The roles of bulk and surface thermodynamics in the selective adsorption of a confined azeotropic mixture

Katie L. Y. Zhou, Anna T. Bui, Stephen J. Cox · 2026

Read The roles of bulk and surface thermodynamics in the selective adsorption of a confined azeotropic mixture by Katie L. Y. Zhou, Anna T. Bui, Stephen J. Cox (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-21591.txt.

Introduction

Katie L. Y. Zhou, Anna T. Bui, and Stephen J. Cox study how a trapped fluid mixture sorts itself near walls. Some mixtures form an azeotrope, a blend that simple boiling cannot separate. Trapping such a blend in a tiny space may sort it in a way bulk methods cannot.

The authors use a density theory boosted by machine learning on a simple model mixture. They weigh the roles of the bulk fluid and the surfaces that confine it. Selective adsorption at walls could offer a new route to purify stubborn blends. That matters for chemical plants, where separation eats a huge share of energy.

Catalog record

CREATORKatie L. Y. Zhou, Anna T. Bui, Stephen J. Cox
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-21591

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The roles of bulk and surface thermodynamics in the selective adsorption of a confined azeotropic mixture Katie L. Y. Zhou, Anna T. Bui, Stephen J. Cox Abstract Fluid mixtures that exhibit an azeotrope cannot be purified by simple bulk distillation. Consequently, there is strong motivation to understand the behavior of azeotropic mixtures under confinement. We address this problem using a machine-learning-enhanced classical density functional theory applied to a binary Lennard–Jones mixture that exhibits azeotropic phase behavior. As proof-of-principle of a “train once, learn many” strategy, our approach combines a neural functional trained on a single-component repulsive reference system with a mean-field treatment of attractive interactions, derived within the framework of hyperdensity functional theory (hyper-DFT). The theory faithfully describes capillary condensation and results from grand canonical Monte Carlo simulations. Moreover, by taking advantage of a known accurate equation of state, the theory we present well-describes bulk thermodynamics by construction. Exploiting the computational efficiency of hyper-DFT, we systematically evaluate adsorption selectivity across a wide range of compositions, pressures, temperatures, and wall–fluid affinities. In cases where the wall–fluid interaction is the same for both species, we find that the pore becomes completely unselective at the bulk azeotropic composition. Strikingly, this unselective point persists far from liquid–vapor coexistence, including in the supercritical regime. Analysis of the bulk equation of state across a wide range of thermodynamic state points shows that the azeotropic composition coincides with equal partial molar volumes and an extremum in the isothermal compressibility. A complementary thermodynamic analysis demonstrates that unselective adsorption corresponds to an aneotrope (a point of zero relative adsorption) and an extremum in the interfacial free energy. We also find that the two interfaces of the slit pore behave independently down to remarkably small slits. I Introduction A detailed understanding of the thermophysical properties of fluid mixtures under confinement is crucial for a wide range of applications, including industrial separation processes, gas storage, and healthcare 79 ; 29 ; 11 . Specific examples span oil recovery 4 , hydrogen storage 14 ; 13 , and microfluidic systems 100 ; 44 . From a theoretical standpoint, molecular simulations are widely regarded as the method of choice 97 ; 36 ; 37 ; 26 ; 92 ; 78 , offering microscopic insight that is often inaccessible to experiment alone 43 ; 5 . However, traditional approaches rooted in liquid state theory are experiencing a renaissance 6 ; 98 . In particular, classical density functional theory (cDFT) is attracting renewed interest in physical chemistry, driven by the ability of machine learning (ML) to deliver highly accurate and computationally efficient approximations 72 ; 9 ; 65 ; 74 ; 8 ; 1 ; 12 ; 99 ; 50 . In this article, our aim is to demonstrate a recently proposed ML-based approach to cDFT for a simple Lennard–Jones (LJ) fluid mixture. While similar ideas have been explored recently in Ref. 65 , our approach differs in two key respects. First, methodologically, we train the ML model on a repulsive single-component reference system and incorporate attractive interactions through a simple mean-field treatment—an approach that enhances transferability among systems. Second, from an application standpoint, we focus on a binary mixture with asymmetric interactions, which exhibits azeotropic phase behavior. As we will see, the existence of an azeotrope appears to strongly influence the fluid’s behavior—both in bulk and under confinement—across a broad range of thermodynamic conditions. The potential appeal of cDFT over molecular simulations becomes immediately evident when considering its fundamental formalism 23 ; 33 . The central object in cDFT is the grand potential functional \varOmega_{V}\big([\{\varrho_{\alpha}\}],T\big)=\mathcal{F}^{(\rm id)}_{\text{intr}}\big([\{\varrho_{\alpha}\}],T\big)+\mathcal{F}^{(\rm ex)}_{\text{intr}}\big([\{\varrho_{\alpha}\}],T\big)\\ +\sum_{\alpha}\int\!\text{d}\bm{\mathit{r}}\,\varrho_{\alpha}(\bm{\mathit{r}})\big[V_{\alpha}(\bm{\mathit{r}})-\mu_{\alpha}\big], (1) where \mathcal{F}^{(\rm id)}_{\text{intr}} and \mathcal{F}^{(\rm ex)}_{\text{intr}} are, respectively, the ideal and excess intrinsic Helmholtz free energy functionals. A particle of species \alpha , with chemical potential \mu_{\alpha} , experiences an external one-body potential V_{\alpha} , while \varrho_{\alpha} denotes its average one-body density, though not necessarily at equilibrium. Functional minimization of \varOmega_{V} with respect to \varrho_{\alpha} yields the equilibrium one-body density of species \alpha , \rho_{\alpha} , which satisfies the Euler–Lagrange equation, \Lambda^{3}_{\alpha}\rho_{\alpha}(\bm{\mathit{r}})=\exp\left(-\beta\big[V_{\alpha}(\bm{\mathit{r}})-\mu_{\alpha}\big]+c^{(1)}_{\alpha}(\bm{\mathit{r}};[\{\rho_{\alpha}\}],T)\right) (2) with \Lambda_{\alpha} denoting the thermal de Broglie wavelength of species \alpha , \beta=1/k_{\rm B}T ( k_{\rm B} is the Boltzmann constant), and c^{(1)}_{\alpha}(\bm{\mathit{r}};[\{\varrho_{\alpha}\}],T)=-\frac{\delta\beta\mathcal{F}^{(\rm ex)}_{\rm intr}([\{\varrho_{\alpha}\}],T)}{\delta\varrho_{\alpha}(\bm{\mathit{r}})}, (3) is the one-body direct correlation functional. In the context of the Euler–Lagrange equation, c^{(1)}_{\alpha} acts as a self-consistent one-body potential that accounts for the effects of correlations on the structure of the fluid. With the equilibrium densities from Eq. 2 , the grand potential of the system readily follows: \Omega=\varOmega_{V}([\{\rho_{\alpha}\}],T) . The procedure for obtaining the structure and thermodynamics of the system can, therefore, be succinctly put: minimize a functional to obtain the equilibrium structure, then evaluate the functional at equilibrium to obtain the thermodynamic potential of the system. This is a far simpler operation than explicitly sampling the many-body equilibrium distribution function with molecular simulations. What, then, has prevented cDFT from becoming the method of choice for understanding the equilibrium properties of fluids? The answer is straightforward. While a rigorous theoretical framework, in practice cDFT relies upon approximations to \mathcal{F^{(\rm ex)}_{\rm intr}} . (Only for hard rods in a single dimension is \mathcal{F^{(\rm ex)}_{\rm intr}} known exactly 56 .) In the case of hard spheres, very good approximations founded on Rosenfeld’s fundamental measure theory (FMT) have been available for several decades 68 ; 70 ; 69 ; 34 . When combined with an ad hoc mean-field treatment of attractive interactions, cDFT has been shown to be a powerful practical method for understanding the physics of simple liquids. In particular, the physics of capillarity 21 ; 22 , surface drying and wetting 24 ; 25 , and solvophobicity 17 have been extensively studied in single-component systems, owing to the ability of the mean-field approximation to qualitatively capture liquid–gas phase coexistence. While less intensely studied compared to single-component fluids, such a mean-field approach has also provided qualitative insight into the behavior of simple mixtures 88 ; 46 ; 45 ; 87 ; 39 ; 60 ; 18 ; 53 , though the richer phase behavior of multicomponent fluids—including liquid–liquid coexistence, azeotropy and heteroazeotropy 91 —pushes the limits of analytical approximations for \mathcal{F}^{(\mathrm{ex})}_{\mathrm{intr}} . Going beyond hard spheres and a mean-field treatment for attractive interactions has, however, proven a major challenge. Naturally, several groups have recently turned to ML. For example, early works by Oettel and co-workers focused on one-dimensional systems 50 ; 77 and anisotropic patchy particles 81 , while Cats et al. used ML to improve the standard mean-field approximation for the three-dimensional LJ fluid 12 . One ML scheme that is gaining significant traction is “neural functional theory” 72 introduced by Sammüller et al. The basis of this physics-informed approach is the Euler–Lagrange equation (Eq. 2 ). For known \{V_{\alpha}\} and \{\mu_{\alpha}\} , with \{\rho_{\alpha}\} obtained from grand canonical Monte Carlo (GCMC) simulation, \{c^{(1)}_{\alpha}\} at equilibrium are determined by rearranging Eq. 2 : c^{(1)}_{\alpha}(\bm{\mathit{r}};[\{\rho_{\alpha}\}],T)=\ln\Lambda_{\alpha}^{3}\rho_{\alpha}(\bm{\mathit{r}})+\beta\big(V_{\alpha}(\bm{\mathit{r}})-\mu_{\alpha}\big). (4) By obtaining \{c^{(1)}_{\alpha}\} from many different GCMC simulations with different \{V_{\alpha}\} and \{\mu_{\alpha}\} , a training set is established to learn the local functional dependence of c^{(1)}_{\alpha} on \{\varrho_{\alpha}\} with a neural network. Originally developed for the single-component hard-sphere fluid—where it even outperformed FMT-based functionals 72 —the neural cDFT approach has since been extended to more complex systems. For example,

Read the complete text