Anisotropic cells 231
asdh 0 =V−^3 dh. In addition, det(h 0 ) = det(h)/V. Thus, substituting the cell-matrix
transformation into eqn. (5.6.4) yields
∆(N,P,T) =
1
V 0
∫∞
0
dV
∫
dhV−^3 e−βPVQ(N,h,T)δ
(
1
V
det(h)− 1
)
=
1
V 0
∫∞
0
dV
∫
dhV−^3 e−βPVQ(N,h,T)V δ(det(h)−V)
=
1
V 0
∫∞
0
dV
∫
dhV−^2 e−βPVQ(N,h,T)δ(det(h)−V), (5.6.5)
where the dependence ofQonVandh 0 has been expressed as an equivalent depen-
dence only onh. Performing the integration over the volume using theδ-function, we
obtain for the partition function
∆(N,P,T) =
1
V 0
∫
dh[det(h)]−^2 e−βPdet(h)Q(N,h,T). (5.6.6)
In an arbitrary numberdof spatial dimensions, the transformation ish=V^1 /dh 0 ,
and the partition function becomes
∆(N,P,T) =
1
V 0
∫
dh[det(h)]^1 −de−βPdet(h)Q(N,h,T). (5.6.7)
Before describing the generalization of the virial theorems of Section 5.4, we note
that the internal pressure of a canonical ensemble with a fixed cellmatrixhdescribing
an anisotropic system cannot be described by a single scalar quantity as is possible for
an isotropic system. Rather, atensoris needed; this tensor is known as thepressure
tensor,P(int). Since the Helmholtz free energyA=A(N,h,T) depends on the full cell
matrix, the pressure tensor, which is a 3×3 (or rank 2) tensor, has components given
by
Pαβ(int)=−
1
det(h)
∑^3
γ=1
hβγ
(
∂A
∂hαγ
)
N,T
, (5.6.8)
which can be expressed in terms of the canonical partition functionas
Pαβ(int)=
kT
det(h)
∑^3
γ=1
hβγ
(
∂lnQ
∂hαγ
)
N,T
. (5.6.9)
In Section 5.7, an appropriate microscopic estimator for the pressure tensor will be
derived.
If we now consider the average of the pressure tensor in the isothermal-isobaric
ensemble, atensorialversion of virial theorem can be proved for an applied isotropic
external pressureP. The average of the internal pressure tensor is
〈Pαβ(int)〉=
1
∆(N,P,T)
∫
dh[det(h)]−^2 e−βPdet(h)
kTQ(N,h,T)
det(h)
∑^3
γ=1
hβγ
(
∂lnQ
∂hαγ
)
N,T