Blue moon ensemble 339
of the transformation is not unity. LetJ(q) =|∂(r 1 ,...,rN)/∂(q 1 ,...,q 3 N)|denote this
Jacobian. Eqn. (8.7.10) then becomes
1
P(s)
dP
ds
=
1
N!λ^3 NQ(N,V,T)
∫
d^3 Nq J(q)e−βU ̃(q)∂s∂δ(q 1 −s)
〈δ(q 1 −s)〉
=
1
N!λ^3 NQ(N,V,T)
∫
d^3 Nqe−β(
U ̃(q)−kTlnJ(q))∂
∂sδ(q^1 −s)
〈δ(q 1 −s)〉
, (8.7.11)
where, in the last line, the Jacobian has been exponentiated. Changing the derivative
∂/∂sto∂/∂q 1 and performing the integration by parts as was done in eqn. (8.7.7),we
obtain
1
P(s)
dP
ds
=
1
N!λ^3 NQ(N,V,T)
∫
d^3 Nq∂q∂ 1 e−β(
U ̃(q)−kTlnJ(q))
δ(q 1 −s)
〈δ(q 1 −s)〉
=−
β
N!λ^3 NQ(N,V,T)
×
∫
d^3 Nq
[
∂U ̃
∂q 1 −kT
∂
∂q 1 lnJ(q)
]
e−β(
U ̃(q)−kTlnJ(q))
δ(q 1 −s)
〈δ(q 1 −s)〉
=−β
〈[
∂U ̃
∂q 1
−kT
∂
∂q 1
lnJ(q)
]〉cond
s
. (8.7.12)
Therefore, the free energy profile becomes
A(q) =A(s(i)) +
∫q
s(i)
ds
〈[
∂U ̃
∂q 1
−kT
∂
∂q 1
lnJ(q)
]〉cond
s
. (8.7.13)
The derivative ofU ̃, the transformed potential, can be computed form the original
potentialUusing the chain rule
∂U ̃
∂q 1
=
∑N
i=1
∂U
∂ri
·
∂ri
∂q 1
. (8.7.14)
Eqn. (8.7.13) can be applied straightforwardly to simple reaction coordinates for which
the full transformation to generalized coordinates is known. Let us return to the prob-
lem of computing conditional ensemble averages from constrained molecular dynamics.
We will use this discussion as a vehicle for introducing yet another expression forA(q)
that does not require a coordinate transformation at all.
Recall from Section 1.9 that the equations of motion for a system subject to a
single holonomic constraintσ(r 1 ,...,rN) = 0 are
r ̇i=
pi
mi