Introduction
The diffuse-interface model (DIM) was proposed by Korteweg [55 ] as an attempt to avoid the abrupt change of parameters in the models of liquid/vapour interfaces existing at the time. It is based on two assumption:
(i) The long-range attractive intermolecular force (the van der Waals force) can be modelled by a pair-wise potential, so that the force affecting a molecule is the algebraic sum of those exerted on it by other molecules.
(ii) The characteristic distance over which the van der Waals force acts is much smaller than the interfacial thickness.
The resulting representation of the molecular interaction has been incorporated into models of various phenomena, such as phase transitions in ferroelectric materials (Ginzburg [39 ] ), spinodal decomposition ([18 , 60 ] ), growth of, and oscillations in, crystals ([22 , 92 , 47 , 73 ] ), solidification of alloys ([91 , 71 ] ), phase separation in polymer blends ([93 , 63 ] ), electrowetting (Lu et al. [61 ] ), contact lines ([50 , 74 , 27 , 102 , 101 , 86 , 28 , 17 ] ), contact lines in liquids with surfactants ([105 , 106 ] ), Faraday instability ([16 , 15 ] ), Rayleigh–Taylor instability ([104 , 103 ] ), cavitation ([72 ] ), nucleation and collapse of bubbles ([66 , 65 , 36 , 38 ] ), capillary condensation (Pomeau [76 ] ; Benilov [8 ] ), liquid films ([5 , 8 ] ), tumor growth ([35 , 80 , 24 ] ), classification of high-dimensional data ([13 , 14 ] ), etc.
The present paper is mainly concerned with application of the DIM to evaporation of drops and condensation of vapour on a solid. These phenomena have been examined using the single-component version of the DIM [9 a] , where the fluid/gas interface is modelled by that between the liquid and vapour phases of the same fluid. The evaporation in this case was shown to be due to a flow caused by a weak imbalance of chemical potentials of the liquid and vapour phases.
It is unclear, however, how the results obtained via the single-component DIM are modified by the effect of air, whose density (at, say, 25°C) exceeds that of water vapour by a factor of more than 50. Furthermore, there are three additional physical effects in multicomponent fluids: diffusion (of water vapour in air), the Soret effect (thermodiffusion), and the Dufour effect (heat flux due to density gradient). Only one of these, the diffusion, has been examined before [e.g. 26 , 29 , 30 , 21 , 78 , 70 , 89 , 90 , 51 , 82 , 81 , 79 , 100 ] , but the models employed in these papers do not include the flow-induced evaporative flux discovered via the DIM. Thus, the multicomponent diffuse-interface model appears to be a tool describing all the mechanisms at work in evaporation/condensation of liquids into/from air.
The same can hopefully be said about the dynamics of contact lines, as most of the existing models work for some fluids (including water) only if the so-called slip length – effectively, the interfacial thickness – is set to be unrealistically small [75 , 99 , 77 , 57 , 10 ] .
Before using a new model, one generally needs to examine its basic properties, test it on problems with well-understood physics (to ensure that the mathematics captures it), and parameterise this model for the intended applications. These are the three aims of the present work in the context of the multicomponent DIM.
The following results are reported:
Multicomponent flat interfaces with monotonically-changing densities of the species (components) are all stable. This conclusion follows from the entropy principle and conservation laws.
Several new physical effects are described, the most interesting of which is evaporation of a flat liquid layer in contact with saturated vapour. This phenomenon appears to be similar to evaporation of drops surrounded by saturated vapour [26 , 30 , 6 , 7 , 9 ] , but with one important difference: the drops evaporate because the curvature of their boundary increases the effective saturation pressure (the so-called Kelvin effect), making the saturated vapour effectively undersaturated and, thus, encouraging evaporation. This explanation is clearly inapplicable to layers with flat boundaries – yet they evaporate anyway. The actual mechanism is based on the long-range interaction of the liquid/vapour interface with the substrate – which implies that, for macroscopic liquid films, this effect is weak. It should be important, however, for nano-films whose thickness is comparable to the interfacial thickness.
The multicomponent DIM is parameterised for water/air interfaces at normal conditions, which can be used in the future for modelling evaporation and condensation of water in the Earth’s atmosphere or the dynamics of contact lines of water drops.
The paper has the following structure. In section 2, the problem is formulated mathematically. Section 3 examines the entropy principle and conservation laws. In section 4, the governing equations are nondimensionalised and the main nondimensional parameters are identified. Sections 5–7 examine basic solutions of the DIM, and in section 8, the DIM is parameterised for water/air interfaces. § 9 summarises the results obtained, including the effects partly mentioned in item 2 of the above list.
Formulation
Thermodynamics
When studying hydrodynamics of a compressible fluid, one has to deal with its thermodynamic properties. In this section, they are described briefly and in a self-contained form (for the benefit of readers specializing is incompressible hydrodynamics).
Consider an N N N -component compressible non-ideal fluid, characterised by the temperature T T T and partial mass densities ρ i \rho_i ρ i where i = 1... N i=1...N i = 1... N . The fluid’s thermodynamic properties are fully described by two functions: the internal energy e ( ρ 1 . . . ρ N , T ) e(\rho_1...\rho_N,T) e ( ρ 1 ... ρ N , T ) and entropy s ( ρ 1 . . . ρ N , T ) s(\rho_1...\rho_N,T) s ( ρ 1 ... ρ N , T ) – both specific, i.e., per unit mass. The dependence of the fluid pressure p p p on ( ρ 1 . . . ρ N , T ) (\rho_1...\rho_N,T) ( ρ 1 ... ρ N , T ) , or the equation of state, is defined by
p = ρ ∑ i ρ i ( ∂ e ∂ ρ i − T ∂ s ∂ ρ i ) . (1) p=\rho\sum_i \rho_i\left(\frac{\partial e}{\partial\rho_i}-T\frac{\partial s}{\partial\rho_i}\right). \tag*{(1)} p = ρ i ∑ ρ i ( ∂ ρ i ∂ e − T ∂ ρ i ∂ s ) . (1)
[e.g. 43 ] , where
ρ = ∑ i ρ i (2) \rho=\sum_i \rho_i \tag*{(2)} ρ = i ∑ ρ i (2)
is the total density (here and hereinafter, the summation is implied to be from 1 to N N N unless stated otherwise). The partial chemical potentials, in turn, are given by
G i = ∂ ( ρ e ) ∂ ρ i − T ∂ ( ρ s ) ∂ ρ i . (3) G_i=\frac{\partial(\rho e)}{\partial\rho_i}-T\frac{\partial(\rho s)}{\partial\rho_i}. \tag*{(3)} G i = ∂ ρ i ∂ ( ρ e ) − T ∂ ρ i ∂ ( ρ s ) . (3)
Note that e ( ρ 1 . . . ρ N , T ) e(\rho_1...\rho_N,T) e ( ρ 1 ... ρ N , T ) and s ( ρ 1 . . . ρ N , T ) s(\rho_1...\rho_N,T) s ( ρ 1 ... ρ N , T ) are not fully arbitrary, but should satisfy the fundamental thermodynamic relation, or the Gibbs relation, which can be written in the form
∂ e ∂ T = T ∂ s ∂ T . (4) \frac{\partial e}{\partial T}=T\frac{\partial s}{\partial T}. \tag*{(4)} ∂ T ∂ e = T ∂ T ∂ s . (4)
The equivalence of this equality and the standard form of the Gibbs relation is shown in Appendix § 11 .
Using equations (2.1)–(2.4), one can derive the following identities:
∂ p ∂ T = ∑ i ρ i ∂ G i ∂ T + ρ s , (5) \frac{\partial p}{\partial T}=\sum_i \rho_i\frac{\partial G_i}{\partial T}+\rho s, \tag*{(5)} ∂ T ∂ p = i ∑ ρ i ∂ T ∂ G i + ρ s , (5)
∂ p ∂ ρ j = ∑ i ρ i ∂ G i ∂ ρ j , (6) \frac{\partial p}{\partial\rho_j}=\sum_i \rho_i\frac{\partial G_i}{\partial\rho_j}, \tag*{(6)} ∂ ρ j ∂ p = i ∑ ρ i ∂ ρ j ∂ G i , (6)
then rewrite (2.6) in the form
G i = ∂ ∂ ρ i ( ∑ j ρ j G j − p ) . (7) G_i=\frac{\partial}{\partial\rho_i}\left(\sum_j \rho_jG_j-p\right). \tag*{(7)} G i = ∂ ρ i ∂ ( j ∑ ρ j G j − p ) . (7)
In what follows, one also needs the specific heat capacity at constant volume,
c = ∂ e ∂ T . (8) c=\frac{\partial e}{\partial T}. \tag*{(8)} c = ∂ T ∂ e . (8)
where the traditional subscript V V V is omitted. In this paper, c c c is assumed to be positive, which is indeed the case for neutral fluids [62 ] .
Define also
a i = − ∂ e ∂ ρ i , (9) a_i=-\frac{\partial e}{\partial\rho_i}, \tag*{(9)} a i = − ∂ ρ i ∂ e , (9)
which will be referred to as the generalised van der Waals parameter (of the i i i -th species), and
B = − p − ρ ∑ i ρ i a i . (10) B=-p-\rho\sum_i \rho_i a_i. \tag*{(10)} B = − p − ρ i ∑ ρ i a i . (10)
B B B is not one of the standard thermodynamic functions, but is convenient to use when thermodynamics is coupled to hydrodynamics: as seen later, B B B characterises the release (consumption) of heat due to the fluid’s mechanical compression (expansion). Using equation of state (2.1) and definition (2.9) of the van der Waals parameter, one can rearrange expression (2.10) in the form
B = ρ T ∑ i ρ i ∂ s ∂ ρ i . (11) B=\rho T\sum_i \rho_i\frac{\partial s}{\partial\rho_i}. \tag*{(11)} B = ρT i ∑ ρ i ∂ ρ i ∂ s . (11)
Examples: the Enskog–Vlasov and van der Waals fluids
In the low-density limit, the specific internal energy and entropy of any fluid should tend to those of ideal gas, i.e.,
e ∼ T ρ ∑ i c i ρ i , s ∼ ln T ρ ∑ i c i ρ i − 1 ρ ∑ i R i ρ i ln ρ i as ρ i → 0 , e\sim\frac{T}{\rho}\sum_i c_i\rho_i,\qquad
s\sim\frac{\ln T}{\rho}\sum_i c_i\rho_i-\frac{1}{\rho}\sum_i R_i\rho_i\ln\rho_i
\qquad\text{as}\qquad\rho_i\to0, e ∼ ρ T i ∑ c i ρ i , s ∼ ρ ln T i ∑ c i ρ i − ρ 1 i ∑ R i ρ i ln ρ i as ρ i → 0 ,
where c i c_i c i is the specific partial heat capacity, and R i = k B / m i R_i=k_B/m_i R i = k B / m i is the specific gas constant (m i m_i m i is the molecular mass of the i i i -th species, and
k B = 1.380649 × 10 − 23 m 2 s − 2 k g K − 1 k_B=1.380649\times10^{-23}\,\mathrm{m^2\,s^{-2}\,kg\,K^{-1}} k B = 1.380649 × 1 0 − 23 m 2 s − 2 kg K − 1
is the Boltzmann constant). Note also that one can replace ln T → ln T / T ˉ \ln T\to\ln T/\bar{T} ln T → ln T / T ˉ and ln ρ → ln ρ / ρ ˉ \ln\rho\to\ln\rho/\bar{\rho} ln ρ → ln ρ / ρ ˉ , where are T ˉ \bar{T} T ˉ and ρ ˉ \bar{\rho} ρ ˉ are suitable dimensional scales. This would make the arguments of the logarithms nondimensional (with none of physically measurable parameters depending on T ˉ \bar{T} T ˉ and ρ ˉ \bar{\rho} ρ ˉ ).
A simple description of non-ideal fluids is delivered by the Enskog–Vlasov (EV) model, according to which
e = T ρ ∑ i c i ρ i − 1 ρ ∑ i j a i j ρ i ρ j , (12) e=\frac{T}{\rho}\sum_i c_i\rho_i-\frac{1}{\rho}\sum_{ij}a_{ij}\rho_i\rho_j, \tag*{(12)} e = ρ T i ∑ c i ρ i − ρ 1 ij ∑ a ij ρ i ρ j , (12)
s = ln T ρ ∑ i c i ρ i − 1 ρ ∑ i R i ρ i ln ρ i − Θ ( ρ 1 … ρ N ) , (13) s=\frac{\ln T}{\rho}\sum_i c_i\rho_i-\frac{1}{\rho}\sum_i R_i\rho_i\ln\rho_i-\Theta(\rho_1\ldots\rho_N), \tag*{(13)} s = ρ ln T i ∑ c i ρ i − ρ 1 i ∑ R i ρ i ln ρ i − Θ ( ρ 1 … ρ N ) , (13)
where Θ ( ρ 1 … ρ N ) \Theta(\rho_1\ldots\rho_N) Θ ( ρ 1 … ρ N ) is an arbitrary analytic function, and a i j a_{ij} a ij characterises the long-range attraction between molecules of the i i i -th and j j j -th species (with the implication that a i j = a j i a_{ij}=a_{ji} a ij = a j i ). For a pure fluid, a 11 a_{11} a 11 coincides with the generalised van der Waals parameter defined by (2.9) – hence, the notation.
One can readily verify that expressions (2.12)–(2.13) satisfy the Gibbs relation (2.4). They can be viewed as two-term expansions in small T T T , under an extra assumption that the heat capacity c c c and the generalised van der Waals parameter a i a_i a i are independent of T T T and are zeroth-degree homogeneous functions of ρ i \rho_i ρ i (for a pure fluid, c c c and a 1 a_1 a 1 do not depend on ρ 1 \rho_1 ρ 1 at all).
The EV fluid model originates from the Enskog–Vlasov kinetic theory (de Sobrino 1967; Grmela 1971) and, as such, is naturally suited for the use with the DIM which can be viewed as a hydrodynamic approximation of the EV kinetic equation [41 , 42 ] . Equations (2.12)–(2.13) work very well for inert gases [12 ] , they will be shown to work reasonably well for water, nitrogen, and oxygen (§ 8 of this paper).
The EV involves too many parameters to be used as an illustration of general results – so, in such cases, the simpler van der Waals model will be employed. It is a particular case of equations (2.12)–(2.13) with
Θ = − 1 ρ ∑ i R i ρ i ln ( 1 − ∑ j b j ρ j ) , (14) \Theta=-\frac{1}{\rho}\sum_i R_i\rho_i\ln\left(1-\sum_j b_j\rho_j\right), \tag*{(14)} Θ = − ρ 1 i ∑ R i ρ i ln ( 1 − j ∑ b j ρ j ) , (14)
where b i b_i b i is the reciprocal of the maximum density of the i i i -th species. Physically, 1 / b i 1/b_i 1/ b i can be interpreted as density of the closest packing.
Given expressions (2.12)–(2.14), definition (2.1) of the pressure and definition (2.3) of the chemical potential yield
p = T 1 − ∑ j b j ρ j ∑ i R i ρ i − ∑ i j a i j ρ i ρ j , (15) p=\frac{T}{1-\sum_j b_j\rho_j}\sum_i R_i\rho_i-\sum_{ij}a_{ij}\rho_i\rho_j, \tag*{(15)} p = 1 − ∑ j b j ρ j T i ∑ R i ρ i − ij ∑ a ij ρ i ρ j , (15)
G i = T R i ln ρ i 1 − ∑ j b j ρ j + T b i ∑ j R j ρ j 1 − ∑ j b j ρ j − 2 ∑ j a i j ρ j + T ( R i + c i − c i ln T ) . (16) G_i=TR_i\ln\frac{\rho_i}{1-\sum_j b_j\rho_j}+\frac{Tb_i\sum_j R_j\rho_j}{1-\sum_j b_j\rho_j}-2\sum_j a_{ij}\rho_j+T(R_i+c_i-c_i\ln T). \tag*{(16)} G i = T R i ln 1 − ∑ j b j ρ j ρ i + 1 − ∑ j b j ρ j T b i ∑ j R j ρ j − 2 j ∑ a ij ρ j + T ( R i + c i − c i ln T ) . (16)
For a pure fluid (N = 1 N=1 N = 1 ), equation (2.15) reduces to the classical van der Waals equation of state [96 ] .
Hydrodynamics
Consider a fluid flow characterised by the species densities ρ i ( r , t ) \rho_i(\mathbf{r},t) ρ i ( r , t ) , mass-averaged velocity v ( r , t ) \mathbf{v}(\mathbf{r},t) v ( r , t ) , and temperature T ( r , t ) T(\mathbf{r},t) T ( r , t ) , where r = [ x , y , z ] \mathbf{r}=[x,y,z] r = [ x , y , z ] is the position vector and t t t , the time. Let the species be affected by forces F i \mathbf{F}_i F i (which will be later identified with the van der Waals forces), and the fluid as a whole, affected by viscosity. The shear viscosity μ s \mu_s μ s and bulk viscosity μ b \mu_b μ b depend generally on ρ i \rho_i ρ i and T T T .
Let the flow be governed by following equations:
∂ ρ i ∂ t + ∇ ⋅ ( ρ i v + J i ) = 0 , (17) \frac{\partial\rho_i}{\partial t}+\nabla\cdot(\rho_i\mathbf{v}+\mathbf{J}_i)=0, \tag*{(17)} ∂ t ∂ ρ i + ∇ ⋅ ( ρ i v + J i ) = 0 , (17)
∂ ( ρ v ) ∂ t + ∇ ⋅ ( ρ v v ) = ∇ ⋅ Π + ∑ i ρ i F i − ∇ p , (18) \frac{\partial(\rho\mathbf{v})}{\partial t}+\nabla\cdot(\rho\mathbf{v}\mathbf{v})=\nabla\cdot\boldsymbol{\Pi}+\sum_i\rho_i\mathbf{F}_i-\nabla p, \tag*{(18)} ∂ t ∂ ( ρ v ) + ∇ ⋅ ( ρ vv ) = ∇ ⋅ Π + i ∑ ρ i F i − ∇ p , (18)
∂ ∂ t ( ρ e + 1 2 ρ ∣ v ∣ 2 ) + ∇ ⋅ [ ( ρ e + 1 2 ρ ∣ v ∣ 2 + p ) v − Π ⋅ v + Q ] = ∑ i F i ⋅ ( ρ i v + J i ) . (19) \frac{\partial}{\partial t}\left(\rho e+\frac{1}{2}\rho|\mathbf{v}|^2\right)+\nabla\cdot\left[\left(\rho e+\frac{1}{2}\rho|\mathbf{v}|^2+p\right)\mathbf{v}-\boldsymbol{\Pi}\cdot\mathbf{v}+\mathbf{Q}\right]=\sum_i\mathbf{F}_i\cdot(\rho_i\mathbf{v}+\mathbf{J}_i). \tag*{(19)} ∂ t ∂ ( ρ e + 2 1 ρ ∣ v ∣ 2 ) + ∇ ⋅ [ ( ρ e + 2 1 ρ ∣ v ∣ 2 + p ) v − Π ⋅ v + Q ] = i ∑ F i ⋅ ( ρ i v + J i ) . (19)
Here, the viscous stress tensor is
Π = μ s [ ∇ v + ( ∇ v ) t r − 2 3 I ( ∇ ⋅ v ) ] + μ b I ( ∇ ⋅ v ) , (20) \boldsymbol{\Pi}=\mu_{\mathrm{s}}\left[\nabla\mathbf{v}+(\nabla\mathbf{v})^{\mathrm{tr}}-\frac{2}{3}\mathbf{I}(\nabla\cdot\mathbf{v})\right]+\mu_{\mathrm{b}}\mathbf{I}(\nabla\cdot\mathbf{v}), \tag*{(20)} Π = μ s [ ∇ v + ( ∇ v ) tr − 3 2 I ( ∇ ⋅ v ) ] + μ b I ( ∇ ⋅ v ) , (20)
where the dotless product of two vectors (e.g., ∇ v \nabla\mathbf{v} ∇ v ) produces a second-order tensor, and the superscript t r \mathrm{tr} tr denotes transposition. The diffusion fluxes J i \mathbf{J}_i J i and the heat flux Q \mathbf{Q} Q are related to the forces F i \mathbf{F}_i F i , temperature T T T , and chemical potentials G i G_i G i by
J i = ∑ j D i j [ F j − T ∇ ( G j T ) ] − ζ i T ∇ T , (21) \mathbf{J}_i=\sum_j D_{ij}\left[\mathbf{F}_j-T\nabla\left(\frac{G_j}{T}\right)\right]-\frac{\zeta_i}{T}\nabla T, \tag*{(21)} J i = j ∑ D ij [ F j − T ∇ ( T G j ) ] − T ζ i ∇ T , (21)
Q = ∑ j ζ j [ F j − T ∇ ( G j T ) ] − κ T ∇ T , (22) \mathbf{Q}=\sum_j\zeta_j\left[\mathbf{F}_j-T\nabla\left(\frac{G_j}{T}\right)\right]-\frac{\kappa}{T}\nabla T, \tag*{(22)} Q = j ∑ ζ j [ F j − T ∇ ( T G j ) ] − T κ ∇ T , (22)
where D i j D_{ij} D ij , ζ i \zeta_i ζ i , and κ \kappa κ are the transport coefficients (all three depend generally on ρ i \rho_i ρ i and T T T ).
To understand the physical meaning of expressions (2.21)–(2.22), rewrite them in the form
J i = ∑ j D i j ′ F j − ∑ j D i j ′ ∇ ρ j − ζ i ρ ′ ∇ T , (23) \mathbf{J}_i=\sum_jD_{ij}'\mathbf{F}_j-\sum_jD_{ij}'\nabla\rho_j-\zeta_{i\rho}'\nabla T, \tag*{(23)} J i = j ∑ D ij ′ F j − j ∑ D ij ′ ∇ ρ j − ζ i ρ ′ ∇ T , (23)
Q = ∑ j ζ j F j − ∑ j ζ j ∑ k ∂ G j ∂ ρ k ∇ ρ k − κ ′ ∇ T , (24) \mathbf{Q}=\sum_j\zeta_j\mathbf{F}_j-\sum_j\zeta_j\sum_k\frac{\partial G_j}{\partial\rho_k}\nabla\rho_k-\kappa'\nabla T, \tag*{(24)} Q = j ∑ ζ j F j − j ∑ ζ j k ∑ ∂ ρ k ∂ G j ∇ ρ k − κ ′ ∇ T , (24)
where
D i k ′ = ∑ j D i j ∂ G j ∂ ρ k , ζ i ′ = ζ i T ρ + ∑ j D i j ρ ( ∂ G j ∂ T − G j T ) , D_{ik}'=\sum_jD_{ij}\frac{\partial G_j}{\partial\rho_k},\qquad
\zeta_i'=\frac{\zeta_i}{T\rho}+\sum_j\frac{D_{ij}}{\rho}\left(\frac{\partial G_j}{\partial T}-\frac{G_j}{T}\right), D ik ′ = j ∑ D ij ∂ ρ k ∂ G j , ζ i ′ = T ρ ζ i + j ∑ ρ D ij ( ∂ T ∂ G j − T G j ) ,
κ ′ = κ T + ∑ j ζ j ( ∂ G j ∂ T − G j T ) \kappa'=\frac{\kappa}{T}+\sum_j\zeta_j\left(\frac{\partial G_j}{\partial T}-\frac{G_j}{T}\right) κ ′ = T κ + j ∑ ζ j ( ∂ T ∂ G j − T G j )
are the standard diffusivities, thermodiffusivities, and thermal conductivity, respectively. The second term in expression (2.23) corresponds to the classical Fick Law (the fluxes depend linearly on the density gradients), and the last term in expression (2.24) characterises heat conduction described by the Fourier Law. The last term in (2.23) describes the Soret effect (∇ T \nabla T ∇ T gives rise to diffusion) and the second term in (2.24), the Dufour effect (∇ ρ j \nabla\rho_j ∇ ρ j gives rise to heat conduction).
The same four effects – diffusion, heat conduction, the Soret and Dufour effects – are described, obviously, by the original expressions (2.21)–(2.22), albeit in a form where the terms cannot be matched to a single effect each.
One might think that representing the fluxes in terms of ∇ ρ j \nabla\rho_j ∇ ρ j would be more natural than using ∇ ( G j / T ) \nabla(G_j/T) ∇ ( G j / T ) . Observe, however, that the coefficient of ( ∇ T ) / T (\nabla T)/T ( ∇ T ) / T in expression (2.21) coincides with the coefficient of [ F j − T ∇ ( G j / T ) ] [\mathbf{F}_j-T\nabla(G_j/T)] [ F j − T ∇ ( G j / T )] in (2.22). This symmetry reflects the so-called Onsager reciprocal relations [34 ] , which also imply
D i j = D j i , D_{ij}=D_{ji}, D ij = D j i ,
i.e., the diffusion of an i i i -th species in a j j j -th species occurs the same way as that of the j j j -th species in the i i i -th species. It should also be assumed that the extended transport matrix,
D i j ( e x t ) = [ ζ 1 D i j ⋱ ζ 1 ⋯ ζ N κ ] D_{ij}^{(\mathrm{ext})} =
\begin{bmatrix}
\zeta_1 & & & \\
& D_{ij} & & \\
& & \ddots& \\
\zeta_1 & \cdots& \zeta_N & \kappa
\end{bmatrix} D ij ( ext ) = ζ 1 ζ 1 D ij ⋯ ⋱ ζ N κ
is positive semidefinite (D i j ( e x t ) ⪰ 0 D_{ij}^{(\mathrm{ext})} \succeq0 D ij ( ext ) ⪰ 0 ) – i.e.,
∑ i = 1 N + 1 ∑ j = 1 N + 1 d i D j i ( e x t ) d j ≥ 0 , \sum_{i=1}^{N+1}\sum_{j=1}^{N+1}d_iD_{ji}^{(\mathrm{ext})}d_j\geq0, i = 1 ∑ N + 1 j = 1 ∑ N + 1 d i D j i ( ext ) d j ≥ 0 ,
for all ( N + 1 ) (N+1) ( N + 1 ) -dimensional arrays d i d_i d i . As seen later, this property is essential for the entropy principle to hold.
Furthermore, the transport coefficients should be such that
∑ i D i j = 0 , ∑ i ζ i = 0. (25) \sum_i D_{ij}=0,\qquad\sum_i\zeta_i=0. \tag*{(25)} i ∑ D ij = 0 , i ∑ ζ i = 0. (25)
As a result, the density equation (2.17) and expressions (2.21) for the diffusion fluxes imply that
∂ ρ ∂ t + ∇ ⋅ ( ρ v ) = 0 , (26) \frac{\partial\rho}{\partial t}+\nabla\cdot(\rho\mathbf{v})=0, \tag*{(26)} ∂ t ∂ ρ + ∇ ⋅ ( ρ v ) = 0 , (26)
where ρ \rho ρ is the total density given by (2.2). Observe that, for a pure fluid, restrictions (2.25) can be satisfied only if D 11 = 0 D_{11}=0 D 11 = 0 and ζ 1 = 0 \zeta_1=0 ζ 1 = 0 , which means that pure fluids neither diffuse nor thermodiffuse.
Equations (2.17)–(2.22) have been first derived from the thermodynamics of irreversible processes [68 ] . They were also derived from statistical mechanics [2 , 69 ] and non-equilibrium statistical thermodynamics [53 ] – for more references, see [40 ] . A derivation of the small-density version of (2.17)–(2.22) from the Boltzmann kinetic equation can be found in any textbook on kinetic theory [e.g. 34 ] .
In all these cases, the derived expressions for the transport coefficients automatically satisfy the Onsager relations and the rest of the properties listed above.
Note also that a reduction of the above equations for a binary fluid with no Soret and Dufour effects (N = 2 N=2 N = 2 , ζ i = 0 \zeta_i=0 ζ i = 0 ) was used by Liu et al. [59 ] to show that such a model is able to describe the phase equilibrium for a real binary mixture of CO2 _2 2 and ethanol.
Alternative forms of the momentum and energy equations
Since the transport fluxes (2.21)–(2.22) are expressed in terms of G i G_i G i and T T T (not ρ i \rho_i ρ i and T T T ), it is convenient to do the same for the pressure gradient in the momentum equation (2.18). Recalling identities (2.6)–(2.5) which imply
∇ p = ∑ i ρ i ∇ G i + ρ s ∇ T , \nabla p=\sum_i\rho_i\nabla G_i+\rho_s\nabla T, ∇ p = i ∑ ρ i ∇ G i + ρ s ∇ T ,
and using equation (2.26) to simply the left-hand side of the momentum equation (2.18), one reduces it to
ρ [ ∂ v ∂ t + ( v ⋅ ∇ ) v ] = ∇ ⋅ Π + ∑ i ρ i ( F i − ∇ G i ) − ρ s ∇ T . (27) \rho\left[\frac{\partial\mathbf{v}}{\partial t}+(\mathbf{v}\cdot\nabla)\mathbf{v}\right]
=\nabla\cdot\boldsymbol{\Pi}+\sum_i\rho_i(\mathbf{F}_i-\nabla G_i)-\rho_s\nabla T. \tag*{(27)} ρ [ ∂ t ∂ v + ( v ⋅ ∇ ) v ] = ∇ ⋅ Π + i ∑ ρ i ( F i − ∇ G i ) − ρ s ∇ T . (27)
The energy equation, in turn, can be conveniently rewritten in terms of the temperature (which is a measurable quantity, unlike the internal energy e e e ). Replacing, thus, in Eq. (19) ,
∂ e ∂ t = ∑ i ∂ e ∂ ρ i ∂ ρ i ∂ t + ∂ e ∂ T ∂ T ∂ t , \frac{\partial e}{\partial t} = \sum_i \frac{\partial e}{\partial\rho_i}\frac{\partial\rho_i}{\partial t} + \frac{\partial e}{\partial T}\frac{\partial T}{\partial t}, ∂ t ∂ e = i ∑ ∂ ρ i ∂ e ∂ t ∂ ρ i + ∂ T ∂ e ∂ t ∂ T ,
one should use the density equation to eliminate ∂ ρ i / ∂ t \partial\rho_i/\partial t ∂ ρ i / ∂ t . Using then equations (2.17) and (2.27), and recalling identities (2.8)–(2.9) and (2.11), one obtains
ρ c ( ∂ T ∂ t + v ⋅ ∇ T ) + ∇ ⋅ Q = Π : ( ∇ v ) + B ∇ ⋅ v + ∑ i ( F i − ρ a i ∇ ) ⋅ J i , (28) \rho c\left(\frac{\partial T}{\partial t}+\mathbf{v}\cdot\nabla T\right)+\nabla\cdot\mathbf{Q}=\mathbf{\Pi}:(\nabla\mathbf{v})+B\nabla\cdot\mathbf{v}+\sum_i\left(\mathbf{F}_i-\rho a_i\nabla\right)\cdot\mathbf{J}_i, \tag*{(28)} ρ c ( ∂ t ∂ T + v ⋅ ∇ T ) + ∇ ⋅ Q = Π : ( ∇ v ) + B ∇ ⋅ v + i ∑ ( F i − ρ a i ∇ ) ⋅ J i , (28)
where the symbol “:” denotes double scalar product of two tensors. The first term on the right-hand side of equation (2.28) describes heat production by viscosity and the second term, that by fluid compression or expansion.
The van der Waals force
Assume that a molecule of a j j j -th species exerts on a molecule of an i i i -th species an isotropic force with a potential Φ i j ( r ) \Phi_{ij}(r) Φ ij ( r ) where r = ( x 2 + y 2 + z 2 ) 1 / 2 r=(x^2+y^2+z^2)^{1/2} r = ( x 2 + y 2 + z 2 ) 1/2 . Assuming for simplicity that the fluid is unbounded, one can write the mass-averaged force affecting the i i i -th species in the hydrodynamic equations (2.18)–(2.19) in the form
F i ( r , t ) = ∇ ∑ j ∫ ρ j ( r ′ , t ) m i m j Φ i j ( ∣ r ′ − r ∣ ) d 3 r ′ , (29) \mathbf{F}_i(\mathbf{r},t)=\nabla\sum_j\int\frac{\rho_j(\mathbf{r}',t)}{m_i m_j}\Phi_{ij}(\lvert\mathbf{r}'-\mathbf{r}\rvert)\,\mathrm{d}^3\mathbf{r}', \tag*{(29)} F i ( r , t ) = ∇ j ∑ ∫ m i m j ρ j ( r ′ , t ) Φ ij (∣ r ′ − r ∣) d 3 r ′ , (29)
where m i m_i m i is the molecular mass an, the integration is implied to be over the whole space. To guarantee the convergence of the integral in (2.29) and those arising later, the potential Φ i j ( r ) \Phi_{ij}(r) Φ ij ( r ) is assumed to decay exponentially as r → ∞ r\to\infty r → ∞ .
Next, let the spatial scale of ρ ( r , t ) \rho(\mathbf{r},t) ρ ( r , t ) be much larger than that of Φ i j ( r ) \Phi_{ij}(r) Φ ij ( r ) , in which case expression (2.29) can be simplified asymptotically. To do so, change in it r ′ → r ′ + r \mathbf{r}'\to\mathbf{r}'+\mathbf{r} r ′ → r ′ + r and then expand ρ j ( r ′ + r , t ) \rho_j(\mathbf{r}'+\mathbf{r},t) ρ j ( r ′ + r , t ) about r ′ \mathbf{r}' r ′ , which yields
F i ( r , t ) = ∇ ∑ j ∫ [ ρ j ( r , t ) + r ′ ⋅ ∇ ρ j ( r , t ) + 1 2 r ′ r ′ : ∇ ∇ ρ j ( r , t ) + ⋯ ] Φ i j ( r ′ ) m i m j d 3 r ′ . \mathbf{F}_i(\mathbf{r},t)=\nabla\sum_j\int\left[\rho_j(\mathbf{r},t)+\mathbf{r}'\cdot\nabla\rho_j(\mathbf{r},t)+\frac{1}{2}\mathbf{r}'\mathbf{r}':\nabla\nabla\rho_j(\mathbf{r},t)+\cdots\right]\frac{\Phi_{ij}(r')}{m_i m_j}\,\mathrm{d}^3\mathbf{r}'. F i ( r , t ) = ∇ j ∑ ∫ [ ρ j ( r , t ) + r ′ ⋅ ∇ ρ j ( r , t ) + 2 1 r ′ r ′ : ∇∇ ρ j ( r , t ) + ⋯ ] m i m j Φ ij ( r ′ ) d 3 r ′ .
Given the isotropy of Φ i j ( r ′ ) \Phi_{ij}(\mathbf{r}') Φ ij ( r ′ ) , the second integral in the above expansion vanishes, and one obtains
F i = ∑ j W i j ∇ ρ j + ∑ j K i j ∇ ∇ 2 ρ j + ⋯ , (30) \mathbf{F}_i=\sum_j W_{ij}\nabla\rho_j+\sum_j K_{ij}\nabla\nabla^2\rho_j+\cdots, \tag*{(30)} F i = j ∑ W ij ∇ ρ j + j ∑ K ij ∇ ∇ 2 ρ j + ⋯ , (30)
where
W i j = ∫ Φ i j ( r ′ ) m i m j d 3 r ′ , K i j = ∫ r ′ 2 2 Φ i j ( r ′ ) m i m j d 3 r ′ . W_{ij}=\int\frac{\Phi_{ij}(r')}{m_i m_j}\,\mathrm{d}^3\mathbf{r}',\qquad K_{ij}=\int\frac{r'^2}{2}\frac{\Phi_{ij}(r')}{m_i m_j}\,\mathrm{d}^3\mathbf{r}'. W ij = ∫ m i m j Φ ij ( r ′ ) d 3 r ′ , K ij = ∫ 2 r ′2 m i m j Φ ij ( r ′ ) d 3 r ′ .
Since Newton’s Third Law implies that Φ i j = Φ j i \Phi_{ij}=\Phi_{ji} Φ ij = Φ j i , the matrices W i j W_{ij} W ij and K i j K_{ij} K ij are symmetric.
Once expansion (2.30) is substituted into the hydrodynamic equations (2.18)–(2.19), its first term can be absorbed into the internal energy – i.e., eliminated by the change
e → e + 1 2 ρ ∑ i j W i j ρ i ρ j , G i → G i + ∑ j W i j ρ j . e\to e+\frac{1}{2\rho}\sum_{ij}W_{ij}\rho_i\rho_j,\qquad G_i\to G_i+\sum_j W_{ij}\rho_j. e → e + 2 ρ 1 ij ∑ W ij ρ i ρ j , G i → G i + j ∑ W ij ρ j .
This does not come as a surprise, as the energy associated with potential interactions of molecules can be viewed as a kind of internal energy – in fact, the second term of expansion (2.30) could also be (and sometimes is) absorbed into e e e . This is not done in this paper, however, as it would make e e e a functional (instead of a function), with the implication that all the thermodynamic definitions and identities in Sect. 2.1 would need to be rewritten in terms of functional derivatives.
Thus, without loss of generality, one can set in expression (2.30) W i j = 0 W_{ij}=0 W ij = 0 . Omitting also the small terms hidden in “⋯ \cdots ⋯ ”, one obtains
F i = ∑ j K i j ∇ ∇ 2 ρ j , (31) \mathbf{F}_i = \sum_j K_{ij}\nabla\nabla^2\rho_j, \tag*{(31)} F i = j ∑ K ij ∇ ∇ 2 ρ j , (31)
which is the multicomponent extension of the standard DIM formula for the van der Waals force (e.g., [74 ] ). The matrix K i j K_{ij} K ij is the extension of the so-called Korteweg parameter for pure fluids, and it will be referred to as the Korteweg matrix. It should be positive definite, K i j > 0 K_{ij}>0 K ij > 0 , as the van der Waals force should be attractive, not repulsive.
Since the original representation (2.29) was a phenomenological model and, thus, the pair-wise potential Φ i j \Phi_{ij} Φ ij cannot be measured, the Korteweg matrix should be viewed as a set of adjustable parameters. As seen later, they can be deduced from the measurements of the equation of state and surface tension. One should keep in mind, however, that the Korteweg matrix should not depend on the temperature. Such a dependence would be physically unjustified, as the intermolecular attraction (characterised by K i j K_{ij} K ij ) should not depend on the molecules’ velocities (characterised by T T T ).
Boundary conditions at a solid wall
Let the fluid occupy a domain D \mathcal{D} D , bounded by a smooth surface ∂ D \partial\mathcal{D} ∂ D . For simplicity, the so-called Navier slip is disallowed in this work, so that the boundary condition for the velocity is
v = 0 at r ∈ ∂ D . (32) \mathbf{v} = 0 \qquad\text{at}\qquad\mathbf{r}\in\partial\mathcal{D}. \tag*{(32)} v = 0 at r ∈ ∂ D . (32)
To ensure mass conservation, one should require
n ⋅ J i = 0 at r ∈ ∂ D , (33) \mathbf{n}\cdot\mathbf{J}_i = 0 \qquad\text{at}\qquad\mathbf{r}\in\partial\mathcal{D}, \tag*{(33)} n ⋅ J i = 0 at r ∈ ∂ D , (33)
where n \mathbf{n} n is the outward-pointing unit normal to ∂ D \partial\mathcal{D} ∂ D .
The boundary condition for the temperature, in turn, depends on the problem at hand. Since this paper is concerned, inter alia , with the entropy principle and energy conservation, it will be assumed that no heat escapes through the boundary,
n ⋅ Q = 0 at r ∈ ∂ D . (34) \mathbf{n}\cdot\mathbf{Q} = 0 \qquad\text{at}\qquad\mathbf{r}\in\partial\mathcal{D}. \tag*{(34)} n ⋅ Q = 0 at r ∈ ∂ D . (34)
Boundary conditions (2.32)–(2.34) would be sufficient for the standard compressible multi-component hydrodynamics, but the DIM requires an extra condition [due to the presence of higher-order derivatives of the density field in expression (2.31)].
The most common version of such condition – prescribing a linear combination of the boundary-value of the density and its normal gradient – ascends to the paper by Cahn & Hilliard [19 ] . In application to pure fluids, the Neumann reduction of this condition was proposed by Seppecher [84 ] and the Dirichlet reduction, by Pismen & Pomeau [74 ] . As shown by Benilov [4 ] , the latter follows from the assumptions under which the whole DIM is derived (pair-wise intermolecular interactions, slowly-varying density field), and so it is used in the present paper. Thus, require that
ρ i = ρ 0 , i at r ∈ ∂ D , (35) \rho_i = \rho_{0,i} \qquad\text{at}\qquad\mathbf{r}\in\partial\mathcal{D}, \tag*{(35)} ρ i = ρ 0 , i at r ∈ ∂ D , (35)
where the constant ρ 0 , i \rho_{0,i} ρ 0 , i is specific to the fluid/solid combination under consideration. The general version of the boundary condition for ρ i \rho_i ρ i is discussed briefly in Appendix § 12 .
To clarify the physical meaning of condition (35) , consider the van der Waals forces acting on a fluid molecule located infinitesimally close to the wall: the solid attracts it towards the wall, while the other fluid molecules pull it away . The former force is fixed, whereas the latter grows with the near-wall density – so that the balance is achieved when the density assumes a certain value – namely, the parameter ρ 0 , i \rho_{0,i} ρ 0 , i in condition (35) . This argument suggests that a smaller value of ρ 0 , i \rho_{0,i} ρ 0 , i corresponds to a hydrophobic wall (characterised by a large contact angle) and larger ρ 0 , i \rho_{0,i} ρ 0 , i , to a hydrophilic one.
According to its physical meaning, ρ 0 , i \rho_{0,i} ρ 0 , i does not depend on the temperature. As shown later, its value can be deduced from a measurement of the contact angle.
The entropy principle
Conservation laws and the H-theorem
It can be verified that the governing equations and boundary conditions introduced above conserve the mass of each species
M i = ∫ D ρ i d 3 r , (36) M_i = \int_D \rho_i \,\mathrm{d}^3\mathbf{r}, \tag*{(36)} M i = ∫ D ρ i d 3 r , (36)
and the total energy
E = ∫ D [ ρ e + 1 2 ρ ∣ v ∣ 2 + 1 2 ∑ i j K i j ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r . (37) E = \int_D \left[\rho e + \frac{1}{2}\rho|\mathbf{v}|^2 + \frac{1}{2}\sum_{ij}K_{ij}(\nabla\rho_i)\cdot(\nabla\rho_j)\right] \,\mathrm{d}^3\mathbf{r}. \tag*{(37)} E = ∫ D [ ρ e + 2 1 ρ ∣ v ∣ 2 + 2 1 ij ∑ K ij ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r . (37)
The three terms in expression (37) represent the internal, kinetic, and van der Waals energies.
The governing equations and boundary conditions satisfy also an H-theorem, reflecting the fact that the net entropy of a fluid in a thermally-insulated container cannot decrease. To prove this, consider the following combination of the governing equations:
( 2.28 ) + T ρ ∑ i ∂ s ∂ ρ i × ( 2.17 ) + T s × ( 2.26 ) . (2.28) + T\rho\sum_i\frac{\partial s}{\partial\rho_i}\times(2.17) + Ts\times(2.26). ( 2.28 ) + T ρ i ∑ ∂ ρ i ∂ s × ( 2.17 ) + T s × ( 2.26 ) .
After straightforward algebra involving the use of the thermodynamic identities presented in subsection 2.1 and expressions (2.21)–(2.22) for the fluxes, one obtains
∂ ( ρ s ) ∂ t + ∇ ⋅ ( ρ s v − ∑ i G i T J i + Q T ) = Π : ( ∇ v ) T + { ∑ i j D i j ∣ F j − T ∇ ( G j T ) ∣ 2 + 2 ∑ i ξ i [ F i − T ∇ ( G i T ) ] ⋅ ( − ∇ T T ) + κ T ∣ ∇ T T ∣ 2 } . (38) \begin{aligned}
\frac{\partial(\rho s)}{\partial t}+\nabla\cdot\left(\rho s\mathbf{v}-\sum_i\frac{G_i}{T}\mathbf{J}_i+\frac{\mathbf{Q}}{T}\right)
={}&\frac{\boldsymbol{\Pi}:(\nabla\mathbf{v})}{T}\\
&+\left\{\sum_{ij}D_{ij}\left|\mathbf{F}_j-T\nabla\left(\frac{G_j}{T}\right)\right|^2
+2\sum_i\xi_i\left[\mathbf{F}_i-T\nabla\left(\frac{G_i}{T}\right)\right]\cdot\left(-\frac{\nabla T}{T}\right)\right.\\
&\left.\qquad+\kappa T\left|\frac{\nabla T}{T}\right|^2\right\}. \tag*{(38)}
\end{aligned} ∂ t ∂ ( ρ s ) + ∇ ⋅ ( ρ s v − i ∑ T G i J i + T Q ) = T Π : ( ∇ v ) + { ij ∑ D ij F j − T ∇ ( T G j ) 2 + 2 i ∑ ξ i [ F i − T ∇ ( T G i ) ] ⋅ ( − T ∇ T ) + κ T T ∇ T 2 } . (38)
The first term on the right-hand side of this equation is non-negative due to the following (easily verifiable) identity,
Π : ( ∇ v ) = 1 2 μ s ∣ ( ∇ v ) + ( ∇ v ) t r − 2 3 I ∇ ⋅ v ∣ 2 + μ b ( ∇ ⋅ v ) 2 , (39) \boldsymbol{\Pi}:(\nabla\mathbf{v})=\frac{1}{2}\mu_s\left|(\nabla\mathbf{v})+(\nabla\mathbf{v})^{\mathrm{tr}}-\frac{2}{3}\mathbf{I}\nabla\cdot\mathbf{v}\right|^2+\mu_b(\nabla\cdot\mathbf{v})^2, \tag*{(39)} Π : ( ∇ v ) = 2 1 μ s ( ∇ v ) + ( ∇ v ) tr − 3 2 I ∇ ⋅ v 2 + μ b ( ∇ ⋅ v ) 2 , (39)
whereas the expression in curly brackets is non-negative because the extended transport matrix is positive semidefinite (see subsection 2.3). Thus, integrating equation (38) over D \mathcal{D} D , and using boundary conditions (2.32)–(2.34), one obtains
d S d t ≥ 0 , (40) \frac{\mathrm{d}S}{\mathrm{d}t} \ge0, \tag*{(40)} d t d S ≥ 0 , (40)
where
S = ∫ D ρ s d 3 r . (41) S = \int_{\mathcal{D}} \rho s \,\mathrm{d}^{3}\mathbf{r}. \tag*{(41)} S = ∫ D ρ s d 3 r . (41)
Inequality (40) is the desired H-theorem.
It follows from (3.3)–(3.4) that the exact equality in (40) can only hold if the velocity field is spatially uniform; together with the no-flow boundary conditions, this requirement amounts to v = 0 \mathbf{v} = 0 v = 0 (i.e., the fluid is static).
Stability via the entropy principle
The most common way to examine the stability of a steady solution of a set of equations consist in linearizing these equations with respect to a small perturbation, assuming the harmonic dependence of the perturbation on t t t , and solving the resulting eigenvalue problem. In the problem at hand, however, it is much simpler to examine stability using the entropy principle.
If, at a certain steady state, the total entropy S S S has a local maximum constrained by the conditions of fixed energy E E E and mass M i M_i M i , this state is stable. The inverse is also true: if S S S does not have a maximum, the corresponding state is unstable – because a perturbed solution with a higher entropy cannot evolve ‘back’. Neutrally stable oscillations are also prohibited by the entropy principle – hence, the system can only evolve further away, towards a steady state where the entropy does have a maximum.
Let a fluid be enclosed in a container (which can be later assumed to be infinitely large, if need be) and seek a maximum of S S S , constrained by the conditions of fixed M i M_i M i and E E E . This problem amounts to finding the stationary points of the functional
H [ ρ 1 … ρ N , T , v ] = S + ∑ i η i M i + λ E , (42) H[\rho_1 \ldots\rho_N,T,\mathbf{v}] = S + \sum_i \eta_i M_i + \lambda E, \tag*{(42)} H [ ρ 1 … ρ N , T , v ] = S + i ∑ η i M i + λ E , (42)
where λ \lambda λ and μ i \mu_i μ i are the Lagrange multipliers, and S S S , E E E , and M i M_i M i are given by (41) , (37) , and (36) , respectively. Varying H H H with respect to v \mathbf{v} v and equating the variation to zero, one obtains
v = 0 , (43) \mathbf{v} = 0, \tag*{(43)} v = 0 , (43)
i.e., a steady state must be (unsurprisingly) static. Next, varying H H H with respect to T T T , one obtains
∂ s ∂ T + λ ∂ e ∂ T = 0. \frac{\partial s}{\partial T} + \lambda\frac{\partial e}{\partial T} = 0. ∂ T ∂ s + λ ∂ T ∂ e = 0.
Comparison of this equality with the Gibbs relation (4) yields
λ = − 1 T . (44) \lambda= -\frac{1}{T}. \tag*{(44)} λ = − T 1 . (44)
Since λ \lambda λ is a constant, equation (44) implies that T T T is spatially uniform – i.e., a steady state must be isothermal.
Finally, varying H H H with respect to ρ i \rho_i ρ i , recalling expressions (3.8)–(3.9) for v \mathbf{v} v and λ \lambda λ , and keeping in mind definition (3) of the chemical potential G i G_i G i , one obtains
1 T ( G i − ∑ j K i j ∇ 2 ρ j ) + η i = 0. (45) \frac{1}{T}\left(G_i-\sum_j K_{ij}\nabla^2\rho_j\right)+\eta_i=0. \tag*{(45)} T 1 ( G i − j ∑ K ij ∇ 2 ρ j ) + η i = 0. (45)
This equation describes all steady states of the governing equations, and it will be extensively used in the rest of this paper. The temperature T T T in equation (45) should be treated as a known parameter, whereas the Lagrange multiplier η i \eta_i η i is to be deduced from the boundary conditions. The latter will be illustrated for the case of an infinite domain, plus an assumption that the fluid at infinity be spatially uniform and characterised by a coordinate-independent chemical potential G i = G ∞ , i G_i=G_{\infty,i} G i = G ∞ , i . In this case equation (45) yields
η i = − G ∞ , i T , (46) \eta_i=-\frac{G_{\infty,i}}{T}, \tag*{(46)} η i = − T G ∞ , i , (46)
as required.
Note also that equation (45) can also be recovered by adapting the governing equations for the state of equilibrium. To do so, one should set v = 0 \mathbf{v}=0 v = 0 , T = c o n s t T=\mathrm{const} T = const , and ∂ ρ i / ∂ t = 0 \partial\rho_i/\partial t=0 ∂ ρ i / ∂ t = 0 in equations (17) –(18) and, recalling expression (21) for the diffusion fluxes, obtain (45) , as required.
To examine a solution of equation (45) for stability, one needs to calculate the second variation of H H H . Omitting the algebra (involving the use of identities (2.3)–(2.4), definition (8) of the heat capacity c c c , and expression (44) for λ \lambda λ ), one obtains
δ 2 H = 1 T ∫ D { ∑ i j [ − ∂ G i ∂ ρ j ( δ ρ i ) ( δ ρ j ) − K i j ( ∇ δ ρ i ) ⋅ ( ∇ δ ρ j ) ] − ρ c T ( δ T ) 2 − ρ ∣ δ v ∣ 2 } d 3 r . \delta^2 H=\frac{1}{T}\int_{\mathcal{D}}\left\{\sum_{ij}\left[-\frac{\partial G_i}{\partial\rho_j}(\delta\rho_i)(\delta\rho_j)-K_{ij}(\nabla\delta\rho_i)\cdot(\nabla\delta\rho_j)\right]-\frac{\rho c}{T}(\delta T)^2-\rho|\delta\mathbf{v}|^2\right\}\,\mathrm{d}^3r. δ 2 H = T 1 ∫ D { ij ∑ [ − ∂ ρ j ∂ G i ( δ ρ i ) ( δ ρ j ) − K ij ( ∇ δ ρ i ) ⋅ ( ∇ δ ρ j ) ] − T ρ c ( δ T ) 2 − ρ ∣ δ v ∣ 2 } d 3 r .
Evidently, perturbations of the temperature and velocity are negative and can only lower the total entropy – hence, the type of the stationary point (maximum vs saddle) is fully determined by the variation of the density. Setting, thus, δ T = 0 \delta T=0 δ T = 0 and δ v = 0 \delta\mathbf{v}=0 δ v = 0 , one obtains
δ 2 H = 1 T ∫ D ∑ i j [ − ∂ G i ∂ ρ j ( δ ρ i ) ( δ ρ j ) − K i j ( ∇ δ ρ i ) ⋅ ( ∇ δ ρ j ) ] d 3 r . (47) \delta^2 H=\frac{1}{T}\int_{\mathcal{D}}\sum_{ij}\left[-\frac{\partial G_i}{\partial\rho_j}(\delta\rho_i)(\delta\rho_j)-K_{ij}(\nabla\delta\rho_i)\cdot(\nabla\delta\rho_j)\right]\,\mathrm{d}^3r. \tag*{(47)} δ 2 H = T 1 ∫ D ij ∑ [ − ∂ ρ j ∂ G i ( δ ρ i ) ( δ ρ j ) − K ij ( ∇ δ ρ i ) ⋅ ( ∇ δ ρ j ) ] d 3 r . (47)
Expression (47) is the main tool used in this paper for studying the stability properties of steady states described by equation (45) .
Nondimensionalization and the governing parameters
Introduce a characteristic density scale ϱ \varrho ϱ , a pressure scale P P P , a temperature scale T 0 T_0 T 0 , and a typical value R R R of the specific gas constant R i R_i R i introduced in subsection 2.2. These scales allow one to nondimensionalise all thermodynamics variables and functions introduced in subsection 2.1,
( ρ i ) n d = ρ i ϱ , ρ n d = ρ ϱ , T n d = T T 0 , e n d = ϱ P e , s n d = s R , (\rho_i)_{nd}=\frac{\rho_i}{\varrho},\qquad\rho_{nd}=\frac{\rho}{\varrho},\qquad T_{nd}=\frac{T}{T_0},\qquad e_{nd}=\frac{\varrho}{P}e,\qquad s_{nd}=\frac{s}{R}, ( ρ i ) n d = ϱ ρ i , ρ n d = ϱ ρ , T n d = T 0 T , e n d = P ϱ e , s n d = R s ,
p n d = p P , ( G i ) n d = ϱ P G i , c n d = c R , a n d = ϱ 2 P a , B n d = B P . p_{nd}=\frac{p}{P},\qquad(G_i)_{nd}=\frac{\varrho}{P}G_i,\qquad c_{nd}=\frac{c}{R},\qquad a_{nd}=\frac{\varrho^2}{P}a,\qquad B_{nd}=\frac{B}{P}. p n d = P p , ( G i ) n d = P ϱ G i , c n d = R c , a n d = P ϱ 2 a , B n d = P B .
It is also convenient to nondimensionalise the coefficients which appear in the governing equations. Using their respective scales, one obtains
( K i j ) n d = K i j K , ( μ s ) n d = μ s μ , ( μ b ) n d = μ b μ , (K_{ij})_{nd}=\frac{K_{ij}}{K},\qquad(\mu_s)_{nd}=\frac{\mu_s}{\mu},\qquad(\mu_b)_{nd}=\frac{\mu_b}{\mu}, ( K ij ) n d = K K ij , ( μ s ) n d = μ μ s , ( μ b ) n d = μ μ b ,
( R i ) n d = R i R , κ n d = κ ϰ , ( D i j ) n d = D i j D , ( ζ i ) n d = ζ i ζ (R_i)_{nd}=\frac{R_i}{R},\quad\kappa_{nd}=\frac{\kappa}{\varkappa},\quad(D_{ij})_{nd}=\frac{D_{ij}}{D},\quad(\zeta_i)_{nd}=\frac{\zeta_i}{\zeta} ( R i ) n d = R R i , κ n d = ϰ κ , ( D ij ) n d = D D ij , ( ζ i ) n d = ζ ζ i .
Finally, introduce
r n d = r l , t n d = V t l , v n d = v V , \mathbf{r}_{nd}=\frac{\mathbf{r}}{l},\qquad t_{nd}=\frac{Vt}{l},\qquad\mathbf{v}_{nd}=\frac{\mathbf{v}}{V}, r n d = l r , t n d = l V t , v n d = V v ,
Π n d = l μ V Π , ( F i ) n d = l 3 K ρ F i , ( J i ) n d = J i ρ V , Q n d = Q P V , \boldsymbol{\Pi}_{nd}=\frac{l}{\mu V}\boldsymbol{\Pi},\qquad(\mathbf{F}_i)_{nd}=\frac{l^3}{K\rho}\mathbf{F}_i,\qquad(\mathbf{J}_i)_{nd}=\frac{\mathbf{J}_i}{\rho V},\qquad\mathbf{Q}_{nd}=\frac{\mathbf{Q}}{PV}, Π n d = μ V l Π , ( F i ) n d = K ρ l 3 F i , ( J i ) n d = ρ V J i , Q n d = P V Q ,
where
l = ρ 2 K P , l=\sqrt{\frac{\rho^2K}{P}}, l = P ρ 2 K ,
is the characteristic interfacial thickness, and
V = P l μ V=\frac{Pl}{\mu} V = μ P l
is the velocity scale reflecting the three-way balance of the pressure gradient, viscous stress, and van der Waals force [3 ] . Note also that the characteristic interfacial thickness l l l is on a nano scale [65 , 38 , 4 ] .
Rewriting equations (17) , (2.27)–(2.28), (2.21)–(2.22), and (31) , and omitting the subscript n d nd n d , one obtains
∂ ρ i ∂ t + ∇ ⋅ ( ρ i v + J i ) = 0 , (48) \frac{\partial\rho_i}{\partial t}+\nabla\cdot(\rho_i\mathbf{v}+\mathbf{J}_i)=0, \tag*{(48)} ∂ t ∂ ρ i + ∇ ⋅ ( ρ i v + J i ) = 0 , (48)
α ρ [ ∂ v ∂ t + ( v ⋅ ∇ ) v ] = ∇ ⋅ Π + ∑ i ρ i ( F i − ∇ G i ) − ρ s ∇ T , (49) \alpha\rho\left[\frac{\partial\mathbf{v}}{\partial t}+(\mathbf{v}\cdot\nabla)\mathbf{v}\right]=\nabla\cdot\boldsymbol{\Pi}+\sum_i\rho_i(\mathbf{F}_i-\nabla G_i)-\rho_s\nabla T, \tag*{(49)} α ρ [ ∂ t ∂ v + ( v ⋅ ∇ ) v ] = ∇ ⋅ Π + i ∑ ρ i ( F i − ∇ G i ) − ρ s ∇ T , (49)
τ ρ c ( ∂ T ∂ t + v ⋅ ∇ T ) + ∇ ⋅ Q = Π : ( ∇ v ) − B ∇ ⋅ v + ∑ i ( F i + ρ α i ∇ ) ⋅ J i , (50) \tau\rho c\left(\frac{\partial T}{\partial t}+\mathbf{v}\cdot\nabla T\right)+\nabla\cdot\mathbf{Q}=\boldsymbol{\Pi}:(\nabla\mathbf{v})-B\nabla\cdot\mathbf{v}+\sum_i(\mathbf{F}_i+\rho\alpha_i\nabla)\cdot\mathbf{J}_i, \tag*{(50)} τ ρ c ( ∂ t ∂ T + v ⋅ ∇ T ) + ∇ ⋅ Q = Π : ( ∇ v ) − B ∇ ⋅ v + i ∑ ( F i + ρ α i ∇ ) ⋅ J i , (50)
Π = μ s [ ∇ v + ( ∇ v ) t r − 2 3 ( ∇ ⋅ v ) I ] + μ b ( ∇ ⋅ v ) I , (51) \boldsymbol{\Pi}=\mu_s\left[\nabla\mathbf{v}+(\nabla\mathbf{v})^{\mathrm{tr}}-\frac{2}{3}(\nabla\cdot\mathbf{v})\mathbf{I}\right]+\mu_b(\nabla\cdot\mathbf{v})\mathbf{I}, \tag*{(51)} Π = μ s [ ∇ v + ( ∇ v ) tr − 3 2 ( ∇ ⋅ v ) I ] + μ b ( ∇ ⋅ v ) I , (51)
δ J i = ∑ j D i j [ F j − T ∇ ( G j T ) ] − ν δ β ζ i T ∇ T , (52) \delta\mathbf{J}_i=\sum_jD_{ij}\left[\mathbf{F}_j-T\nabla\left(\frac{G_j}{T}\right)\right]-\frac{\nu\delta}{\beta}\frac{\zeta_i}{T}\nabla T, \tag*{(52)} δ J i = j ∑ D ij [ F j − T ∇ ( T G j ) ] − β ν δ T ζ i ∇ T , (52)
β Q = ν ∑ j ζ j [ F j − T ∇ ( G j T ) ] − κ ∇ T , (53) \beta\mathbf{Q}=\nu\sum_j\zeta_j\left[\mathbf{F}_j-T\nabla\left(\frac{G_j}{T}\right)\right]-\kappa\nabla T, \tag*{(53)} β Q = ν j ∑ ζ j [ F j − T ∇ ( T G j ) ] − κ ∇ T , (53)
F i = ∇ ∑ j K i j ∇ 2 ρ j , (54) \mathbf{F}_i=\nabla\sum_jK_{ij}\nabla^2\rho_j, \tag*{(54)} F i = ∇ j ∑ K ij ∇ 2 ρ j , (54)
where the nondimensional parameters
α = K ρ 3 μ 2 , τ = ρ R T 0 P , β = K P ρ 0 2 μ κ T 0 , ν = ζ P ϰ T 0 ρ 0 , δ = K ρ 4 D μ P , (55) \alpha=\frac{K\rho^3}{\mu^2},\qquad\tau=\frac{\rho RT_0}{P},\qquad\beta=\frac{KP\rho_0^2}{\mu\kappa T_0},\qquad\nu=\frac{\zeta P}{\varkappa T_0\rho_0},\qquad\delta=\frac{K\rho^4}{D\mu P}, \tag*{(55)} α = μ 2 K ρ 3 , τ = P ρR T 0 , β = μ κ T 0 K P ρ 0 2 , ν = ϰ T 0 ρ 0 ζ P , δ = D μ P K ρ 4 , (55)
are placed for better ‘visibility’ in boxes.
Since α \alpha α appears in front of the material derivative in equation (48) , it should be interpreted as the microscopic Reynolds number (associated with the flow near the interface, not the global flow). τ \tau τ is the nondimensional temperature. β \beta β is the isothermality parameter [3 ] : if it is small, the temperature field is close to being spatially uniform (isothermal). The Nusselt number ν \nu ν characterises the importance of heat diffusion relative to heat conduction – see expression (4.6) for the heat flux. Finally, the position of δ \delta δ in equation (52) suggests that this parameter characterises advection relative to diffusion (recall that the flux J i \mathbf{J}_i J i was nondimensionalised using the advection scale ϱ V \varrho V ϱ V ).
One can also introduce the Prandtl and Schmidt numbers,
P r = β T α = μ R ϰ , S c = δ α = μ ρ D P , Pr = \frac{\beta T}{\alpha} = \frac{\mu R}{\varkappa}, \qquad Sc = \frac{\delta}{\alpha} = \frac{\mu\rho}{DP}, P r = α β T = ϰ μ R , S c = α δ = D P μ ρ ,
characterizing viscosity relative to heat conduction and diffusion, respectively.
The nondimensional versions of the boundary conditions look exactly the same as their dimensional counterparts, (2.32)–(2.35) – as do the nondimensional versions of the thermodynamic identities of subsection 2.1 except definition (2.8) of the heat capacity, which becomes
c = 1 τ ∂ e ∂ T . c = \frac{1}{\tau}\frac{\partial e}{\partial T}. c = τ 1 ∂ T ∂ e .
This paper does not aim to present a comprehensive classification of asymptotic regimes of the multicomponent DIM (for the pure-fluid DIM, see Benilov [3 ] ). Only the simplest regime will be described and used later as a qualitative illustration of theoretically-predicted behaviors. It corresponds to the following assumptions:
α ≪ 1 , β ≪ 1 , ν ≪ 1. \alpha\ll1, \qquad\beta\ll1, \qquad\nu\ll1. α ≪ 1 , β ≪ 1 , ν ≪ 1.
The smallness of α \alpha α allows one to take advantage of the slow-flow approximation – whereas the other two assumptions and equations (50) and (53) imply that
T = 1 + O ( β ) , T = 1 + O(\beta), T = 1 + O ( β ) ,
i.e., the fluid is almost isothermal. Setting, thus, T = c o n s t T = \mathrm{const} T = const in expression (4.5) for the diffusion flux and substituting it into the density equation (48) , one obtains
∂ ρ i ∂ t + ∇ ⋅ [ ρ i v + ∑ j D i j ∇ ( ∑ n K j n ∇ 2 ρ n − G j ) ] = 0 , (56) \frac{\partial\rho_i}{\partial t} + \nabla\cdot\left[\rho_i \mathbf{v} + \sum_j D_{ij}\nabla\left(\sum_n K_{jn}\nabla^2 \rho_n - G_j\right)\right] = 0, \tag*{(56)} ∂ t ∂ ρ i + ∇ ⋅ [ ρ i v + j ∑ D ij ∇ ( n ∑ K j n ∇ 2 ρ n − G j ) ] = 0 , (56)
where it was assumed, without loss of generality, that δ = 1 \delta= 1 δ = 1 . Simplifying similarly equation (49) and substituting into it expression (4.7) for F i F_i F i , one obtains
∇ ⋅ Π + ∑ i ρ i ∇ ( ∑ j K i j ∇ 2 ρ j − G i ) = 0. (57) \nabla\cdot\mathbf{\Pi} + \sum_i \rho_i \nabla\left(\sum_j K_{ij}\nabla^2 \rho_j - G_i\right) = 0. \tag*{(57)} ∇ ⋅ Π + i ∑ ρ i ∇ ( j ∑ K ij ∇ 2 ρ j − G i ) = 0. (57)
Observe that equations (56) –(57) do not involve the (small) temperature variations – hence, the temperature equation (50) can be simply omitted.
Equations (56) –(57) and expression (4.4) for the viscous stress form a closed set of approximate equations for the unknowns ρ i \rho_i ρ i and v \mathbf{v} v . The chemical potential G i G_i G i in these equations should be treated as a known function of ρ 1 … ρ N \rho_1\ldots\rho_N ρ 1 … ρ N , and the temperature T T T , as a known parameter.
Unlike the exact set – which describes fast acoustic waves and slow interfacial flow – the approximate equations describe only the latter. This is a clear advantage: in a numerical simulation, for example, waves necessitate a small timestep and, thus, dramatically slow down the computation. At the same time, the two sets of equations have very similar properties: they share the same steady solutions, both conserve mass and energy, and satisfy the H-theorem.
Basic solutions and their stability
Spatially uniform states
Consider a uniform fluid where there is no flow and all species are in vapour phase. If the temperature drops, one of the species may become overcooled, giving rise to rapid condensation. A similar instability may occur when all or some of the species are in liquid phase, and the temperature increases – giving rise to rapid evaporation .
To determine which exactly states are unstable (thus, do not occur in real world), one could perform the usual linear analysis. For a pure fluid (N = 1 N = 1 N = 1 ), this is a straightforward task yielding the following stability criterion:
∂ p ∂ ρ > 0. (58) \frac{\partial p}{\partial\rho} > 0. \tag*{(58)} ∂ ρ ∂ p > 0. (58)
Thus, instability occurs if an increase in density lowers the pressure, so that the flow generated by the pressure gradient brings even more fluid to this region.
For an arbitrary N N N , however, the analysis of linearised equations is extremely cumbersome. Instead, it will be examined via the entropy principle – i.e., using expression (47) . Taking the limit D → R 3 \mathcal{D} \to\mathbb{R}^{3} D → R 3 (unbounded fluid), assuming that ρ i \rho_i ρ i is spatially uniform, and recalling that K i j ≻ 0 K_{ij} \succ0 K ij ≻ 0 and c > 0 c > 0 c > 0 , one can deduce from (47) that H H H has a maximum if
∂ G i ∂ ρ j > 0. (59) \frac{\partial G_i}{\partial\rho_j} > 0. \tag*{(59)} ∂ ρ j ∂ G i > 0. (59)
This is the standard stability criterion for a spatially uniform state of a multicomponent fluid [44 ] .
Four comments are in order:
For a physically meaningful G i ( ρ 1 … ρ N , T ) G_i(\rho_1\ldots\rho_N, T) G i ( ρ 1 … ρ N , T ) , condition (59) holds for a sufficiently rarefied vapour or a sufficiently dense liquid.
The states with marginal stability are sometimes referred to as “spinodal points” and stable vapour, as “subspinodal vapour”.
Interestingly, the viscosity and transport coefficients do not appear in criterion (59) . The corresponding effects can only slow the instability down, but not eliminate it.
To reconcile criterion (59) with its pure-fluid counterpart (58) , note that for N = 1 N = 1 N = 1 , ∂ G / ∂ ρ \partial G/\partial\rho ∂ G / ∂ ρ and ∂ p / ∂ ρ \partial p/\partial\rho ∂ p / ∂ ρ have the same sign [as implied by identity (6) ].
Two-phase saturated states
Consider an interface separating liquid and vapour of the same pure fluid. If they are in equilibrium, their temperatures are equal, and the rest of the parameters satisfy the so-called Maxwell construction [67 ] ,
G ( ρ ( l ) , T ) = G ( ρ ( v ) , T ) , p ( ρ ( l ) , T ) = p ( ρ ( v ) , T ) , (60) G(\rho^{(l)}, T) = G(\rho^{(v)}, T), \qquad p(\rho^{(l)}, T) = p(\rho^{(v)}, T), \tag*{(60)} G ( ρ ( l ) , T ) = G ( ρ ( v ) , T ) , p ( ρ ( l ) , T ) = p ( ρ ( v ) , T ) , (60)
where the superscripts ( l ) (l) ( l ) and ( v ) (v) ( v ) mark the parameters of the liquid and vapour, respectively. The former and latter equalities in (60) guarantee thermodynamic and mechanical equilibria of the interface, respectively.
Assume also that both phases are stable,
( ∂ G ∂ ρ ) ρ = ρ ( l ) > 0 , ( ∂ G ∂ ρ ) ρ = ρ ( v ) > 0 , \left(\frac{\partial G}{\partial\rho}\right)_{\rho=\rho^{(l)}} > 0, \qquad\left(\frac{\partial G}{\partial\rho}\right)_{\rho=\rho^{(v)}} > 0, ( ∂ ρ ∂ G ) ρ = ρ ( l ) > 0 , ( ∂ ρ ∂ G ) ρ = ρ ( v ) > 0 ,
and the density of the liquid exceeds that of the vapour,
ρ ( l ) > ρ ( v ) . \rho^{(l)} > \rho^{(v)}. ρ ( l ) > ρ ( v ) .
Subject to these conditions, the Maxwell construction (5.3) uniquely determines the saturated densities ρ ( l ) \rho^{(l)} ρ ( l ) and ρ ( v ) \rho^{(v)} ρ ( v ) as functions of T T T .
In what follows, the multicomponent version of the Maxwell construction will be shown to follow from the DIM’s entropy principle.
Consider an insulated container with fluid subdivided between two states, liquid and vapour. If the liquid/vapour and fluid/wall interfaces are sufficiently thin, the corresponding full entropy, mass, and energy can be approximately written in the form
S = V ( l ) ρ ( l ) s ( ρ 1 ( l ) … ρ N ( l ) , T ( l ) ) + V ( v ) ρ ( v ) s ( ρ 1 ( v ) … ρ N ( v ) , T ( v ) ) , S = V^{(l)}\rho^{(l)}s(\rho_1^{(l)}\ldots\rho_N^{(l)},T^{(l)}) + V^{(v)}\rho^{(v)}s(\rho_1^{(v)}\ldots\rho_N^{(v)},T^{(v)}), S = V ( l ) ρ ( l ) s ( ρ 1 ( l ) … ρ N ( l ) , T ( l ) ) + V ( v ) ρ ( v ) s ( ρ 1 ( v ) … ρ N ( v ) , T ( v ) ) ,
M i = V ( l ) ρ i ( l ) + V ( v ) ρ i ( v ) , (61) M_i = V^{(l)}\rho_i^{(l)} + V^{(v)}\rho_i^{(v)}, \tag*{(61)} M i = V ( l ) ρ i ( l ) + V ( v ) ρ i ( v ) , (61)
E = V ( l ) ρ ( l ) e ( ρ 1 ( l ) … ρ N ( l ) , T ( l ) ) + V ( v ) ρ ( v ) e ( ρ 1 ( v ) … ρ N ( v ) , T ( v ) ) , (62) E = V^{(l)}\rho^{(l)}e(\rho_1^{(l)}\ldots\rho_N^{(l)},T^{(l)}) + V^{(v)}\rho^{(v)}e(\rho_1^{(v)}\ldots\rho_N^{(v)},T^{(v)}), \tag*{(62)} E = V ( l ) ρ ( l ) e ( ρ 1 ( l ) … ρ N ( l ) , T ( l ) ) + V ( v ) ρ ( v ) e ( ρ 1 ( v ) … ρ N ( v ) , T ( v ) ) , (62)
where V ( l ) V^{(l)} V ( l ) and V ( v ) V^{(v)} V ( v ) are the volumes of the liquid and vapour phases, respectively. Introduce also the full volume of the container,
V = V ( l ) + V ( v ) . (63) V = V^{(l)} + V^{(v)}. \tag*{(63)} V = V ( l ) + V ( v ) . (63)
The Maxwell construction can be derived by maximizing S S S subject to the constraints of fixed M i M_i M i , E E E , and V V V (which are now functions , as opposed to being functionals in the previous subsection). Straightforward calculations show that the maximum of entropy is achieved if T ( l ) = T ( v ) T^{(l)} = T^{(v)} T ( l ) = T ( v ) (isothermality) and
G i ( ρ 1 ( l ) … ρ N ( l ) , T ) = G i ( ρ 1 ( v ) … ρ N ( v ) , T ) , (64) G_i(\rho_1^{(l)}\ldots\rho_N^{(l)},T) = G_i(\rho_1^{(v)}\ldots\rho_N^{(v)},T), \tag*{(64)} G i ( ρ 1 ( l ) … ρ N ( l ) , T ) = G i ( ρ 1 ( v ) … ρ N ( v ) , T ) , (64)
p ( ρ 1 ( l ) … ρ N ( l ) , T ) = p ( ρ 1 ( v ) … ρ N ( v ) , T ) . (65) p(\rho_1^{(l)}\ldots\rho_N^{(l)},T) = p(\rho_1^{(v)}\ldots\rho_N^{(v)},T). \tag*{(65)} p ( ρ 1 ( l ) … ρ N ( l ) , T ) = p ( ρ 1 ( v ) … ρ N ( v ) , T ) . (65)
Four comments are in order:
The Maxwell construction can sometimes yield a solution with negative V ( l ) V^{(l)} V ( l ) or V ( v ) V^{(v)} V ( v ) . In such cases, the two-phase equilibrium is irrelevant, and the fluid evolves towards the one-phase state with the same masses of the species and total energy.
The multicomponent Maxwell construction (5.7)–(5.8) comprises N + 1 N + 1 N + 1 equations for 2 N 2N 2 N unknowns – hence, does not uniquely fix the liquid and vapour densities.
To close set (5.7)–(5.8), one should assume the masses M i M_i M i and the total energy E E E to be known and view equalities (5.4)–(5.6) as additional equations. They bring the total number of equations to 2 N + 3 2N + 3 2 N + 3 (making the ‘expanded’ Maxwell construction appear overdetermined) – but in this formulation, V ( l ) V^{(l)} V ( l ) , V ( v ) V^{(v)} V ( v ) , and T T T should also be viewed as unknowns.
Physically, if a certain amount of fluid, with a certain amount of energy, is placed in a box, the entropy principle uniquely determines the final temperature and the proportion in which the box is subdivided between the liquid and vapour phases.
The closure of the Maxwell construction described in the previous bullet is, obviously, inapplicable to containers of infinite volume. To understand how conditions (5.7)–(5.8) should be closed in this case, consider the interface between the Earth’s atmosphere and ocean. For this setting, one should prescribe the (atmospheric) pressure and composition
of dry air above the ocean’s surface. With these parameters given, (5.7)–(5.8) yield the saturated moisture content of the air and the saturated amounts of nitrogen, oxygen, etc. dissolved in the oceanic water.
To illustrate the use of the Maxwell construction, consider a van der Waals fluid, whose pressure and chemical potential are described by expressions (2.15)–(2.16). Assume for simplicity that the fluid is pure (N = 1 N = 1 N = 1 ) and monatomic (c i = 3 R i / 2 c_i = 3R_i/2 c i = 3 R i /2 ) – and let the scales ϱ \varrho ϱ , P P P , and R R R used for nondimensionalization be such that
a 11 = 1 , b 1 = 1 , R 1 = 1. (66) a_{11} = 1,\qquad b_1 = 1,\qquad R_1 = 1. \tag*{(66)} a 11 = 1 , b 1 = 1 , R 1 = 1. (66)
If T < 8 / 27 T < 8/27 T < 8/27 (which is a subcritical temperature of the van der Waals fluid), two states exist representing the liquid and vapour phases, with some (but not all) of the states in between being unstable (see Figure 1 ). If the temperature is supercritical , only one phase exists and interfaces do not.
Figure 1. The pressure [panel (a)] and chemical potential [panel (b)] of a pure van der Waals fluid as functions of the density, for three values of the temperature: (1) T = 0.26 T = 0.26 T = 0.26 (subcritical); (2) T = 8 / 27 T = 8/27 T = 8/27 (critical), (3) T = 0.33 T = 0.33 T = 0.33 (supercritical). The critical point on curve (2) is marked with a filled circle; the saturation and spinodal points on curve (1) are shown by filled and empty squares, respectively. The labels “undersat[urated] vapour”, “oversat[urated] v[apour]”, etc. apply only to curve (1).
The simplest model describing a water/air interface is that with N = 2 N = 2 N = 2 . The Maxwell construction in this case should be complemented with one extra condition, setting the pressure above the interface equal to its atmospheric value,
p ( ρ 1 ( v ) , ρ 2 ( v ) , T ) = p A . p(\rho_1^{(v)},\rho_2^{(v)},T)=p_A. p ( ρ 1 ( v ) , ρ 2 ( v ) , T ) = p A .
The difference between a pure and a multicomponent fluids is illustrated in Figure 2 for parameters (5.9) and
Figure 2. The saturated densities [panel (a)] and pressure [panel (b)] vs the temperature, for a pure (solid line) and two-component (dotted line) van der Waals fluids, with the parameters (5.9)–(5.10). The empty circles mark the boiling point, the filled circle marks the critical point. The saturated densities of the second species (of the two-component fluid) are not shown.
a 22 = 0.2 , a 12 = 0 , b 2 = 1 , R 2 = 0.6 , p A = 0.03. (67) a_{22}=0.2,\qquad a_{12}=0,\qquad b_2=1,\qquad R_2=0.6,\qquad p_A=0.03. \tag*{(67)} a 22 = 0.2 , a 12 = 0 , b 2 = 1 , R 2 = 0.6 , p A = 0.03. (67)
These values reflect a compromise between simplicity, illustrative purposes (the curves with N = 1 N=1 N = 1 and N = 2 N=2 N = 2 should be visibly different), and an attempt to loosely match the parameters of water and air (loosely, because the van der Waals model includes few adjustable constants). In particular, parameters (5.10) make the critical temperature of the second species noticeably smaller than that of the first species. As a result, a temperature range exists where the former is definitely vapour, whereas the latter can be either vapour or liquid (as is indeed the case with water under normal conditions).
The most important difference between a multicomponent fluid at fixed pressure and a pure fluid is that the former exhibits a boiling point. It occurs when the saturated pressure of the liquid-phase species matches the applied (atmospheric) pressure: as a result, the second species is completely replaced by the vapour of the first species – and so the fluid becomes pure. If it is heated beyond the boiling point, the pressure can no longer be kept fixed, but ought to increase (to match the saturated pressure of the first species, which grows with T T T ).
One should keep in mind that the small-T T T part of figure 2 is physically meaningless due to freezing (which is not described by the DIM).
One-dimensional steady states
Flat liquid/vapour interfaces
Consider an unbounded fluid involving liquid and vapour phases in equilibrium, separated by a static flat interface. Its spatial structure is described by equation (45) derived from the entropy principle. To adapt it specifically for a flat interface, let ρ i \rho_i ρ i depend on a single coordinate – say, z z z – so that the interface is horizontal. Let the fluid below the interface be liquid and above, vapour,
ρ i → ρ i ( l ) as z → − ∞ , (68) \rho_i \to\rho_i^{(l)} \qquad\text{as} \qquad z \to-\infty, \tag*{(68)} ρ i → ρ i ( l ) as z → − ∞ , (68)
ρ i → ρ i ( v ) as z → + ∞ . (69) \rho_i \to\rho_i^{(v)} \qquad\text{as} \qquad z \to+\infty. \tag*{(69)} ρ i → ρ i ( v ) as z → + ∞. (69)
Taking in equation (45) the limit z → + ∞ z \to+\infty z → + ∞ , one can determine the constant η i \eta_i η i and rewrite (3.10) in the form
∑ j K i j d 2 ρ j d z 2 = G i − G i ( ρ 1 ( v ) … ρ N ( v ) , T ) . (70) \sum_j K_{ij}\frac{\mathrm{d}^2\rho_j}{\mathrm{d}z^2}=G_i-G_i(\rho_1^{(v)}\ldots\rho_N^{(v)},T). \tag*{(70)} j ∑ K ij d z 2 d 2 ρ j = G i − G i ( ρ 1 ( v ) … ρ N ( v ) , T ) . (70)
Boundary-value problem (6.1)–(6.3) is invariant with respect to the change z → z + c o n s t z \to z+\mathrm{const} z → z + const – hence, its solution is not unique. To make it such, an extra condition is needed – say,
ρ 1 = 1 2 ( ρ 1 ( l ) + ρ 1 ( v ) ) at z = 0. (71) \rho_1=\frac{1}{2}\left(\rho_1^{(l)}+\rho_1^{(v)}\right) \qquad\text{at} \qquad z=0. \tag*{(71)} ρ 1 = 2 1 ( ρ 1 ( l ) + ρ 1 ( v ) ) at z = 0. (71)
As shown before, the parameters ρ i ( l ) \rho_i^{(l)} ρ i ( l ) and ρ i ( v ) \rho_i^{(v)} ρ i ( v ) are not arbitrary but have to satisfy the Maxwell construction, i.e., conditions (5.7)–(5.8). Thus, they must be intrinsic to boundary-value problem (6.1)–(6.4) and can be derived directly from it, as a condition for existence of solution.
Indeed, consider equation (6.3) in the limit z → − ∞ z \to-\infty z → − ∞ and recall (6.1) – which immediately yields the first half of the Maxwell construction, equation (5.7). To derive the second half, consider
∑ i ∫ − ∞ ∞ ( 6.3 ) × d ρ i d z d z . \sum_i\int_{-\infty}^{\infty}(6.3)\times\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\,\mathrm{d}z. i ∑ ∫ − ∞ ∞ ( 6.3 ) × d z d ρ i d z .
The integral in this equation can be evaluated via identity (2.7), and the constant of integration can be fixed via boundary condition (6.2). After straightforward algebra, one obtains
1 2 ∑ i j K i j d ρ i d z d ρ j d z = ∑ i ρ i [ G i − G i ( ρ 1 ( v ) … ρ N ( v ) , T ) ] − p + p ( ρ 1 ( v ) … ρ N ( v ) , T ) . (72) \frac{1}{2}\sum_{ij}K_{ij}\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\frac{\mathrm{d}\rho_j}{\mathrm{d}z}
=\sum_i\rho_i\left[G_i-G_i(\rho_1^{(v)}\ldots\rho_N^{(v)},T)\right]-p+p(\rho_1^{(v)}\ldots\rho_N^{(v)},T). \tag*{(72)} 2 1 ij ∑ K ij d z d ρ i d z d ρ j = i ∑ ρ i [ G i − G i ( ρ 1 ( v ) … ρ N ( v ) , T ) ] − p + p ( ρ 1 ( v ) … ρ N ( v ) , T ) . (72)
Taking in this equation the limit z → − ∞ z \to-\infty z → − ∞ and recalling boundary condition (6.1), one recovers the second part of the Maxwell construction, equation (5.8), as required.
Boundary-value problem (6.1)–(6.4) was solved numerically for a two-component van der Waals fluid, using the MATLAB function BVP4c based on the three-stage Lobatto IIIa formula [54 ] . Examples of numerical solutions with parameters (5.9)–(5.10) and
K 11 = K 22 = 1 , K 12 = 0 , (73) K_{11} = K_{22} = 1,\qquad K_{12} = 0, \tag*{(73)} K 11 = K 22 = 1 , K 12 = 0 , (73)
are shown in Figure 3 . Observe that an increase in temperature makes the interface thicker, and it becomes infinitely thick when the critical temperature is reached.
Figure 3. Comparison between liquid/vapour interfaces in pure (solid line) and two-component (dotted line) van der Waals fluids with parameters (5.9)–(5.10) and (6.6). Curves (1)–(3) correspond to T = 0.17 T = 0.17 T = 0.17 , 0.23 0.23 0.23 , 0.28132 0.28132 0.28132 . The last of these values is the (approximate) boiling point, so that the solid and dotted curves virtually coincide and ρ 2 \rho_2 ρ 2 is virtually zero.
As shown in Appendix § 13 .§ 1 , all solutions describing liquid/vapour interfaces, are stable as long as d ρ i / d z d\rho_i/dz d ρ i / d z never vanishes (ρ i ( z ) \rho_i(z) ρ i ( z ) is strictly monotonic) for all i i i .
One-dimensional drops and bubbles
Apart from liquid/vapour interfaces, equation (3.10) admits spatially localised solutions, describing drops or bubbles floating in vapour or liquid, respectively. In this subsection, the simplest – one-dimensional – solutions of this kind are discussed, describing a flat layer of increased or decreased density. They both correspond to the following boundary condition
ρ i → ρ ∞ , i as z → ± ∞ , (74) \rho_i \to\rho_{\infty,i}\qquad\text{as}\qquad z \to\pm\infty, \tag*{(74)} ρ i → ρ ∞ , i as z → ± ∞ , (74)
where ρ ∞ , i \rho_{\infty,i} ρ ∞ , i is the density of the vapour (liquid). Unlike the interface solutions examined above, the fluid outside the drop (bubble) does not have to be saturated vapour (liquid); it can actually be any stable or metastable state.
The 1D reduction of the steady-state equation (45) and expression (46) for η i \eta_i η i can be written in the form
∑ j K i j d 2 ρ j d z 2 = G i − G ∞ , i , (75) \sum_j K_{ij} \frac{\mathrm{d}^2 \rho_j}{\mathrm{d}z^2} = G_i - G_{\infty,i}, \tag*{(75)} j ∑ K ij d z 2 d 2 ρ j = G i − G ∞ , i , (75)
where G ∞ , i = G i ( ρ ∞ , 1 … ρ ∞ , N , T ) G_{\infty,i} = G_i(\rho_{\infty,1} \ldots\rho_{\infty,N},T) G ∞ , i = G i ( ρ ∞ , 1 … ρ ∞ , N , T ) is the chemical potential at infinity.
Figure 4 illustrates typical solutions of boundary-value problem (6.7)–(6.8) computed for a pure fluid. The following features can be observed in panel (a) depicting some 1D drop solutions:
Figure 4. Solutions of boundary-value problem (6.7)–(6.8) for a pure van der Waals fluid with T = 0.2 T = 0.2 T = 0.2 . (a) Drop solutions with ρ ∞ , 1 = 0.1 \rho_{\infty,1} = 0.1 ρ ∞ , 1 = 0.1 , 0.05 0.05 0.05 , 0.03587 0.03587 0.03587 , 0.0358485 0.0358485 0.0358485 (curves (1)–(4), respectively). (b) Bubble solutions with ρ ∞ , 1 = 0.63 \rho_{\infty,1} = 0.63 ρ ∞ , 1 = 0.63 , 0.68 0.68 0.68 , 0.7286 0.7286 0.7286 , 0.728671 0.728671 0.728671 (curves (1)–(4), respectively). The labels “u[ndersaturated] v[apour]”, “o[versaturated] v[apour]”, etc. characterise ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 .
Drop solutions exist only if ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 corresponds to oversaturated vapour.
As ρ ∞ , 1 → ρ 1 ( v ) + 0 \rho_{\infty,1} \to\rho_1^{(v)} + 0 ρ ∞ , 1 → ρ 1 ( v ) + 0 , the drop becomes increasingly thick.
Once ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 passes ρ 1 ( v ) \rho_1^{(v)} ρ 1 ( v ) , drop solutions cease to exist. Physically, this is because liquid drops surrounded by undersaturated vapour evaporate.
Similar tendencies have been observed for 1D bubble solutions, illustrated in Figure 4 (b).
Most importantly, all drops (bubbles) in oversaturated vapour (undersaturated liquid) are likely to be unstable. For pure fluids (N = 1 N = 1 N = 1 ), the instability can be proven rigorously: as shown in Appendix § 13 .§ 2 , the entropy has a saddle (not maximum) on these solutions.
To understand how slightly perturbed drops and bubbles evolve, they were simulated numerically using the simplest asymptotic version of the DIM – the one based on equations (4.9)–(4.10) and (4.4). It is adapted for N = 1 N=1 N = 1 and a single spatial coordinate in Appendix D, which also outlines the numerical method used.
Various initial conditions have been simulated for the van der Waals pure fluid, with only two patterns of evolution observed. If the initial condition includes a steady-drop solution ρ 1 ( s d ) ( z ) \rho_1^{(sd)}(z) ρ 1 ( s d ) ( z ) plus some extra fluid, spontaneous condensation is typically triggered off, giving rise to two shock waves propagating away from the drop’s center. This behavior is illustrated in Figure 5 (a) for
Figure 5. Evolution of a perturbed drop, described by equations (134) –(135) and boundary conditions (D5)–(D6), for a pure van der Waals fluid and parameters (6.9). (a) Initial condition (6.10) (the steady solution plus extra fluid); the time t t t of a snapshot and the corresponding curve number n n n are inter-related via t = 5000 n t=5000n t = 5000 n . (b) Initial condition (6.11) (the steady solution minus extra fluid); t = 500 n t=500n t = 500 n .
T = 0.2 , ρ ∞ , 1 = 0.05 , (76) T=0.2,\qquad\rho_{\infty,1}=0.05, \tag*{(76)} T = 0.2 , ρ ∞ , 1 = 0.05 , (76)
and the initial condition
ρ 1 = ρ 1 ( s d ) ( 1.01 z ) , w = 0 at t = 0 , (77) \rho_1=\rho_1^{(sd)}(1.01z),\qquad w=0\qquad\text{at}\qquad t=0, \tag*{(77)} ρ 1 = ρ 1 ( s d ) ( 1.01 z ) , w = 0 at t = 0 , (77)
where w w w is the vertical velocity (the other two velocity components are zero due to the 1D nature of the flow).
If, however, the initial condition includes less fluid than ρ 1 ( n m ) ( z ) \rho_1^{(nm)}(z) ρ 1 ( nm ) ( z ) , the drop would typically evaporate. This pattern is illustrated in Figure 5 (b) for parameters (6.9) and the following initial condition,
ρ 1 = ρ 1 ( s d ) ( 0.99 z ) , w = 0 at t = 0. (78) \rho_1=\rho_1^{(sd)}(0.99z),\qquad w=0\qquad\text{at}\qquad t=0. \tag*{(78)} ρ 1 = ρ 1 ( s d ) ( 0.99 z ) , w = 0 at t = 0. (78)
The two patterns can be understood physically on the basis of the fact that the steadydrop solutions exist only for drops surrounded by oversaturated metastable vapour – which is stable with respect to infinitesimally small perturbations, but may be unstable with respect to finite ones. One can thus assume that the drop solution provides a lower bound for the mass of perturbations capable of triggering off instability of the surrounding vapour.
It is also interesting to see how a drop would evolve if it is surrounded by saturated (not oversaturated ) vapour.
The fact that the thickness of a steady drop becomes infinite as ρ ∞ , 1 → ρ 1 ( v ) \rho_{\infty,1} \to\rho_1^{(v)} ρ ∞ , 1 → ρ 1 ( v ) suggests that a finite-size liquid layer, surrounded by saturated vapour, evaporates. The mechanism of this evaporation, however, is not immediately clear. 2D and 3D drops, for example, evaporate because the curvature of their boundary gives rise to a liquid-to-vapour mass flux [6 , 7 , 9 ] – but the boundaries of 1D drops are flat. One can only assume that they evaporate due to a long-range interaction of the drop’s upper and lower interfaces.
To explore this hypothesis, 1D drops floating in saturated vapour were simulated numerically (using the same model and numerical method as before). It has turned out that these drops do evaporate, and their evolution can be subdivided into three distinct stages:
The boundaries of the drop rapidly assume the profile of a steady liquid/vapour interface – of the kind examined in the previous subsection.
The drop begins to get thinner, but the density of the drop’s core remains close to ρ 1 ( l ) \rho_1^{(l)} ρ 1 ( l ) .
Once the thickness of the drop becomes comparable to the interfacial thickness, the density at the drop’s center begins to rapidly decrease, and the drop disappears.
The main characteristic of such behavior is the evaporation time t e t_e t e , which can be defined as the interval over which the density at the drop’s center falls by a factor of, say, 10 – i.e.,
ρ 1 ( 0 , t e ) ρ 1 ( 0 , 0 ) = 0.1. \frac{\rho_1(0,t_e)}{\rho_1(0,0)}=0.1. ρ 1 ( 0 , 0 ) ρ 1 ( 0 , t e ) = 0.1.
t e t_e t e was computed as a function of the drop’s initial size W W W , for the following initial condition:
ρ 1 = ρ 1 ( v ) + ρ 1 ( l ) − ρ 1 ( v ) 1 + ( 2 z W ) 6 , w = 0 at t = 0. \rho_1=\rho_1^{(v)}+\frac{\rho_1^{(l)}-\rho_1^{(v)}}{1+\left(\frac{2z}{W}\right)^6}, \qquad w=0 \qquad\text{at} \qquad t=0. ρ 1 = ρ 1 ( v ) + 1 + ( W 2 z ) 6 ρ 1 ( l ) − ρ 1 ( v ) , w = 0 at t = 0.
It has turned out that, for W W W changing from 0.2 to 10, the evaporation time t e t_e t e grows from 48.5 to 10 6 10^6 1 0 6 – i.e., exponentially. This agrees with the hypothesis that the drop evaporation is caused by long-range interaction of the drop’s boundaries and the fact that the density in liquid/vapour interfaces tends to ρ 1 ( v ) \rho_1^{(v)} ρ 1 ( v ) and ρ 1 ( l ) \rho_1^{(l)} ρ 1 ( l ) (as z → ± ∞ z \to\pm\infty z → ± ∞ ) exponentially quickly.
Solid/fluid interfaces
Let the fluid be bounded below by a flat horizontal wall (substrate) located at z = 0 z=0 z = 0 , so that boundary condition (2.35) reduces to
ρ i = ρ 0 , i at z = 0. (79) \rho_i=\rho_{0,i} \qquad\text{at} \qquad z=0. \tag*{(79)} ρ i = ρ 0 , i at z = 0. (79)
Far above the substrate, the fluid is homogeneous,
ρ i → ρ ∞ , i as z → + ∞ , (80) \rho_i \to\rho_{\infty,i} \qquad\text{as} \qquad z \to+\infty, \tag*{(80)} ρ i → ρ ∞ , i as z → + ∞ , (80)
where ρ ∞ , i \rho_{\infty,i} ρ ∞ , i corresponds to a stable or metastable state. Finally, equation (6.8) used in the previous subsection for 1D drops and bubbles applies to the present case as well.
Boundary-value problem (6.8), (6.12)–(6.13) was solved numerically for a van der Waals pure fluid with
T = 0.2 , ρ 0 , 1 = 0.3. (81) T=0.2,\qquad\rho_{0,1}=0.3. \tag*{(81)} T = 0.2 , ρ 0 , 1 = 0.3. (81)
Its typical solutions are illustrated in figure 6 (ρ 0 , i \rho_{0,i} ρ 0 , i varies, ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 is fixed) and in figure 7 (ρ 0 , i \rho_{0,i} ρ 0 , i is fixed, ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 varies). The following features can be observed.
(i) For a given ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 , there exists a certain pool of ρ 0 , i \rho_{0,i} ρ 0 , i which can be ‘connected’ to this ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 . If, for example, ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 equals the saturated vapour density (the case illustrated in figure 6a), than ρ 0 , 1 \rho_{0,1} ρ 0 , 1 must be smaller than the saturated liquid density (and vice versa : if ρ ∞ , 1 = ρ 1 ( l ) \rho_{\infty,1}=\rho_1^{(l)} ρ ∞ , 1 = ρ 1 ( l ) , then ρ 0 , i > ρ 1 ( v ) \rho_{0,i}>\rho_1^{(v)} ρ 0 , i > ρ 1 ( v ) – see figure 6b). When ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 approaches the pool’s boundary, a clearly visible liquid/vapour interface emerges in the solution and is gradually moving away from the substrate.
(ii) It follows from the above that, for some pairs ( ρ 0 , 1 , ρ ∞ , 1 ) (\rho_{0,1},\rho_{\infty,1}) ( ρ 0 , 1 , ρ ∞ , 1 ) , no solution exists. As illustrated in figure 7, some of such pairs involve unstable values of ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 (and, thus, are unimportant), but there are also ones with a metastable ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 .
To clarify what happens in such cases, the evolution was simulated numerically (using the model and method used in the previous two subsections). It has turned out that, if boundary-value problem (6.8), (6.12)–(6.13) does not have a solution, the substrate triggers off a spontaneous phase change. This result can be readily interpreted physically: if a sufficiently hydrophilic substrate (with a sufficiently small ρ 0 , 1 \rho_{0,1} ρ 0 , 1 ) touches oversaturated vapour, it triggers off spontaneous condensation, whereas a sufficiently hydrophobic substrate touching undersaturated liquid triggers off evaporation. In either case, a shock wave of phase change propagates away from the substrate, and the solid/fluid interface cannot be stationary.
(iii) For some pairs ρ 0 , 1 \rho_{0,1} ρ 0 , 1 , ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 , there are two different solutions – a monotonic and non-monotonic ones (see Figure 7 ). It is unclear which of the two occurs in reality.
Figure 7. Examples of solid/fluid interfaces for a pure van der Waals fluid with parameters (6.14). The fluid at infinity is: (1) undersaturated vapour, (2) oversaturated vapour with two possible solutions (the unstable one is shown in thin line), (3) oversaturated vapour with a single solution, (4) undersaturated liquid with a single solution, etc. Observe that there are no solutions between curves (3) and (4).
It has turned out that only monotonic solutions of boundary-value problem (79) –(80) , (75) maximise the entropy, whereas non-monotonic solutions do not (the former claim is proved in the general case, but the latter, only for N = 1 N = 1 N = 1 – see Appendix § 13 .§ 3 ). That is, non-monotonic solutions cannot be ruled out with certainty for N ≥ 2 N \ge2 N ≥ 2 – yet the mere fact that the general stability proof that works for monotonic solutions cannot be extended to non-monotonic ones seems to resolve the dilemma in favour of the former.
(iv) Another physically important conclusion follows from the fact that boundary-value problem (75) , (79) –(80) has no more than one stable solution – and, thus, does not admit solutions describing a liquid layer of a finite thickness on a substrate, with saturated vapour above it. All such layers, regardless of their thickness, evaporate – and the Kelvin effect cannot be responsible for this effect (because the liquid/vapour interface is flat). The evaporation in this case can only be caused by long-range interaction between the interface and the substrate.
(v) It is interesting to compare the interfacial profiles shown in Figure 6 –Figure 7 to those computed by Evans et al. [33 ] via the density functional theory and Monte Carlo method for a Lennard-Jones fluid bounded by a single wall or contained between two parallel walls. The single-wall profiles of Evans et al. [33 ] (see their Figure 3 ) are qualitatively similar to those computed in this paper, but their two-wall profiles are riddled with short-scale oscillations (see Figure 6 , 14, and 20). It is not clear whether the oscillations are caused by ‘interference’ of the walls: on the one hand, the distance between the walls exceeds the spatial scale of wall–molecule interaction by a factor of 30 (hence, the ‘interference’ should be weak) – but, on the other hand, no oscillations occur when this distance is infinite.
Figure 6. Examples of solid/fluid interfaces for a pure van der Waals fluid with parameters (6.14), for a given ρ ∞ , 1 \rho_{\infty,1} ρ ∞ , 1 : (a) ρ ∞ , 1 = ρ 1 ( v ) \rho_{\infty,1}=\rho_1^{(v)} ρ ∞ , 1 = ρ 1 ( v ) , ρ 0 , 1 \rho_{0,1} ρ 0 , 1 varies from 0 to ρ 1 ( l ) \rho_1^{(l)} ρ 1 ( l ) ; (b) ρ ∞ , 1 = ρ 1 ( l ) \rho_{\infty,1}=\rho_1^{(l)} ρ ∞ , 1 = ρ 1 ( l ) , ρ 0 , 1 \rho_{0,1} ρ 0 , 1 varies from 1 to ρ 1 ( v ) \rho_1^{(v)} ρ 1 ( v ) . The arrows show the direction towards the boundary of the pool of allowable values of ρ 0 , 1 \rho_{0,1} ρ 0 , 1 .
Whatever the nature of the oscillations is, one should not expect the DIM to describe them due to the omission of the higher-order derivatives of ρ i \rho_i ρ i when obtaining expression (2.31) for the van der Waals force.
Surface tension and contact angle
Consider a fluid bounded below by a horizontal substrate, and an oblique liquid/vapour interface intersecting the substrate – see Figure 8 . Note that this figure depicts to a hydrophilic substrate, such that the contact angle θ \theta θ is smaller than π / 2 \pi/2 π /2 .
Figure 8. A schematic illustrating boundary conditions (7.4)–(7.5) for a static contact line. θ \theta θ is the contact angle, ϕ i ( s / v ) ( z ) \phi^{(s/v)}_i(z) ϕ i ( s / v ) ( z ) describes the solid/vapour interface, etc.
If in equilibrium, the setting outlined is described by the two-dimensional reduction of equation (45) . Given expression (3.11) for η i \eta_i η i , one obtains
G i − ∑ j K i j ( ∂ 2 ρ j ∂ x 2 + ∂ 2 ρ j ∂ z 2 ) + G ∞ , i = 0. (82) G_i-\sum_j K_{ij}\left(\frac{\partial^2\rho_j}{\partial x^2}+\frac{\partial^2\rho_j}{\partial z^2}\right)+G_{\infty,i}=0. \tag*{(82)} G i − j ∑ K ij ( ∂ x 2 ∂ 2 ρ j + ∂ z 2 ∂ 2 ρ j ) + G ∞ , i = 0. (82)
Impose also the boundary condition
ρ i → ρ i ( v ) as z → + ∞ , (83) \rho_i\to\rho_i^{(v)} \qquad\text{as}\qquad z\to+\infty, \tag*{(83)} ρ i → ρ i ( v ) as z → + ∞ , (83)
which implies that the constant in (7.1) is G ∞ , i = G i ( ρ 1 ( v ) … ρ N ( v ) , T ) G_{\infty,i}=G_i(\rho_1^{(v)}\ldots\rho_N^{(v)},T) G ∞ , i = G i ( ρ 1 ( v ) … ρ N ( v ) , T ) . At the substrate, the standard boundary condition,
ρ i = ρ 0 , i at z = 0 , (84) \rho_i=\rho_{0,i}\qquad\text{at}\qquad z=0, \tag*{(84)} ρ i = ρ 0 , i at z = 0 , (84)
is assumed.
To close boundary-value problem (7.1)–(7.3), one should set boundary conditions as x → ± ∞ x \to\pm\infty x → ± ∞ . The setting depicted in figure 8 corresponds to
ρ i → ρ i ( s / v ) ( z ) as x → − ∞ . (85) \rho_i \to\rho_i^{(s/v)}(z) \qquad\text{as} \qquad x \to-\infty. \tag*{(85)} ρ i → ρ i ( s / v ) ( z ) as x → − ∞. (85)
ρ i → ρ i ( s / l ) ( z ) + ρ i ( l / v ) ( ξ ) as x → + ∞ . (86) \rho_i \to\rho_i^{(s/l)}(z) + \rho_i^{(l/v)}(\xi) \qquad\text{as} \qquad x \to+\infty. \tag*{(86)} ρ i → ρ i ( s / l ) ( z ) + ρ i ( l / v ) ( ξ ) as x → + ∞. (86)
Here, the function ρ i ( s / v ) ( z ) \rho_i^{(s/v)}(z) ρ i ( s / v ) ( z ) describes a solid/vapour interface (i.e., satisfies boundary-value problem (6.8), (6.12)–(6.13) with ρ ∞ , i = ρ i ( v ) \rho_{\infty,i} = \rho_i^{(v)} ρ ∞ , i = ρ i ( v ) ); the function ρ i ( s / l ) ( z ) \rho_i^{(s/l)}(z) ρ i ( s / l ) ( z ) describes a solid/liquid interface; and ρ i ( l / v ) ( ξ ) \rho_i^{(l/v)}(\xi) ρ i ( l / v ) ( ξ ) (where ξ = z cos θ − x sin θ \xi= z \cos\theta- x \sin\theta ξ = z cos θ − x sin θ ) describes a liquid/vapour interface tilted at an angle θ \theta θ (it satisfies boundary-value problem (6.1)–(6.3) with z z z changed to ξ \xi ξ ).
In the framework of the DIM, the contact angle θ \theta θ is not arbitrary, but is fully determined by the fluid’s thermodynamic properties, the Korteweg matrix K i j K_{ij} K ij , and the near-wall density ρ 0 , i \rho_{0,i} ρ 0 , i . To derive an expression for θ \theta θ (similar to that derived by Pismen & Pomeau [74 ] for pure fluids), consider
∑ i ∫ 0 ∞ ( 7.1 ) × ∂ ρ i ∂ x d z . \sum_i \int_0^\infty(7.1) \times\frac{\partial\rho_i}{\partial x}\,\mathrm{d}z. i ∑ ∫ 0 ∞ ( 7.1 ) × ∂ x ∂ ρ i d z .
After straightforward algebra involving integration by parts and the use of boundary conditions (7.2)–(7.3) and identity (2.7), one obtains
∂ ∂ x ∫ 0 ∞ [ ∑ j ρ j ( G j − G j , ∞ ) − p + p ∞ + 1 2 ∑ i j K i j ( ∂ ρ i ∂ z ∂ ρ j ∂ z − ∂ ρ i ∂ x ∂ ρ j ∂ x ) ] d z = 0 , \frac{\partial}{\partial x}\int_0^\infty\left[\sum_j \rho_j(G_j-G_{j,\infty})-p+p_\infty+\frac{1}{2}\sum_{ij}K_{ij}\left(\frac{\partial\rho_i}{\partial z}\frac{\partial\rho_j}{\partial z}-\frac{\partial\rho_i}{\partial x}\frac{\partial\rho_j}{\partial x}\right)\right]\mathrm{d}z=0, ∂ x ∂ ∫ 0 ∞ [ j ∑ ρ j ( G j − G j , ∞ ) − p + p ∞ + 2 1 ij ∑ K ij ( ∂ z ∂ ρ i ∂ z ∂ ρ j − ∂ x ∂ ρ i ∂ x ∂ ρ j ) ] d z = 0 ,
where p ∞ = p ( ρ 1 ( v ) … ρ N ( v ) , T ) p_\infty=p(\rho_1^{(v)}\ldots\rho_N^{(v)},T) p ∞ = p ( ρ 1 ( v ) … ρ N ( v ) , T ) . Finally, integrating the above expression from x = − ∞ x=-\infty x = − ∞ to x = + ∞ x=+\infty x = + ∞ and recalling boundary conditions (7.4)–(7.5), one obtains
∫ 0 ∞ { [ ∑ i ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( s / v ) + 1 2 ∑ i j K i j d ρ i ( s / v ) d z d ρ j ( s / v ) d z } d z = ∫ 0 ∞ { [ ∑ i ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( s / l ) + 1 2 ∑ i j K i j d ρ i ( s / l ) d z d ρ j ( s / l ) d z } d z + ∫ 0 ∞ { [ ∑ i ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( l / v ) + 1 2 ( cos 2 θ − sin 2 θ ) ∑ i j K i j d ρ i ( l / v ) d ξ d ρ j ( l / v ) d ξ } d ξ cos θ . \begin{aligned}
\int_0^\infty\Bigg\{ \left[\sum_i \rho_i(G_i-G_{\infty,i})-p+p_\infty\right]_{\rho_i=\rho_i^{(s/v)}}+\frac{1}{2}\sum_{ij}K_{ij}\frac{\mathrm{d}\rho_i^{(s/v)}}{\mathrm{d}z}\frac{\mathrm{d}\rho_j^{(s/v)}}{\mathrm{d}z}\Bigg\}\,\mathrm{d}z
={}&\int_0^\infty\Bigg\{ \left[\sum_i \rho_i(G_i-G_{\infty,i})-p+p_\infty\right]_{\rho_i=\rho_i^{(s/l)}}+\frac{1}{2}\sum_{ij}K_{ij}\frac{\mathrm{d}\rho_i^{(s/l)}}{\mathrm{d}z}\frac{\mathrm{d}\rho_j^{(s/l)}}{\mathrm{d}z}\Bigg\}\,\mathrm{d}z \\
&+\int_0^\infty\Bigg\{ \left[\sum_i \rho_i(G_i-G_{\infty,i})-p+p_\infty\right]_{\rho_i=\rho_i^{(l/v)}}+\frac{1}{2}\left(\cos^2\theta-\sin^2\theta\right)\sum_{ij}K_{ij}\frac{\mathrm{d}\rho_i^{(l/v)}}{\mathrm{d}\xi}\frac{\mathrm{d}\rho_j^{(l/v)}}{\mathrm{d}\xi}\Bigg\}\frac{\mathrm{d}\xi}{\cos\theta}.
\end{aligned} ∫ 0 ∞ { [ i ∑ ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( s / v ) + 2 1 ij ∑ K ij d z d ρ i ( s / v ) d z d ρ j ( s / v ) } d z = ∫ 0 ∞ { [ i ∑ ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( s / l ) + 2 1 ij ∑ K ij d z d ρ i ( s / l ) d z d ρ j ( s / l ) } d z + ∫ 0 ∞ { [ i ∑ ρ i ( G i − G ∞ , i ) − p + p ∞ ] ρ i = ρ i ( l / v ) + 2 1 ( cos 2 θ − sin 2 θ ) ij ∑ K ij d ξ d ρ i ( l / v ) d ξ d ρ j ( l / v ) } cos θ d ξ .
This equality can be simplified using identity (6.5) (with ρ i \rho_i ρ i changed to ρ i ( l / v ) \rho_i^{(l/v)} ρ i ( l / v ) ), and similar identities for ρ i ( s / v ) \rho_i^{(s/v)} ρ i ( s / v ) and ρ i ( s / l ) \rho_i^{(s/l)} ρ i ( s / l ) . After straightforward algebra involving re-denoting ξ → z \xi\to z ξ → z , one obtains the DIM version of Young’s formula,
cos θ = σ ( s / v ) − σ ( s / l ) σ ( l / v ) . (87) \cos\theta=\frac{\sigma^{(s/v)}-\sigma^{(s/l)}}{\sigma^{(l/v)}}. \tag*{(87)} cos θ = σ ( l / v ) σ ( s / v ) − σ ( s / l ) . (87)
where
σ ( s / v ) = ∑ i j K i j ∫ 0 ∞ d ρ i ( s / v ) d z d ρ j ( s / v ) d z d z , σ ( s / l ) = ∑ i j K i j ∫ 0 ∞ d ρ i ( s / l ) d z d ρ j ( s / l ) d z d z , σ ( l / v ) = ∑ i j K i j ∫ − ∞ ∞ d ρ i ( l / v ) d z d ρ j ( l / v ) d z d z (88) \begin{aligned}
\sigma^{(s/v)} &= \sum_{ij} K_{ij} \int_{0}^{\infty} \frac{\mathrm{d}\rho_i^{(s/v)}}{\mathrm{d}z}\frac{\mathrm{d}\rho_j^{(s/v)}}{\mathrm{d}z}\,\mathrm{d}z, \qquad
\sigma^{(s/l)} = \sum_{ij} K_{ij} \int_{0}^{\infty} \frac{\mathrm{d}\rho_i^{(s/l)}}{\mathrm{d}z}\frac{\mathrm{d}\rho_j^{(s/l)}}{\mathrm{d}z}\,\mathrm{d}z, \\
\sigma^{(l/v)} &= \sum_{ij} K_{ij} \int_{-\infty}^{\infty} \frac{\mathrm{d}\rho_i^{(l/v)}}{\mathrm{d}z}\frac{\mathrm{d}\rho_j^{(l/v)}}{\mathrm{d}z}\,\mathrm{d}z \tag*{(88)}
\end{aligned} σ ( s / v ) σ ( l / v ) = ij ∑ K ij ∫ 0 ∞ d z d ρ i ( s / v ) d z d ρ j ( s / v ) d z , σ ( s / l ) = ij ∑ K ij ∫ 0 ∞ d z d ρ i ( s / l ) d z d ρ j ( s / l ) d z , = ij ∑ K ij ∫ − ∞ ∞ d z d ρ i ( l / v ) d z d ρ j ( l / v ) d z (88)
are the surface tension coefficients of the solid/vapour, solid/liquid, and liquid/vapour interfaces, respectively. Treating hydrophobic substrates (θ > 1 2 π \theta> \frac{1}{2}\pi θ > 2 1 π ) in a similar fashion, one can show that formula (87) applies to that case as well.
To calculate the surface tension – say, σ ( s / l ) \sigma^{(s/l)} σ ( s / l ) – one first needs to solve boundary-value problem (6.1)–(6.4) which determines the function ρ i = ρ i ( l / v ) ( z ) \rho_i = \rho_i^{(l/v)}(z) ρ i = ρ i ( l / v ) ( z ) . For a pure fluid, however, the expression for σ ( s / l ) \sigma^{(s/l)} σ ( s / l ) can be rewritten as a closed-form integral. To do so, let N = 1 N = 1 N = 1 in equation (72) , which yields
d ρ 1 d z = − 2 K 11 { ρ 1 [ G 1 ( ρ 1 , T ) − G 1 ( ρ 1 ( v ) , T ) ] − p ( ρ 1 , T ) + p ( ρ 1 ( v ) , T ) } \frac{\mathrm{d}\rho_1}{\mathrm{d}z}=-\sqrt{\frac{2}{K_{11}}\left\{\rho_1\left[G_1(\rho_1,T)-G_1(\rho_1^{(v)},T)\right]-p(\rho_1,T)+p(\rho_1^{(v)},T)\right\}} d z d ρ 1 = − K 11 2 { ρ 1 [ G 1 ( ρ 1 , T ) − G 1 ( ρ 1 ( v ) , T ) ] − p ( ρ 1 , T ) + p ( ρ 1 ( v ) , T ) }
Using this equation to change the variable of integration in the expression for σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) in (88) , one obtains
σ ( l / v ) = 2 K 11 ∫ ρ 1 ( v ) ρ 1 ( l ) ρ 1 [ G 1 ( ρ 1 , T ) − G 1 ( ρ 1 ( v ) , T ) ] − p ( ρ 1 , T ) + p ( ρ 1 ( v ) , T ) d ρ (89) \sigma^{(l/v)}=\sqrt{2K_{11}}\int_{\rho_1^{(v)}}^{\rho_1^{(l)}}\sqrt{\rho_1\left[G_1(\rho_1,T)-G_1(\rho_1^{(v)},T)\right]-p(\rho_1,T)+p(\rho_1^{(v)},T)}\,\mathrm{d}\rho \tag*{(89)} σ ( l / v ) = 2 K 11 ∫ ρ 1 ( v ) ρ 1 ( l ) ρ 1 [ G 1 ( ρ 1 , T ) − G 1 ( ρ 1 ( v ) , T ) ] − p ( ρ 1 , T ) + p ( ρ 1 ( v ) , T ) d ρ (89)
Thus, to calculate σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) , one should first use the given p ( ρ 1 , T ) p(\rho_1,T) p ( ρ 1 , T ) and G 1 ( ρ 1 , T ) G_1(\rho_1,T) G 1 ( ρ 1 , T ) to find ρ 1 ( l ) \rho_1^{(l)} ρ 1 ( l ) and ρ 1 ( v ) \rho_1^{(v)} ρ 1 ( v ) (via the Maxwell construction) and then evaluate integral (89) .
Parameterizing the diffuse–interface model for water/air interfaces
To use the DIM in applications, one needs the following external parameters: the fluid’s thermodynamic properties, the dependence of the viscosity and transport coefficients on ( ρ 1 … ρ N , T ) (\rho_1\ldots\rho_N,T) ( ρ 1 … ρ N , T ) , and the Korteweg matrix. In this section, an approach to specifying these parameters is described and applied to water/air interfaces.
For reasons described in subsection 2.2, the fluid’s thermodynamic properties will be approximated by the Enskog–Vlasov (EV) model. It involves an undetermined matrix a i j a_{ij} a ij and an undetermined function Θ ( ρ 1 … ρ N ) \Theta(\rho_1\ldots\rho_N) Θ ( ρ 1 … ρ N ) , which should be fixed as the best fits of the empiric characteristics of the fluid under consideration.
In subsections 8.1–8.2, it will be explained how the fitting should be carried out for a pure fluid (in application to water, nitrogen, and oxygen). An approach to determining K 11 K_{11} K 11 for a pure fluid will be outlined in subsection 8.3. Subsection 8.4 describes how a i j a_{ij} a ij , Θ \Theta Θ , and K i j K_{ij} K ij can be determined for water–air mixture, and its viscosity and transport coefficients are dealt with in subsection 8.5.
All these tasks will be carried out in terms of the original (dimensional) variables.
The van der Waals parameter of pure water, nitrogen, and oxygen
For a pure fluid, the EV expression (12) for the internal energy yields (the subscript 1 1 1 is omitted)
e = c T − a ρ , e=cT-a\rho, e = c T − a ρ ,
which suggests that the van der Waals parameter a a a can be determined as the linear fit of the empiric dependence of c T − e cT-e c T − e on ρ \rho ρ . The heat capacity c c c in this expression should be the same as that in the Enskog–Vlasov kinetic theory – i.e., 3 R 3R 3 R for water and 5 R / 2 5R/2 5 R /2 for nitrogen and oxygen. For simplicity, the fitting was carried out using only the data on the critical isobar p = p c r p=p_{\mathrm{cr}} p = p cr [as done by Benilov & Benilov [12 ] ], but the resulting straight line fits the isobars p = p c r / 2 p=p_{\mathrm{cr}}/2 p = p cr /2 and p = 2 p c r p=2p_{\mathrm{cr}} p = 2 p cr reasonably well too (see the top panels of Figure 9 ). The parameters of the critical points for the fluids under consideration, as well as the other parameters needed here and hereinafter, can be found in Table 1 –Table 2 .
Figure 9. Thermodynamic properties of H 2 O \mathrm{H_2O} H 2 O (left-hand panels) and N 2 \mathrm{N_2} N 2 (right-hand panels). The non-connected symbols show the empiric data from Lindstrom & Mallard [58 ] presented in isobaric form, for three values of the pressure p p p relative to its critical value, p c r p_{\mathrm{cr}} p cr (see the legend). The gap between ρ ( v ) \rho^{(v)} ρ ( v ) and ρ ( l ) \rho^{(l)} ρ ( l ) in the empiric data for p = p c r / 2 p=p_{\mathrm{cr}}/2 p = p cr /2 reflects the impossibility (difficulty) of measurements in an unstable (metastable) fluid. The dotted lines show the parameters calculated via the EV fluid model.
Table 1. The parameters of the triple and critical points (subscripts t r tr t r and c r cr cr , respectively) of H 2 O \mathrm{H_2O} H 2 O , N 2 \mathrm{N_2} N 2 , and O 2 \mathrm{O_2} O 2 Lindstrom & Mallard [58 ] .
Table 2. The parameters of H2 _2 2 O, N2 _2 2 , O2 _2 2 , and air: m m m is the molecular mass, R R R is the specific gas constant, a a a is the van der Waals parameter, K K K is the Korteweg constant. The parameters of air are calculated as the 79-21 weighted averages of the corresponding parameters of nitrogen and oxygen, respectively.
The resulting values of the van der Waals parameter a a a for H2 _2 2 O, N2 _2 2 , and O2 _2 2 are presented in Table 2 . Interestingly, a a a of water exceeds significantly those of nitrogen and oxygen. This is likely to be caused by the difference in the molecular structure of these fluids: the water molecule has a non-zero dipolar moment, whereas nitrogen and oxygen molecules are symmetric – hence, do not. Since the van der Waals force is of electric nature, one can assume that dipolar molecules interact stronger than neutral ones.
The equation of state of pure H2 _2 2 O, N2 _2 2 , and O2 _2 2
Pure fluids will be described by the EV model {(2.12)–(2.13)} with
Θ = R [ − q ( 0 ) ln ( 1 − 0.99 ρ ρ t p ) + ∑ n = 1 4 q ( n ) ( ρ ρ t p ) n ] (90) \Theta= R\left[-q^{(0)}\ln\left(1-0.99\frac{\rho}{\rho_{\mathrm{tp}}}\right)+\sum_{n=1}^{4}q^{(n)}\left(\frac{\rho}{\rho_{\mathrm{tp}}}\right)^n\right] \tag*{(90)} Θ = R [ − q ( 0 ) ln ( 1 − 0.99 ρ tp ρ ) + n = 1 ∑ 4 q ( n ) ( ρ tp ρ ) n ] (90)
where q ( 0 ) … q ( 4 ) q^{(0)}\ldots q^{(4)} q ( 0 ) … q ( 4 ) are undetermined coefficients and ρ t p \rho_{\mathrm{tp}} ρ tp is the fluid’s density at the triple point (ρ t p \rho_{\mathrm{tp}} ρ tp is just a convenient density scale; the fact that, at the triple point, the liquid and vapour are in equilibrium with the solid phase is irrelevant, as solids are not described by the DIM). Note that the first term in expression (90) sets the maximum density at ρ t p / 0.99 \rho_{\mathrm{tp}}/0.99 ρ tp /0.99 .
The coefficients q ( n ) q^{(n)} q ( n ) were determined for H2 _2 2 O, N2 _2 2 , and O2 _2 2 by ensuring that the expressions for p ( ρ , T ) p(\rho,T) p ( ρ , T ) and G ( ρ , T ) G(\rho,T) G ( ρ , T ) corresponding to (90) yield the ‘correct’ – i.e., empiric – values for the critical density, temperature, and pressure, as well as the liquid and vapour densities at the triple-point temperature T = T t p T=T_{\mathrm{tp}} T = T tp (five equations for the five unknown coefficients). All the necessary empiric data can be found in Table 1 –Table 2 , and the computed values of q ( n ) q^{(n)} q ( n ) are listed in Table 3. Such an approach to calibrating the EV fluid model is a slight modification of that of Benilov [4 ] – which is, in turn, a modification of the approach of Benilov & Benilov [11 ] .
To illustrate how well the EV model describes real fluids, the isobaric (with p p p fixed) dependence of the temperature on the density is plotted for H2 _2 2 O and N2 _2 2 in the lower panels of Figure 9 . The results for O2 _2 2 (not presented) are similar to those for N2 _2 2 .
Note that the DIM has been coupled with realistic equations of state before (e.g. Caupin [20 ] ; Magaletti et al. [64 ] ); for pure water, this was typically done using the IAPWS-95 equation (Wagner & Pruss [97 ] ). The EV model used here is undoubtedly
caption: tbl:3 : The coefficients of the equation of state (8.1) for H2 _2 2 O, N2 _2 2 , and O2 _2 2 .
:::
less accurate than the IAPWS-95 model – but it allows one to describe consistently all of the fluids involved (H2 _2 2 O, N2 _2 2 , and O2 _2 2 ).
The Korteweg parameter of pure H2 _2 2 O, N2 _2 2 , and O2 _2 2
[4 ] proposed to deduce K K K from the requirement that the DIM predict the correct value of the surface tension σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) of liquid/vapour interface at the triple point. The same was done in the present work: σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) was calculated via the DIM formula (89) with G G G and p p p of the EV fluid, and the value of K K K was chosen for which (89) yields the same result as the empiric formula of [88 ] . The resulting K K K for H2 _2 2 O, N2 _2 2 , and O2 _2 2 can be found in Table 2 .
It should be emphasised that the values of K K K computed via the above approach depend on the chosen fluid model. This explains the difference between the result for K K K in this paper and the one computed by Benilov [4 ] : the former is based on the EV fluid model and the latter, on the van der Waals model. The resulting 30% difference in σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) reflects the fact that the latter model is much less accurate.
As a test of the whole approach based on the DIM and EV models, the theoretical values for the saturation characteristics and σ ( l / v ) \sigma^{(l/v)} σ ( l / v ) have been compared to their empiric counterparts for the whole temperature range where liquid/vapour interfaces exist, i.e., between the triple and critical points. The results are shown in Figure 10 : evidently, the theoretical prediction are reasonably accurate.
Figure 10. Comparison of the results obtained via the DIM for an EV fluid (solid line) with the corresponding empiric data (dashed line): (a) the densities of saturated vapour and liquid, (b) saturated pressure, (c) surface tension of liquid/vapour interface. The empiric data in panels (a) and (b) are from [58 ] , and those in panel (c) are from [88 ] .
Parameters of water–air interaction
Generally, the parameters of a fluid should be deduced from measurements of its equation of state, surface tension, etc. – but these are rarely known for multicomponent fluids. The water–air mixture at normal conditions is an exception in this respect: the density of air is small in this case, and its equation of state is close to that of an ideal gas. In addition, air will be treated as a pure fluid with parameters equal to the weighted averages of those of nitrogen and oxygen (see Table 2 ).
Thus, let the function Θ \Theta Θ (the non-ideal part of the entropy of an EV fluid) be independent of the air density ρ 2 \rho_{2} ρ 2 , so that expressions (2.15)–(2.16) with N = 2 N=2 N = 2 yield
p = T [ R 1 ρ 1 + R 2 ρ 2 + ( ρ 1 + ρ 2 ) d Θ ( ρ 1 ) d ρ 1 ] − a 11 ρ 1 2 − 2 a 12 ρ 1 ρ 2 (91) p = T\left[R_{1}\rho_{1}+R_{2}\rho_{2}+(\rho_{1}+\rho_{2})\frac{\mathrm{d}\Theta(\rho_{1})}{\mathrm{d}\rho_{1}}\right]-a_{11}\rho_{1}^{2}-2a_{12}\rho_{1}\rho_{2} \tag*{(91)} p = T [ R 1 ρ 1 + R 2 ρ 2 + ( ρ 1 + ρ 2 ) d ρ 1 d Θ ( ρ 1 ) ] − a 11 ρ 1 2 − 2 a 12 ρ 1 ρ 2 (91)
G 1 = T [ R 1 ln ρ 1 + Θ ( ρ 1 ) + ( ρ 1 + ρ 2 ) d Θ ( ρ 1 ) d ρ 1 ] − 2 ( a 11 ρ 1 + a 12 ρ 2 ) + T [ R 1 + c 1 ( 1 − ln T ) ] (92) G_{1}=T\left[R_{1}\ln\rho_{1}+\Theta(\rho_{1})+(\rho_{1}+\rho_{2})\frac{\mathrm{d}\Theta(\rho_{1})}{\mathrm{d}\rho_{1}}\right]-2(a_{11}\rho_{1}+a_{12}\rho_{2})+T\left[R_{1}+c_{1}(1-\ln T)\right] \tag*{(92)} G 1 = T [ R 1 ln ρ 1 + Θ ( ρ 1 ) + ( ρ 1 + ρ 2 ) d ρ 1 d Θ ( ρ 1 ) ] − 2 ( a 11 ρ 1 + a 12 ρ 2 ) + T [ R 1 + c 1 ( 1 − ln T ) ] (92)
G 2 = T [ R 2 ln ρ 2 + Θ ( ρ 1 ) ] − 2 a 12 ρ 1 + T [ R 2 + c 2 ( 1 − ln T ) ] . (93) G_2 = T [R_2 \ln\rho_2 + \Theta(\rho_1)] - 2a_{12}\rho_1 + T [R_2 + c_2 (1 - \ln T)] . \tag*{(93)} G 2 = T [ R 2 ln ρ 2 + Θ ( ρ 1 )] − 2 a 12 ρ 1 + T [ R 2 + c 2 ( 1 − ln T )] . (93)
Here, the function Θ ( ρ 1 ) \Theta(\rho_1) Θ ( ρ 1 ) is the one given by formula (8.1), with ρ \rho ρ changed to ρ 1 \rho_1 ρ 1 , and with its coefficients corresponding to water.
Before using expressions (8.2)–(8.4), one needs to fix the non-diagonal term a 12 a_{12} a 12 of the matrix a i j a_{ij} a ij , responsible for the interaction of water and air molecules. It can be deduced from ρ 2 ( l ) \rho_2^{(l)} ρ 2 ( l ) (the amount of air dissolved in water): one should choose such a 12 a_{12} a 12 that the Maxwell construction based on (8.2)–(8.4) predicts the correct value of ρ 2 ( l ) \rho_2^{(l)} ρ 2 ( l ) .
Since a 12 a_{12} a 12 is supposed to not depend on T T T , such a calculation should be done for a single temperature and the atmospheric pressure. For T = 25 ∘ C T = 25^\circ\mathrm{C} T = 2 5 ∘ C and p = 1 a t m p = 1\ \mathrm{atm} p = 1 atm , for example, ρ 2 ( l ) = 0.0227 k g m − 3 \rho_2^{(l)} = 0.0227\,\mathrm{kg\,m^{-3}} ρ 2 ( l ) = 0.0227 kg m − 3 [31 ] . It can then be deduced that the Maxwell construction based on (91) –(93) yields the correct value of ρ 2 ( l ) \rho_2^{(l)} ρ 2 ( l ) if
a 12 = 208 m 5 s − 2 k g − 1 , a_{12} = 208\,\mathrm{m^5\,s^{-2}\,kg^{-1}}, a 12 = 208 m 5 s − 2 k g − 1 ,
i.e., approximately equal to a 22 a_{22} a 22 and ten times smaller than a 11 a_{11} a 11 .
The accuracy of expressions (91) –(93) can be illustrated by the corresponding value of the boiling point (T ≈ 109 ∘ C T \approx109^\circ\mathrm{C} T ≈ 10 9 ∘ C , as opposed to the exact value of T = 100 ∘ C T = 100^\circ\mathrm{C} T = 10 0 ∘ C ). Furthermore, at the “room temperature” (say, T = 25 ∘ C T = 25^\circ\mathrm{C} T = 2 5 ∘ C ), the error should be even smaller, because the room temperature is close to the triple point of water (T ≈ 0 ∘ C T \approx0^\circ\mathrm{C} T ≈ 0 ∘ C ) where (91) –(93) were calibrated.
It still remains to determine the non-diagonal term K 12 K_{12} K 12 of the Korteweg matrix K i j K_{ij} K ij . Since K i j K_{ij} K ij is supposed to be positive definite, K 12 K_{12} K 12 should satisfy
− 1 < K 12 ( K 11 K 22 ) 1 / 2 < 1. (94) -1 < \frac{K_{12}}{(K_{11}K_{22})^{1/2}} < 1. \tag*{(94)} − 1 < ( K 11 K 22 ) 1/2 K 12 < 1. (94)
One would think that K 12 K_{12} K 12 could be determined by comparing the surface tension of liquid-water/air interface to that of liquid-water/vapour-water interface. It turns out, however, that the difference between these is so small that the existing experimental techniques cannot detect it (see Vargaftik et al. [94 ] , and references therein). This can be due to smallness of K 12 K_{12} K 12 – but more likely, because the density of air is small. Either way, the determination of K 12 K_{12} K 12 via measurements of surface tension is impossible.
To at least confirm that air cannot affect significantly the surface tension of water, boundary-value problem (68) –(71) was solved for the parameters of water and air at T = 25 ∘ C T = 25^\circ\mathrm{C} T = 2 5 ∘ C , and K 12 K_{12} K 12 varying through range (94) . It has turned out that the resulting variation of the surface tension is only 7%.
Even though the dependence of the surface tension (which is a global characteristic ) on K 12 K_{12} K 12 is weak, this parameter does influence the local structure of the interface.
Figure 11 depicts the density profiles ρ 1 ( z ) \rho_1(z) ρ 1 ( z ) and ρ 2 ( z ) \rho_2(z) ρ 2 ( z ) for the water/air interface at room temperature [i.e., the above-mentioned solution of boundary-value problem (68) –(71) ]. One can see that, as K 12 K_{12} K 12 approaches the right endpoint of range (94) , the profile of water density is getting steeper, and a layer of high air density is developing inside the interface. For curve 5 of figure 11(b) – which is half-way through range (94) – the maximum of ρ 2 ( z ) \rho_2(z) ρ 2 ( z ) exceeds the atmospheric air density by a factor of approximately 20. Since the amount of air drawn into the interface grows with K 12 K_{12} K 12 , one concludes that the air is ‘pulled in’ by the van der Waals attraction exerted by the bulk of the liquid.
This suggests a possibility of deducing K 12 K_{12} K 12 from empiric data on evaporation of drops (which depends on both global and local characteristics of the interface (Benilov [9 ] ). Such an approach, however, requires an extension of the evaporation model of Benilov [9 ] to multicomponent fluids, which is currently in progress.
Observe that the interfaces depicted in figure 11 are such that ρ 2 ( z ) \rho_2(z) ρ 2 ( z ) is not monotonic – hence, the sufficient stability criterion from subsection 6.1 and Appendix C.1 does not apply. This does not, however, mean that this interface is unstable. It is, in fact, difficult to imagine that microscopic non-monotonicity of the density of air dissolved in water could destabilise the whole interface – but this issue is still of interest theoretically and, thus, deserves further investigation.
The viscosity and thermal conductivity of water–air mixture
There is a large body of work on the viscosity of multicomponent fluids, with even the simplest theoretical models – e.g., that of Enskog–Chapman for dilute gases [e.g. 34 ] – yielding a fairly complicated dependence of μ s \mu_s μ s and μ b \mu_b μ b on ρ i \rho_i ρ i . Phenomenological results, on the other hand, tend to include many ad hoc parameters specific to the fluid under consideration ([e.g. 25 and references therein] ). Overall, the simplest option seems to be the expression for the shear viscosity proposed on a phenomenological basis by Hind et al. [48 ] and justified, under certain approximations, via statistical mechanics by Bearman & Jones [1 ] ,
Figure 11. The spatial structure of water/air interfaces, for T = 25 ∘ C T = 25^{\circ}\mathrm{C} T = 2 5 ∘ C and K 12 = ( K 11 K 22 ) 1 / 2 × n / 10 K_{12} = (K_{11}K_{22})^{1/2} \times n/10 K 12 = ( K 11 K 22 ) 1/2 × n /10 , where n n n is the curve number: (a) density of water; (b) density of air. The horizontal scales of panel (a) and (b) differ by a factor of 5; the latter and the horizontal scale of the shaded inset in panel (b) differ by a factor of 1500. The inset shows that the concentration of the gas dissolved in liquid is non-zero.
μ s = μ s , 1 f 1 2 + ( μ s , 1 + μ s , 2 ) f 1 f 2 + μ s , 2 f 2 2 . (95) \mu_s = \mu_{s,1} f_1^2 + (\mu_{s,1} + \mu_{s,2}) f_1 f_2 + \mu_{s,2} f_2^2. \tag*{(95)} μ s = μ s , 1 f 1 2 + ( μ s , 1 + μ s , 2 ) f 1 f 2 + μ s , 2 f 2 2 . (95)
where μ s , i \mu_{s,i} μ s , i is the shear viscosity of the i i i -th species and
f i = ρ i / m i ρ 1 / m 1 + ρ 2 / m 2 f_i = \frac{\rho_i/m_i}{\rho_1/m_1 + \rho_2/m_2} f i = ρ 1 / m 1 + ρ 2 / m 2 ρ i / m i
is its mole fraction. Expression (8.6) does not include any fluid-specific parameters, but is capable of predicting the shear viscosity with a reasonable accuracy. If, for example, air is treated as a mixture of N 2 \mathrm{N}_2 N 2 and O 2 \mathrm{O}_2 O 2 , the error of (8.6) for the range T = 0 T = 0 T = 0 –100 ∘ C 100^{\circ}\mathrm{C} 10 0 ∘ C is less than 1%. A similar “mixture rule” can be assumed for the bulk viscosity and thermal conductivity,
μ b = μ b , 1 f 1 2 + ( μ b , 1 + μ b , 2 ) f 1 f 2 + μ b , 2 f 2 2 . (96) \mu_b = \mu_{b,1}f_1^2 + (\mu_{b,1}+\mu_{b,2})f_1f_2 + \mu_{b,2}f_2^2. \tag*{(96)} μ b = μ b , 1 f 1 2 + ( μ b , 1 + μ b , 2 ) f 1 f 2 + μ b , 2 f 2 2 . (96)
κ = κ 1 f 1 2 + ( κ 1 + κ 2 ) f 1 f 2 + κ 2 f 2 2 . (97) \kappa= \kappa_1f_1^2 + (\kappa_1+\kappa_2)f_1f_2 + \kappa_2f_2^2. \tag*{(97)} κ = κ 1 f 1 2 + ( κ 1 + κ 2 ) f 1 f 2 + κ 2 f 2 2 . (97)
Generally, various mechanical properties of a multicomponent fluid are often described by the same mixture rule, in which case it is referred to as “generalised”.
It still remains to fix the dependence of μ s , i \mu_{s,i} μ s , i , μ b , i \mu_{b,i} μ b , i , and κ i \kappa_i κ i on ( ρ 1 , ρ 2 , T ) (\rho_1,\rho_2,T) ( ρ 1 , ρ 2 , T ) . In application to air – which can be treated as a dilute gas – one can assume these parameter to depend only on T T T (as predicted by the kinetic theory of dilute gases), i.e.,
μ s , 2 ( ρ 2 , T ) = μ s , 2 ( 0 , T ) , μ b , 2 ( ρ 2 , T ) = μ b , 2 ( 0 , T ) , κ 2 ( ρ 2 , T ) = κ 2 ( 0 , T ) , (98) \mu_{s,2}(\rho_2,T)=\mu_{s,2}(0,T), \qquad\mu_{b,2}(\rho_2,T)=\mu_{b,2}(0,T), \qquad\kappa_2(\rho_2,T)=\kappa_2(0,T), \tag*{(98)} μ s , 2 ( ρ 2 , T ) = μ s , 2 ( 0 , T ) , μ b , 2 ( ρ 2 , T ) = μ b , 2 ( 0 , T ) , κ 2 ( ρ 2 , T ) = κ 2 ( 0 , T ) , (98)
where μ s , 2 ( 0 , T ) \mu_{s,2}(0,T) μ s , 2 ( 0 , T ) , μ b , 2 ( 0 , T ) \mu_{b,2}(0,T) μ b , 2 ( 0 , T ) , and κ 2 ( 0 , T ) \kappa_2(0,T) κ 2 ( 0 , T ) are the small-density limiting values of the corresponding parameters.
For water, whose liquid phase cannot be treated as a dilute gas, such an approximation is inapplicable. Aiming again for simplicity, one can approximate both viscosities by a parabola passing through two reference points – the zero-density limit and the saturated liquid state,
μ s , 1 ( ρ 1 , T ) = μ s , 1 ( 0 , T ) + [ μ s , 1 ( ρ 1 ( l ) , T ) − μ s , 1 ( 0 , T ) ] [ ρ 1 ρ 1 ( l ) ( T ) ] 2 , (99) \mu_{s,1}(\rho_1,T)=\mu_{s,1}(0,T)+\left[\mu_{s,1}\left(\rho_1^{(l)},T\right)-\mu_{s,1}(0,T)\right]\left[\frac{\rho_1}{\rho_1^{(l)}(T)}\right]^2, \tag*{(99)} μ s , 1 ( ρ 1 , T ) = μ s , 1 ( 0 , T ) + [ μ s , 1 ( ρ 1 ( l ) , T ) − μ s , 1 ( 0 , T ) ] [ ρ 1 ( l ) ( T ) ρ 1 ] 2 , (99)
μ b , 1 ( ρ 1 , T ) = μ b , 1 ( 0 , T ) + [ μ b , 1 ( ρ 1 ( l ) , T ) − μ b , 1 ( 0 , T ) ] [ ρ 1 ρ 1 ( l ) ( T ) ] 2 . (100) \mu_{b,1}(\rho_1,T)=\mu_{b,1}(0,T)+\left[\mu_{b,1}\left(\rho_1^{(l)},T\right)-\mu_{b,1}(0,T)\right]\left[\frac{\rho_1}{\rho_1^{(l)}(T)}\right]^2. \tag*{(100)} μ b , 1 ( ρ 1 , T ) = μ b , 1 ( 0 , T ) + [ μ b , 1 ( ρ 1 ( l ) , T ) − μ b , 1 ( 0 , T ) ] [ ρ 1 ( l ) ( T ) ρ 1 ] 2 . (100)
A similar approximation, but by a linear function, will be adopted for the thermal conductivity,
κ 1 ( ρ 1 , T ) = κ 1 ( 0 , T ) + [ κ 1 ( ρ 1 ( l ) , T ) − κ 1 ( 0 , T ) ] ρ 1 ρ 1 ( l ) ( T ) . (101) \kappa_1(\rho_1,T)=\kappa_1(0,T)+\left[\kappa_1\left(\rho_1^{(l)},T\right)-\kappa_1(0,T)\right]\frac{\rho_1}{\rho_1^{(l)}(T)}. \tag*{(101)} κ 1 ( ρ 1 , T ) = κ 1 ( 0 , T ) + [ κ 1 ( ρ 1 ( l ) , T ) − κ 1 ( 0 , T ) ] ρ 1 ( l ) ( T ) ρ 1 . (101)
The accuracy of the above approximations for μ s , 1 \mu_{s,1} μ s , 1 and κ 1 \kappa_1 κ 1 is illustrated in Figure 12 .
Figure 12. The parameters of pure water at T = 327 ∘ C T = 327^{\circ}\mathrm{C} T = 32 7 ∘ C : (a) shear viscosity, (b) thermal conductivity. The empiric data [58 ] are shown in solid line and approximations (8.10) and (8.12), in dashed line. The dots mark the reference points. The gap in the empiric data reflects the impossibility (difficulty) of measurements in an unstable (metastable) fluid.
To use formulae (8.6)–(8.12), one needs to know how the viscosities and thermal conductivities of water and air depend on T T T . For μ s \mu_s μ s and κ \kappa κ , such data are widely available [e.g. 58 , 98 ] . Measurements of μ b \mu_b μ b , on the other hand, are scarce, but can still be found in [49 ] , [23 ] , and [85 ] for liquid water, vapour water, and air, respectively.
The empiric formulae proposed in these papers will not be discussed in detail. The characteristic values they yield for normal conditions are listed in Table 4 .
Table 4. The empiric viscosity and thermal conductivity of liquid water, water vapour, and air at 25 ∘ C 25^{\circ}\mathrm{C} 2 5 ∘ C , and the corresponding references.
The transport coefficients
Recalling restrictions (25) , one can deduce that the extended transport matrix for a two-component fluid is
D i j ( e x t ) = [ D 11 − D 11 ζ 1 − D 11 D 11 − ζ 1 ζ 1 − ζ 1 κ ] D_{ij}^{(ext)} =
\begin{bmatrix}
D_{11} & -D_{11} & \zeta_1 \\
-D_{11} & D_{11} & -\zeta_1 \\
\zeta_1 & -\zeta_1 & \kappa
\end{bmatrix} D ij ( e x t ) = D 11 − D 11 ζ 1 − D 11 D 11 − ζ 1 ζ 1 − ζ 1 κ
Evidently, it involves only three independent coefficients, one of which (the thermal conductivity κ \kappa κ ) has already been discussed in subsection 8.5. The other two, D 11 D_{11} D 11 and ζ 1 \zeta_1 ζ 1 , are discussed below.
It can be safely assumed that diffusion is of importance only where the water density is
comparable to that of air. In the region where the former is high, the latter is miniscule, and so is its influence on the global dynamics. This effectively means that the diffusivities can be represented using the Enskog–Chapman method: even though it does not apply to liquid water, the error due to using it anyway is negligible.
According to the leading-order Enskog–Chapman formula (e.g., [34 ] ),
D 11 = ρ 1 ρ 2 ( ρ 1 m 2 + ρ 2 m 1 ) ( ρ 1 + ρ 2 ) 2 k B T D ( T ) . (102) D_{11}=\frac{\rho_1\rho_2\left(\rho_1m_2+\rho_2m_1\right)}{(\rho_1+\rho_2)^2 k_{\mathrm{B}}T}\mathcal{D}(T). \tag*{(102)} D 11 = ( ρ 1 + ρ 2 ) 2 k B T ρ 1 ρ 2 ( ρ 1 m 2 + ρ 2 m 1 ) D ( T ) . (102)
where D ( T ) \mathcal{D}(T) D ( T ) does not depend on ρ 1 \rho_1 ρ 1 and ρ 2 \rho_2 ρ 2 . Its dependence on T T T can be found in [32 ] , from which the following characteristic value can be deduced,
D ( 25 ∘ C ) = 2.49 × 10 − 5 m 2 s − 1 . (103) \mathcal{D}(25^\circ\mathrm{C})=2.49\times10^{-5}\,\mathrm{m^2\,s^{-1}}. \tag*{(103)} D ( 2 5 ∘ C ) = 2.49 × 1 0 − 5 m 2 s − 1 . (103)
The Enskog–Chapman expression for the thermodiffusivity is fairly bulky – as a result, one often uses the simpler formula of [46 ] ,
ζ 1 = ( ρ 1 + ρ 2 ) ( ρ 1 / m 1 ) ( ρ 2 / m 2 ) ( ρ 1 / m 1 + ρ 2 / m 2 ) 2 U ( T ) . (104) \zeta_1=(\rho_1+\rho_2)\frac{(\rho_1/m_1)(\rho_2/m_2)}{(\rho_1/m_1+\rho_2/m_2)^2}\mathcal{U}(T). \tag*{(104)} ζ 1 = ( ρ 1 + ρ 2 ) ( ρ 1 / m 1 + ρ 2 / m 2 ) 2 ( ρ 1 / m 1 ) ( ρ 2 / m 2 ) U ( T ) . (104)
Unfortunately, no data on U ( T ) \mathcal{U}(T) U ( T ) for water–air mixture are available in the literature; the author of this paper was able to find only an estimate for a single temperature value ([56 ] ),
U ( 21 ∘ C ) = − 4.98 × 10 − 6 m 2 s − 1 . (105) \mathcal{U}(21^\circ\mathrm{C})=-4.98\times10^{-6}\,\mathrm{m^2\,s^{-1}}. \tag*{(105)} U ( 2 1 ∘ C ) = − 4.98 × 1 0 − 6 m 2 s − 1 . (105)
It is shown later, however, that thermodiffusion in water-air interfaces is negligible, so the lack of data on U ( T ) \mathcal{U}(T) U ( T ) is unimportant.
In the next subsection, characteristic values of the coefficients D 11 D_{11} D 11 and ζ 1 \zeta_1 ζ 1 will be needed. These will be estimated for the particular case ρ 1 = ρ 2 = 1.17 k g m − 3 \rho_1=\rho_2=1.17\,\mathrm{kg\,m^{-3}} ρ 1 = ρ 2 = 1.17 kg m − 3 – i.e., when the water density matches that of air at normal conditions. Substituting these values and the molecular masses of water and air from Table 2 into formulae (8.13)–(8.16), one obtains
D 11 ≈ 1.37 × 10 − 10 m 3 s k g − 1 , ζ 1 ≈ − 2.76 × 10 − 6 m − 1 s − 1 k g . (106) D_{11}\approx1.37\times10^{-10}\,\mathrm{m^3\,s\,kg^{-1}},\qquad\zeta_1\approx-2.76\times10^{-6}\,\mathrm{m^{-1}\,s^{-1}\,kg}. \tag*{(106)} D 11 ≈ 1.37 × 1 0 − 10 m 3 s k g − 1 , ζ 1 ≈ − 2.76 × 1 0 − 6 m − 1 s − 1 kg . (106)
Even though these values apply to different temperatures (T = 25 ∘ C T=25^\circ\mathrm{C} T = 2 5 ∘ C and T = 21 ∘ C T=21^\circ\mathrm{C} T = 2 1 ∘ C ), they will be used simultaneously in the same qualitative estimate (under the implied assumption that the four-degree difference does not matter).
The nondimensional parameters
To understand which effects are important for water/air interfaces under normal conditions, one should estimate the nondimensional parameters (55) . They will be calculated using the characteristics of water from Table 2 –Table 4 and estimates (106) . The characteristic pressure scale will be assumed to be P = a ρ 2 P=a\rho^2 P = a ρ 2 where ρ = 997 k g m − 3 \rho=997\,\mathrm{kg\,m^{-3}} ρ = 997 kg m − 3 is the density of liquid water at 25 ∘ C 25^\circ\mathrm{C} 2 5 ∘ C and a a a equals the van der Waals parameter of water from Table 2 . The following expression for the viscosity scale will be used:
μ = 4 3 μ s + μ b , \mu=\frac{4}{3}\mu_s+\mu_b, μ = 3 4 μ s + μ b ,
which arises naturally in problems involving liquid films ([5 , 8 ] ) and drops ([9 ] ). In the estimates for this paper, the viscosities μ s \mu_s μ s and μ b \mu_b μ b are those for water at 25 ∘ C 25^\circ\mathrm{C} 2 5 ∘ C .
The following values of parameters (55) have been obtained:
α ≈ 0.00139 , τ ≈ 0.0653 , β ≈ 0.0593 , ν ≈ 0.0321 , δ ≈ 0.0176. (107) \alpha\approx0.00139,\qquad\tau\approx0.0653,\qquad\beta\approx0.0593,\qquad\nu\approx0.0321,\qquad\delta\approx0.0176. \tag*{(107)} α ≈ 0.00139 , τ ≈ 0.0653 , β ≈ 0.0593 , ν ≈ 0.0321 , δ ≈ 0.0176. (107)
The smallness of the microscopic Reynolds number α \alpha α suggests that inertia plays little role in the dynamics of water/air interfaces (hence, one can use the Stokes approximation). The smallness of the nondimensional temperature τ \tau τ does not have physical implications, but enables one to use asymptotic tools when calculating, say, the profile of the equilibrium interface (Benilov [5 ] ). The smallness of the Nusselt number ν \nu ν implies that the Soret and Dufour effects are negligible, and so the lack of empiric data on them for water and air is unimportant. The smallness of β \beta β , in turn, implies that the flow is almost isothermal, whereas the small value of δ \delta δ suggests that diffusion dominates advection.
Similar estimates have also been carried for the parameters of air at p = 1 p = 1 p = 1 atm and T = 25 T = 25 T = 25 °C. It has turned out that α \alpha α , β \beta β , ν \nu ν , and δ \delta δ are even smaller than those for water, but the nondimensional temperature τ \tau τ for air is order-one. The latter is not surprising, as the room temperature is much higher than the freezing temperatures of nitrogen and oxygen, but is close to that for water.
All these observations should be helpful when using the DIM to examine asymptotically the problems of evaporation of drops and moving contact lines.
One should keep in mind, however, that parameters (55) are sensitive to the choice of the viscosity scale μ \mu μ . Benilov [3 ] , for example, chose μ \mu μ equal the half-sum of the shear viscosities of liquid water and vapour water – as a result, α \alpha α and β \beta β were noticeably larger than those calculated above. More generally, one should choose the viscosity scale, as well as the other characteristic scales, to reflect the essentially physics of the setting at hand.
Note also that estimates (107) apply to water under normal conditions – but in industrial applications, the governing nondimensional parameters can be very different. In steam turbines, for example, the temperature can be as high as 540°C, and the pressure can exceed 230 atm ([95 ] ).
Concluding remarks
The results obtained in this paper can be briefly summarised as follows:
(i) The entropy principle and conservation laws of the multicomponent diffuse-interface model have been used to examine the stability of liquid/vapour interfaces. Several physically important results are reported:
For certain values of the near-substrate density (which is an external parameter in DIM, linked to the contact angle of the substrate), no steady solution exists describing a solid/vapour interfaces. Physically, such substrates are too hydrophilic and, thus, trigger off spontaneous condensation of adjacent vapour. Similarly, some substrates are too hydrophobic and trigger off spontaneous evaporation of adjacent liquid.
A liquid layer between a flat substrate and a semi-space filled with vapour can exist as a steady state only if the vapour is oversaturated . All such layers are unstable, however: depending on the perturbation, they either fully evaporate or grow indefinitely due to condensation of vapour on its upper boundary. If the vapour above the layer is saturated or undersaturated , the liquid evaporates and no steady solution exists.
Similar conclusions apply to 1D vapour layers between a flat substrate and a semi-space filled with liquid.
Similar conclusions to those in the previous bullet apply to a liquid layer with vapour both above and below it, and a vapour layer with liquid below and above it. Such solutions can be viewed as 1D drops and bubbles, respectively.
(ii) The multicomponent DIM has been fully parameterised for water/air interfaces under normal conditions. It is shown that the Soret and Dufour effects are weak in this case, which agrees with the results of Jiang et al. [52 ] . It is also argued that the interfacial flow in this case is isothermal.
These are of importance when studying evaporation of water drops or the dynamics of their contact lines.
Declaration of interests
The author reports no conflict of interests.
Appendix A
Introduce the volumetric densities of energy U ( ρ 1 … ρ N , T ) U(\rho_1 \ldots\rho_N,T) U ( ρ 1 … ρ N , T ) , entropy S ( ρ 1 … ρ N , T ) S(\rho_1 \ldots\rho_N,T) S ( ρ 1 … ρ N , T ) , and partial chemical potentials G i ( ρ 1 … ρ N , T ) \mathcal{G}_i(\rho_1 \ldots\rho_N,T) G i ( ρ 1 … ρ N , T ) – related to the corresponding specific quantities by
U = e ρ , S = s ρ , G i = G i . (108) U=e\rho,\qquad S=s\rho,\qquad\mathcal{G}_i=G_i. \tag*{(108)} U = e ρ , S = s ρ , G i = G i . (108)
Now, definition (3) of G i \mathcal{G}_i G i can be rewritten in the form
G i = ∂ U ∂ ρ i − T ∂ S ∂ ρ i . (109) \mathcal{G}_i=\frac{\partial U}{\partial\rho_i}-T\frac{\partial S}{\partial\rho_i}. \tag*{(109)} G i = ∂ ρ i ∂ U − T ∂ ρ i ∂ S . (109)
The standard volumetric version of the Gibbs relation (e.g., Eq. (2.4) of Giovangigli & Matuszewski [43 ] ) is
d U = T d S + ∑ i G i d ρ i . \mathrm{d}U=T\,\mathrm{d}S+\sum_i \mathcal{G}_i\,\mathrm{d}\rho_i. d U = T d S + i ∑ G i d ρ i .
Substituting (A 1)–(A 2) into this equality and rewriting it in terms of d T \mathrm{d}T d T and d ρ i \mathrm{d}\rho_i d ρ i (instead of d S \mathrm{d}S d S and d ρ i \mathrm{d}\rho_i d ρ i ), one obtains (2.4) as required.
Appendix B
Consider the following generalization of boundary condition (2.35):
∑ j C i j ( n ⋅ ∇ ρ j ) + ρ i = ρ 0 , i at r ∈ ∂ D . (110) \sum_j C_{ij}\left(\mathbf{n}\cdot\nabla\rho_j\right)+\rho_i=\rho_{0,i}
\qquad\text{at}\qquad\mathbf{r}\in\partial D. \tag*{(110)} j ∑ C ij ( n ⋅ ∇ ρ j ) + ρ i = ρ 0 , i at r ∈ ∂ D . (110)
where C i j C_{ij} C ij is a symmetric matrix, depending generally on ρ 1 … ρ N \rho_1 \ldots\rho_N ρ 1 … ρ N . (B 1) is a multicomponent extension of a boundary condition often used for pure fluids (e.g. Madruga & Thiele [63 ] ; Gallo et al. [37 ] ).
To understand how the energy conservation law is affected by switching to a different boundary condition, observe that the governing equations (17) –(22) and the other boundary conditions, (2.32)–(2.34), imply that
d d t ∫ D [ ρ e + 1 2 ρ ∣ v ∣ 2 + 1 2 ∑ i j K i j ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r + ∫ ∂ D ∑ i ( n ⋅ ∇ ρ i ) ∑ j K i j ∂ ρ j ∂ t d A = 0 , (111) \frac{\mathrm{d}}{\mathrm{d}t}\int_{\mathcal{D}}\left[\rho e+\frac{1}{2}\rho|\mathbf{v}|^2+\frac{1}{2}\sum_{ij}K_{ij}(\nabla\rho_i)\cdot(\nabla\rho_j)\right]\mathrm{d}^3r+\int_{\partial\mathcal{D}}\sum_i(\mathbf{n}\cdot\nabla\rho_i)\sum_jK_{ij}\frac{\partial\rho_j}{\partial t}\mathrm{d}A=0, \tag*{(111)} d t d ∫ D [ ρ e + 2 1 ρ ∣ v ∣ 2 + 2 1 ij ∑ K ij ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r + ∫ ∂ D i ∑ ( n ⋅ ∇ ρ i ) j ∑ K ij ∂ t ∂ ρ j d A = 0 , (111)
where dA is the elemental aria on ∂ D \partial\mathcal{D} ∂ D . The ‘old’ boundary condition (2.35) implies that the second integral in this equality vanishes, making the integral in the first term be a conserved quantity (the energy) in this case.
Next, assume that the ‘new’ boundary condition (B1) is imposed. Differentiate it with respect to t t t , change the indices – first j → k j\to k j → k , then i → j i\to j i → j – and use the resulting equality to rearrange (B2) in the form
d d t ∫ D [ ρ e + 1 2 ρ ∣ v ∣ 2 + 1 2 ∑ i j K i j ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r − ∫ ∂ D ∑ i j K i j ′ ( n ⋅ ∇ ρ i ) ∂ ( n ⋅ ∇ ρ j ) ∂ t d A = 0 , (112) \frac{\mathrm{d}}{\mathrm{d}t}\int_{\mathcal{D}}\left[\rho e+\frac{1}{2}\rho|\mathbf{v}|^2+\frac{1}{2}\sum_{ij}K_{ij}(\nabla\rho_i)\cdot(\nabla\rho_j)\right]\mathrm{d}^3r-\int_{\partial\mathcal{D}}\sum_{ij}K'_{ij}(\mathbf{n}\cdot\nabla\rho_i)\frac{\partial(\mathbf{n}\cdot\nabla\rho_j)}{\partial t}\mathrm{d}A=0, \tag*{(112)} d t d ∫ D [ ρ e + 2 1 ρ ∣ v ∣ 2 + 2 1 ij ∑ K ij ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r − ∫ ∂ D ij ∑ K ij ′ ( n ⋅ ∇ ρ i ) ∂ t ∂ ( n ⋅ ∇ ρ j ) d A = 0 , (112)
where
K i j ′ = ∑ k K i k C k j . K'_{ij}=\sum_k K_{ik}C_{kj}. K ij ′ = k ∑ K ik C k j .
Equation (B3) implies that the following quantity is conserved:
E = ∫ D [ ρ e + 1 2 ρ ∣ v ∣ 2 + 1 2 ∑ i j K i j ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r − 1 2 ∫ ∂ D ∑ i j K i j ′ ( n ⋅ ∇ ρ i ) ( n ⋅ ∇ ρ j ) d A . E=\int_{\mathcal{D}}\left[\rho e+\frac{1}{2}\rho|\mathbf{v}|^2+\frac{1}{2}\sum_{ij}K_{ij}(\nabla\rho_i)\cdot(\nabla\rho_j)\right]\mathrm{d}^3r-\frac{1}{2}\int_{\partial\mathcal{D}}\sum_{ij}K'_{ij}(\mathbf{n}\cdot\nabla\rho_i)(\mathbf{n}\cdot\nabla\rho_j)\mathrm{d}A. E = ∫ D [ ρ e + 2 1 ρ ∣ v ∣ 2 + 2 1 ij ∑ K ij ( ∇ ρ i ) ⋅ ( ∇ ρ j ) ] d 3 r − 2 1 ∫ ∂ D ij ∑ K ij ′ ( n ⋅ ∇ ρ i ) ( n ⋅ ∇ ρ j ) d A .
The second term in this expression is the surface energy corresponding to the new boundary condition (B1).
By comparison to the Dirichlet boundary condition (2.35), the mixed condition (B1) includes additional N ( N − 1 ) / 2 N(N-1)/2 N ( N − 1 ) /2 adjustable parameters. Even though the extra parameters may enable the DIM to be more accurate quantitatively, one should think that a physics-based model does not need many adjustable parameters to capture the qualitative nature of the phenomenon it is applied to.
Appendix C
In this appendix, the stability of three families of 1D solutions are examined:
(i) liquid/vapour interfaces,
(ii) 1D drops and bubbles,
(iii) solid/fluid interfaces.
Liquid/vapour interfaces
As mentioned before, equation (45) describes all steady states in the problem at hand, and expression (47) describes the corresponding second variation of the entropy. To adapt the latter for flat liquid/vapour interfaces, one needs to replace the 3D integral over the domain D D D with a 1D integral over the domain − ∞ < z < ∞ -\infty< z < \infty − ∞ < z < ∞ ,
δ 2 H = 1 T ∫ − ∞ ∞ ∑ i j [ − ∂ G i ∂ ρ j ( δ ρ i ) ( δ ρ j ) − K i j d ( δ ρ i ) d z d ( δ ρ j ) d z ] d z . \delta^2 H = \frac{1}{T} \int_{-\infty}^{\infty} \sum_{ij} \left[-\frac{\partial G_i}{\partial\rho_j}(\delta\rho_i)(\delta\rho_j) - K_{ij}\frac{\mathrm{d}(\delta\rho_i)}{\mathrm{d}z}\frac{\mathrm{d}(\delta\rho_j)}{\mathrm{d}z}\right]\mathrm{d}z. δ 2 H = T 1 ∫ − ∞ ∞ ij ∑ [ − ∂ ρ j ∂ G i ( δ ρ i ) ( δ ρ j ) − K ij d z d ( δ ρ i ) d z d ( δ ρ j ) ] d z .
This expression can be rewritten in the form
δ 2 H = 1 T ∫ − ∞ ∞ ∑ i j δ ρ i O ^ i j δ ρ j d z , (113) \delta^2 H = \frac{1}{T}\int_{-\infty}^{\infty}\sum_{ij}\delta\rho_i\widehat{O}_{ij}\delta\rho_j\,\mathrm{d}z, \tag*{(113)} δ 2 H = T 1 ∫ − ∞ ∞ ij ∑ δ ρ i O ij δ ρ j d z , (113)
where
O ^ i j = − ∂ G i ∂ ρ j + K i j d 2 d z 2 (114) \widehat{O}_{ij}=-\frac{\partial G_i}{\partial\rho_j}+K_{ij}\frac{\mathrm{d}^2}{\mathrm{d}z^2} \tag*{(114)} O ij = − ∂ ρ j ∂ G i + K ij d z 2 d 2 (114)
is a second-order differential operator. Since the matrices ∂ G i / ∂ ρ j \partial G_i/\partial\rho_j ∂ G i / ∂ ρ j and K i j K_{ij} K ij are symmetric (see definition (3) of G i G_i G i and subsection 2.5, respectively), O ^ i j \widehat{O}_{ij} O ij is self-adjoint – hence, its spectrum is real. Note also that O ^ i j \widehat{O}_{ij} O ij depends on the steady state ρ i ( z ) \rho_i(z) ρ i ( z ) via the coefficient ∂ G i / ∂ ρ j \partial G_i/\partial\rho_j ∂ G i / ∂ ρ j .
It follows from equation (C.1) that the functional H H H has a maximum at ρ i ( z ) \rho_i(z) ρ i ( z ) if and only if the corresponding operator O ^ i j \widehat{O}_{ij} O ij is negative definite – or, equivalently, its spectrum (the set of all discrete and continuous eigenvalues) is negative.
Theorem 1. If the solution ρ i ( z ) \rho_i(z) ρ i ( z ) of boundary-value problem (6.1)–(6.3) is monotonic for all i i i , the spectrum of the corresponding operator O ^ i j \widehat{O}_{ij} O ij is negative.
Proof. Let Λ \Lambda Λ be an eigenvalue of the discrete spectrum (if any) and ψ i \psi_i ψ i , the corresponding eigenfunction,
∑ j ( − ∂ G i ∂ ρ j ψ j + K i j d 2 ψ j d z 2 ) = Λ ψ i , (115) \sum_j\left(-\frac{\partial G_i}{\partial\rho_j}\psi_j+K_{ij}\frac{\mathrm{d}^2\psi_j}{\mathrm{d}z^2}\right)=\Lambda\psi_i, \tag*{(115)} j ∑ ( − ∂ ρ j ∂ G i ψ j + K ij d z 2 d 2 ψ j ) = Λ ψ i , (115)
ψ i → 0 as z → ± ∞ . (116) \psi_i\to0\qquad\text{as}\qquad z\to\pm\infty. \tag*{(116)} ψ i → 0 as z → ± ∞. (116)
Introduce
ϕ i = ( d ρ i d z ) − 1 ψ i , (117) \phi_i=\left(\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\right)^{-1}\psi_i, \tag*{(117)} ϕ i = ( d z d ρ i ) − 1 ψ i , (117)
and observe that, since ρ i ( z ) \rho_i(z) ρ i ( z ) is monotonic, d ρ i / d z \mathrm{d}\rho_i/\mathrm{d}z d ρ i / d z never vanishes and ϕ i ( z ) \phi_i(z) ϕ i ( z ) is non-singular.
Rewriting boundary-value problem (C.3)–(C.4) in terms of ϕ i \phi_i ϕ i and keeping in mind that ρ i ( z ) \rho_i(z) ρ i ( z ) satisfies equation (70) , one obtains
∑ j K i j ( d 2 ρ j d z 2 d ϕ j d z + d ρ j d z d 2 ϕ j d z 2 ) = Λ d ρ i d z ϕ i , (118) \sum_j K_{ij}\left(\frac{\mathrm{d}^2\rho_j}{\mathrm{d}z^2}\frac{\mathrm{d}\phi_j}{\mathrm{d}z}+\frac{\mathrm{d}\rho_j}{\mathrm{d}z}\frac{\mathrm{d}^2\phi_j}{\mathrm{d}z^2}\right)=\Lambda\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\phi_i, \tag*{(118)} j ∑ K ij ( d z 2 d 2 ρ j d z d ϕ j + d z d ρ j d z 2 d 2 ϕ j ) = Λ d z d ρ i ϕ i , (118)
d ρ i d z ϕ i → 0 as z → ± ∞ . (119) \frac{\mathrm{d}\rho_i}{\mathrm{d}z}\phi_i\to0\qquad\text{as}\qquad z\to\pm\infty. \tag*{(119)} d z d ρ i ϕ i → 0 as z → ± ∞. (119)
Let ( K i j ) − 1 (K_{ij})^{-1} ( K ij ) − 1 be the inverse matrix to K i j K_{ij} K ij (the latter is positive definite – hence, invertible) and rewrite equation (C 6) in the form
d d z [ ( d ρ j d z ) 2 d ϕ j d z ] = Λ ∑ i ( K i j ) − 1 d ρ j d z d ρ i d z ϕ i . \frac{\mathrm{d}}{\mathrm{d}z}\left[\left(\frac{\mathrm{d}\rho_j}{\mathrm{d}z}\right)^2\frac{\mathrm{d}\phi_j}{\mathrm{d}z}\right]=\Lambda\sum_i(K_{ij})^{-1}\frac{\mathrm{d}\rho_j}{\mathrm{d}z}\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\phi_i. d z d [ ( d z d ρ j ) 2 d z d ϕ j ] = Λ i ∑ ( K ij ) − 1 d z d ρ j d z d ρ i ϕ i .
Multiplying this equation by ϕ j \phi_j ϕ j , summing it with respect to j j j , integrating from z = 0 z=0 z = 0 to z = ∞ z=\infty z = ∞ , integrating the left-hand side by parts, and recalling boundary conditions (C 7), one obtains
− ∫ − ∞ ∞ ∑ j ( d ρ j d z ) 2 ( d ϕ j d z ) 2 d z = Λ ∫ − ∞ ∞ ∑ i j ( K i j ) − 1 ( ϕ i d ρ i d z ) ( ϕ j d ρ j d z ) d z . (120) -\int_{-\infty}^{\infty}\sum_j\left(\frac{\mathrm{d}\rho_j}{\mathrm{d}z}\right)^2\left(\frac{\mathrm{d}\phi_j}{\mathrm{d}z}\right)^2\mathrm{d}z=\Lambda\int_{-\infty}^{\infty}\sum_{ij}(K_{ij})^{-1}\left(\phi_i\frac{\mathrm{d}\rho_i}{\mathrm{d}z}\right)\left(\phi_j\frac{\mathrm{d}\rho_j}{\mathrm{d}z}\right)\mathrm{d}z. \tag*{(120)} − ∫ − ∞ ∞ j ∑ ( d z d ρ j ) 2 ( d z d ϕ j ) 2 d z = Λ ∫ − ∞ ∞ ij ∑ ( K ij ) − 1 ( ϕ i d z d ρ i ) ( ϕ j d z d ρ j ) d z . (120)
It can be shown that the integrands of both integrals in equation (C 8) decay exponentially as z → + ∞ z\to+\infty z → + ∞ – hence, the integrals converge. Keeping also in mind that K i j K_{ij} K ij is positive definite (hence, ( K i j ) − 1 (K_{ij})^{-1} ( K ij ) − 1 is such), one concludes that equality (C 8) implies that Λ < 0 \Lambda<0 Λ < 0 , as required.
Next, let Λ \Lambda Λ be an eigenvalue of the continuous spectrum and ψ ( z ) \psi(z) ψ ( z ) , the corresponding eigenfunction. They satisfy equation (C 3) and the boundary conditions
ψ i ∼ A i ± e i k ± z as z → ± ∞ , (121) \psi_i\sim A_{i\pm}e^{\mathrm{i}k_{\pm}z}\qquad\text{as}\qquad z\to\pm\infty, \tag*{(121)} ψ i ∼ A i ± e i k ± z as z → ± ∞ , (121)
where A i ± A_{i\pm} A i ± and k ± k_{\pm} k ± are undetermined constants (the latter is real). Substituting asymptotics (C9) into equation (C3), one obtains
∑ j [ − ( ∂ G i ∂ ρ j ) ρ 1 = ρ 1 ( l ) … ρ N = ρ N ( l ) − k − 2 K i j ] A j − = Λ A i − , \sum_j\left[-\left(\frac{\partial G_i}{\partial\rho_j}\right)_{\rho_1=\rho_1^{(l)}\ldots\rho_N=\rho_N^{(l)}}-k_-^2K_{ij}\right]A_{j-}=\Lambda A_{i-}, j ∑ [ − ( ∂ ρ j ∂ G i ) ρ 1 = ρ 1 ( l ) … ρ N = ρ N ( l ) − k − 2 K ij ] A j − = Λ A i − ,
∑ j [ − ( ∂ G i ∂ ρ j ) ρ 1 = ρ 1 ( v ) … ρ N = ρ N ( v ) − k + 2 K i j ] A j + = Λ A i + . \sum_j\left[-\left(\frac{\partial G_i}{\partial\rho_j}\right)_{\rho_1=\rho_1^{(v)}\ldots\rho_N=\rho_N^{(v)}}-k_+^2K_{ij}\right]A_{j+}=\Lambda A_{i+}. j ∑ [ − ( ∂ ρ j ∂ G i ) ρ 1 = ρ 1 ( v ) … ρ N = ρ N ( v ) − k + 2 K ij ] A j + = Λ A i + .
Evidently, Λ \Lambda Λ is a common eigenvalue of the matrices
− G i j − ′ − k − 2 K i j and − G i j + ′ − k + 2 K i j -G'_{ij-}-k_-^2K_{ij}\qquad\text{and}\qquad-G'_{ij+}-k_+^2K_{ij} − G ij − ′ − k − 2 K ij and − G ij + ′ − k + 2 K ij
where
G i j − ′ = ( ∂ G i ∂ ρ j ) ρ 1 = ρ 1 ( l ) … ρ N = ρ N ( l ) , G i j + ′ = ( ∂ G i ∂ ρ j ) ρ 1 = ρ 1 ( v ) … ρ N = ρ N ( v ) . G'_{ij-}=\left(\frac{\partial G_i}{\partial\rho_j}\right)_{\rho_1=\rho_1^{(l)}\ldots\rho_N=\rho_N^{(l)}},\qquad G'_{ij+}=\left(\frac{\partial G_i}{\partial\rho_j}\right)_{\rho_1=\rho_1^{(v)}\ldots\rho_N=\rho_N^{(v)}}. G ij − ′ = ( ∂ ρ j ∂ G i ) ρ 1 = ρ 1 ( l ) … ρ N = ρ N ( l ) , G ij + ′ = ( ∂ ρ j ∂ G i ) ρ 1 = ρ 1 ( v ) … ρ N = ρ N ( v ) .
Recall that, at infinity, the liquid and vapour are supposed to be stable, which implies that G i j ± ′ G'_{ij\pm} G ij ± ′ are both positive definite. The Korteweg matrix K i j K_{ij} K ij is also positive definite – hence, all eigenvalues of − G i j ± ′ − k ± 2 K i j -G'_{ij\pm}-k_{\pm}^2K_{ij} − G ij ± ′ − k ± 2 K ij are negative, as required.
1D drops and bubbles
The entropy maximization problem in this case can again be reduced to the analysis of the operator O ^ i j \hat{O}_{ij} O ^ ij . Furthermore, the second part of the proof of Theorem 1 can be applied to 1D drops and bubbles without modifications, and so the continuous spectrum of O ^ i j \hat{O}_{ij} O ^ ij is, again, negative.
The discrete -spectrum part of Theorem 1, however, cannot be generalised for non-monotonic ρ i ( z ) \rho_i(z) ρ i ( z ) . Indeed, if d ρ i / d z \mathrm{d}\rho_i/\mathrm{d}z d ρ i / d z vanishes somewhere, the function ϕ i ( z ) \phi_i(z) ϕ i ( z ) defined by (C 5) is singular, and the integral on the left-hand side of (C 8) diverges. Thus, for 1D drops and bubbles, O ^ i j \hat{O}_{ij} O ^ ij may have positive discrete eigenvalues, but proving that it definitely does is not easy. In what follows, such a proof is presented only for N = 1 N=1 N = 1 .
Theorem 2. The operator O ^ 11 \hat{O}_{11} O ^ 11 with ρ 1 ( z ) \rho_1(z) ρ 1 ( z ) satisfying boundary-value problem (74) –(75) with N = 1 N=1 N = 1 , has at least one positive discrete eigenvalue (corresponding to an even eigenfunction).
Proof. For N = 1 N=1 N = 1 , boundary-value problem (74) –(75) becomes
K 11 d 2 ρ 1 d z 2 = G 1 − G ∞ , 1 , (122) K_{11}\frac{\mathrm{d}^2\rho_1}{\mathrm{d}z^2}=G_1-G_{\infty,1}, \tag*{(122)} K 11 d z 2 d 2 ρ 1 = G 1 − G ∞ , 1 , (122)
ρ 1 → ρ ∞ , 1 as z → ± ∞ . (123) \rho_1\to\rho_{\infty,1}\quad\text{as}\quad z\to\pm\infty. \tag*{(123)} ρ 1 → ρ ∞ , 1 as z → ± ∞. (123)
It can be readily shown that
ρ 1 ( z ) ∼ ρ ∞ , 1 + A e G ′ z as z → − ∞ , (124) \rho_1(z)\sim\rho_{\infty,1}+A e^{\sqrt{G'}z}\quad\text{as}\quad z\to-\infty, \tag*{(124)} ρ 1 ( z ) ∼ ρ ∞ , 1 + A e G ′ z as z → − ∞ , (124)
where G ′ = ( ∂ G 11 / ∂ ρ 1 ) ρ 1 = ρ ∞ , 1 G'=\left(\partial G_{11}/\partial\rho_1\right)_{\rho_1=\rho_{\infty,1}} G ′ = ( ∂ G 11 / ∂ ρ 1 ) ρ 1 = ρ ∞ , 1 and A A A is a real constant (which can be expressed via a certain integral, but its exact value is unimportant). Note that A A A is positive for drops [i.e., solutions of the kind illustrated in Figure 4 (a)] and negative for bubbles [solutions illustrated in Figure 4 (b)].
Next, eigenvalue problem (C 3)–(C 4) with N = 1 N=1 N = 1 has the form
− ∂ G 1 ∂ ρ 1 ψ 1 + K 11 d 2 ψ 1 d z 2 = Λ ψ 1 , (125) -\frac{\partial G_1}{\partial\rho_1}\psi_1+K_{11}\frac{\mathrm{d}^2\psi_1}{\mathrm{d}z^2}=\Lambda\psi_1, \tag*{(125)} − ∂ ρ 1 ∂ G 1 ψ 1 + K 11 d z 2 d 2 ψ 1 = Λ ψ 1 , (125)
ψ 1 → 0 as z → − ∞ , (126) \psi_1\to0\quad\text{as}\quad z\to-\infty, \tag*{(126)} ψ 1 → 0 as z → − ∞ , (126)
ψ 1 → 0 as z → + ∞ . (127) \psi_1\to0\quad\text{as}\quad z\to+\infty. \tag*{(127)} ψ 1 → 0 as z → + ∞. (127)
Note that, for 1D drops an bubbles, ρ 1 ( z ) \rho_1(z) ρ 1 ( z ) is even – hence, so is ∂ G 1 / ∂ ρ 1 \partial G_1/\partial\rho_1 ∂ G 1 / ∂ ρ 1 which appears in equation (C 13). Since all other coefficients in this (linear second-order) equation are constants, one concludes that ψ 1 \psi_1 ψ 1 is either even or odd. Assuming the latter, one can reduce the interval where equation (C 13) is to be solved to ( 0 , ∞ ) (0,\infty) ( 0 , ∞ ) and replace boundary condition (C 15) with
d ψ 1 d z = 0 at z = 0. (128) \frac{\mathrm{d}\psi_1}{\mathrm{d}z}=0\quad\text{at}\quad z=0. \tag*{(128)} d z d ψ 1 = 0 at z = 0. (128)
Next, define an auxiliary function χ ( z ) \chi(z) χ ( z ) which satisfies equation (C 13) and boundary condition (C 14) – but not necessarily condition (C 16). To still have two boundary conditions, require that
χ ( z ) ∼ A G ′ + Λ e G ′ + Λ z as z → − ∞ , (129) \chi(z)\sim A\sqrt{G'}+\Lambda e^{\sqrt{G'+\Lambda}z}\quad\text{as}\quad z\to-\infty, \tag*{(129)} χ ( z ) ∼ A G ′ + Λ e G ′ + Λ z as z → − ∞ , (129)
where A A A is the constant from asymptotics (C10). Note that χ ( z ) \chi(z) χ ( z ) exists for any value Λ \Lambda Λ (depends on it as a parameter). In particular,
χ ( z ) = d ρ 1 d z for Λ = 0 , (130) \chi(z)=\frac{\mathrm{d}\rho_1}{\mathrm{d}z}\quad\text{for}\quad\Lambda=0, \tag*{(130)} χ ( z ) = d z d ρ 1 for Λ = 0 , (130)
which can be verified by comparing equation (C 13) with the derivative of equation (C 10) and asymptotics (C 17) with the derivative of asymptotics (C 12). For a large Λ \Lambda Λ , on the other hand, it can be deduced from equation (C 13) and boundary condition (C 17) that
χ ( z ) ∼ A Λ e Λ z as Λ → + ∞ . (131) \chi(z)\sim A\sqrt{\Lambda}e^{\sqrt{\Lambda}z}\quad\text{as}\quad\Lambda\to+\infty. \tag*{(131)} χ ( z ) ∼ A Λ e Λ z as Λ → + ∞. (131)
This asymptotics applies to all finite z z z .
Now, consider a drop solution [e.g., one of those illustrated in Figure 4 (a). It is evident from the figure (and common sense) that ( d 2 ρ 1 d z 2 ) z = 0 < 0 \left(\frac{\mathrm{d}^{2}\rho_{1}}{\mathrm{d}z^{2}}\right)_{z=0}<0 ( d z 2 d 2 ρ 1 ) z = 0 < 0 and A > 0 A>0 A > 0 – hence, (C18)–(C19) imply
( d χ d z ) z = 0 < 0 for Λ = 0 , \left(\frac{\mathrm{d}\chi}{\mathrm{d}z}\right)_{z=0}<0 \qquad\text{for} \qquad\Lambda=0, ( d z d χ ) z = 0 < 0 for Λ = 0 ,
( d χ d z ) z = 0 > 0 as Λ → + ∞ . \left(\frac{\mathrm{d}\chi}{\mathrm{d}z}\right)_{z=0}>0 \qquad\text{as} \qquad\Lambda\to+\infty. ( d z d χ ) z = 0 > 0 as Λ → + ∞.
These two inequalities indicate that there exists a Λ ∈ ( 0 , + ∞ ) \Lambda\in(0,+\infty) Λ ∈ ( 0 , + ∞ ) such that ( d χ / d z ) z = 0 = 0 (\mathrm{d}\chi/\mathrm{d}z)_{z=0}=0 ( d χ / d z ) z = 0 = 0 – hence, the corresponding function χ ( z ) \chi(z) χ ( z ) satisfies boundary condition (C16) [as well as condition (C14) and equation (C13)]. The corresponding (positive) Λ \Lambda Λ is, obviously, an eigenvalue, as required.
The case of bubble solutions is similar to that of drops (examined above) and yields the same conclusion.
Solid/fluid interfaces
In this case, ρ i ( z ) \rho_i(z) ρ i ( z ) satisfies equation (6.8) and boundary conditions (6.12)–(6.13).
Theorem 3. The spectrum of the operator O ^ i j \hat{O}_{ij} O ^ ij corresponding to a monotonic ρ i ( z ) \rho_i(z) ρ i ( z ) is real and negative.
The proof of this theorem is similar to that of Theorem 1. The only difference is in the boundary conditions: since the density is constrained to a fixed value at the substrate, its variation is zero – and the eigenfunctions of the operator O ^ i j \hat{O}_{ij} O ^ ij should satisfy
ψ i = 0 at z = 0. \psi_i=0 \qquad\text{at} \qquad z=0. ψ i = 0 at z = 0.
This boundary condition should hold for both discrete and continuous spectra.
Theorem 4. If N = 1 N=1 N = 1 , the operator O ^ 11 \hat{O}_{11} O ^ 11 with non-monotonic ρ 1 ( z ) \rho_1(z) ρ 1 ( z ) has at least one positive discrete eigenvalue.
The proof of this theorem is similar to that of Theorem 2
1D pure-fluid reduction of equations (4.9)–(4.10) and (4.4)
To adapt asymptotic equation (4.9)–(4.10) for a pure fluid, recall that the transport coefficients D 11 D_{11} D 11 and ζ 1 \zeta_1 ζ 1 are zero in this case (because this is the only possibility to satisfy restrictions (2.25) for N = 1 N=1 N = 1 ). Thus, one obtains
∂ ρ 1 ∂ t + ∇ ⋅ ( ρ 1 v ) = 0 , (132) \frac{\partial\rho_1}{\partial t}+\nabla\cdot(\rho_1\mathbf{v})=0, \tag*{(132)} ∂ t ∂ ρ 1 + ∇ ⋅ ( ρ 1 v ) = 0 , (132)
∇ ⋅ Π + ρ 1 ∇ ( K 11 ∇ 2 ρ 1 − G 1 ) = 0. (133) \nabla\cdot\mathbf{\Pi}+\rho_1\nabla\left(K_{11}\nabla^2\rho_1-G_1\right)=0. \tag*{(133)} ∇ ⋅ Π + ρ 1 ∇ ( K 11 ∇ 2 ρ 1 − G 1 ) = 0. (133)
Let the flow be 1D, so that the unknowns depend only on z z z and t t t , and the velocity has only one component v = [ 0 , 0 , w ] \mathbf{v}=[0,0,w] v = [ 0 , 0 , w ] . Then, equations (D1)–(D2) and expression (4.4) for the viscous stress yield
∂ ρ 1 ∂ t + ∂ ( ρ 1 w ) ∂ z = 0 , (134) \frac{\partial\rho_1}{\partial t}+\frac{\partial(\rho_1w)}{\partial z}=0, \tag*{(134)} ∂ t ∂ ρ 1 + ∂ z ∂ ( ρ 1 w ) = 0 , (134)
∂ ∂ z [ ( 4 3 μ s + μ b ) ∂ w ∂ z ] + ρ 1 ∂ ∂ z ( K 11 ∂ 2 ρ 1 ∂ z 2 − G 1 ) = 0. (135) \frac{\partial}{\partial z}\left[\left(\frac{4}{3}\mu_s+\mu_b\right)\frac{\partial w}{\partial z}\right]+\rho_1\frac{\partial}{\partial z}\left(K_{11}\frac{\partial^2\rho_1}{\partial z^2}-G_1\right)=0. \tag*{(135)} ∂ z ∂ [ ( 3 4 μ s + μ b ) ∂ z ∂ w ] + ρ 1 ∂ z ∂ ( K 11 ∂ z 2 ∂ 2 ρ 1 − G 1 ) = 0. (135)
Let the substrate be located at z = 0 z = 0 z = 0 , so that boundary conditions (32) and (35) yield
w = 0 , ρ 1 → ρ 0 , 1 at z = 0. (136) w = 0,\qquad\rho_1 \to\rho_{0,1}\quad\text{at}\quad z = 0. \tag*{(136)} w = 0 , ρ 1 → ρ 0 , 1 at z = 0. (136)
At infinity, the density tends to a fixed value and there should be no stress – hence,
ρ 1 → ρ ∞ , 1 , ∂ w ∂ z → 0 as z → + ∞ . (137) \rho_1 \to\rho_{\infty,1},\qquad\frac{\partial w}{\partial z} \to0\quad\text{as}\quad z \to+\infty. \tag*{(137)} ρ 1 → ρ ∞ , 1 , ∂ z ∂ w → 0 as z → + ∞. (137)
Using identity (6) , one can replace in equation (D4) ρ 1 ∂ G 1 / ∂ z → ∂ p / ∂ z \rho_1 \partial G_1/\partial z \to\partial p/\partial z ρ 1 ∂ G 1 / ∂ z → ∂ p / ∂ z and then integrate this equation with respect to z z z . Fixing the constant of integration via boundary conditions (D6)–(D7), one obtains
( 4 3 μ s + μ b ) ∂ w ∂ z + K 11 [ ρ 1 ∂ 2 ρ 1 ∂ z 2 − 1 2 ( ∂ ρ 1 ∂ z ) 2 ] = 0. (138) \left(\frac{4}{3}\mu_s+\mu_b\right)\frac{\partial w}{\partial z}+K_{11}\left[\rho_1\frac{\partial^2\rho_1}{\partial z^2}-\frac{1}{2}\left(\frac{\partial\rho_1}{\partial z}\right)^2\right]=0. \tag*{(138)} ( 3 4 μ s + μ b ) ∂ z ∂ w + K 11 [ ρ 1 ∂ z 2 ∂ 2 ρ 1 − 2 1 ( ∂ z ∂ ρ 1 ) 2 ] = 0. (138)
Given a suitable initial condition, equations (D3) and (D7), with boundary conditions (D5) fully determine ρ 1 ( z , t ) \rho_1(z,t) ρ 1 ( z , t ) and w ( z , t ) w(z,t) w ( z , t ) .
For numerical simulations, equations (D3) and (D7) were discretised in z z z using central differences, and the resulting set of ordinary differential equations were solved using the MATLAB function ODE23tb (which can handle stiff problems). This approach is usually referred to as the “method of lines” [83 ] .