
● Fundamental statistical mechanics thermodynamics link↔
Ab initioAb initio
● Probabilistic interpretation of free energy
Free energy, one quantity, many definitions
Freeenergy evaluation:
Harmonic approximation (solids)
Thermodynamic integration. Phase diagrams
Thermodynamics perturbation (overlap, umbrella sampling)
Accelerated sampling, metadynamics.
Replica Exchange MD: finding the optimal dimensionality reduction
Outline
T1
T2
T3
T4
Configurational coordinate
Ene
rgy
Parallel tempering: the concept
Exchange rule, ensuring canonical sampling at all temperatures:
Parallel tempering: the concept
Parallel tempering: the conceptParallel tempering: the concept
Overlap necessary: the smaller (the system size) the better
Swap 2Swap 1M
D o
r M
C
run
1
MD
or
MC
ru
n 2
T3
T1
MD
or
MC
ru
n 3
T2
T4
T5
Parallel tempering: the implementationParallel tempering: the conceptParallel tempering: the implementation
Parallel tempering: monitoringParallel tempering: the implementationParallel tempering: the conceptParallel tempering: monitoring
Pote
ntia
l Ene
rgy
(eV
)
(degree)
100—620 K 100 K PT (100 ps) 100 K serial (100 ps)
Au4: coexistence of several isomers
0.00 eV
0.04 eV
0.36 eV
Parallel Tempering
biasing potential
means sampling according to distribution:
Replica exchange: other than temperature
Assuming each count in each histogram as independent, then likelihood of observing the ith histrogram:
Replica exchange: temperature – weighted histogram analysis
If all histograms are independent:
Maximum likelihood estimate:
Angle [degree]Angle [degree]
Free
Ene
rgy
[eV
]Pr
obab
ility Pa
rtit
ion
Func
tion
(i
nteg
rate
d pr
obab
ility
)
0.00 eV 0.04 eV
Au4, relative population
Free Energy (Landau): F/kBT = ln [ P (Q) ]
The quest for the “right variable(s)”:Sketch maps
(thanks to Michele Ceriotti, Oxford, for providing figures)
Ala12 landscape
24D Gaussian of stdev 0.5
24D uniform distribution Distribution between pair of ala2 configurations
Sketch map of folded lanscape of ala12
Is pointwise (relative) free energy invariant upon dimensionality reduction?No, only for regions:
Field representation
This field replaces the usual representation based on ddimensional points x. The overlap between fields, which measures their similarity, replaces the distance.