betaFeedback

WorksPhysics and Astronomy

Mixtures of a fluid with competing interactions and a simple solvent near a hard wall

Abstract

We study a binary mixture of soft particles near a planar surface. The main component interacts through a competing short-range attraction and long-range repulsion (SALR) potential, modelled by a sum of three Yukawa terms, while the solvent and the cross interactions are described by two Yukawa terms. Using a field-theoretical mean-field approach we derive the coupled equations for the density profiles of both species, formulate the contact theorem for the two-component system, and obtain closed-form profiles in the mesoscopically homogeneous region by linearizing the theory and solving the associated eigenvalue problem. Their amplitudes are fixed by exact boundary conditions following from the continuity of the Yukawa potentials at the wall, which supply as many relations as there are unknowns. Although these conditions are implicit in the derivation of the contact theorem, they have not previously been used to fix the amplitudes of the linearized profiles, and they bring those profiles into quantitative agreement with the nonlinear solution. The equations are solved numerically for a fluid confined by hard walls. The theory predicts mixing of the two species far from the surface at high temperatures, and their microphase separation into alternating layers upon cooling or upon increasing the density. The mean-field profiles agree well with molecular dynamics data once the mean-field overestimate of the onset temperature of microsegregation is taken into account, and the analytical ones reproduce them closely except in the immediate vicinity of the wall.

Introduction

Biological and soft-matter systems are usually spatially inhomogeneous and multicomponent, since they are frequently composed of a mixture of charged macromolecules or nanoparticles immersed in a medium of much smaller solvent molecules. Despite the substantial progress achieved in the study of one-component fluids with competing interactions, mixtures of particles with competing interactions started to attract attention only recently [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]. In particular, it has been proposed that membraneless organelles, i.e., liquid-like intracellular condensates, may form by a mechanism involving fluids with short-range attraction and long-range repulsion (SALR) rather than by conventional liquid–liquid phase separation [1]. Monte Carlo simulations of a SALR model showed that giant SALR clusters can form at a model membrane surface even when they are unstable in the bulk, mimicking the puncta formed at the cell membrane by the actin-regulatory protein WASP (Wiskott–Aldrich syndrome protein) together with the adaptor protein Nck [13]. Such multicomponent biological systems motivate the extension of SALR models from one-component fluids to mixtures. At the single-component level, a SALR potential provides a coarse-grained, effective description of the interactions between complex mesoscopic particles, which are often mediated by a solvent. It is by now well established that such interactions drive the spontaneous formation of periodically modulated phases, i.e., microphase separation [14]. The resulting cluster, cylindrical, lamellar and gyroid-like morphologies, and the thermodynamic conditions under which they arise, have been studied in considerable detail by theory and simulations [15, 16, 17, 18, 19, 20, 21, 22, 23]. The subject remains very active, with recent extensions to, for instance, self-propelled (active) SALR suspensions [24]. Introducing a second, solvent-like species enlarges this picture considerably and can give access to a broader family of self-assembled morphologies depending on model parameters and thermodynamic conditions. For example, binary mixtures of oppositely charged colloids self-assemble into ionic colloidal crystals [25], while the interplay of a short-range attraction and a long-range repulsion can arrest bulk phase separation into equilibrium clusters and colloidal gels [26]. Understanding the mechanisms of clustering in such mixtures is also important from an applied point of view, as it may open additional possibilities for tailoring the properties of novel nanomaterials. Moreover, cluster formation in mixtures of fluids with competing interactions may take place in the region of the thermodynamic parameters where it is absent for the one-component fluid [6, 10]. The way in which the cross-interaction between the two species controls clustering, secondary nucleation and microsegregation has been addressed in several recent studies of binary SALR mixtures [4, 5], and a soft (penetrable) cross-interaction has been argued to be, by itself, sufficient to drive microphase separation in a binary colloidal system [27]. Investigating the features brought about by a binary mixture in comparison with the one-component case is thus of both fundamental and practical interest.

Given the large number of open questions concerning the properties of mixtures with competing interactions, the influence of a spatial confinement on such systems has been studied even less. At the same time, computer simulation data indicate the possibility of non-trivial adsorption effects for such mixtures near a planar surface, for which the density profile of the component with competing interactions differs substantially from the one-component case [9, 28]. It has been shown that the adsorption and the distribution of the particles near a wall depend essentially on the fluid density and on the amplitude of the surface-particle interaction. Recent simulations further underline the richness of the structures formed by confined systems with competing interactions. A mixture of nanoparticles with competing interactions adsorbed onto a spherical colloid was found to self-assemble into alternating layers, tori or spirals of the two components [29]. Competing-interaction colloids enclosed in spherical cavities form cluster, lamellar or perforated morphologies depending on the curvature of the confining wall [30], and confinement in ordered porous materials selects different structures depending on the pore geometry [31]. In the related two-dimensional geometry, binary mixtures of particles with competing interactions confined to a monolayer spontaneously organize into a variety of ordered patterns [10].

In this work we take as our starting point the one-component soft-core competing-interaction fluid at a hard wall, introduced in [32], and generalize it to a two-component mixture. Our main goal is to understand how a solvent species modifies the structure of a competing-interaction fluid near a hard surface. We pay particular attention to the possibility of microsegregation of the two species and to the way in which the concentrations and the temperature control it.

The theoretical description used here rests on a classical field-theoretical formalism for inhomogeneous fluids. In this formalism the partition function is represented as a functional integral over the density field, whose saddle point yields the mean-field density profiles and, if required, may be supplemented by Gaussian fluctuations. The scheme was originally devised for ionic systems and electric double layers by Di Caprio, Stanev and Badiali [33, 34, 35, 36]. It was subsequently transferred to fluids with Yukawa-type interactions confined by a hard wall to a single-Yukawa fluid [37, 38], a two-Yukawa fluid [39] and an oscillating Yukawa fluid [40], as well as to nematogenic fluids with an orientation-dependent Maier-Saupe interaction [41, 42, 43, 44] – and, more recently, to the one-component fluid with a three-Yukawa (SALR) competing interaction [32], for which closed-form density profiles were obtained. Here we extend this single-component treatment to a binary mixture confined by a hard wall. In doing so we also revisit the boundary conditions that fix the amplitudes of the linearized profiles: in place of the contact theorem imposed on each Yukawa potential separately, we use the exact conditions that follow from the continuity of the potentials at the wall. In the mean-field approximation this field-theoretical description is equivalent to the standard mean-field density functional theory [45]. The mean-field predictions are complemented by our own computer simulations. Related exact contact relations for confining surfaces of non-planar geometry were recently established [46].

The remainder of the paper is structured as follows. In section 2 we introduce the model. In section 3 we develop the field-theoretical mean-field description of the mixture, derive the equations for the density profiles, and formulate the contact theorem. Section 4 presents the numerical mean-field results for the bulk region and for a hard wall, together with a comparison with molecular dynamics. In section 5 we derive the analytical density profiles in the mesoscopically homogeneous region, bounded by the λ\lambda-surface, examine in detail the boundary conditions at the wall, and study the qualitative behaviour of the profiles. Conclusions are drawn in § 6.

Model

We consider a binary mixture of soft particles of two species. The interparticle interaction of the main component (species 1) is modelled by a competing SALR potential written as a sum of three Yukawa terms (3Y), while the interaction between the solvent particles (species 2) and the cross interaction are described by two Yukawa terms, one repulsive and one attractive (2Y):

ν11(r)=A11(1)rexp⁡(−α11(1)r)+A11(2)rexp⁡(−α11(2)r)+A11(3)rexp⁡(−α11(3)r),(1)\nu_{11}(r)=\frac{A_{11}^{(1)}}{r}\exp\left(-\alpha_{11}^{(1)}r\right)+\frac{A_{11}^{(2)}}{r}\exp\left(-\alpha_{11}^{(2)}r\right)+\frac{A_{11}^{(3)}}{r}\exp\left(-\alpha_{11}^{(3)}r\right), \tag*{(1)}
ν12(r)=A12(1)rexp⁡(−α12(1)r)+A12(2)rexp⁡(−α12(2)r),(2)\nu_{12}(r)=\frac{A_{12}^{(1)}}{r}\exp\left(-\alpha_{12}^{(1)}r\right)+\frac{A_{12}^{(2)}}{r}\exp\left(-\alpha_{12}^{(2)}r\right), \tag*{(2)}
ν22(r)=A22(1)rexp⁡(−α22(1)r)+A22(2)rexp⁡(−α22(2)r),(3)\nu_{22}(r)=\frac{A_{22}^{(1)}}{r}\exp\left(-\alpha_{22}^{(1)}r\right)+\frac{A_{22}^{(2)}}{r}\exp\left(-\alpha_{22}^{(2)}r\right), \tag*{(3)}

where rr is the distance between two particles, Aij(k)A_{ij}^{(k)} are the amplitudes, and αij(k)\alpha_{ij}^{(k)} the inverse ranges of the corresponding Yukawa terms; throughout, the subscripts i,ji,j label the two species, while the superscript (k)(k) numbers the individual Yukawa terms. For the main component the amplitudes and ranges are tuned so that species 1 realizes the competing SALR form: A11(1)>0A_{11}^{(1)}>0, A11(2)<0A_{11}^{(2)}<0, A11(3)>0A_{11}^{(3)}>0 with A11(1)>∣A11(2)∣>A11(3)A_{11}^{(1)}>|A_{11}^{(2)}|>A_{11}^{(3)} and α11(1)>α11(2)>α11(3)\alpha_{11}^{(1)}>\alpha_{11}^{(2)}>\alpha_{11}^{(3)}, i.e., a short-range attraction followed by a long-range repulsion. Since the amplitudes do not sum to zero, the potentials diverge at the origin only as ∑kAij(k)/r\sum_k A_{ij}^{(k)}/r. This weak, integrable Coulomb-like core is much softer than the steep repulsion of hard-sphere or Lennard-Jones particles, so that two particles can overlap partially. Such a soft core makes the model suitable for coarse-grained descriptions of mesoscopic particles that can interpenetrate, such as microgels, star and grafted polymers or globular proteins, and also allows efficient molecular dynamics sampling. The numerical values of the parameters used in this work are collected in Table 1. The simplified model used in the analytical treatment of § 5 is defined by equation (20).

A11(1)=150.6561472A_{11}^{(1)}=150.6561472A11(2)=−122.6130119A_{11}^{(2)}=-122.6130119A11(3)=27.11810649A_{11}^{(3)}=27.11810649
α11(1)=1.92325375\alpha_{11}^{(1)}=1.92325375α11(2)=1.261150\alpha_{11}^{(2)}=1.261150α11(3)=0.756690\alpha_{11}^{(3)}=0.756690
A12(1)=413.5305236A_{12}^{(1)}=413.5305236A12(2)=−275.6870157A_{12}^{(2)}=-275.6870157
α12(1)=3.405465108\alpha_{12}^{(1)}=3.405465108α12(2)=3.0\alpha_{12}^{(2)}=3.0
A22(1)=413.5305236A_{22}^{(1)}=413.5305236A22(2)=−275.6870157A_{22}^{(2)}=-275.6870157
α22(1)=3.405465108\alpha_{22}^{(1)}=3.405465108α22(2)=3.0\alpha_{22}^{(2)}=3.0

Table 1. Parameters of the pair potentials (2.1)–(2.3).

Throughout this work, all energy quantities are in units of the depth of the attractive well of the main component, i.e., ∣ν11(rmin⁡)∣\lvert\nu_{11}(r_{\min})\rvert, lengths are in units of the distance r0r_0 at which ν11(r0)=0\nu_{11}(r_0)=0, densities are in units of r0−3r_0^{-3}, and the reduced temperature is kBT/∣ν11(rmin⁡)∣k_{\mathrm{B}}T/\lvert\nu_{11}(r_{\min})\rvert, where kBk_{\mathrm{B}} is the Boltzmann constant.

To describe the presence of a surface, we introduce external potentials W1(z)W_1(z) and W2(z)W_2(z) acting on the particles of species 1 and 2, respectively, which describe a single hard wall or a slit-like pore bounded by two hard walls,

Wi(z)={+∞,z<0,0,z≥0,W~i(z)=Wi(z)+Wi(Lz−z),i=1,2,(4)W_i(z)= \begin{cases} +\infty, & z<0,\\ 0, & z\ge0, \end{cases} \qquad \widetilde{W}_i(z)=W_i(z)+W_i(L_z-z),\qquad i=1,2, \tag*{(4)}

where zz is the distance from the wall and LzL_z is the width of the slit-like pore (the second term describes the second wall; a single wall is recovered as Lz→∞L_z \to\infty). The walls are thus purely hard: they exert no force on either species and only exclude the particles from the region behind the wall. The pair potentials (2.1)–(2.3) are shown in figure 1.

Figure 1. The interparticle pair potentials (2.1)–(2.3) with the parameters of table 1: ν11(r)\nu_{11}(r) (competing SALR, solid red line) and ν12(r)=ν22(r)\nu_{12}(r)=\nu_{22}(r) of the full 3Y+2Y model (dashed blue line). Both vanish at r=1r=1, which serves as the soft-core particle diameter. The dash-dotted green line shows the purely repulsive ν12(r)=ν22(r)\nu_{12}(r)=\nu_{22}(r) of the simplified 3Y+1Y model (5.1) used in section 5.

For the single-component version of this model, the extensive Monte Carlo and molecular dynamics simulations reported in [32] show that the competing potential (2.1) drives a rich sequence of self-assembled mesophases: the fluid is disordered at high temperature and, on cooling, organizes into lamellar, gyroid, hexagonally packed cylindrical and cubic cluster (and void) structures, the selected morphology depending on the density and the temperature. These simulation snapshots are not reproduced here. The corresponding phase diagram and illustrations of the mesophases can be found in [32]. Spatial confinement is known to reinforce this tendency to order, the near-wall structuring surviving up to temperatures higher than those at which the bulk fluid has already become homogeneous. In what follows we build on this one-component picture and ask how it is modified once a solvent species is added.

Field-theoretical mean-field description

Hamiltonian and mean-field approximation

In the field-theoretical formulation one starts from a Hamiltonian H[ρ(r)]H[\rho(\mathbf{r})] that depends functionally on the density field ρ(r)\rho(\mathbf{r}) and separates into an entropic (ideal-gas) part and an interaction part. For a single-component fluid,

βH[ρ(r1)]=∫ρ(r1)(ln⁡[ρ(r1)Λ3]−1) dr1+β2∫ν(r12)[ρ(r1)ρ(r2)−ρ(r1)δ(r1−r2)] dr1dr2,(5)\beta H[\rho(\mathbf{r}_1)] = \int\rho(\mathbf{r}_1)\left(\ln[\rho(\mathbf{r}_1)\Lambda^3]-1\right)\,\mathrm{d}\mathbf{r}_1 + \frac{\beta}{2}\int\nu(r_{12})\left[\rho(\mathbf{r}_1)\rho(\mathbf{r}_2)-\rho(\mathbf{r}_1)\delta(\mathbf{r}_1-\mathbf{r}_2)\right]\,\mathrm{d}\mathbf{r}_1\mathrm{d}\mathbf{r}_2, \tag*{(5)}

where β=1/(kBT)\beta=1/(k_{\mathrm{B}}T) is the inverse temperature, ρ(r)\rho(\mathbf{r}) is the number density and Λ\Lambda the de Broglie wavelength. Taking into account the potentials (2.1)–(2.3), the Hamiltonian of the two-component mixture reads

βH[ρ1,ρ2]=∑i=12∫ρi(r1)(ln⁡[ρi(r1)Λ3]−1) dr1+β2∑i,j=12∫νij(r12)[ρi(r1)ρj(r2)−ρi(r1)δ(r1−r2)δij] dr1dr2,(6)\beta H[\rho_1,\rho_2] = \sum_{i=1}^{2}\int\rho_i(\mathbf{r}_1)\left(\ln[\rho_i(\mathbf{r}_1)\Lambda^3]-1\right)\,d\mathbf{r}_1 + \frac{\beta}{2}\sum_{i,j=1}^{2}\int\nu_{ij}(r_{12})\left[\rho_i(\mathbf{r}_1)\rho_j(\mathbf{r}_2)-\rho_i(\mathbf{r}_1)\delta(\mathbf{r}_1-\mathbf{r}_2)\delta_{ij}\right]\,d\mathbf{r}_1d\mathbf{r}_2, \tag*{(6)}

where ρi(r)\rho_i(\mathbf{r}) denotes the density field of species ii. The partition function and the Helmholtz free energy are given by

Z[ρ]=∫Dρexp⁡{−βH[ρ]},βF=−ln⁡Z,(7)Z[\rho] = \int\mathcal{D}\rho\exp\{-\beta H[\rho]\}, \qquad\beta F=-\ln Z, \tag*{(7)}

where the functional integration runs over all density distributions with a fixed total number of particles of each species. In the mean-field approximation (MFA) the free energy is obtained by minimizing the Hamiltonian with respect to the density fields,

δβHδρ1∣ρMF=λ1,δβHδρ2∣ρMF=λ2,(8)\left.\frac{\delta\beta H}{\delta\rho_1}\right|_{\rho_{\mathrm{MF}}}=\lambda_1,\qquad \left.\frac{\delta\beta H}{\delta\rho_2}\right|_{\rho_{\mathrm{MF}}}=\lambda_2, \tag*{(8)}

λi\lambda_i being Lagrange multipliers, so that βFMF=βHMF\beta F^{\mathrm{MF}}=\beta H^{\mathrm{MF}}.

Bulk: partial pressures

In the spatially homogeneous (bulk) case the density fields are constant, ρi(r)=ρib\rho_i(\mathbf{r})=\rho_{ib}, and the mean-field free energy becomes

βFMF=βFid+V2[ρ1b2ν^11+2ρ1bρ2bν^12+ρ2b2ν^22],ν^ij=β∫νij(r) dr=4πβ∑kAij(k)(αij(k))2,(9)\beta F^{\mathrm{MF}}=\beta F_{\mathrm{id}}+\frac{V}{2}\left[\rho_{1b}^{2}\hat{\nu}_{11}+2\rho_{1b}\rho_{2b}\hat{\nu}_{12}+\rho_{2b}^{2}\hat{\nu}_{22}\right],\qquad \hat{\nu}_{ij}=\beta\int\nu_{ij}(r)\,d\mathbf{r}=4\pi\beta\sum_k\frac{A_{ij}^{(k)}}{(\alpha_{ij}^{(k)})^2}, \tag*{(9)}

where βFid=∑iNi(ln⁡[ρibΛ3]−1)\beta F_{\mathrm{id}}=\sum_iN_i(\ln[\rho_{ib}\Lambda^3]-1) is the ideal-gas free energy, Ni=ρibVN_i=\rho_{ib}V, and VV is the volume. Introducing the screening-like parameters

χij(k)2=4πβρjbAij(k),(10)\chi_{ij}^{(k)2}=4\pi\beta\rho_{jb}A_{ij}^{(k)}, \tag*{(10)}

the bulk values of the mean-field potentials [defined in equation (3.11) below] are βVijb(k)=χij(k)2/(αij(k))2\beta V_{ijb}^{(k)}=\chi_{ij}^{(k)2}/(\alpha_{ij}^{(k)})^2. Differentiation of (3.5) yields the partial pressures,

βP1MF=ρ1b[1+12∑k=13χ11(k)2(α11(k))2+12∑k=12χ12(k)2(α12(k))2],(11)\beta P_1^{\mathrm{MF}}=\rho_{1b}\left[1+\frac{1}{2}\sum_{k=1}^{3}\frac{\chi_{11}^{(k)2}}{(\alpha_{11}^{(k)})^2}+\frac{1}{2}\sum_{k=1}^{2}\frac{\chi_{12}^{(k)2}}{(\alpha_{12}^{(k)})^2}\right], \tag*{(11)}
βP2MF=ρ2b[1+12∑k=12χ22(k)2(α22(k))2+12∑k=12χ21(k)2(α21(k))2],(12)\beta P_2^{\mathrm{MF}}=\rho_{2b}\left[1+\frac{1}{2}\sum_{k=1}^{2}\frac{\chi_{22}^{(k)2}}{(\alpha_{22}^{(k)})^2}+\frac{1}{2}\sum_{k=1}^{2}\frac{\chi_{21}^{(k)2}}{(\alpha_{21}^{(k)})^2}\right], \tag*{(12)}

with the total pressure βPMF=βP1MF+βP2MF\beta P^{\mathrm{MF}}=\beta P_1^{\mathrm{MF}}+\beta P_2^{\mathrm{MF}}. These bulk relations are used below in the contact theorem and in the comparison with the near-surface profiles.

Density profiles near a surface and the contact theorem

In the presence of a surface, the minimization conditions (3.4) lead to the following equations for the density profiles. Because the system remains uniform in the two directions parallel to the wall, every position-dependent quantity is a function of the single coordinate zz measured perpendicular to the surface:

ρ1(z)=ρ1bexp⁡[−β∑k=13(V11(k)(z)−V11b(k))−β∑k=12(V12(k)(z)−V12b(k))−βW~1(z)].(13)\rho_1(z)=\rho_{1b}\exp\left[-\beta\sum_{k=1}^{3}\left(V_{11}^{(k)}(z)-V_{11b}^{(k)}\right)-\beta\sum_{k=1}^{2}\left(V_{12}^{(k)}(z)-V_{12b}^{(k)}\right)-\beta\widetilde{W}_1(z)\right]. \tag*{(13)}
ρ2(z)=ρ2bexp⁡[−β∑k=12(V22(k)(z)−V22b(k))−β∑k=12(V21(k)(z)−V21b(k))−βW~2(z)].(14)\rho_2(z)=\rho_{2b}\exp\left[-\beta\sum_{k=1}^{2}\left(V_{22}^{(k)}(z)-V_{22b}^{(k)}\right)-\beta\sum_{k=1}^{2}\left(V_{21}^{(k)}(z)-V_{21b}^{(k)}\right)-\beta\widetilde{W}_2(z)\right]. \tag*{(14)}

where the mean-field potential created by species jj and felt by species ii through the kk-th Yukawa term, and the corresponding field, are

Vij(k)(r1)=∫ρj(r2)Aij(k)r12exp⁡(−αij(k)r12)dr2,Eij(k)(r1)=−∇Vij(k)(r1).(15)V_{ij}^{(k)}(\mathbf{r}_1)=\int\rho_j(\mathbf{r}_2)\frac{A_{ij}^{(k)}}{r_{12}}\exp\left(-\alpha_{ij}^{(k)}r_{12}\right)\mathrm{d}\mathbf{r}_2,\qquad E_{ij}^{(k)}(\mathbf{r}_1)=-\nabla V_{ij}^{(k)}(\mathbf{r}_1). \tag*{(15)}

Acting on (15) with the Laplacian and using the property of the Yukawa potential, (Δ−(αij(k))2)Vij(k)=−4πAij(k)ρj\left(\Delta-\left(\alpha_{ij}^{(k)}\right)^2\right)V_{ij}^{(k)}=-4\pi A_{ij}^{(k)}\rho_j, together with the gradient of the logarithm of equations (3.9)–(3.10), one obtains two coupled systems of first-order differential equations for the unknown functions ρi(z)\rho_i(z), Vij(k)(z)V_{ij}^{(k)}(z) and Eij(k)(z)E_{ij}^{(k)}(z):

ρ1′(z)=ρ1(z)[β∑k=13E11(k)(z)+β∑k=12E12(k)(z)−βW~1′(z)],\rho_1'(z)=\rho_1(z)\left[\beta\sum_{k=1}^{3}E_{11}^{(k)}(z)+\beta\sum_{k=1}^{2}E_{12}^{(k)}(z)-\beta\widetilde{W}_1'(z)\right],
Vij(k)′(z)=−Eij(k)(z),Eij(k)′(z)=−(αij(k))2Vij(k)(z)+χij(k)2βρjbρj(z),(16)V_{ij}^{(k)\prime}(z)=-E_{ij}^{(k)}(z),\qquad E_{ij}^{(k)\prime}(z)=-\left(\alpha_{ij}^{(k)}\right)^2V_{ij}^{(k)}(z)+\frac{\chi_{ij}^{(k)2}}{\beta\rho_{jb}}\rho_j(z), \tag*{(16)}

for the pairs ij={11,12}ij=\{11,12\}, and an analogous conjugate system for ρ2(z)\rho_2(z) with ij={22,21}ij=\{22,21\}. This set of equations is solved numerically in section 4 and analytically in section 5.

The system (16) admits a first integral, which provides an exact sum rule. For hard walls one can verify that

ddz[ρ1(z)+ρ2(z)+∑i=12∑kρib2χii(k)2{(αii(k))2(βVii(k))2−(βEii(k))2}+∑k14πβA12(k){(α12(k))2β2V12(k)V21(k)−β2E12(k)E21(k)}]=0.(17)\frac{\mathrm{d}}{\mathrm{d}z}\left[\rho_1(z)+\rho_2(z)+\sum_{i=1}^{2}\sum_k\frac{\rho_{ib}}{2\chi_{ii}^{(k)2}}\left\{\left(\alpha_{ii}^{(k)}\right)^2\left(\beta V_{ii}^{(k)}\right)^2-\left(\beta E_{ii}^{(k)}\right)^2\right\}+\sum_k\frac{1}{4\pi\beta A_{12}^{(k)}}\left\{\left(\alpha_{12}^{(k)}\right)^2\beta^2V_{12}^{(k)}V_{21}^{(k)}-\beta^2E_{12}^{(k)}E_{21}^{(k)}\right\}\right]=0. \tag*{(17)}

where 1/(4πβA12(k))=ρ1b/χ21(k)2=ρ2b/χ12(k)21/(4\pi\beta A_{12}^{(k)})=\rho_{1b}/\chi_{21}^{(k)2}=\rho_{2b}/\chi_{12}^{(k)2}.

Two features distinguish (17) from its one-component counterpart [32], and both are essential. The densities enter unscaled, and the cross fields appear through the bilinear products V12(k)V21(k)V_{12}^{(k)}V_{21}^{(k)} and E12(k)E21(k)E_{12}^{(k)}E_{21}^{(k)} rather than through separate squares. The reason is that V12(k)V_{12}^{(k)} is created by the solvent and V21(k)V_{21}^{(k)} by the main component, so that the two are different functions of zz, and their contributions cancel only in the combination above due to the identity

ρ1βE12(k)+ρ2βE21(k)=14πβA12(k)ddz[β2E12(k)E21(k)−(α12(k))2β2V12(k)V21(k)].(18)\rho_1\beta E_{12}^{(k)}+\rho_2\beta E_{21}^{(k)}=\frac{1}{4\pi\beta A_{12}^{(k)}}\frac{\mathrm{d}}{\mathrm{d}z}\left[\beta^2E_{12}^{(k)}E_{21}^{(k)}-\left(\alpha_{12}^{(k)}\right)^2\beta^2V_{12}^{(k)}V_{21}^{(k)}\right]. \tag*{(18)}

which follows from (16). Replacing the densities by ρi/ρib\rho_i/\rho_{ib} and the bilinear by separate squares, as would be natural by analogy with the one-component case, leaves a residual (∑kβE12(k)−∑kβE21(k))(ρ1/ρ1b−ρ2/ρ2b)\left(\sum_k\beta E_{12}^{(k)}-\sum_k\beta E_{21}^{(k)}\right)(\rho_1/\rho_{1b}-\rho_2/\rho_{2b}), which does not vanish across the interface.

Evaluating the conserved quantity in the bulk, where ρi(z)→ρib\rho_i(z)\to\rho_{ib}, Eij(k)→0E_{ij}^{(k)}\to0 and Vij(k)→Vijb(k)V_{ij}^{(k)}\to V_{ijb}^{(k)}, and at contact (z=0+z=0+), and using (3.7), we obtain the contact theorem for the two-component system:

βP1MF+βP2MF=ρ1(0+)+ρ2(0+).(19)\beta P_1^{\mathrm{MF}}+\beta P_2^{\mathrm{MF}}=\rho_1(0+)+\rho_2(0+). \tag*{(19)}

which generalizes the one-component relation ρ(0+)=βPMF\rho(0+) = \beta P^{\mathrm{MF}} [32]. At contact every field term in (17) vanishes, since the absence of fluid behind the wall makes βEij(k)(0)=−αij(k)βVij(k)(0)\beta E_{ij}^{(k)}(0) = -\alpha_{ij}^{(k)}\beta V_{ij}^{(k)}(0). The same conditions are used in section 5.2 to fix the amplitudes of the linearized profiles, so that the contact theorem and the boundary conditions (28) are one and the same statement. The sum rule (19) relates the total contact density to the total mean-field pressure and is satisfied exactly by the numerical solution. The species-resolved relations βPiMF=ρi(0+)\beta P_i^{\mathrm{MF}} = \rho_i(0+) do not follow from it and are not imposed in the analytical treatment of section 5: there the amplitudes are fixed by the exact conditions (28), and the contact density of each species follows from the solution.

Numerical results in the mean-field approximation

Bulk

We first solve the mean-field equations for the mixture in the bulk, i.e., without walls, by imposing periodic boundary conditions on ρ1(z)\rho_1(z) and ρ2(z)\rho_2(z) over a period LzL_z, and follow the evolution of the density profiles with temperature. At high temperatures both species are homogeneously mixed. Upon cooling, the main component spontaneously develops periodic density modulations, and below a certain temperature the mixture separates into alternating layers rich in species 1 and in species 2. Where the density of one species is close to its maximum, the density of the other is nearly zero. This sequence is qualitatively similar to the one found in simulations of the one-component fluid [32] (see § 2). The onset temperature of this microsegregation depends on the composition. For equal concentrations ρ1b=ρ2b=0.20\rho_{1b}=\rho_{2b}=0.20 the modulations appear below T≈5T\approx5 (Figure 2 a–Figure 2c), whereas for a larger fraction of the main component (ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20) they persist up to T≈10T\approx10 (Figure 2 d–Figure 2f). In the mean-field approximation, the full extent of the modulated region is given by the λ\lambda-surface Tλ(ρ1b,ρ2b)T_\lambda(\rho_{1b},\rho_{2b}) of the 3Y+2Y model shown in Figure 3. Above this surface the homogeneous mixture is stable, while below it the mixture is microsegregated into alternating layers.

Figure 2. Bulk mean-field density profiles of the mixture at equal concentrations ρ1b=ρ2b=0.20\rho_{1b} = \rho_{2b} = 0.20 for (a) T=2T = 2, (b) T=3T = 3, (c) T=4T = 4 and at unequal concentrations ρ1b=0.80\rho_{1b} = 0.80, ρ2b=0.20\rho_{2b} = 0.20 for (d) T=6T = 6, (e) T=8T = 8, (f) T=9T = 9. Solid red lines: species 1; dashed blue lines: species 2.

Figure 3. The λ\lambda-surface of the full 3Y+2Y model: the temperature TλT_\lambda below which the bulk mixture ceases to be mesoscopically homogeneous and develops periodic density modulations. (a) TλT_\lambda as a function of the density of the main component ρ1b\rho_{1b} at fixed solvent densities ρ2b=0.05\rho_{2b}=0.05, 0.10, 0.20 and 0.30; (b) TλT_\lambda over the (ρ1b,ρ2b)(\rho_{1b},\rho_{2b}) plane (colour scale). The mixture is mesoscopically homogeneous above this surface and microsegregated below it.

It is worth noting that, within the mean-field approximation, the one-component fluid with the potential (2.1) shows no modulational instability at all, since ν^11(q)>0\hat{\nu}_{11}(q)>0 for all qq (see § 5.1). The microsegregation found here is therefore induced by the second component: the solvent shifts the onset of periodic ordering into a region where the pure competing interaction fluid remains homogeneous at the mean-field level, whereas in simulations the one-component fluid orders only at lower temperatures [32].

These mean-field results can be compared with direct simulations of the same model. Figure 4 shows equilibrated configurations from molecular dynamics in the bulk at ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20, the composition of Figure 2 d–Figure 2f, over the range T=2.0T=2.0–4.5. At T=2.0T=2.0 the mixture has segregated into a clear lamellar structure of alternating slabs of species 1 and species 2. The same structure survives at T=2.5T=2.5–3.5, with progressively more diffuse interfaces between the slabs. At T=4.0T=4.0 the layers begin to lose their shape, and at T=4.5T=4.5 they are destroyed altogether. The onset of microsegregation in the simulated system thus lies between T=4.0T=4.0 and T=4.5T=4.5, well below the mean-field estimate T≈10T\approx10 (see § 4.3).

Equilibrated configurations from molecular dynamics for the bulk mixture at $\rho_{1b}=0.8$, $\rho_{2b}=0.2$ at six temperatures

Figure 4. Equilibrated configurations from molecular dynamics for the bulk mixture at ρ1b=0.8\rho_{1b}=0.8, ρ2b=0.2\rho_{2b}=0.2: (a) T=2.0T=2.0, (b) T=2.5T=2.5, (c) T=3.0T=3.0, (d) T=3.5T=3.5, (e) T=4.0T=4.0 and (f) T=4.5T=4.5. The lamellar structure formed at low temperature becomes progressively more diffuse on heating and is destroyed between T=4.0T=4.0 and T=4.5T=4.5.

Hard wall

Figures 5 and 6 show the density profiles of the mixture confined between two hard walls at several temperatures: figure 5 for equal bulk concentrations ρ1b=ρ2b=0.20\rho_{1b}=\rho_{2b}=0.20 (T=2T=2–8), and figure 6 for ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20. For the latter composition, at the lowest temperatures (T≲9T\lesssim9) the two components form alternating layers in which the concentration of one species is close to zero while that of the other reaches its maximal value. These oscillations do not decay towards the middle of the slit. At higher temperatures, the mixture becomes homogeneous far from the wall where the densities of the species reach their bulk values. The ordering is destroyed across the slit only at T≈12T\approx12, i.e., at a higher temperature than in the corresponding bulk system (T≈10T\approx10). The confining wall thus promotes the ordering – it acts as a nucleation site for the near-surface structure, which persists at all temperatures considered. Near the surface the structure of the mixture differs from the bulk in all cases: the main component is adsorbed, its contact value substantially exceeding the bulk one, and a layer a few effective diameters wide is formed in which the solvent is almost absent. The behaviour of the main component near the wall is reminiscent of the one-component case studied in [32, 39], in particular the two-peak structure that develops immediately in front of the wall, whereas the solvent is expelled from the near-surface region by the main component at all temperatures considered. A qualitatively similar near-surface structure, consisting of separate, partially mixed and fully mixed layers with a contact film of the main component, was observed in molecular simulations of a related self-assembling binary mixture near a wall [9].

Comparison with computer simulations

The theoretical predictions have been verified and complemented by molecular dynamics simulations of the mixture of species 1 and 2 in the canonical ensemble [47], carried out with the LAMMPS package [48], in which the pair interactions (2.1)–(2.3) were implemented specifically for this purpose. The bulk runs used a cubic cell of side L=36L=36 with periodic boundary conditions in all three directions, containing N1=ρ1bL3N_1=\rho_{1b}L^3 and N2=ρ2bL3N_2=\rho_{2b}L^3 particles of the two species. In the slit geometry a cell of the same size was used, with the walls placed at z=0z=0 and z=Lz=Lz=L_z=L and implemented as reflecting hard walls, and periodic boundary conditions were applied only in the xx and yy directions. All particles have the same mass m=1m=1, and time is measured in units of τ=r0m/∣ν11(rmin⁡)∣\tau=r_0\sqrt{m/|\nu_{11}(r_{\min})|}. The equations of motion were integrated with the velocity Verlet algorithm at a time step δt=0.01τ\delta t=0.01\tau. The temperature was controlled by a Nosé–Hoover thermostat with a relaxation time of 1.0τ1.0\tau, and the pair potentials were truncated at rc=12.0r_c=12.0. Starting from a random placement of particles of both species, each system was equilibrated over 10510^5 steps, after which the density profiles of both species were accumulated over a further 10510^5 steps, with configurations sampled every 100 steps.

Figure 5. Density profiles of the binary mixture between two hard walls for bulk densities ρ1b=ρ2b=0.20\rho_{1b}=\rho_{2b}=0.20 and temperatures (a) T=2T=2, (b) T=4T=4, (c) T=6T=6 and (d) T=8T=8. Solid red lines: species 1; dashed blue lines: species 2.

For the bulk mixture, the simulation results are presented only as the representative snapshots of Figure 4. A quantitative comparison is made for the 3Y+2Y mixture in the slit geometry. As noted in § 4.1, the simulations place the onset of microsegregation in the bulk mixture with ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20 between T=4.0T=4.0 and T=4.5T=4.5, whereas the mean-field theory gives T≈10T\approx10. This reflects a general feature of the mean-field approximation, which systematically overestimates the temperature at which microsegregation sets in, so that the λ\lambda-surface obtained from the theory lies at higher temperatures than the boundary of the modulated region observed in simulation. A comparison of the two profiles at one and the same temperature could therefore place the system on opposite sides of this boundary and would not be meaningful. Instead, the mean-field and simulation profiles are compared at shifted temperatures, the simulation temperature being taken lower than the mean-field one. The shift is larger for the denser mixture, in line with the larger mean-field overestimate found there.

Figure 7 shows three such pairs for each of the two mixtures confined by hard walls, with equal densities ρ1b=ρ2b=0.20\rho_{1b}=\rho_{2b}=0.20 (figure 7 a–c) and with ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20 (figure 7 d–f). For the first mixture, the mean-field profiles at T=4T=4 are compared with the simulation data at T=2T=2 in figure 7 a. Both species show undamped oscillations across the whole slit, and the theory reproduces their wavelength and the positions of the maxima and minima. The agreement is close for the competing-interaction component, whereas for the solvent the theory considerably overestimates the amplitude of the oscillations, which are only weak in simulation. As the temperature is raised (mean field at T=5T=5 versus simulation at T=3T=3, and mean field at T=6T=6 versus simulation at T=4T=4), the profiles become only weakly structured except near the walls, where theory and simulation agree well for both components (Figure 7 b,Figure 7c).

Figure 7. Density profiles from the mean-field theory (MF, lines) and from molecular dynamics simulations (MD, symbols) for ρ1b=ρ2b=0.20\rho_{1b}=\rho_{2b}=0.20: (a) MF at T=4T=4 versus MD at T=2T=2, (b) MF at T=5T=5 versus MD at T=3T=3, and (c) MF at T=6T=6 versus MD at T=4T=4 and for ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20: (d) MF at T=10T=10 versus MD at T=5T=5, (e) MF at T=11T=11 versus MD at T=6T=6, and (f) MF at T=12T=12 versus MD at T=7T=7. The simulation temperature is taken lower than the mean-field one to compensate for the mean-field overestimation of the onset temperature of microsegregation (see text). Red: species 1; blue: species 2.

Figure 6. Density profiles of the binary mixture between two hard walls for bulk densities ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20 and temperatures (a) T=7T=7, (b) T=9T=9, (c) T=11T=11 and (d) T=12T=12. Solid red lines: species 1; dashed blue lines: species 2.

The agreement is even better for the denser mixture, ρ1b=0.80\rho_{1b}=0.80, ρ2b=0.20\rho_{2b}=0.20 (Figure 7 d–Figure 7f). At the lowest pair of temperatures (mean field at T=10T=10 versus simulation at T=5T=5) both descriptions show alternating layers of the two species that do not decay towards the middle of the slit. The positions and heights of the maxima of the main component practically coincide, and the solvent is concentrated in the minima of the main-component profile in both cases. Near each wall, theory and simulation consistently show a strong adsorption of the main component with a large contact value, the two-peak structure in front of the wall, and a layer a few diameters wide from which the solvent is expelled. The mean-field theory only slightly overestimates the height of the first solvent maximum. On heating (mean field at T=11T=11 and 1212 versus simulation at T=6T=6 and 77), the oscillations in the middle of the slit are progressively damped in both descriptions at a similar rate, while the near-wall structure remains almost unchanged. Thus, once the shift of the onset temperature is taken into account, the mean-field theory provides a quantitatively reasonable description of the simulated profiles, particularly for the main component and for the denser mixture.

Analytical density profiles in the mesoscopically homogeneous region

Two restrictions underlie the closed-form results of this section, and we state them explicitly at the outset. First, the analytical treatment is confined to the mesoscopically homogeneous part of the phase diagram, i.e., to temperatures above the λ\lambda-surface, T>TλT>T_\lambda, where the near-wall modulations decay into the bulk. There, sufficiently far from TλT_\lambda, the deviations of the profiles from their bulk values are small and the linearization of equations (13)–(14) is justified. The microsegregated regime below TλT_{\lambda} is accessible only through the numerical solution of § 4. Second, to obtain the profiles in radicals we adopt a simplified version of the model in which the solvent–solvent and cross interactions are reduced to a single repulsive Yukawa term identical to the leading term of ν11\nu_{11},

A12(1)=A22(1)=A11(1),α12(1)=α22(1)=α11(1),A12(2)=A22(2)=0,(20)A_{12}^{(1)}=A_{22}^{(1)}=A_{11}^{(1)},\qquad\alpha_{12}^{(1)}=\alpha_{22}^{(1)}=\alpha_{11}^{(1)},\qquad A_{12}^{(2)}=A_{22}^{(2)}=0, \tag*{(20)}

cf. equations (2)–(3). For the full 3Y+2Y model the characteristic equation is of seventh degree, cannot in general be solved in radicals, and must be treated numerically. Following the shorthand of § 2, we refer to the full model as the 3Y+2Y model and to this reduced version as the 3Y+1Y model, after the number of Yukawa terms in ν11\nu_{11} and in ν12=ν22\nu_{12}=\nu_{22} respectively. Under these two restrictions the problem becomes analytically tractable. The scheme presented below extends to the binary mixture the one-component derivation of [32]. Here we outline the main steps and quote the final results, while the details of the calculation are given in appendix A.

Linearization and the eigenvalue problem

In the mesoscopically homogeneous region the density profiles can be obtained in closed form. The route to this is to linearize the exponentials in equations (3.9)–(3.10),

ρ1(z)=ρ1b[1−β∑k=13(V11(k)(z)−V11b(k))−β∑k=12(V12(k)(z)−V12b(k))],ρ2(z)=ρ2b[1−β∑k=12(V22(k)(z)−V22b(k))−β∑k=12(V21(k)(z)−V21b(k))],(21)\begin{aligned} \rho_1(z) &= \rho_{1b}\left[1-\beta\sum_{k=1}^{3}\left(V_{11}^{(k)}(z)-V_{11b}^{(k)}\right)-\beta\sum_{k=1}^{2}\left(V_{12}^{(k)}(z)-V_{12b}^{(k)}\right)\right],\\ \rho_2(z) &= \rho_{2b}\left[1-\beta\sum_{k=1}^{2}\left(V_{22}^{(k)}(z)-V_{22b}^{(k)}\right)-\beta\sum_{k=1}^{2}\left(V_{21}^{(k)}(z)-V_{21b}^{(k)}\right)\right], \tag*{(21)} \end{aligned}

which turns the system (16) into a set of linear equations for the fields Eij(k)(z)E_{ij}^{(k)}(z). Eliminating the densities and potentials, one arrives at a system of second-order differential equations that can be written in matrix form,

E′′=AE,(22)\mathbf{E}''=\mathbf{A}\mathbf{E}, \tag*{(22)}

where E\mathbf{E} collects the seven independent fields E11(1,2,3)E_{11}^{(1,2,3)}, E12(1,2)E_{12}^{(1,2)}, E21(1,2)E_{21}^{(1,2)}, and A\mathbf{A} is a 7×77 \times7 matrix built from the parameters αij(k)\alpha_{ij}^{(k)} and ϰij(k)\varkappa_{ij}^{(k)}. The decay coefficients of the profiles are the square roots of the eigenvalues Λn\Lambda_n of A\mathbf{A}, i.e., the roots of the characteristic septic polynomial det⁡(A−ΛI)=0\det(\mathbf{A}-\Lambda\mathbf{I})=0,

Λ7+a6Λ6+a5Λ5+a4Λ4+a3Λ3+a2Λ2+a1Λ+a0=0.(23)\Lambda^7+a_6\Lambda^6+a_5\Lambda^5+a_4\Lambda^4+a_3\Lambda^3+a_2\Lambda^2+a_1\Lambda+a_0=0. \tag*{(23)}

The septic equation (23) cannot be solved in radicals in the general case. However, it reduces to a cubic equation in the eigenvalue Λ\Lambda (equivalently, a bicubic equation in the decay coefficient λ=Λ\lambda=\sqrt{\Lambda}) if three conditions are met simultaneously: (i) the solvent–solvent and cross interactions are purely repulsive, A12(2)=A22(2)=0A_{12}^{(2)}=A_{22}^{(2)}=0; (ii) the range of the remaining Yukawa term coincides with that of the leading term of the main component, α1≡α11(1)=α12(1)=α22(1)\alpha_1\equiv\alpha_{11}^{(1)}=\alpha_{12}^{(1)}=\alpha_{22}^{(1)}; and (iii) so does its amplitude, A12(1)=A22(1)=A11(1)A_{12}^{(1)}=A_{22}^{(1)}=A_{11}^{(1)}. Together they define the simplified model (20) treated in what follows; without condition (iii) the equation would be quartic. Under these conditions V21(1)=V11(1)V_{21}^{(1)}=V_{11}^{(1)} and V22(1)=V12(1)V_{22}^{(1)}=V_{12}^{(1)}, i.e., both species feel the leading Yukawa term through the total density ρ1+ρ2\rho_1+\rho_2. Introducing the superposition of the two fields with the common range, E~(z)=E11(1)(z)+E12(1)(z)\widetilde{E}(z)=E_{11}^{(1)}(z)+E_{12}^{(1)}(z), that is the sum of the two fields of range α1\alpha_1 felt by the main component, the problem reduces to a closed system of three second-order equations for E=(E~,E12,E13)T\mathbf{E}=(\widetilde{E},E_{12},E_{13})^{\mathsf{T}}, with

A=(α12+ϰ112+ϰs2ϰ122ϰ122ϰ122α22+ϰ122ϰ122ϰ132ϰ132α32+ϰ132),(24)\mathbf{A}= \begin{pmatrix} \alpha_1^2+\varkappa_{11}^2+\varkappa_s^2 & \varkappa_{12}^2 & \varkappa_{12}^2\\ \varkappa_{12}^2 & \alpha_2^2+\varkappa_{12}^2 & \varkappa_{12}^2\\ \varkappa_{13}^2 & \varkappa_{13}^2 & \alpha_3^2+\varkappa_{13}^2 \end{pmatrix}, \tag*{(24)}

where we abbreviated ϰ1k2≡ϰ11(k)2\varkappa_{1k}^2\equiv\varkappa_{11}^{(k)2} and αk≡α11(k)\alpha_k\equiv\alpha_{11}^{(k)} (k=1,2,3k=1,2,3), and ϰs2=4πβρ2bA11(1)=(ρ2b/ρ1b)ϰ11(1)2\varkappa_s^2=4\pi\beta\rho_{2b}A_{11}^{(1)}=(\rho_{2b}/\rho_{1b})\varkappa_{11}^{(1)2} is the coupling through which the solvent enters the (1,1)(1,1) entry of A\mathbf{A}. The eigenvalues now solve the cubic equation

Λ3−bΛ2+cΛ−d=0,(25)\Lambda^3-b\Lambda^2+c\Lambda-d=0, \tag*{(25)}

with

b=ϰ~11 2+ϰ~12 2+ϰ~13 2,c=ϰ~11 2ϰ~12 2+ϰ~11 2ϰ~13 2+ϰ~12 2ϰ~13 2−ϰ11 2ϰ12 2−ϰ11 2ϰ13 2−ϰ12 2ϰ13 2,d=ϰ~11 2(ϰ12 2ϰ13 2−ϰ12 2ϰ13 2)−ϰ11 2ϰ12 2ϰ13 2−ϰ11 2ϰ13 2ϰ12 2+2ϰ11 2ϰ12 2ϰ13 2,(26)\begin{aligned} b &= \widetilde{\varkappa}_{11}^{\,2}+\widetilde{\varkappa}_{12}^{\,2}+\widetilde{\varkappa}_{13}^{\,2},\\ c &= \widetilde{\varkappa}_{11}^{\,2}\widetilde{\varkappa}_{12}^{\,2}+\widetilde{\varkappa}_{11}^{\,2}\widetilde{\varkappa}_{13}^{\,2}+\widetilde{\varkappa}_{12}^{\,2}\widetilde{\varkappa}_{13}^{\,2}-\varkappa_{11}^{\,2}\varkappa_{12}^{\,2}-\varkappa_{11}^{\,2}\varkappa_{13}^{\,2}-\varkappa_{12}^{\,2}\varkappa_{13}^{\,2},\\ d &= \widetilde{\varkappa}_{11}^{\,2}\left(\varkappa_{12}^{\,2}\varkappa_{13}^{\,2}-\varkappa_{12}^{\,2}\varkappa_{13}^{\,2}\right)-\varkappa_{11}^{\,2}\varkappa_{12}^{\,2}\varkappa_{13}^{\,2}-\varkappa_{11}^{\,2}\varkappa_{13}^{\,2}\varkappa_{12}^{\,2}+2\varkappa_{11}^{\,2}\varkappa_{12}^{\,2}\varkappa_{13}^{\,2}, \tag*{(26)} \end{aligned}

where χ~11=χ112+χs2+α12\tilde{\chi}_{11}=\chi_{11}^{2}+\chi_{\mathrm{s}}^{2}+\alpha_{1}^{2}, χ~12=χ122+α22\tilde{\chi}_{12}=\chi_{12}^{2}+\alpha_{2}^{2}, χ~13=χ132+α32\tilde{\chi}_{13}=\chi_{13}^{2}+\alpha_{3}^{2}. For the range of parameters considered here, equation (5.6) yields a single real solution Λ0=λ02\Lambda_{0}=\lambda_{0}^{2} together with a complex-conjugate pair Λ1,2=λ1,22\Lambda_{1,2}=\lambda_{1,2}^{2}, λ1,2=λ±iμ\lambda_{1,2}=\lambda\pm i\mu, all following in closed form from the standard formulae for the roots of a cubic equation. It is this reduction that makes the linearized theory fully analytic, and it is the reason we use the 3Y+1Y model to test the analytical description against the numerical mean-field solution. For the full 3Y+2Y model the septic equation (5.4) has to be solved numerically. Its linearized profiles at hard and attractive walls will be considered elsewhere. The general solution for each field is a superposition of a monotonically decaying term and an exponentially damped oscillation,

E~(z)=c10e−λ0z+C1e(−λ+iμ)z+C2e(−λ−iμ)z,(27)\tilde{E}(z)=c_{10}e^{-\lambda_{0}z}+C_{1}e^{(-\lambda+i\mu)z}+C_{2}e^{(-\lambda-i\mu)z}, \tag*{(27)}

where only the terms decaying at z→∞z\to\infty are retained, and C1C_{1}, C2C_{2} are complex conjugate. Analogous expressions hold for E12(z)E_{12}(z) and E13(z)E_{13}(z). Since the system is linear and homogeneous, all coefficients can be expressed through a single set of unknowns using the eigenvector relations.

The linearized description presupposes that the bulk mixture is mesoscopically homogeneous, and it is worth recording where that is the case. The boundary of the homogeneous region is the λ\lambda-surface – the counterpart, for a two-component mixture, of the λ\lambda-line of a one-component fluid – reached when a complex-conjugate pair λ±iμ\lambda\pm i\mu attains λ=0\lambda=0, so that the damping of the oscillations disappears while their wavelength stays finite. Equivalently, it is the locus at which min⁡qdet⁡[I+βρjbν^ij(q)]\min_{q}\det[\mathbf{I}+\beta\rho_{j\mathrm{b}}\hat{\nu}_{ij}(q)] first vanishes, at a non-zero wavenumber q∗q^{*}. Figure 8 shows the resulting Tλ(ρ1b,ρ2b)T_{\lambda}(\rho_{1\mathrm{b}},\rho_{2\mathrm{b}}) for the simplified model. Since the determinant depends on the densities and the temperature only through the products βρib\beta\rho_{i\mathrm{b}}, TλT_{\lambda} is a homogeneous function of degree one of the two densities, Tλ(sρ1b,sρ2b)=sTλ(ρ1b,ρ2b)T_{\lambda}(s\rho_{1\mathrm{b}},s\rho_{2\mathrm{b}})=sT_{\lambda}(\rho_{1\mathrm{b}},\rho_{2\mathrm{b}}), and q∗q^{*} depends only on their ratio ρ2b/ρ1b\rho_{2\mathrm{b}}/\rho_{1\mathrm{b}}.

Figure 8. The λ\lambda-surface of the simplified model: the temperature TλT_{\lambda} below which the bulk mixture ceases to be mesoscopically homogeneous. (a) λ\lambda-lines, TλT_{\lambda} as a function of ρ1b\rho_{1\mathrm{b}} at fixed solvent densities ρ2b=0.05,0.10,0.20,0.30\rho_{2\mathrm{b}}=0.05, 0.10, 0.20, 0.30; (b) TλT_{\lambda} over the (ρ1b,ρ2b)(\rho_{1\mathrm{b}},\rho_{2\mathrm{b}}) plane (colour scale), with contour lines spaced by 10. The linearized theory of this section applies above this surface.

Two features are relevant to what follows. First, within the present soft-core mean-field treatment ν^11(q)>0\hat{\nu}_{11}(q)>0 and the pure fluid shows no λ\lambda-line. Excluded-volume effects, not included here, make it microsegregate at lower temperatures [32]. The instability of the mixture is therefore produced by the coupling to the solvent rather than by the competing interaction alone. Second, because the cross and solvent interactions of the simplified model are purely repulsive, and both strong and long-ranged, TλT_{\lambda} rises steeply with either density. The analytical theory is accordingly useful at relatively dilute densities for a given temperature, and all the states considered below lie above the surface of figure 8.

Boundary conditions at the wall

The eigenvalue problem of § 5.1 fixes the shapes e−λ0ze^{-\lambda_0 z} and e−λz±iμze^{-\lambda z\pm i\mu z} of the admissible solutions, but not their amplitudes and these must be supplemented with boundary conditions at the wall. Such conditions follow from the geometry alone. For z<0z<0 both densities vanish, so each potential obeys Vi′′−αi2ViV_i''-\alpha_i^2V_i there and, being bounded as z→−∞z\to-\infty, decays as Vi(z)=Vi(0)eαizV_i(z)=V_i(0)e^{\alpha_i z}. Continuity of ViV_i and of Ei=−Vi′E_i=-V_i' at z=0z=0 eliminates the single free constant of this exterior solution and leaves one relation per field,

Vi′(0)=αiVi(0),equivalentlyEi(0)=−αiVi(0),i=1,2,3.(28)V_i'(0)=\alpha_iV_i(0),\qquad\text{equivalently}\qquad E_i(0)=-\alpha_iV_i(0),\qquad i=1,2,3. \tag*{(28)}

Equations (28) are exact: nothing has been assumed about the density profile, only that there is no fluid behind the wall and that the potential of a bounded density distribution is continuously differentiable. They are in fact the same statement that yields the contact theorem in § 3.3, at the step where αi2Vi2−Ei2\alpha_i^2V_i^2-E_i^2 is seen to vanish at z=0+z=0+. Since there is one such relation for every Yukawa field, the number of conditions equals the number of unknown amplitudes and nothing further is required. In particular the contact densities are not prescribed: they follow from the solution.

Evaluating Vi(0)V_i(0) and Ei(0)E_i(0) from the general solution (see appendix § 8.§ 4), equations (28) become a linear system for the amplitudes,

(λ0+αi)rˉi0+(λ2+μ2+αiλ)rˉi1−αiμrˉi2=−χi2αi,i=1,2,3.(29)(\lambda_0+\alpha_i)\bar r_{i0}+(\lambda^2+\mu^2+\alpha_i\lambda)\bar r_{i1}-\alpha_i\mu\bar r_{i2}=-\frac{\chi_i^2}{\alpha_i},\qquad i=1,2,3. \tag*{(29)}

where the index ii labels the Yukawa term and not the species: αi\alpha_i is the range of the iith term of ν11\nu_{11}, and the corresponding couplings are χ12=χ112+χ32\chi_1^2=\chi_{11}^2+\chi_3^2, χ22=χ122\chi_2^2=\chi_{12}^2 and χ32=χ132\chi_3^2=\chi_{13}^2. As in the eigenvalue problem, the two-component character enters only through the solvent coupling χs2\chi_s^2 contained in χ12\chi_1^2. The elimination of rˉ20\bar r_{20}, rˉ30\bar r_{30} and rˉlm\bar r_{lm} and the resulting closed-form amplitudes are given in appendix § 8.§ 4.

It is worth contrasting (28) with the prescription used in the earlier treatments of one-, two- and three-Yukawa fluids at a hard wall [32, 37, 38, 39], where the contact theorem is imposed on each of the three potentials separately,

Vib−Vi(0)=χi22αi2,i=1,2,3.(30)V_{ib}-V_i(0)=\frac{\chi_i^2}{2\alpha_i^2},\qquad i=1,2,3. \tag*{(30)}

Summed over ii, equations (30) reproduce the contact theorem exactly. Taken separately they say more than it does: each states that a given Yukawa potential at the wall is one half of its bulk value, which is what one obtains if the fluid retains its bulk density right up to the surface. The contact densities are then fixed at βPMF\beta P^{\mathrm{MF}} and βPMF\beta P^{\mathrm{MF}}, a species-resolved split that the exact nonlinear solution does not reproduce. The two prescriptions differ at second order in the interaction: for a one-component fluid, (5.11) gives ρ(0+)/ρb=1+Γ/2\rho(0+)/\rho_b=1+\Gamma/2 with Γ=∑iVib\Gamma=\sum_iV_{ib}, whereas (5.9) yields ρ(0+)/ρb=1+Γ\rho(0+)/\rho_b=\sqrt{1+\Gamma}. The two agree to first order, which is the order to which the linearization (5.2) is valid, but 1+Γ/21+\Gamma/2 is a value imported from the nonlinear theory and imposed on the linearized equations, while 1+Γ\sqrt{1+\Gamma} is what those equations produce on their own. Neither prescription is free of objection, and in what follows we examine both, comparing each with the numerical solution of the nonlinear equations. It should be borne in mind that a linearized theory cannot satisfy the contact theorem identically in any case: the relation (3.15) follows from the first integral of the nonlinear equations, and the truncation (5.2) has already discarded the terms needed to reproduce it. Retaining higher orders in the expansion of the exponentials would bring the linearized contact densities closer to it. Conditions (5.11) thus restore the summed relation by construction, but at the price of a further assumption, namely that the contact density is split between the species according to their partial pressures. This assumption is supported neither by the contact theorem nor by the exact solution.

Density profiles

Integrating the fields and substituting into the linearized equations (21), the density profiles of the two species take the form

ρ1(z)ρ1b=1−[rˉ10+rˉ20+rˉ30]e−λ0z−e−λz[{λ(rˉ11+rˉ21+rˉ31)−μ(rˉ12+rˉ22+rˉ32)}cos⁡μz−{μ(rˉ11+rˉ21+rˉ31)+λ(rˉ12+rˉ22+rˉ32)}sin⁡μz],(31)\frac{\rho_1(z)}{\rho_{1b}} = 1 - \left[\bar{r}_{10}+\bar{r}_{20}+\bar{r}_{30}\right]e^{-\lambda_0 z} - e^{-\lambda z}\left[\left\{\lambda(\bar{r}_{11}+\bar{r}_{21}+\bar{r}_{31})-\mu(\bar{r}_{12}+\bar{r}_{22}+\bar{r}_{32})\right\}\cos\mu z - \left\{\mu(\bar{r}_{11}+\bar{r}_{21}+\bar{r}_{31})+\lambda(\bar{r}_{12}+\bar{r}_{22}+\bar{r}_{32})\right\}\sin\mu z\right], \tag*{(31)}
ρ2(z)ρ2b=1−rˉ10e−λ0z−e−λz[{λrˉ11−μrˉ12}cos⁡μz−{μrˉ11+λrˉ12}sin⁡μz],(32)\frac{\rho_2(z)}{\rho_{2b}} = 1 - \bar{r}_{10}e^{-\lambda_0 z} - e^{-\lambda z}\left[\left\{\lambda\bar{r}_{11}-\mu\bar{r}_{12}\right\}\cos\mu z-\left\{\mu\bar{r}_{11}+\lambda\bar{r}_{12}\right\}\sin\mu z\right], \tag*{(32)}

where rˉi0\bar{r}_{i0}, rˉi1\bar{r}_{i1}, rˉi2\bar{r}_{i2} (i=1,2,3i=1,2,3) are the amplitudes of the three mean-field potentials Vi(z)V_i(z) and they are lengthy but explicit functions of the pair-potential parameters, the densities and the temperature, fixed by the eigenvector relations of A\mathbf{A} together with the boundary conditions (28) of section 5.2. Their closed-form expressions are derived in appendix A.4. A notable feature of equations (31)–(32) is that the two species share the same decay coefficients λ0\lambda_0, λ\lambda, and μ\mu: the profile of the main component collects all three Yukawa fields, while that of the solvent involves only the field with the range α1\alpha_1 common to both species. Each profile combines a monotonic contribution that decays with the decay coefficient λ0\lambda_0 with an oscillatory contribution that decays with the decay coefficient λ\lambda and oscillates with the wavelength 2π/μ2\pi/\mu. This shared oscillatory structure, with amplitudes of opposite sign for the two species, is precisely what produces their alternating layering.

Comparison with the numerical results

Throughout this comparison the bulk densities are small enough for the near-wall modulation to remain weak, which is the regime in which the linearization (21) is meaningful. Where the profile is strongly structured, the linearization itself fails, and neither prescription for the amplitudes is reliable.

The role of the boundary conditions is shown directly in Figure 9. Close to the boundary of the mesoscopically homogeneous region the two prescriptions differ substantially. At T=15T=15, i.e., T/Tλ=1.05T/T_\lambda=1.05, the first maximum of the main component lies 0.054 above the bulk density in the numerical solution, 0.077 above it with the continuity conditions (28), and 0.152 above it, nearly three times the numerical value, with the contact-value conditions (30). Over 0.5<z<250.5<z<25, the root-mean-square deviation from the numerical profiles obtained with (28) is smaller than that obtained with (30) by a factor of four for the main component and of three for the solvent. Both linearized solutions take slightly negative solvent densities near z≈2z\approx2 at this temperature, −0.009-0.009 with (28) against −0.037-0.037 with (30), which shows that the linearization itself is close to its limit so near TλT_\lambda. The difference between the two prescriptions falls off rapidly with temperature: at T=18T=18 the corresponding factors are eight and three, and by T=23T=23 the three sets of curves are indistinguishable on the scale of Figure 9, panels (a)–(d).

Figure 9. Effect of the boundary conditions on the density profiles at ρ1b=0.1\rho_{1b}=0.1, ρ2b=0.02\rho_{2b}=0.02 (Tλ=14.3T_\lambda=14.3): (a) T=10T=10, (b) T=15T=15, (c) T=18T=18, (d) T=23T=23. In (a), below TλT_\lambda, the mean-field solution in a slit of width Lz=40L_z=40 develops alternating layers throughout and the linearized description does not apply, so only the MF profiles are shown. In (b)-(d), at a single hard wall: solid lines, numerical mean-field solution (MF); dashed lines, linearized profiles from the continuous conditions (5.9); dotted lines, from the contact-value conditions (5.11) imposed on each potential separately. Red lines refer to the main component (sort 1), blue lines to the solvent (sort 2).

The two prescriptions differ correspondingly in their behaviour at contact. Equations (30) reproduce the summed contact theorem (19) by construction, at the cost of the profile away from the wall. With the exact conditions (28) the contact densities are outputs rather than inputs, and their sum falls below βPMF\beta P^{\mathrm{MF}} by 14, 12 and 9 percent at T=15T=15, 18 and 23 respectively. This is the second-order discrepancy discussed in section 5.2. It diminishes as the temperature rises and the linearization (21) becomes more accurate. In neither case are the species-resolved relations βPiMF=ρi(0+)\beta P_i^{\mathrm{MF}}=\rho_i(0_+) obeyed, and the numerical solution does not obey them either: only the summed relation (19) is exact.

Qualitative behaviour with temperature and solvent density

The analytical expressions (31)–(32) provide a convenient tool to explore the qualitative behaviour of the profiles as functions of the thermodynamic parameters. As the temperature is varied at fixed densities ρ1b=0.1\rho_{1b}=0.1, ρ2b=0.02\rho_{2b}=0.02 (Figure 9), on cooling towards TλT_\lambda the amplitude of the near-surface structure grows and the decay length 1/λ1/\lambda increases, so that the alternating oscillations of the two species extend further into the bulk. On heating, the structure becomes weak and the mixture is essentially uniform already a few diameters away from the wall.

The role of the solvent concentration is illustrated in Figure 10, where the density of the main component is fixed at ρ1b=0.1\rho_{1b}=0.1 and T=23T=23 while the solvent density is increased from ρ2b=0.02\rho_{2b}=0.02 up to ρ2b=0.065\rho_{2b}=0.065 (the last value lying just above the λ\lambda-line) in four steps (all states still above the λ\lambda-surface of Figure 8, so that the linearization applies). At the lowest solvent density shown (ρ2b=0.02\rho_{2b}=0.02) the near-surface structure is weak and the three descriptions nearly coincide. As the solvent density grows, the structure of both components becomes progressively more pronounced: the profile of the main component develops a marked maximum near the surface, and the two species arrange themselves in an increasingly clear microsegregated pattern. This shows that the solvent, although interacting through a simple potential, actively promotes the near-surface ordering and the microphase separation of the mixture. Each panel of Figure 10 also compares the two boundary-condition prescriptions with the numerical mean-field solution: at low ρ2b\rho_{2b} all three coincide, but as ρ2b\rho_{2b} grows and the state approaches the λ\lambda-line the contact-value prescription overshoots the near-wall maximum. For the state closest to the λ\lambda-surface, ρ2b=0.065\rho_{2b}=0.065 (Figure 9d), it develops large oscillations and negative solvent densities, whereas the continuity conditions continue to track the mean-field solution. This is the same trend seen with temperature in Figure 9.

Figure 10. Density profiles for ρ1b=0.1\rho_{1b}=0.1 at T=23T=23 and increasing solvent density: (a) ρ2b=0.02\rho_{2b}=0.02, (b) ρ2b=0.04\rho_{2b}=0.04, (c) ρ2b=0.06\rho_{2b}=0.06 and (d) ρ2b=0.065\rho_{2b}=0.065 (just above the λ\lambda**-line). In each panel the numerical mean-field solution (MF, solid) is compared with the linearized profiles from the continuous conditions (5.9) (dashed) and from the contact-value conditions (5.11) (dotted). Red lines refer to species 1 (sort 1), blue lines to species 2 (sort 2).

Conclusions

We have developed a field-theoretical mean-field description of a binary mixture of soft particles near a planar surface, in which the main component interacts through a competing SALR potential (a sum of three Yukawa terms), while the solvent and the cross interactions are modelled by two Yukawa terms. We derived the coupled equations for the density profiles of both species, obtained a first integral of these equations, and used it to formulate the contact theorem for the two-component system, which reduces to the known one-component relation in the appropriate limit.

The mean-field equations were solved numerically in the bulk and for a fluid confined by hard walls. At the mean-field level, in the temperature range considered here, the pure SALR fluid is homogeneous – it develops its own mesostructures only at lower temperatures [32], so the solvent shifts the microsegregation onset up into the region where the one-component fluid is still uniform. At high temperatures the two species are homogeneously mixed, whereas upon cooling or upon increasing the density they separate into alternating layers rich in one or the other component, the onset temperature depending on the composition. Near a hard wall the main component is strongly adsorbed and retains the two-peak structure typical of the one-component fluid [32], while the solvent is expelled from a near-surface layer a few particle diameters wide. The wall promotes the ordering, so that the layering persists across a slit up to temperatures at which the bulk mixture is already homogeneous. Molecular dynamics simulations confirm this picture. The mean-field theory overestimates the onset temperature of microsegregation by roughly a factor of two, but oncethis shift is taken into account the mean-field profiles agree well with the simulation data, particularly for the main component and for the denser mixture.

In the mesoscopically homogeneous region, above the λ\lambda-surface, we obtained closed-form density profiles by linearizing the theory and solving the resulting eigenvalue problem. When the solvent–solvent and cross interactions are purely repulsive and coincide in range and amplitude with the leading Yukawa term of the main component, the problem reduces to a cubic equation with a single real root and a complex-conjugate pair, giving a monotonic plus an exponentially damped oscillatory contribution. Both species share the same decay lengths and oscillation period, with oscillatory amplitudes of opposite sign, which is the origin of their alternating layering. The analytical profiles correctly describe the dependence on temperature and solvent concentration, in particular the solvent-induced enhancement of the near-surface ordering. The amplitudes of these profiles are fixed by boundary conditions that follow from the continuity of the Yukawa potentials and of their derivatives at the wall. These conditions are exact and supply one relation per Yukawa field, so that their number matches the number of unknown amplitudes. The same continuity underlies the contact theorem itself, which, however, yields only one summed relation. Imposing it on each potential separately, as done previously, is therefore an additional assumption, which in a mixture also prescribes an unjustified split of the contact density between the species. The continuity conditions require no such assumption and leave the contact densities to be determined by the solution. With these conditions the linearized profiles agree quantitatively with the numerical mean-field solution except very close to the wall, the agreement deteriorating only in the immediate vicinity of the λ\lambda-surface.

These results extend the field-theoretical description of one-component Yukawa fluids [32, 39], including the competing-interaction (SALR) case [32], to two-component mixtures, and complement recent studies of self-assembling mixtures near surfaces and in monolayers [9, 10]. Natural extensions of the present work include the linearized treatment of the full 3Y+2Y model at hard and attractive walls, for which the seventh-degree characteristic equation must be solved numerically.

Acknowledgements

IK and TP acknowledge financial support from the National Research Foundation of Ukraine (project No. 2020.02/0317). OP acknowledges the support of the Ministry of Education and Science of Ukraine (Agreement No. PH/16-2023). Computer time for the reported simulations was provided by the Interdisciplinary Center for Computer Simulations (Lviv), which is supported by the NRFU Grant No. 2023.05/0019.

Derivation of the linearized density profiles

Here we give the details of the calculation summarized in § 5. The scheme is a two-component generalization of the one-component derivation of [32], and is carried out for the simplified model (20), in which the solvent and cross interactions are described by a single Yukawa term identical to the first term of v11v_{11}, i.e., with the same amplitude A11(1)A_{11}^{(1)} and range α1\alpha_1. Throughout we use the abbreviations of § 5.1, i.e., αk\alpha_k and χ1k2\chi_{1k}^{2} (k=1,2,3k = 1, 2, 3) for the ranges and coupling parameters of the three Yukawa terms of the main component, and χs2\chi_{s}^{2} for the contribution of the solvent through the shared term. Compared with the one-component derivation of [32], the two-component character enters in three places: the solvent coupling χs2\chi_{s}^{2}, which appears in the (1,1)(1,1) entry of A\mathbf{A} in equation (24) and in the first boundary condition below; the field superposition E~=E11(1)+E12(1)\tilde{E} = E_{11}^{(1)} + E_{12}^{(1)} that closes the reduced system; and the species-resolved profiles, (31)–(32), in which the main component collects all three Yukawa fields while the solvent couples only to the shared one. The remaining algebra parallels the one-component case, and is quoted compactly below.

Roots of the characteristic cubic

The eigenvalues of the matrix A\mathbf{A} in equation (24), written as Λ0=λ02\Lambda_0 = \lambda_0^2 and Λ1,2=λ1,22\Lambda_{1,2} = \lambda_{1,2}^2, are the roots of the cubic (25) with the coefficients (26). Introducing

Δ0=b2−3c,Δ1=2b3−9bc+27d,C=[Δ1+Δ12−4Δ032]1/3(33)\Delta_0 = b^2 - 3c,\qquad\Delta_1 = 2b^3 - 9bc + 27d,\qquad C = \left[\frac{\Delta_1 + \sqrt{\Delta_1^2 - 4\Delta_0^3}}{2}\right]^{1/3} \tag*{(33)}

where CC is taken to be the real positive cube root, the real eigenvalue and the complex-conjugate pair follow from the standard (Cardano) solution as

λ02=13(b+C+Δ0C),λ1,22=M1±iM2(34)\lambda_0^2 = \frac{1}{3}\left(b+C+\frac{\Delta_0}{C}\right), \qquad\lambda_{1,2}^2=M_1\pm iM_2 \tag*{(34)}
M1=13(b−C2−Δ02C),M2=36(C−Δ0C).(35)M_1=\frac{1}{3}\left(b-\frac{C}{2}-\frac{\Delta_0}{2C}\right), \qquad M_2=\frac{\sqrt{3}}{6}\left(C-\frac{\Delta_0}{C}\right). \tag*{(35)}

For the parameter range considered here λ02\lambda_0^2 is real and positive, while λ1,22\lambda_{1,2}^2 are complex conjugate. Writing λ1,2=λ±iμ\lambda_{1,2}=\lambda\pm i\mu one obtains

λ0=λ02,λ=12[M12+M22+M1]1/2,μ=sgn⁡(M2)2[M12+M22−M1]1/2,(36)\lambda_0=\sqrt{\lambda_0^2}, \qquad\lambda=\frac{1}{\sqrt{2}}\left[\sqrt{M_1^2+M_2^2}+M_1\right]^{1/2}, \qquad\mu=\frac{\operatorname{sgn}(M_2)}{\sqrt{2}}\left[\sqrt{M_1^2+M_2^2}-M_1\right]^{1/2}, \tag*{(36)}

so that λ0\lambda_0 and λ\lambda are the (positive) decay coefficients of monotonic and oscillatory decay, respectively, and 2π/μ2\pi/\mu is the oscillation wavelength.

Field amplitudes

Besides the superposition E~(z)≡E11(1)(z)+E12(1)(z)\widetilde{E}(z)\equiv E_{11}^{(1)}(z)+E_{12}^{(1)}(z), the two remaining fields associated with the second and third Yukawa terms of the main component are denoted E12(z)E_{12}(z) and E13(z)E_{13}(z). Retaining only the contributions that decay as z→∞z\to\infty, their general form is

E~(z)=c10e−λ0z+(r11+ir12)e(−λ+iμ)z+(r11−ir12)e(−λ−iμ)z,E12(z)=c20e−λ0z+(r21+ir22)e(−λ+iμ)z+(r21−ir22)e(−λ−iμ)z,E13(z)=c30e−λ0z+(r31+ir32)e(−λ+iμ)z+(r31−ir32)e(−λ−iμ)z.(37)\begin{aligned} \widetilde{E}(z)&=c_{10}e^{-\lambda_0z}+(r_{11}+ir_{12})e^{(-\lambda+i\mu)z}+(r_{11}-ir_{12})e^{(-\lambda-i\mu)z},\\ E_{12}(z)&=c_{20}e^{-\lambda_0z}+(r_{21}+ir_{22})e^{(-\lambda+i\mu)z}+(r_{21}-ir_{22})e^{(-\lambda-i\mu)z},\\ E_{13}(z)&=c_{30}e^{-\lambda_0z}+(r_{31}+ir_{32})e^{(-\lambda+i\mu)z}+(r_{31}-ir_{32})e^{(-\lambda-i\mu)z}. \tag*{(37)} \end{aligned}

Because equation (5.3)-(5.5) is homogeneous, only three of these amplitudes are independent, and we take c10c_{10}, r11r_{11}, r12r_{12}. Substituting (A.5) into E′′=AEE''=AE and equating the coefficients of the same zz-dependent function separately in their real and imaginary parts gives two decoupled linear systems. For the oscillatory terms, with κˉ12=λ2−μ2−α12−ϰ112\bar{\kappa}_1^2=\lambda^2-\mu^2-\alpha_1^2-\varkappa_{11}^2 and κˉj2=λ2−μ2−αj2−ϰ1j2\bar{\kappa}_j^2=\lambda^2-\mu^2-\alpha_j^2-\varkappa_{1j}^2 (j=2,3j=2,3),

κˉ12r11+2λμr12−ϰ112(r21+r31)=0,κˉ12r12−2λμr11−ϰ112(r22+r32)=0,κˉ22r21+2λμr22−ϰ122(r11+r31)=0,κˉ22r22−2λμr21−ϰ122(r12+r32)=0,κˉ32r31+2λμr32−ϰ132(r11+r21)=0,κˉ32r32−2λμr31−ϰ132(r12+r22)=0.(38)\begin{aligned} \bar{\kappa}_1^2r_{11}+2\lambda\mu r_{12}-\varkappa_{11}^2(r_{21}+r_{31})&=0, & \bar{\kappa}_1^2r_{12}-2\lambda\mu r_{11}-\varkappa_{11}^2(r_{22}+r_{32})&=0,\\ \bar{\kappa}_2^2r_{21}+2\lambda\mu r_{22}-\varkappa_{12}^2(r_{11}+r_{31})&=0, & \bar{\kappa}_2^2r_{22}-2\lambda\mu r_{21}-\varkappa_{12}^2(r_{12}+r_{32})&=0,\\ \bar{\kappa}_3^2r_{31}+2\lambda\mu r_{32}-\varkappa_{13}^2(r_{11}+r_{21})&=0, & \bar{\kappa}_3^2r_{32}-2\lambda\mu r_{31}-\varkappa_{13}^2(r_{12}+r_{22})&=0. \tag*{(38)} \end{aligned}

Any two of these pairs suffice. Solving them expresses the remaining oscillatory amplitudes linearly through r11r_{11} and r12r_{12}, rlm=Klm11r11+Klm12r12r_{lm}=K_{lm}^{11}r_{11}+K_{lm}^{12}r_{12} (l=2,3l=2,3), while matching the monotonic terms gives c20=K20c10c_{20}=K_{20}c_{10} and c30=K30c10c_{30}=K_{30}c_{10}. The coefficients Klm11K_{lm}^{11}, Klm12K_{lm}^{12}, K20K_{20} and K30K_{30} are rational functions of {αk,ϰ1k2,λ,μ}\{\alpha_k,\varkappa_{1k}^2,\lambda,\mu\}; their explicit form is obtained exactly as in the one-component case [32].

Potentials, boundary conditions and coefficients

Integrating the fields once yields the mean-field potentials

Vi(z)=Vib+rˉi0e−λ0z+e−λz[(λrˉi1−μrˉi2)cos⁡μz−(μrˉi1+λrˉi2)sin⁡μz],(39)V_i(z)=V_{ib}+\bar{r}_{i0}e^{-\lambda_0z}+e^{-\lambda z}\left[(\lambda\bar{r}_{i1}-\mu\bar{r}_{i2})\cos\mu z-(\mu\bar{r}_{i1}+\lambda\bar{r}_{i2})\sin\mu z\right], \tag*{(39)}

where rˉi0=+ci0/λ0\bar{r}_{i0}=+c_{i0}/\lambda_0 and rˉij=2rij/(λ2+μ2)\bar{r}_{ij}=2r_{ij}/(\lambda^2+\mu^2) (i=1,2,3i=1,2,3, j=1,2j=1,2); both follow from Vi(z)=Vib+∫z∞Ei(t) dtV_i(z)=V_{ib}+\int_z^\infty E_i(t)\,\mathrm{d}t, which is the integral of (3.12) with Vi→VibV_i\to V_{ib} as z→∞z\to\infty. The relations of section A.2 carry over to the barred amplitudes, so that rˉ20=K20rˉ10\bar{r}_{20}=K_{20}\bar{r}_{10}, rˉ30=K30rˉ10\bar{r}_{30}=K_{30}\bar{r}_{10} and rˉlm=Klm11rˉ11+Klm12rˉ12\bar{r}_{lm}=K_{lm}^{11}\bar{r}_{11}+K_{lm}^{12}\bar{r}_{12}. The three remaining unknowns rˉ10\bar{r}_{10}, rˉ11\bar{r}_{11}, rˉ12\bar{r}_{12} are fixed by the boundary conditions at the wall. The conditions (5.9) used in this work are treated in section A.4. For comparison we first record the earlier prescription, in which the contact theorem is imposed on each potential separately. Setting z=0z = 0 in the linearized relations gives Vib−Vi(0)=χi2/(2αi2)V_{ib} - V_i(0) = \chi_i^2/(2\alpha_i^2), i.e.,

−χ112+χs22α12=rˉ10+λrˉ11−μrˉ12,χ1222α22=rˉ20+λrˉ21−μrˉ22,-\frac{\chi_{11}^{2}+\chi_s^{2}}{2\alpha_1^{2}}=\bar r_{10}+\lambda\bar r_{11}-\mu\bar r_{12},\qquad \frac{\chi_{12}^{2}}{2\alpha_2^{2}}=\bar r_{20}+\lambda\bar r_{21}-\mu\bar r_{22},
−χ1322α32=rˉ30+λrˉ31−μrˉ32.(40)-\frac{\chi_{13}^{2}}{2\alpha_3^{2}}=\bar r_{30}+\lambda\bar r_{31}-\mu\bar r_{32}. \tag*{(40)}

Together with the amplitude relations of section A.2, the three equations (A.8) form a closed linear system for rˉ10\bar r_{10}, rˉ11\bar r_{11}, rˉ12\bar r_{12}, from which these coefficients are obtained in closed form. Note that the two-component nature of the problem is confined to the first of equations (A.8), through the term χs2\chi_s^2. The solution otherwise follows the same algebra as the one-component case [32]. The amplitudes used for the profiles (5.12) and (5.13) of the main text are those of section A.4, obtained from the exact conditions (5.9) instead.

Amplitudes for the corrected boundary conditions

We collect here the algebra behind the conditions (5.9) introduced in section 5.2. Setting z=0z = 0 in (A.7) gives Vi(0)−Vib=rˉi0+λrˉi1−μrˉi2V_i(0)-V_{ib}=\bar r_{i0}+\lambda\bar r_{i1}-\mu\bar r_{i2}, which is what produced (A.8), while differentiating (A.7) at z=0z=0 gives

Vi′(0)=−λ0rˉi0−(λ2+μ2)rˉi1,i.e.Ei(0)=λ0rˉi0+(λ2+μ2)rˉi1.(41)V_i'(0)=-\lambda_0\bar r_{i0}-(\lambda^2+\mu^2)\bar r_{i1},\qquad\text{i.e.}\qquad E_i(0)=\lambda_0\bar r_{i0}+(\lambda^2+\mu^2)\bar r_{i1}. \tag*{(41)}

Substituting both into (5.9) and using V1b=(χ112+χs2)/α12V_{1b}=(\chi_{11}^2+\chi_s^2)/\alpha_1^2, V2b=χ122/α22V_{2b}=\chi_{12}^2/\alpha_2^2 and V3b=χ132/α32V_{3b}=\chi_{13}^2/\alpha_3^2 yields the system (5.10) of the main text, which replaces (A.8).

Eliminating rˉ20\bar r_{20}, rˉ30\bar r_{30} and rˉlm\bar r_{lm} by the amplitude relations of section A.2 reduces (5.10) to a closed 3×33\times3 system for the three remaining unknowns,

airˉ10+birˉ11+cirˉ12=di,i=1,2,3,(42)a_i\bar r_{10}+b_i\bar r_{11}+c_i\bar r_{12}=d_i,\qquad i=1,2,3, \tag*{(42)}

with Wi=λ2+μ2+αiλW_i=\lambda^2+\mu^2+\alpha_i\lambda and

ai=(λ0+αi)Ki0,bi=WiKi111−αiμKi211,a_i=(\lambda_0+\alpha_i)K_{i0},\qquad b_i=W_iK_{i1}^{11}-\alpha_i\mu K_{i2}^{11},
ci=WiKi112−αiμKi212,di=−χi2αi.(43)c_i=W_iK_{i1}^{12}-\alpha_i\mu K_{i2}^{12},\qquad d_i=-\frac{\chi_i^2}{\alpha_i}. \tag*{(43)}

where the conventions K10=K1111=K1212=1K_{10}=K_{11}^{11}=K_{12}^{12}=1, K1112=K1211=0K_{11}^{12}=K_{12}^{11}=0 complete the set of amplitude coefficients of section A.2. For comparison, the earlier conditions (A.8) are the same system (A.10) with ai=Ki0a_i=K_{i0}, bi=λKi111−μKi211b_i=\lambda K_{i1}^{11}-\mu K_{i2}^{11}, ci=λKi112−μKi212c_i=\lambda K_{i1}^{12}-\mu K_{i2}^{12} and di=−χi2/(2αi2)d_i=-\chi_i^2/(2\alpha_i^2).

Since (A.10) is a 3×33\times3 linear system with explicit coefficients, Cramer's rule gives the amplitudes in closed form,

rˉ10=D1D,rˉ11=D2D,rˉ12=D3D.(44)\bar r_{10}=\frac{D_1}{D},\qquad\bar r_{11}=\frac{D_2}{D},\qquad\bar r_{12}=\frac{D_3}{D}. \tag*{(44)}

where DD is the determinant of the coefficient matrix of (A.10) and D1D_1, D2D_2, D3D_3 are obtained from it by replacing its first, second and third column, respectively, by the column (d1,d2,d3)T(d_1,d_2,d_3)^{\mathsf T},

D=∣a1b1c1a2b2c2a3b3c3∣,D1=∣d1b1c1d2b2c2d3b3c3∣,D2=∣a1d1c1a2d2c2a3d3c3∣,D3=∣a1b1d1a2b2d2a3b3d3∣.(45)D=\begin{vmatrix}a_1&b_1&c_1\\a_2&b_2&c_2\\a_3&b_3&c_3\end{vmatrix},\qquad D_1=\begin{vmatrix}d_1&b_1&c_1\\d_2&b_2&c_2\\d_3&b_3&c_3\end{vmatrix},\qquad D_2=\begin{vmatrix}a_1&d_1&c_1\\a_2&d_2&c_2\\a_3&d_3&c_3\end{vmatrix},\qquad D_3=\begin{vmatrix}a_1&b_1&d_1\\a_2&b_2&d_2\\a_3&b_3&d_3\end{vmatrix}. \tag*{(45)}

The remaining amplitudes follow from the relations of section A.2, and the profiles from (5.12)–(5.13) as before. Because the cubic (25) is itself solvable in radicals through (A.2)–(A.4), the entire solution remains explicit. Setting z=0z = 0 in (5.12)–(5.13), the contact densities that result are

ρ1(0+)ρ1b=1−∑i=13(rˉi0+λrˉi1−μrˉi2),ρ2(0+)ρ2b=1−(rˉ10+λrˉ11−μrˉ12)(46)\frac{\rho_1(0^+)}{\rho_{1b}} = 1 - \sum_{i=1}^{3} \left(\bar{r}_{i0} + \lambda\bar{r}_{i1} - \mu\bar{r}_{i2}\right), \qquad\frac{\rho_2(0^+)}{\rho_{2b}} = 1 - \left(\bar{r}_{10} + \lambda\bar{r}_{11} - \mu\bar{r}_{12}\right) \tag*{(46)}

neither of which is imposed.

References

  1. [1]Sweatman M. B., Lue L., The giant SALR cluster fluid: A review, Adv. Theory Simul., 2019, 2, 19000025, doi:10.1002/adts.201900025.DOI
  2. [2]von Konigslow K., Cardenas-Mendez E. D., Thompson R. B., Rasmussen K. Ø., The self-assembly of particles with isotropic interactions, J. Phys.: Condens. Matter, 2013, 25, 325101, doi:10.1088/0953-8984/25/32/325101.DOI
  3. [3]Ferreiro-Rangel C. A., Sweatman M. B., Cluster formation in binary fluids with competing short-range and long-range interactions, Mol. Phys., 2018, 116, 3231–3244, doi:10.1080/00268976.2018.1489082.
  4. [4]Tan J., Afify N. D., Ferreiro-Rangel C. A., Fan X., Sweatman M. B., Cluster formation in symmetric binary SALR mixtures, J. Chem. Phys., 2021, 154, 074504, doi:10.1063/5.0036046.DOI
  5. [5]Tan J., Sweatman M. B., Secondary nucleation in symmetric binary SALR mixtures, Phys. Chem. Chem. Phys., 2024, 26, 17057, doi:10.1039/D3CP05765H.DOI
  6. [6]Munaò G., Prestipino S., Bomont J.-M., Costa D., Clustering in mixtures of SALR particles and hard spheres with cross attraction, J. Phys. Chem. B, 2022, 126, 2027, doi:10.1021/acs.jpcb.1c09758.DOI
  7. [7]Munaò G., Costa D., Malescio G., Bomont J.-M., Prestipino S., Competition between clustering and phase separation in binary mixtures containing SALR particles, Soft Matter, 2022, 18, 6453–6464, doi:10.1039/d2sm00944g.DOI
  8. [8]Patsahan O., Litniowski M., Ciach A., Self-assembly in mixtures with competing interactions, Soft Matter, 2021, 17, 2883, doi:10.1039/D0SM02072A.DOI
  9. [9]Patsahan O., Meyra A., Ciach A., Effect of a confining surface on a mixture with spontaneous inhomogeneities, J. Mol. Liq., 2022, 363, 119844, doi:10.1016/j.molliq.2022.119844.
  10. [10]Patsahan O., Meyra A., Ciach A., Spontaneous pattern formation in monolayers of binary mixtures with competing interactions, Soft Matter, 2024, 20, 1410, doi:10.1039/D3SM01537H.DOI
  11. [11]Prestipino S., Pini D., Costa D., Malescio G., Munaò G., A density functional theory and simulation study of stripe phases in symmetric colloidal mixtures, J. Chem. Phys., 2023, 159, No. 20, 204902, doi:10.1063/5.0177209.DOI
  12. [12]Pini D., Prestipino S., Costa D., Malescio G., Munaò G., A density-functional theory study of vapor, liquid, and mesophases in a symmetric, hard-sphere mixture with cross attraction at non-equimolar concentrations, J. Chem. Phys., 2025, 163, No. 16, 164905, doi:10.1063/5.0291013.DOI
  13. [13]Sweatman M. B., Insall R., Assembly of the actin catalyst WASP by giant SALR cluster formation, Advanced Theory and Simulations, 2019, 2, 1800203, doi:10.1002/adts.201800203.DOI
  14. [14]Archer A. J., Wilding N. B., Phase behavior of a fluid with competing attractive and repulsive interactions, Phys. Rev. E, 2007, 76, 031501, doi:10.1103/PhysRevE.76.031501.DOI
  15. [15]Ciach A., Universal sequence of ordered structures obtained from mesoscopic description of self-assembly, Phys. Rev. E, 2008, 78, No. 6, 061505, doi:10.1103/physreve.78.061505.DOI
  16. [16]Ciach A., Góźdź W., Mesoscopic description of network-forming clusters of weakly charged colloids, Condens. Matter Phys., 2010, 13, No. 2, 23603, doi:10.5488/cmp.13.23603.DOI
  17. [17]Ciach A., Pękalski J., Góźdź W. T., Origin of similarity of phase diagrams in amphiphilic and colloidal systems with competing interactions, Soft Matter, 2013, 9, No. 27, 6301, doi:10.1039/c3sm50668a.DOI
  18. [18]Zhuang Y., Zhang K., Charbonneau P., Equilibrium phase behavior of a continuous-space microphase former, Phys. Rev. Lett., 2016, 116, 098301, doi:10.1103/PhysRevLett.116.098301.DOI
  19. [19]Edelmann M., Roth R., Gyroid phase of fluids with spherically symmetric competing interactions, Phys. Rev. E, 2016, 93, No. 6, 062146, doi:10.1103/physreve.93.062146.DOI
  20. [20]Pini D., Parola A., Pattern formation and self-assembly driven by competing interactions, Soft Matter, 2017, 13, 9259, doi:10.1039/C7SM02125A.DOI
  21. [21]Zhuang Y., Charbonneau P., Recent advances in the theory and simulation of model colloidal microphase formers, J. Phys. Chem. B, 2016, 120, 7775, doi:10.1021/acs.jpcb.6b05471.DOI
  22. [22]Bretonnet J.-L., Competing interactions in colloidal suspensions, AIMS Mater. Sci., 2019, 6, 509, doi:10.3934/matersci.2019.4.509.DOI
  23. [23]Costa D., Munaò G., Bomont J.-M., Malescio G., Palatella A., Prestipino S., Microphase versus macrophase separation in the square-well fluid: A theoretical and computational study, Phys. Rev. E, 2023, 108, 034602, doi:10.1103/PhysRevE.108.034602.DOI
  24. [24]Serna H., Martín-Roca J., Meyra A. G., Noya E. G., Activity enhances transport while competing interactions preserve structure in colloidal microphase formers, J. Chem. Phys., 2026, 165, 054902, doi:10.1063/5.0343447.DOI
  25. [25]Leunissen M. E., Christova C. G., Hyninnen A.-P., Royall C. P., Campbell A. I., Imhof A., Dijkstra M., van Roij R., van Blaaderen A., Ionic colloidal crystals of oppositely charged particles, Nature, 2005, 437, 235, doi:10.1038/nature03946.DOI
  26. [26]Campbell A. I., Anderson V. J., van Duijneveldt J. S., Bartlett P., Dynamical arrest in attractive colloids: The effect of long-range repulsion, Phys. Rev. Lett., 2005, 94, 208301, doi:10.1103/PhysRevLett.94.208301.DOI
  27. [27]Dhumal U., Cross-interaction softness as a route to microphase separation in binary colloidal systems, 2026, preprint, Preprint arXiv:2601.00099.
  28. [28]Litniewski K., Ciach M., Adsorption in mixtures with competing interactions, Molecules, 2021, 26, 4532, doi:10.3390/molecules26154532.DOI
  29. [29]Litniewski M., Gόdź W. T., Ciach A., Adsorption on a spherical colloidal particle from a mixture of nanoparticles with competing interactions, Molecules, 2024, 29, 3170, doi:10.3390/molecules29133170.DOI
  30. [30]Li N., Li J., Qing L., Ma S., Li Y., Li B., Self-assembly of colloids with competing interactions confined in spheres, Soft Matter, 2024, 20, 304, doi:10.1039/D3SM01227A.DOI
  31. [31]Serna H., Noya E. G., Gόdź W. T., Confinement of colloids with competing interactions in ordered porous materials, J. Phys. Chem. B, 2020, 124, 10567, doi:10.1021/acs.jpcb.0c08148.DOI
  32. [32]Kravtsiv I., Patsahan T., Holovko M., Di Caprio D., Soft core fluid with competing interactions at a hard wall, J. Mol. Liq., 2022, 362, 119652, doi:10.1016/j.molliq.2022.119652.DOI
  33. [33]Di Caprio D., Stafiej J., Badiali J. P., A field theory study of the effect of specific interactions in ionic systems: A simple model, J. Chem. Phys., 1998, 108, 8572, doi:10.1063/1.476286.DOI
  34. [34]Di Caprio D., Stafiej J., A field theoretical approach to the Debye-Hückel electrolyte differential capacitance in a slab, Condens. Matter Phys., 2001, 4, 715, doi:10.5488/CMP.4.4.715.DOI
  35. [35]Di Caprio D., Stafiej J., Badiali J. P., Field theoretical approach to inhomogeneous ionic systems: thermodynamic consistency with the contact theorem, Gibbs adsorption and surface tension, Mol. Phys., 2003, 101, 2545, doi:10.1080/0026897031000154293.DOI
  36. [36]Di Caprio D., Badiali J. P., A formally exact field theory for classical systems at equilibrium, J. Phys. A: Math. Theor., 2008, 41, 125401, doi:10.1088/1751-8113/41/12/125401.DOI
  37. [37]Holovko M. F., Kravtsiv I. Y., Soviak E. M., Spatially confined system interacting with Yukawa potential, Condens. Matter Phys., 2009, 12, 137, doi:10.5488/CMP.12.2.137.
  38. [38]Di Caprio D., Stafiej J., Holovko M., Kravtsiv I., Yukawa fluid at a hard wall: field theory description, Mol. Phys., 2011, 109, 695, doi:10.1080/00268976.2010.547524.DOI
  39. [39]Kravtsiv I., Patsahan T., Holovko M., Di Caprio D., Two-Yukawa fluid at a hard wall: Field theory treatment, J. Chem. Phys., 2015, 142, 194708, doi:10.1063/1.4921242.DOI
  40. [40]Di Caprio D., Kravtsiv I., Patsahan T., Holovko M., Oscillating Yukawa fluid at a hard wall: field theory description, Mol. Phys., 2016, 114, 2500, doi:10.1080/00268976.2016.1174787.DOI
  41. [41]Holovko M., Di Caprio D., Kravtsiv I., Maier-Saupe nematogenic fluid: field theoretical approach, Condens. Matter Phys., 2011, 14, 33605, doi:10.5488/CMP.14.3.33605.DOI
  42. [42]Kravtsiv I., Holovko M., Di Caprio D., Maier-Saupe nematogenic fluid interacting with an isotropic and an anisotropic Yukawa potentials: field theory description, Mol. Phys., 2013, 111, 1023, doi:10.1080/00268976.2012.762615.DOI
  43. [43]Holovko M., Kravtsiv I., Di Caprio D., Nematic fluid at a hard wall in the mean field approximation, Condens. Matter Phys., 2013, 16, 14002, doi:10.5488/CMP.16.14002.DOI
  44. [44]Holovko M., Patsahan T., Kravtsiv I., Di Caprio D., Maier-Saupe nematogenic fluid with isotropic Yukawa repulsion at a hard wall: Mean field approximation, Condens. Matter Phys., 2016, 19, 13608, doi:10.5488/CMP.19.13608.DOI
  45. [45]Evans R., The nature of the liquid-vapour interface and other topics in the statistical mechanics of non-uniform, classical fluids, Adv. Phys., 1979, 28, 143, doi:10.1080/00018737900101365.DOI
  46. [46]Holovko M., Vlachy V., Di Caprio D., On the contact conditions for the density and charge profiles in the theory of electrical double layer: From planar to spherical and cylindrical geometry, J. Mol. Liq., 2023, 371, 121040, doi:10.1016/j.molliq.2022.121040.DOI
  47. [47]Allen M. P., Tildesley D. J., Computer simulation of liquids, Oxford University Press, 2017.
  48. [48]Thompson A. P., Aktulga H. M., Berger R., Bolintineanu D. S., Brown W. M., Crozier P. S., in’t Veld P. J., Kohlmeyer A., Moore S. G., Nguyen T. D., et al., LAMMPS – a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales, Comput. Phys. Commun., 2022, 271, 108171.DOI

Paper details

vlnc:W_wyf9nmtq
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents