Upload
srokkam
View
226
Download
0
Embed Size (px)
Citation preview
7/30/2019 Thermostat MD
1/29
Thermostats in Molecular Dynamics Simulations
Alden Johnson, Teresa Johnson, Aimee KhanUniversity of Massachusetts Amherst
December 6th, 2012
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)
Thermostats in MD Simulations December 6th, 2012 1 / 29
7/30/2019 Thermostat MD
2/29
An Interest in Thermostats
Thermostat: A modification of the Newtonian MD scheme with thepurpose of generating a statistical ensemble at a constant temperature.
Match experimental conditions
Manipulate temperatures in algorithms such as simulated annealing
Avoid energy drifts caused by accumulation of numerical errors.
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)
Thermostats in MD Simulations December 6th, 2012 2 / 29
7/30/2019 Thermostat MD
3/29
Statistical Ensembles
Ensemble: a large collection of microscopically defined states of asystem, with certain constant macroscopic properties
Microcanonical (NVE)
Arises in Newtonian MD simulation Conserves total energy
Canonical (NVT)
Implement thermostats to sample from here Relevant to real behavior in experiment
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)
Thermostats in MD Simulations December 6th, 2012 3 / 29
7/30/2019 Thermostat MD
4/29
Microcanonical Ensemble
Consider an isolated system
Microstate: complete description of a state of the system,microscopically
The probabilities of being in a certain microstate for themicrocanonical ensemble are uniform over all possible states
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 4 / 29
7/30/2019 Thermostat MD
5/29
Canonical Ensemble
Follows a Gibbs distribution for probability pj of being in a givenmicrostate j with energy Ej
pj =eEj
Z , Z =
j
e
Ej
=1
kBT
Derivation follows from maximizing entropy
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 5 / 29
7/30/2019 Thermostat MD
6/29
Ergodic Hypothesis
The ergodic hypothesis says the long time average of an observable fcoincides with an ensemble average of the observable f.
f = limt
1
tt
0 f(x(s)) ds
f =
f(x) d(x)
where is the ensemble measure and is the phase space of theobservable.
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 6 / 29
7/30/2019 Thermostat MD
7/29
Molecular Dynamics Simulation
Phase space is collection of positions q and momenta p of particles insystem
The Hamiltonian Formdqt = pH(qt, pt) dtdpt = qH(qt, pt) dt
H(q, p) = Ekin
(p) + V(q), Ekin
(p) =1
2pTM1p
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 7 / 29
7/30/2019 Thermostat MD
8/29
Sampling from Ensembles
A canonical ensemble (constant average energy) is a distribution ofmicrocanonical ensembles (constant energy)
To sample from the canonical ensemble, the following thermostatsmodulate the energy entering and leaving the boundaries of thesystem
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 8 / 29
7/30/2019 Thermostat MD
9/29
Velocity Rescaling
Velocities are described by Maxwell-Boltzmann distribution
P(vi,) = m
2kBT12
e
mv2i,2kBT
Adjust instantaneous temperature by scaling all velocities
Average Ekin per degree of freedom related to T via the equipartitiontheorem
mv2i,2
=1
2 kBT
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 9 / 29
7/30/2019 Thermostat MD
10/29
Velocity Rescaling
Ensemble average average over velocities of all particles: defineinstantaneous temperature Tc for a finite system
kBTc =1
Nfi,
mv2
i,
Tc = T until rescaling
v
i, =T
Tcvi,
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 10 / 29
7/30/2019 Thermostat MD
11/29
Velocity Rescaling
Disadvantages
Results do not correspond to any ensemble
Does not allow the proper temperature fluctuations
Localized correlation not removed
Not time reversible
Advantages
Straightforward to implement
Good for use in warmup / initialization phase
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 11 / 29
7/30/2019 Thermostat MD
12/29
Nose-Hoover
Based on extended Lagrangian formalism
Deterministic trajectory
Simulated system contains virtual variables related to real variables
Coordinates qi = qi
Momenta pi = pi/s
Time t =t
0dts
s: additional degree of freedom, acts as external system
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 12 / 29
7/30/2019 Thermostat MD
13/29
Nose-Hoover
Hamiltonian given by
H =Ni=1
p2i
2mis2+ V(q) +
Qs2
2+ (3N + 1)kBTln s
Logarithmic term required for proper time scaling: canonical ensemble
Effective mass Q associated with s Determines thermostat strength Q too small: system not canonical
Q too large: temperature control inefficientMicrocanonical dynamics on extended system give canonicalproperties
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 13 / 29
7/30/2019 Thermostat MD
14/29
Nose-Hoover
Disadvantages
Extended system not guaranteed to be ergodic
Advantages
Easy to implement and use
Implement as a chain
Each link: apply thermostatting to the previous thermostat variable
Increasing Q lengthens decay time of response to instantaneous
temperature jump
Deterministic and time reversible
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 14 / 29
7/30/2019 Thermostat MD
15/29
Langevin
Consider the motion of large particles through a continuum of smallerparticles
dqidt
=pi
mi
dpidt
= V(q)qi
pi + Gi
Viscous drag force proportional to velocity pi Smaller particles give random pushes to large particle
Fluctuation-dissipation relation
2 = 2mikBT
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 15 / 29
7/30/2019 Thermostat MD
16/29
Langevin
Disadvantages
Difficult to implement drag for non-spherical particles: related toparticle radius
Momentum transfer lost: cannot compute diffusion coefficients
Advantages
Damping + random force correct canonical ensemble
Ergodic Can use larger time step
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 16 / 29
A d
7/30/2019 Thermostat MD
17/29
Andersen
Couple a system to a heat bath to impose desired temperature
Equations of motion are Hamiltonian with stochastic collision term
Strength of coupling specified by , the stochastic collision frequency
When particle has collision, new velocity is sampled from N(0,T)
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 17 / 29
A d
7/30/2019 Thermostat MD
18/29
Andersen
Disadvantages
Newtonian dynamics + stochastic collisions Markov chain
Algorithm randomly decorrelates velocity: dynamics are not physical
Advantages
Allows sampling from canonical ensemble
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 18 / 29
Th S
7/30/2019 Thermostat MD
19/29
Thermostats Summary
Thermostat Description Canonical? Stochastic?
Velocity Rescaling KE fixed to match T no no
Nose-Hoover extra degree of freedomacts as thermal reser-voir
yes no
Langevin noise and drag balanceto give correct T
yes yes
Andersen momenta occasionally
re-randomized
yes yes
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 19 / 29
A li d M th M t P j t
7/30/2019 Thermostat MD
20/29
Applied Math Masters Project
Molecular Dynamics Simulation
Implemented in MATLAB and C++
Simulations in 2 and 3 dimensions
Periodic and walled boundary conditions
External fields such as gravity
Optimization
OpenMP
CUDA
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 20 / 29
L d J P t ti l
7/30/2019 Thermostat MD
21/29
Lennard-Jones Potential
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 21 / 29
Numeric Integration
7/30/2019 Thermostat MD
22/29
Numeric Integration
Verlet Algorithm
pn+1/2 = pn t2V(qn)
qn+1 = qn + tM1pn+1/2
p
n+1
= p
n+1/2
t
2 V(qn+1
)
Preserves modified Hamiltonian
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 22 / 29
Thermostat Implementation
7/30/2019 Thermostat MD
23/29
Thermostat Implementation
Velocity Rescaling
pi
T
Tcpi
Anderson = 1%, 0.1%, 0.05% Velocities are sampled from N(0,T)
Langevin
BBK algorithmpn+1/2 = pn t
2V(qn) t
2(qn)M1pn +
t2(qn)Gn
qn+1 = qn + tM1pn+1/2
pn+1
= pn+1/2
t
2 V(qn+1
)t
2 (qn+1
)M1
pn+1
+t
2 (qn+1
)Gn+1
chosen to be constant =
2MkBT
G sampled from a standard normal distribution
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 23 / 29
Velocity Rescaling Instantaneous Temperature
7/30/2019 Thermostat MD
24/29
Velocity Rescaling - Instantaneous Temperature
0 2000 4000 6000 8000 100000
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
Temp
time step
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 24 / 29
Anderson Instantaneous Temperature
7/30/2019 Thermostat MD
25/29
Anderson - Instantaneous Temperature
0 2000 4000 6000 8000 100000
1
2
3
4
5
6
Temp
time step
= 1%
= 0.1%
= 0.05%
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 25 / 29
Anderson - Histogram of Velocity Distribution
7/30/2019 Thermostat MD
26/29
Anderson - Histogram of Velocity Distribution
10 5 0 5 100
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
velocity
p(v)
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 26 / 29
Langevin Instantaneous Temperature
7/30/2019 Thermostat MD
27/29
Langevin Instantaneous Temperature
0 2 4 6 8 10
x 104
1
2
3
4
5
6
7
Temp
time step
= 0.1
= 0.01
= 0.001
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 27 / 29
Langevin - Histogram of Velocity Distribution
7/30/2019 Thermostat MD
28/29
Langevin Histogram of Velocity Distribution
10 5 0 5 100
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
p(v)
velocity
Alden Johnson, Teresa Johnson, Aimee Khan (UMass Amherst)Thermostats in MD Simulations December 6th, 2012 28 / 29
7/30/2019 Thermostat MD
29/29