Professional Documents
Culture Documents
Wu (2006) PDF
Wu (2006) PDF
JOURNAL REVIEW
DOI 10.1002/aic.10713
Published online October 28, 2005 in Wiley InterScience (www.interscience.wiley.com).
冓冘
phase and interfacial properties of fluids and soft materials.
Following a brief introduction to the basic concepts and new
developments toward a unified free-energy functional reflect-
ing various components of intermolecular forces, the main text
i共R兲 ⫽
Ni
冔
␦(r1 , r2 , . . . , rm i) (1)
where the subscript j stands for an atom and the subscript i for
a molecule. If concern is restricted to atomic fluids such as
argon, the molecular density is simply a function of a normal
three-dimensional vector standing for the position of an atom.
Even though, in a uniform atomic fluid, the spatial average of
the atomic density (r) becomes identical to the bulk density
av, the local density (r) can be infinitely large when the atoms
are localized such as in an idealized crystalline solid. For
example, in a one-component hard-sphere system at the close-
packed limit, the average reduced density is av3 ⫽ 公2,
where denotes the hard-sphere diameter. In this case, the hard
spheres are arranged onto a face-centered cubic lattice and the
local density is infinite at the lattice sites and zero otherwise. Figure 2. DFT is applicable to both uniform and inhomo-
The second law of thermodynamics requires that, for an open geneous systems.
system, the grand potential ⍀ must be minimized at equilib- Even in a uniform fluid such as liquid argon, the average local
rium. The grand potential is also a functional of the molecular density near an arbitrarily tagged molecule is inhomogeneous.
density i(R), defined as The ratio of the local density (r) and the average density av
defines a radial distribution function, an important quantity for
describing the microscopic structure. A similar concept can be
⫽
k BT
12 冕 drr2 c共r, 兲 (6) F⫽ 冕 dr⌽关共r兲兴 ⫹
kB T
4 冕 dr1 dr2 c
冕
exact, a precise expression of the intrinsic Helmholtz energy as
F⫽ drd共r, 兲关ln 共r, 兲 ⫺ 1兴 a functional of the molecular density profiles is unknown for
most systems of practical interest.†† Formulation of the Helm-
冕 冕
holtz energy functional is a task essentially equivalent to enu-
kB T meration of the statistical partition function for the particular
⫺ dr1 d1 dr2 d2 共r1 , 1 兲共r2 , 2 兲 f共r1 ⫺ r2 , 1 , 1 兲 system under investigation. A viable approach, practiced in
2
classical thermodynamics for more than 100 years, is to divide
(7) the Helmholtz energy into an ideal part and an excess part. The
ideal part represents the contribution of an ideal gas where all
where f ⫽ e⫺u ⫺ 1 is the Mayer function,  ⫽ 1/(kBT), and u nonbonded interactions are turned off; the excess part accounts
stands for the pair potential between the anisotropic molecules. for interactions leading to the thermodynamic nonideality.
The first term on the right-hand side of Eq. 7 represents the For an ideal-gas system that is free of nonbonded interac-
Helmholtz energy functional of noninteracting (ideal) rigid tions, the Helmholtz energy functional is known exactly:
冕 再
nection between free energy and molecular structure (density
profiles) rather than on the overwhelming data generated by the n1 n2 ⫺ nV1nV2
trajectories of all constituent particles in molecular simulation.
ex
F hs ⫽ k BT dr ⫺n0 ln(1 ⫺ n3 ) ⫹
1 ⫺ n3
A microscopic state of a many-body system entails 6N degrees
of freedom (where N is the number of spherical particles in the
system), whereas the density of a spherical object is a simple
function that depends solely on the three-dimensional vector r.
⫹
1
36 冋
n3 ln(1 ⫺ n3 ) ⫹
n32
(1 ⫺ n3 )2 册
(n23 ⫺ 3n2 nV2nV2)
n33 冎 (11)
冘 冕
spherical approximation (MSA).79 Unlike various mean-field
ex
F att ⫽ F att
ex
共 i0兲 ⫹ att
dr⌬i 共r兲 theories derived from the Poisson–Boltzmann (PB) equation,
i
i the quadratic approximation is often sufficient to capture a
冘 冕冕
number of counterintuitive electrostatic phenomena observed
kB T in solutions containing multivalent ions such as charge inver-
⫺ drdrⴕcijatt 共兩r ⫺ rⴕ兩兲⌬i 共r兲⌬j 共rⴕ兲 (12) sion of macroions and attraction between like charges.80,81 The
2
i, j limitation of the PB equation arises from its neglect of the size
of small ions and the correlation of charge distributions.
where F exatt ( i ) is the attractive part of the excess Helmholtz
0
Correlations Attributed to Chain Connectivity. Applica-
energy of the reference bulk fluid and 0i is the average density tion of classical DFT to polymeric systems was first discussed
of an atomic component i. This “semiquadratic” approach by Chandler, McCoy, and Singer (CMS) in 1986.82 The past
requires the excess chemical potential ex att and the direct cor- few years have witnessed enormous growth in this area.83-88
relation function c att ij (r) of a uniform atomic fluid as input. For Early versions of polymer DFT were heavily influenced by the
that purpose, the analytical correlation functions derived from self-consistent field theory and by Landau expansions for the
the first-order mean-spherical approximation (FMSA) are par- selection of the reference system or for the formulation of the
ticularly useful.55-60 It has been demonstrated that, at least for free-energy functional. These influences remain evident in dif-
relatively simple models such as Lennard–Jones systems and ferent versions of dynamic DFT.14,30 Most recent applications
coarse-grained models of polymers, excellent agreement be- of DFT, however, adopt segment-level intermolecular forces
tween theory and simulation results can be attained.61 Because following either the CMS theory or the generalized thermody-
the analytical expressions of both the excess chemical poten- namic perturbation theory.89 The former is based on a quadratic
tials and direct correlation functions are readily available from density expansion of the Helmholtz energy functional with
FMSA, the numerical implementation and computational cost respect to that for a system of ideal chains, much like the
of the quadratic approximation are very comparable to those methods used for simple or charged fluids.90-92 This approach
for a mean-field approach. requires as input the direct correlation functions from the
Weak Association. Another important component in a con- polymer integral-equation theory (that is, PRISM93) and the
ventional force field is the formation of chemical or hydrogen intramolecular correlation functions from a single-chain Monte
bonds as in associating fluids. For fluid-phase equilibrium Carlo simulation. Similar to the hypernetted-chain approxima-
calculations, the thermodynamic perturbation theory62,63 has tion for simple fluids, CMS theory is unable to describe phase
been successfully used to develop the statistical associating transitions such as liquid–vapor coexistence. The generalized
fluid theory (SAFT), a generic equation of state for associating thermodynamic perturbation theory was initially introduced by
fluids and also for polymers.64 By incorporation of basic con- Kierlik and Rosinburg,89 who built on earlier work by Wood-
cepts from DFT, various extensions of SAFT have been ap- ward.94,95 In this approach, the Helmholtz energy functional
plied to inhomogeneous associating fluids near a hard wall, in includes an exact formalism for the ideal chains that retains the
slit pores, and at vapor–liquid or liquid–liquid interfaces.65-73 A details of bond connectivity and an excess part accounting for
relatively simple version of the excess Helmholtz energy func- the contributions from all nonbonded inter- and intramolecular
tional was recently derived by introducing the scalar- and interactions. The excess Helmholtz energy functional is ex-
vector-weighted densities of FMT into the SAFT equation of pressed in terms of a weighted-density approximation for short-
state: range forces and a first-order perturbation theory for chain
correlations.61,96,97 For systems containing only homopoly-
冘 冕 drn 关ln X
mers, the contribution of chain connectivity to the excess
共 A兲
ex
F ass ⫽ k BT 0 i i ⫺ Xi共 A兲 /2 ⫹ 1/2兴 (13) Helmholtz energy functional is
where the subscript i stands for the atomic species i with the
ex
F ch ⫽ k BT i 0,i i ii ␣ (15)
association site A. The inhomogeneous factor i is related to i
Liquid–liquid interface
Demixing of a liquid mixture may occur at both high and
low temperatures. For a binary liquid mixture consisting of
nonpolar molecules of similar size, demixing occurs only be-
low an upper critical solution temperature (UCST). However,
Figure 4. Surface tension of polydimethylsiloxane (PDMS) the situation is much more complicated if the molecules of two
from DFT (lines) and from experiment (symbols).
(A) The effect of temperature on surface tension for two
monodisperse fractions of PDMS with molecular weights
32,000 and 770 Daltons, respectively. (B) The surface tension
of a PDMS blend vs. the weight fraction of the low molecular
weight polymer. The experimental data are from Dee and
Sauer113 and the lines are calculated by Kierlik et al.107
冕 冕 冕
proposed by Stoyanov and coworkers.124 In this method, the
1 excess surface Helmholtz energy functional consists of a local
P T共 z兲/k BT ⫽ 共 z兲 ⫺ dr 12 d1 d2 f共r12 , 1 , 2 兲 density approximation for the chain elastic energy of surfactant
2
molecules and the surface osmotic pressure124
⫻ 冕
0
1
d共z ⫺ z12 , 1 兲关 z ⫹ 共1 ⫺ 兲z12 , 2 兴 (17) F ex ⫽ 冕 dz关K sz 2 ⫹ 共1 ⫺ 2 兲k BT 2/ 2v solv兴 (18)
of these classical methods, DFT is able to describe quantita- has also been used for characterization of amorphous materials
tively all six types of adsorption isotherm within a unified such as activated carbons, oxides, or silica gels where we lack
framework. For example, Balbuena and Gubbins131 investi- definite knowledge of the pore structure.135
gated adsorption of an argon-like fluid in a graphite-like slit The mean-field approximation for representing the van der
pore using a nonlocal density functional theory (NLDFT). Waals attractions between gas molecules becomes inadequate
Their model system consists of Lennard–Jones molecules ad- when the surfaces are weakly attractive. As an improvement, a
sorbed in a single slit pore where the wall potential is specified number of more sophisticated versions of DFT have been
by Steele’s 10-4-3 potential proposed in the past few years.136-138 To illustrate, Figure 10
shows density profiles for a Lennard–Jones fluid near a hard
冋冉 冊 冉 冊
2 w w w4
册
10 4
wall and in an attractive slit pore calculated from Monte Carlo
s共 z兲 ⫽ w ⫺ ⫺ (20)
5 z z 3⌬( z ⫹ 0.61⌬) 3 simulations and from two different versions of DFT.54 Al-
though the mean-field theory is excellent for adsorption in the
where the parameters w, w, and ⌬ are related to the properties attractive slit pore, its performance near a hard wall is inferior
of the solid substrate. The excess Helmholtz energy functional to that of a non-mean-field version that takes appropriate ac-
includes a hard-sphere term proposed by Tarazona132 and an count of correlation effects.
attractive term represented by the mean-field approximation In addition to adsorption and capillary condensation of sim-
ple fluids, DFT is useful for describing liquid–liquid demixing
冕冕
under confinement. For simple binary mixtures of nonpolar
1
F ex关 共r兲兴 ⫽ Fhs
ex
关共r兲兴 ⫹ dr1 dr2 uatt 共r1 ⫺ r2 兲共r1 兲共r2 兲 liquids in a slit-like pore, classical DFT predicts that, in gen-
2 eral, the confinement and the wall attraction depress the UCST
(21) and the region of immiscibility.117 If the confining surface
preferentially attracts one component in the binary mixture, the
Balbuena and Gubbins131 studied the variation of adsorption confinement leads to a shift of the coexistence curve toward the
behavior as a function of surface energy and pore width. They phase that is rich in the preferred component.139 Similar cal-
found that NLDFT is able to reproduce all six types of adsorp- culations have been reported for binary mixtures of associating
tion isotherms according to the IUPAC classification. More liquids that exhibit closed-loop immiscibility.116 For polymer
recently, Neimark and coworkers133 also demonstrated that, as mixtures, a DFT calculation indicates that, when the confining
reproduced in Figure 9, DFT properly bridges the gap between surface is attractive, confinement in slit-like pores leads to a
molecular simulations and phenomenological equations, two-step demixing transition.140 The first step of demixing
thereby providing a description of capillary condensation/evap- occurs within a few layers adjacent to the attractive surface and
oration at all length scales. the second step occurs in the entire pore. The multistep demix-
Application of DFT for gas adsorption is not limited to ing transitions of confined Gaussian-core fluids are strikingly
nanostructured materials such as zeolites, MCM-41, or alu- similar to the condensation and layer transitions of associating
minophosphates where the microscopic structure can be deter- fluids and water in attractive micropores.141,142 The effects of
mined by X-ray or neutron diffraction.134 In recent years, DFT confinement on liquid–liquid separations revealed by DFT
Li
exp共⫺␣兩r ⫺ Li 兩2 兲 (22)
recently.238,239 Using a nonlocal density functional for the free
energy of effective hard spheres in the Parson–Lee model,
Somoza and Tarazona240 predicted first- and second-order
where the inverse width of the Gaussian parameter ␣ provides phase transitions among nematic, smectic-A, and smectic-C
a measure of the density localization in the crystal and {Li} phases in hard spherocylinders. Figure 19 schematically shows
represnts the real-space lattice vectors. With different methods the phase behavior of hard cylinders. It has been demonstrated
for the paramerization of the solid density, DFT has been that the Somoza–Tarazona theory is also applicable to inho-
successfully applied to representing fluid–solid and solid–solid mogeneous spherocylinders and their mixtures.241-244 Applica-
transitions in hard spheres,206-209 simple liquids,210,211 anisotro- tion of Rosenfeld’s fundamental measure theory to binary
pic fluids,212-214 and various models of colloidal dispersions.215- hard-platelet fluids reveals that the bulk phase diagram includes
221 DFT is useful not only for freezing/melting transitions of
an isotropic phase, one or two nematic phases of different
one-component systems but also for mixtures where exotic composition, and a columnar phase.245
fluid–solid and solid–solid transitions have been identi-
fied.222-224 Application of DFT to a fluid–solid interface (with Gay–Berne model
explicit consideration of solid inhomogeneity) is able to cap- Liquid crystals are often divided into two basic classifica-
ture the partially periodic microscopic structure and the orien- tions: thermotropic and lyotropic. The phase transitions of
tation variation of the interfacial tension.225,226 More recently, lyotropic liquid crystals mainly depend on concentration,
DFT has been used to explore structure formation in colloids whereas those of thermotropic liquid crystals depend on both
under an external field.227-229 The periodic potential-induced temperature and concentration. Whereas hard spherocylinders
freezing and reentrant melting of colloidal particles at a surface and platelets provide a good representation of lyotropic liquid
predicted by DFT are all in good agreement with experi- crystals, they are not useful for thermotropic liquid crystals
ment.230,231 because the attractions between these particles are not in-
cluded. Much theoretical work on the phase transitions of
Phase Behavior of Liquid Crystals thermotropic liquid crystals is based on the Gay–Berne mod-
Liquid crystals are characterized by orientational anisotropy el.246-248 The interaction potential between two axially asym-
on a macroscopic scale arising from partial ordering of non- metric molecules is
spherical molecules. The equilibrium density profile in a liquid
再冋 0
册
crystal depends on both the distribution of the center of mass 12