Reformulation of Ensemble Averages via Coordinate Mapping for

fluid phase properties

Akshara Goyal, Andrew J. Schultz and David A. Kofke
Department of Chemical and Biological Engineering, University at Buffalo, SUNY

Objectives Coordinate Mapping Mapping Framework Conclusions

1 Develop coordinate mapping as an alternative Coordinate Mapping is a mathematical technique • Calculate FE derivative w.r.t. a parameter (ν): 1 Preliminary results show that mapped averaging
approach to perturbation theory for calculating developed by Kofke group at UB to improve free- Aν = − < Jν > + < U ν > (7) technique is efficient over conventional
ensemble average of fluid-phase properties. energy (FE) calculations. It applies Jarzynski’s method at low density range.
2 Calculate energy of a Lennard-Jones fluid FE perturbation technique and combines it with Jν = ∇ · V ν (8) 2 Inconsistencies at higher densities could be

system via mapped averaging technique. atomic displacements to calculate FE derivatives. U ν = uν + V ν · ∇u (9) either due to potential truncation corrections or
• Traditional Free Energy difference by Zwanzig: • Calculate Mapping field (V ) by solving: ν finite size effects.
−∆(βU )
D E    
∆(βA) = −ln e (3) ∂ p p ν
Background +∇· V  =0 (10)
• Method converges poorly if little overlap in ∂ν q q Future Research
Z
• What is an Ensemble? configuration space between ensembles 1 and 2 p=e −U
,q = pdx
1 Current equation to calculate Mapped Potential
An ensemble is a collection of different microstates • Targeted Free-Energy difference by Jarzynski: 2 ∂q(β)
(ρV ) ∂β can be refined for long range corrections.
(set of molecular positions and velocities) of a −∆(βU ) <u>=−
D E
∆(βA) = −ln Je (4) 2 q(β) 2 Existing model is based on spherically symmetric
molecular system. For example, Canonical ensem- 
The FE derivatives hence obtained can be used to βij
h ∂u(rij ) 1  and pairwise-additive potential. It can be
ble (NVT) describes all microstates with same total − + (Fi − Fj ) · r̂ij 
X
+ βij v
estimate thermodynamic properties. i<j ∂rij 2 further developed for multi-body interactions.
number of molecules, volume and temperature.
3 The application of Coordinate Mapping can be

extended towards calculation of radial

Motivation Preliminary Results distribution function g(r), virial
coefficients and other thermodynamic
• Results
from NVT MC simulations of LJ fluid Low density Plot
properties.
developed by Schultz et. al. [1] 0.002 T = 1.0
T = 1.5
Figure 1: An ensemble of six microstates in a system of fixed T = 2.0

volume and no. of molecules

U-Umap (LJ Units)

References
0

• What is Ensemble averaging? [1] Andrew J. Schultz, Sabry G. Moustafa, Weisong Lin,
Steven J. Weinstein, and David A. Kofke.
Consider a thermodynamic quantity A(Γ) defined -0.002 Reformulation of ensemble averages via coordinate
for each point in phase space, for example total mapping.
intermolecular energy. For a discrete set of mi- 0.02 0.04 0.06
Density (LJ Units)
0.08 0.1 Journal of Chemical Theory and Computation,
crostates, the ensemble average of A is: 12(4):1491–1498, 2016.
X Figure 3: Conventional vs Mapped Energy (ρ : 0.01 − 0.1) PMID: 26950263.
< A >= Ai π i (1) [2] Sabry G. Moustafa, Andrew J. Schultz, and David A.
High density Plot
For example: 0.03 Kofke.
T = 1.5 Very fast averaging of thermal properties of crystals by
U1 = 3, U2 = 5, U3 = 7, U4 = 2, U5 = 1, U6 = 9 T = 2.0
molecular simulation.
Assuming all states are equally probable: Figure 2: Conventional vs Mapped average pressure 0.015
Phys. Rev. E, 92:043303, Oct 2015.
U-Umap (LJ Units)

3+5+7+2+1+9 q̂ 2 β

< U >= = 4.5 βP = ρ − ρ +
X
< 0
r[v 2(r) 0
− u 2(r)] >
0

6 2 3V 
i<j
In the continuum phase space, for canonical ensem- *
1
+
-0.015
s Contact Information

0
+ v (r)[v 2(r) + (Fi − Fj ) · rˆji]  (5)
ble: 2
1 1 Z NZ N N N −βE
< A >= dp dr A(p , r )e -0.03
• Email: aksharag@buffalo.edu
dN
Qh N ! q̂ = −2B2 (6) 0.2 0.4 0.6
Density (LJ Units)
0.8 1

• Phone: +1 (716) 868 6397

(2) B2 is second virial coefficient Figure 4: Conventional vs Mapped Energy (ρ : 0.1 − 1.0)