1549380323-Statistical Mechanics Theory and Molecular Simulation

(jair2018) #1

500 Classical time-dependent statistical mechanics


this approximation. Interestingly, however, the approximation oflinear response the-
ory proves to be remarkably robust (Bianucciet al., 1996). Within linear response
theory, eqn. (13.2.12) reduces to


(

∂t
+iL 0

)


∆f(x,t) =−i∆L(t)f 0 (H(x)), (13.2.13)

which follows from the facts that∂f 0 /∂t= 0 andiL 0 f 0 (H(x)) = 0. In order to solve
eqn. (13.2.13), we take the driving force to be 0 fort <0, so that att= 0, the ensemble
is described byf(x,0) =f 0 (H(x)), and ∆f(x,0) = 0. Eqn. (13.2.13) is a simple first-
order inhomogeneous differential equation that can be solved usingthe unperturbed
classical propagator exp(iL 0 t) as an integrating factor. The solution that satisfies the
initial condition is


∆f(x,t) =−

∫t

0

dse−iL^0 (t−s)i∆L(s)f 0 (H(x)). (13.2.14)

In order to simplify eqn. (13.2.14), we note that


i∆L(s)f 0 (H(x)) = (iL(s)−iL 0 )f 0 (H(x))

=iL(s)f 0 (H(x))

= ̇x(s)·∇xf 0 (H(x)) (13.2.15)

sinceiL 0 f 0 (H(x)) = 0. However,


̇x(s)·∇xf 0 (H(x)) = ̇x(s)·
∂f 0
∂H

∂H


∂x

=


∂f 0
∂H

∑^3 N


i=1

[


p ̇i(s)

∂H


∂pi

+ ̇qi(s)

∂H


∂qi

]


=


∂f 0
∂H

∑^3 N


i=1

[


∂H


∂pi

(



∂H


∂qi

+Di(x)Fe(s)

)


+


∂H


∂qi

(


∂H


∂pi

+Ci(x)Fe(s)

)]


=


∂f 0
∂H

∑^3 N


i=1

[


Di(x)

∂H


∂pi

+Ci(x)

∂H


∂qi

]


Fe(s). (13.2.16)

The quantity


j(x) =−

∑^3 N


i=1

[


Di(x)

∂H


∂pi

+Ci(x)

∂H


∂qi

]


(13.2.17)


appearing in eqn. (13.2.16) is known as thedissipative flux. In terms of this quantity,
we have

Free download pdf