Solvation in atomic liquids: connection between Gaussian field theory and density functional theory

Solvation in atomic liquids: connection between Gaussian field theory and density functional theory

For the problem of molecular solvation, formulated as a liquid submitted to the external potential field created by a molecular solute of arbitrary shape dissolved in that solvent, we draw a connection between the Gaussian field theory derived by David Chandler [Phys. Rev. E, 1993, 48, 2898] and c...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Datum:2017
Hauptverfasser: Sergiievskyi, V., Levesque, M., Rotenberg, B., Borgis, D.
Format: Artikel
Sprache:English
Veröffentlicht: Інститут фізики конденсованих систем НАН України 2017
Schriftenreihe:Condensed Matter Physics
Online Zugang:http://dspace.nbuv.gov.ua/handle/123456789/157011
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Solvation in atomic liquids: connection between Gaussian field theory and density functional theory / V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis // Condensed Matter Physics. — 2017. — Т. 20, № 3. — С. 33005: 1–14. — Бібліогр.: 75 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id irk-123456789-157011
record_format dspace
spelling irk-123456789-1570112019-06-21T01:26:57Z Solvation in atomic liquids: connection between Gaussian field theory and density functional theory Sergiievskyi, V. Levesque, M. Rotenberg, B. Borgis, D. For the problem of molecular solvation, formulated as a liquid submitted to the external potential field created by a molecular solute of arbitrary shape dissolved in that solvent, we draw a connection between the Gaussian field theory derived by David Chandler [Phys. Rev. E, 1993, 48, 2898] and classical density functional theory. We show that Chandler’s results concerning the solvation of a hard core of arbitrary shape can be recovered by either minimising a linearised HNC functional using an auxiliary Lagrange multiplier field to impose a vanishing density inside the core, or by minimising this functional directly outside the core — indeed a simpler procedure. Those equivalent approaches are compared to two other variants of DFT, either in the HNC, or partially linearised HNC approximation, for the solvation of a Lennard-Jones solute of increasing size in a Lennard-Jones solvent. Compared to Monte-Carlo simulations, all those theories give acceptable results for the inhomogeneous solvent structure, but are completely out-of-range for the solvation free-energies. This can be fixed in DFT by adding a hard-sphere bridge correction to the HNC functional. Для проблеми молекулярної сольватацiї, що формулюється як рiдина в зовнiшньому потенцiальному полi, створеному молекулами довiльної форми, якi розчиненi в розчиннику, ми приводимо зв’язок мiж теорiєю гауссового поля, виведеною Давидом Чандлером [Phys. Rev. E, 1993, 48, 2898] i класичною теорiєю функцiоналу густини (DFT). Ми показуємо, що результати Чандлера щодо сольватацiї твердого кору довiльної форми можуть бути зрегенерованi або шляхом мiнiмiзацiї лiнеаризованого HNC функцiоналу, використовуючи допомiжне поле множникiв Лагранжа для накладання умови зникаючої густини всерединi кору, або мiнiмiзацiєю цього функцiоналу напряму в областi зовнi кору, що є насправдi простiшою процедурою. Цi еквiвалентнi пiдходи порiвнюються з двома варiантами DFT, або в наближеннi HNC, або в наближеннi частково лiнеаризованого HNC, для сольватацiї розчиненої речовини iз взаємодiєю Леннарда-Джонса зi зростаючим розмiром в розчиннику iз леннард-джонсiвською взаємодiєю. Щодо порiвняння з моделюванням методом Монте Карло, всi цi теорiї дають прийнятнi результати для неоднорiдної структури розчинника, але є повнiстю поза дiапазоном для сольватацiйних вiльних енергiй. Це може бу 2017 Article Solvation in atomic liquids: connection between Gaussian field theory and density functional theory / V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis // Condensed Matter Physics. — 2017. — Т. 20, № 3. — С. 33005: 1–14. — Бібліогр.: 75 назв. — англ. 1607-324X PACS: 05.20.Jj, 11.10.-z, 82.60.Lf, 64.75.Bc DOI:10.5488/CMP.20.33005 arXiv:1708.01299 http://dspace.nbuv.gov.ua/handle/123456789/157011 en Condensed Matter Physics Інститут фізики конденсованих систем НАН України
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
language English
description For the problem of molecular solvation, formulated as a liquid submitted to the external potential field created by a molecular solute of arbitrary shape dissolved in that solvent, we draw a connection between the Gaussian field theory derived by David Chandler [Phys. Rev. E, 1993, 48, 2898] and classical density functional theory. We show that Chandler’s results concerning the solvation of a hard core of arbitrary shape can be recovered by either minimising a linearised HNC functional using an auxiliary Lagrange multiplier field to impose a vanishing density inside the core, or by minimising this functional directly outside the core — indeed a simpler procedure. Those equivalent approaches are compared to two other variants of DFT, either in the HNC, or partially linearised HNC approximation, for the solvation of a Lennard-Jones solute of increasing size in a Lennard-Jones solvent. Compared to Monte-Carlo simulations, all those theories give acceptable results for the inhomogeneous solvent structure, but are completely out-of-range for the solvation free-energies. This can be fixed in DFT by adding a hard-sphere bridge correction to the HNC functional.
format Article
author Sergiievskyi, V.
Levesque, M.
Rotenberg, B.
Borgis, D.
spellingShingle Sergiievskyi, V.
Levesque, M.
Rotenberg, B.
Borgis, D.
Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
Condensed Matter Physics
author_facet Sergiievskyi, V.
Levesque, M.
Rotenberg, B.
Borgis, D.
author_sort Sergiievskyi, V.
title Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
title_short Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
title_full Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
title_fullStr Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
title_full_unstemmed Solvation in atomic liquids: connection between Gaussian field theory and density functional theory
title_sort solvation in atomic liquids: connection between gaussian field theory and density functional theory
publisher Інститут фізики конденсованих систем НАН України
publishDate 2017
url http://dspace.nbuv.gov.ua/handle/123456789/157011
citation_txt Solvation in atomic liquids: connection between Gaussian field theory and density functional theory / V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis // Condensed Matter Physics. — 2017. — Т. 20, № 3. — С. 33005: 1–14. — Бібліогр.: 75 назв. — англ.
series Condensed Matter Physics
work_keys_str_mv AT sergiievskyiv solvationinatomicliquidsconnectionbetweengaussianfieldtheoryanddensityfunctionaltheory
AT levesquem solvationinatomicliquidsconnectionbetweengaussianfieldtheoryanddensityfunctionaltheory
AT rotenbergb solvationinatomicliquidsconnectionbetweengaussianfieldtheoryanddensityfunctionaltheory
AT borgisd solvationinatomicliquidsconnectionbetweengaussianfieldtheoryanddensityfunctionaltheory
first_indexed 2025-07-14T09:21:29Z
last_indexed 2025-07-14T09:21:29Z
_version_ 1837613596295561216
fulltext Condensed Matter Physics, 2017, Vol. 20, No 3, 33005: 1–14 DOI: 10.5488/CMP.20.33005 http://www.icmp.lviv.ua/journal Solvation in atomic liquids: connection between Gaussian field theory and density functional theory ∗ V. Sergiievskyi1, M. Levesque1, B. Rotenberg2, D. Borgis1,3 1 Sorbonne Universités, UPMC Univ Paris 06, ENS, CNRS, UMR 8640 PASTEUR, 75005 Paris, France 2 Sorbonne Universités, UPMC Univ Paris 06, CNRS, UMR 8234 PHENIX, 4 Place Jussieu, 75005 Paris, France 3 Maison de la Simulation, CEA, CNRS, Univ. Paris-Sud, UVSQ, Université Paris-Saclay, 91191 Gif-sur-Yvette, France Received June 23, 2017, in final form July 19, 2017 For the problem of molecular solvation, formulated as a liquid submitted to the external potential field created by a molecular solute of arbitrary shape dissolved in that solvent, we draw a connection between the Gaussian field theory derived by David Chandler [Phys. Rev. E, 1993, 48, 2898] and classical density functional theory. We show that Chandler’s results concerning the solvation of a hard core of arbitrary shape can be recovered by either minimising a linearised HNC functional using an auxiliary Lagrange multiplier field to impose a vanishing density inside the core, or by minimising this functional directly outside the core— indeed a simpler procedure. Those equivalent approaches are compared to two other variants of DFT, either in the HNC, or partially linearised HNC approximation, for the solvation of a Lennard-Jones solute of increasing size in a Lennard-Jones solvent. Compared to Monte-Carlo simulations, all those theories give acceptable results for the inhomogeneous solvent structure, but are completely out-of-range for the solvation free-energies. This can be fixed in DFT by adding a hard-sphere bridge correction to the HNC functional. Key words: statistical mechanics, classical fluids, 3-dimensional systems, density functional theory, gaussian field theory PACS: 05.20.Jj, 11.10.-z, 82.60.Lf, 64.75.Bc 1. Introduction In a world of hard-core numerical simulations on huge computers where most problems in solution chemistry are formulated in terms of molecular dynamics simulations and subsequent data analysis, it is wise to keep simpler methods that make it possible to derive analytical results or to perform the calculations with reasonable computer resources. Such methods rely on the statistical mechanics of atomic and molecular liquids that has been developed in the second half of the last century and are found by now in classical textbooks [1–3]. Along this vein is the beautiful and appealing recent theoretical work of Dung Di Caprio and Jean-Pierre Badiali who were able to formulate the description of classical fluids at equilibrium as a formally exact field theory [4–8]; this formalism was applied to model atomic and molecular fluids at solid interfaces [9–12]. Other more traditional approaches include molecular integral equation theories in the reference interaction site (RISM) [13–16], molecular [17–24], or mixed [25, 26] picture and the density functional theory (DFT) in its atomic [27–29] or molecular version [30–38]. The basic theoretical principles of classical DFT can be found in the seminal paper by Evans [27] and subsequent excellent reviews by him [27–29] and other authors [39]. The advent in the late 1980’s of a quasi-exact DFT for inhomogeneous hard sphere mixtures, the fundamental measure theory (FMT) [40–45], has recently promoted a great deal of applications to atomic-like fluids in bulk or confined ∗This contribution is dedicated to Prof. Jean-Pierre Badiali, who did much for promoting an original field theoretical picture of classical fluids. This work is licensed under a Creative Commons Attribution 4.0 International License . Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI. 33005-1 https://doi.org/10.5488/CMP.20.33005 http://www.icmp.lviv.ua/journal http://creativecommons.org/licenses/by/4.0/ V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis conditions or at interfaces. Classical “atomic” DFT can be nowadays considered as a method of choice for many chemical engineering problems [46, 47]. Much less applications exist for molecular fluids for which solvent orientations should be considered. The description has been generally limited to generic dipolar solvents or dipolar solvent/ions mixtures [32–35]; such an approach may be already considered as “civilized” compared to primitive continuum models [34]. We have proposed an extension of molecular DFT to arbitrary fluid/solvents (the so-called MDFT method) with the goal of describing the solvation of three-dimensional molecular object in those solvents [36–38, 48–56]. Note that a 3D-version of the RISM equations [57–62], as well as a RISM-based DFT approach [63, 64] have also been recently developed with the same goal. In this paper, we also elaborate on a field theoretical approach that is different from the one by di Caprio and Badiali — and certainly starts from a less fundamental ground. We refer to the Gaussian field theory (GFT) of fluids developed by Chandler and collaborators [65–68]. Our main focus will be to draw a connection between the GFT approach of Chandler and our favorite classical DFT in the context of molecular solvation, i.e., a liquid submitted to an external potential field v(r) created by a molecular solute of arbitrary shape dissolved at infinite dilution in it. For simplicity, we restrict the discussion to atomic or pseudo-atomic solvents (such as CCl4) modelled by spherical Lennard-Jones particles for which only the position r matters. 2. Density functional theory and HNC approximation We begin by recalling the basis of the density functional theory of liquids submitted to an external potential field v(r). The grand potential density functional for a fluid having an inhomogeneous density ρ(r) in the presence of an external field v(r) can be defined as [27, 28] Ω[ρ] = F[ρ] − µs ∫ ρ(r)dr, (2.1) where F[ρ] is the Helmholtz free energy functional and µs is the chemical potential. The grand potential can be evaluated relatively to a reference homogeneous fluid having the same chemical potential µs and particle density ρ0 Ω[ρ] = Ω[ρ0] + F [ρ]. (2.2) Following the general theoretical scheme introduced by Evans [27, 28], the density functional F [ρ] can be split into three contributions: an ideal term, an external potential term and an excess free-energy term accounting for the intrinsic interactions within the fluid, F [ρ] = Fid[ρ] + Fext[ρ] + Fexc[ρ], (2.3) with the following expressions of the first two terms Fid[ρ] = kBT ∫ dr { ρ (r) ln [ ρ (r) ρ0 ] − ρ (r) + ρ0 } , (2.4) Fext[ρ] = ∫ dr v(r)ρ (r). (2.5) There are several ways of arriving at an exact expression of the excess free-energy, i.e., using an adiabatic perturbation of the pair potential (the so-called adiabatic connection route in electronic DFT), of the external potential, or of the density itself. A conventional approximation is to express the excess term as an expansion around the homogeneous density ρ0 Fexc[ρ] = − kBT 2 ∫ dr1dr2 c(r12; ρ0)∆ρ(r1)∆ρ(r2) + FB[ρ]. (2.6) The first term is the (two-body) direct correlation function (DCF) of the homogeneous solvent, that depends on r12 = |r2 − r1 |, and can be thus denoted as c(r12; ρ0). We define the so-called bridge 33005-2 Gaussian field and density functional theory functional in terms of the higher-order direct correlation functions FB[ρ] = − kBT 6 ∫ dr1dr2dr3 c(3)(r1, r2, r3; ρ0)∆ρ(r1)∆ρ(r2)∆ρ(r3) +O(∆ρ4), (2.7) which thus starts with a cubic term in ∆ρ. Setting FB[ρ] = 0 corresponds to the so-called homogeneous reference fluid (HRF) approximation. It can be shown to be equivalent to the hypernetted chain (HNC) approximation in integral equation theories [29]. The input of the theory is thus a direct correlation function of the pure solvent, which can be extracted from simulation or experimental data by measuring the total correlation function h(r) = g(r) − 1 and solving subsequently the Ornstein-Zernike equation, i.e., in Fourier space: 1 − ρ0c(k) = [1 + ρ0h(k)]−1 = χ−1(k). (2.8) χ(r) is the structure factor, or the density susceptibility, measuring density-density correlations at a given distance in the fluid. The excess free energy can thus be also expressed in terms of the inverse susceptibility Fexc[ρ] = kBT 2 ∫ dr1dr2 χ −1(r12)∆ρ(r1)∆ρ(r2) − kBT 2ρ0 ∫ dr∆ρ(r)2 + FB[ρ]. (2.9) Minimization of equation (2.3) with respect to ρ gives the equilibrium density ρ(r1) = ρ0 exp [ −βv(r1) − ∫ dr2 χ −1(r12)∆ρ(r2) + ∆ρ(r1) ρ0 − δ(βFB) δρ (r1) ] . (2.10) 3. Chandler’s Gaussian field theory Along the same lines as above, Chandler considered the case of a liquid of density ρ0, characterised by its intrinsic density susceptibility χ(r), containing a solute creating an external potential v(r) outside a hard core that defines an inside volume Vin where the density ρ(r) is zero and where by convention v(r) = 0. Chandler writes a gaussian field Hamiltonian for the pure fluid HB = kBT 2 ∫ dr1dr2 ∆ρ(r1) χ −1(r12)∆ρ(r2), (3.1) and the partition function of the fluid + solute system as a field integral Z = ∫ Dρ [ ∏ r inside δ ( ρ(r) ) ] exp [ −βHB − ∫ dr βv(r) ρ(r) ] , (3.2) where the product of delta-functions imposes the constraint of zero-density inside the core. Performing the Gaussian integral exactly, Chandler arrives at the expression of the solvation free energy βFeq = − log Z = ρ0 ∫ dr βv(r) − 1 2 ∫ out dr1 ∫ out dr2 βv(r1)χ(|r2 − r1 |)βv(r2) + 1 2 ∫ in dr3 ∫ in dr4 χ −1 in (r3, r4) ρ0 − ∫ out dr1 χ(|r3 − r1 |)βv(r1)  ρ0 − ∫ out dr2 χ(|r4 − r2 |)βv(r2)  + 1 2 ln (det χin) , (3.3) and, by functional differentiation with respect to the external potential, at the one-particle equilibrium density ρeq(r) = ρ0 − ∫ out dr1 χ(|r − r1 |)βv(r1) − ∫ in dr1 ∫ in dr2 χ −1 in (r1, r2)χ(|r1 − r|) + ρ0 − ∫ out dr3 χ(|r2 − r3 |)βv(r3)  . (3.4) 33005-3 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis We stick here to Chandler’s notations, with his u(r) equal to −βv(r). Note that χ−1 in should be understood as (χin)−1. One of the main results in Chandler’s paper is that the susceptibility of the medium, defined as χ(r1, r2) = δ〈ρ(r1)〉/δv(r2), is altered by the presence of the hard core and changed from χ(r1, r2) = χ(|r1 − r2 |) for the infinite medium to an effective susceptibility χeff(r1, r2) = χ(|r1 − r2 |) − ∫ in dr3 ∫ in dr4 χ(|r1 − r3 |) χ −1 in (r3, r4) χ(|r4 − r2 |) (3.5) that is not translationally invariant anymore. 4. Linearised and partially-linearised HNC approximations and connec- tion to Gaussian field theory The linearised HNC approximation consists in expanding the ideal term in equation (2.4) at dominant order in ∆ρ Fid[ρ] = kBT 2ρ0 ∫ dr∆ρ(r)2 (4.1) so that the functional to be minimised becomes βF [ρ] = 1 2 ∫ dr1dr2 χ −1(r12)∆ρ(r1)∆ρ(r2) + ∫ dr βv(r)ρ (r). (4.2) In the presence of a solute with a hard repulsive core [very positive values of the potential v(r)], such approximationwill obviously fail to give an exponentially vanishing density inside the core. As considered by Chandler above, this approximation should be complemented by constraints imposing ρ(r) = 0 within the inside volume Vin. There are two ways to impose those constraints. The first one, not necessarily the easiest one, is to introduce an auxiliary Lagrange multiplier field λ(r) and minimise the following constrained functional with respect to ρ(r) and λ(r) βFc[ρ] = 1 2 ∫ dr1dr2 χ −1(r12)∆ρ(r1)∆ρ(r2) + ∫ dr βv(r)ρ (r) − ∫ in dr λ(r)ρ (r). (4.3) Thus, the minimisation equations are as follows: δ(βF ) δλ(r) = ρ (r) = 0, r ∈ Vin , (4.4) δ(βF ) δρ(r) = λ(r), (4.5) λ(r) = 0, r ∈ Vout. (4.6) These equations can be readily solved by linear algebra to give an equilibrium density that is equivalent to the one in equation (3.4). Replacement in equation (4.3) does give the equilibrium free energy of equation (3.3), except the last log-of-determinant term that includes a measure of the fluctuations that is absent in the functional approach. Numerical estimations shows that it can be safely neglected with respect to the other terms. We conclude that the Chandler’s Gaussian field approach is, up to a small log-term correction in the energy, equivalent to a DFT approach with a linearised HNC approximation. From a DFT perspective, however, a natural way to account for the constraint is to minimise the functional outside the core only, i.e., for r ∈ Vout. The functional can thus be limited to the outside region and written as βF [ρ] = 1 2 ∫ out dr1 ∫ out dr2 χ −1(r12)∆ρ(r1)∆ρ(r2) + ∫ out dr βv(r)ρ (r) − ρ0 ∫ out dr ∫ in dr1 χ −1(|r − r1 |) ρ(r) − 1 2 ρ2 0 ∫ in dr1 ∫ in dr2 χ −1(r12). (4.7) 33005-4 Gaussian field and density functional theory This functional can be easily numerically minimised on a three-dimensional grid using for example a quasi-Newton minimiser such as L-BFGS [69] to yield the equilibrium density ρeq and the associated free energy. Since the above functional is bilinear in ρ(r), the formal solution can be also obtained by matrix inversion, i.e., outside the core ρeq(r) = ρ0 + ∫ out dr1(χ −1 out) −1(r, r1) ρ0 ∫ in dr2 χ −1(|r2 − r1 |) − βv(r1)  . (4.8) This solution looks quite different from that in equation (3.4); in the appendix below it is shown that the two formulas are in fact equivalent. Thus, we arrive at the main conclusion of this paper: the rather involved formal solutions of the Gaussian field approach (equivalent to a functional minimization with Lagrange multipliers, as seen above), which involves the necessity to numerically invert the matrix χin inside the core and then to perform a double multiplication of this matrix with χ, can be replaced by a simple numerical minimisation of the LHNC functional (4.7) outside the hard core. The basic input is the homogeneous bulk inverse susceptibility χ−1(r12) [or equivalently, the homogeneous bulk DCF c(r12; ρ0)], with no interference whatsoever with the introduction of hard-core conditions. The bulk inverse susceptibility applies everywhere, inside and outside the hard core. The fact that, as noted by Chandler, the introduction of such hard-core boundaries modifies the apparent susceptibility of the medium outside the core is a consequence that applies to the LHNC-DFT approach as it does for the GFT one. It should be also valid at a HNC level; this effect can be measured numerically as χ(r1, r2) = δ 〈ρ(r1)〉 /δv(r2)— indeed not an easy task on a 3D spatial grid. We note that an approximation between HNC and LHNC, referred to as the partially linearised HNC approximation (PLHNC), can be obtained by writing the ideal free energy as βFid[ρ] = ∫ dr fid ( ρ(r) ) with fid ( ρ(r) ) = ∆ρ(r)2 ρ0 (4.9) for ∆ρ(r) > 0 and the full expression in equation (2.4) fid ( ρ(r) ) = ρ(r) ln [ 1 + ∆ρ(r) ρ0 ] − ∆ρ(r) (4.10) for ∆ρ(r) < 0. The overall function remains continuous at ρ(r) = ρ0. In the following we test the HNC, LHNC (equivalent to Gaussian field theory), and PLHC for the solvation of a Lennard-Jones sphere of an increasing diameter in a Lennard-Jones liquid, in comparison with the reference Monte-Carlo generated by Lazaridis [70]. The LJ solvent is characterised by a particle diameter σ0 and reduced thermodynamic conditions ρ∗ = 0.85, T∗ = 0.88. In figure 1, we display the solvent structure for 3 solute diameters, σ/σ0 = 0.2, 1, and 2, respectively. The DFT results were obtained by direct functional minimisation using a home-made spherical 1D code. The hard-core volume for LHNC was identified to the void region obtained after HNC minimisation [ρ(r) < ρmin, a fixed, very small value]. The first observation is that none of the approximations is either perfect or clearly off. Apart from the smaller solute, it is seen that the HNC approximation tends to underestimate the first-peak position and overestimate its height. The second observation is that, surprisingly, LHNC and PLHNC give undistinguishable results; for both, the first peak appears now too low for the smaller solutes and has a correct height but with a shift in position for the biggest, as in HNC. PLHNC can be qualified as a better theory since the hard core is defined and handled automatically by the functional. The situation gets really worse when going to the solvation free energies. In figure 2, we compare the results of the 3 approximations when increasing progressively σ/σ0 to the simulation results of Lazaridis. All of them are off by a large factor and in nearly the same way. The problem has been clearly identified [56, 71, 72]: all those HNC variants give an apparent pressure which is way too high with respect to the exact pressure, Pexact, of the LJ fluid, and thus a spurious ∆P∆V contribution where ∆P = PHNC − Pexact, and ∆V is the solute partial molar volume— close to, but not identical to the inside volumeVin of the solute. This can be corrected by adding an empirical pressure correction, −∆P∆V , to the DFT-HNC (or LHNC, or PLHNC) free energy [56, 71]. Herein below we switch to a more fundamental correction for Lennard-Jones that involves a hard-sphere bridge functional. 33005-5 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis Figure 1. (Color online) Reduced solvent density around LJ solutes of different relative diameters ob- tained by DFT with the HNC, LHNC, and PLHNC approximations (grey, black and dashed black lines, respectively), and compared to Molecular dynamics results in red. The LHNC and PLHNC results are indistinguishable on the scale of the figure. Figure 2. (Color online) Solvation free-energy obtained by DFT in different approximations for a Lennard-Jones solute of increasing relative diameter. The blue triangles are the Monte-Carlo results of Lazaridis [70]. 33005-6 Gaussian field and density functional theory 5. Hard-sphere bridge correction Building the thermodynamics of the Lennard-Jones fluid by taking a suitable hard-sphere fluid as a reference is indeed a classic in liquid-state theory and is at the basis of the Van der Waals theory of fluids. A variant of this idea is to approximate the bridge functional in equation (2.6) by a hard sphere bridge (HSB) functional introduced by Rosenfeld as a universal bridge function [73–75]. F HS B [ρ(r)] = FHS exc [ρ(r)] − FHS exc [ρ0] − δFHS exc [ρ] δρ(r) ���� ρ0 ∫ dr∆ρ(r) + kBT 2 ∫ dr1dr2 cHS(r12; ρ0)∆ρ(r1)∆ρ(r2). (5.1) Here, FHS exc [ρ(r)] represents the one-component hard-sphere excess functional which, up to a very good approximation, can be taken as the fundamental measure theory (FMT) functional of Rosenfeld [40] and Kierlik and Rosinberg [41, 42]. The fourth term involves the direct correlation function of the HS fluid at the same density, i.e., cHS(|r1 − r2 |; ρ0) = − δ2βFHS exc [ρ] δρ(r1)δρ(r2) ���� ρ0 . (5.2) Note that defined as in equation (5.1), F HB B [ρ(r)] carries an expansion in ∆ρ of the order 3 and higher which corrects the second order expansion of the excess free energy in equation (2.6). We show in figure 3 that this HNC+HSB theory works much better than the HNC variants for the prediction of solvation properties of dissolved molecular objects. There we again compare the solvation free energy of the growing LJ sphere to the Monte-Carlo results of Lazaridis [70] using different HS diameters, d. It can be seen that the results are extremely sensitive to the choice of d, and that the best agreement is obtained for d = 1.014σ (indeed close to 1, that would be the initial guess value). For that value, we have plotted in figure 4 the solvent density, g(r) = ρ(r)/ρ0, obtained for solute of different sizes by direct MD simulations that we have generated, or by DFT in the HNC or HNC+HSB approximation. It can be seen that the addition of the hard-sphere bridge greatly improves the results compared to the HNC approximation and furthermore yields a very good structure. Figure 3. (Color online) Solvation free-energy obtained by DFT using the hard-sphere bridge functional of equation (5.1) with different HS diameters, compared to the Monte-Carlo results of Lazaridis [70]. 33005-7 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis Figure 4. (Color online) Reduced solvent density around LJ solutes of different diameters, using the HNC approximation, or adding a hard-sphere bridge functional with d = 1.014σ. The red curves were generated by molecular dynamics. 6. Conclusion In this paper we have shown a close connection between the Gaussian field theory of solvation introduced by Chandler in [65] and density functional theory in a linearised HNC approximation. Chandler’s formulae for the solvation density around hard solutes and the associated solvation free energies can be recovered by minimising the LHNC functional with constraints imposed through an auxiliary Lagrange multiplier field. A simpler but equivalent formulation arises when minimising the functional outside the hard core only. Both theories share with the full HNC approximation, or the intermediate PLHNC approximation, the same caveat of greatly overestimating the solvation free energy of dissolved objects. Chandler was indeed aware of these limitations and provided further improvements based on the coupling of GFT at the microscopic scale to a lattice gas model having a correct macroscopic behaviour at larger scales [68]. In DFT, improvements can be made by considering a bridge functional beyond the second order expansion in density. For the Lennard-Jones solvent, the natural bridge that emerges is that of a reference hard fluid, whose hard-sphere diameter should be optimised. The extension of such an approach to molecular liquids, such as water, has been proposed with some success [51, 63]. This remains to be further explored and improved — since a water molecule is definitely not a spherical entity. The interlink between density functional theories and other versions of liquid-state field theories, such as those developed by Jean-Pierre Badiali and his Parisian and Ukrainian collaborators along the years, is also a very interesting subject that merits to be explored in depth in the future. 33005-8 Gaussian field and density functional theory Acknowledgements We are grateful to late Prof. David Chandler for insightful discussions during a visit in Paris and for attracting our attention to the problem tackled in this paper. VS was supported by a grant from the Fondation Pierre-Gilles de Gennes. A. Connection between “inner” and “outer” DFT formulations, and Chandler’s GFT A.1. Notation Herein below we will use a discrete matrix notation for the fields and associated functionals. Let V be the liquid volume and be decomposed into an inside volume Vin occupied by the hard-sphere solute and the remaining volume Vout. We define the functions on a finite three-dimensional grid. Let m points lie inside the solute and n points outside. In that case, the one-variable functions, like density, can be represented as vectors of size (m + n) × 1 (e.g. ρ). The two-variables functions, e.g., the susceptibility function χ(r1, r2) are represented as matrices (m + n) × (m + n) (e.g X). Then, the convolution can be represented as a matrix multiplication, e.g.,∫ ρ(r1)χ(r1, r2)dr1 ⇐⇒ ∆vρTX, (A.1) where ∆v is the elementary volume which corresponds to each discretisation point. For simplicity, we will take below ∆v = 1. Let the density inside the solute be ρin (m×1 vector), the density outside the solute ρout (n×1 vector). The free energy functional can be defined as follows: F [ρin, ρout] = 1 2 ∆ρT in(X−1)in∆ρin + 1 2 ∆ρT out(X−1)out∆ρ T out + ∆ρ T in(X−1)inter∆ρout + βvToutρout. (A.2) Here, ∆ρ = ρ − ρ0, X is a susceptibility matrix X = ( Xin Xinter XT inter Xout ) , X−1 = ( (X−1)in (X−1)inter (X−1)Tinter (X−1)out ) , (A.3) where Xin is m × m, Xinter is m × n, Xout is n × n. It is important to note that, for example, (Xin) −1 , (X−1)in. (A.4) The above functional should be minimised with the constraint ρin = 0, ∆ρin = −ρ0. There are two approaches to do this: Lagrange multiplier minimisation or restrained minimisation in the outer volume. We show herein below that the two approaches are equivalent to each other and give the same results as Chandler’s Gaussian field theory in [65]. A.2. Lagrange multipliers minimization To perform the minimisation using Lagrange multipliers we add −λρin to the functional: F [ρin, ρout] = 1 2 ∆ρT in(X−1)in∆ρin+ 1 2 ∆ρT out(X−1)out∆ρ T out+∆ρ T in(X−1)inter∆ρout+βvToutρout−λT inρin. (A.5) From the necessary minimization conditions ∂F ∂λin = ρin = 0 (A.6) 33005-9 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis and ∂F ∂ρin = (X−1)in∆ρin + (X−1)inter∆ρout = λin , ∂F ∂ρout = (X−1)Tinter∆ρin + (X−1)out∆ρout = −βvout , (A.7) the last two equations can be rewritten as X−1 ∆ρ = [ λin −βvout ] . (A.8) From this, we find ∆ρ ∆ρ = ( Xin Xinter XT inter Xout ) · [ λin −βvout ] (A.9) and the relations: ∆ρin = Xinλin − Xinterβvout = −ρ0 in , (A.10) ∆ρout = Xinterλin − Xoutβvout. (A.11) Using the first equation we find λin = (Xin) −1 ( −ρ0 in + Xinterβvout ) . (A.12) Inserting this into the second equation: ∆ρout = −XT interX−1 in ρ 0 in + XT interX−1 in Xinterβvout − Xoutβvout = −XT inter (Xin) −1 ( −ρ0 in + Xinterβvout ) − Xoutβvout. (A.13) This is exactly Chandler’s Gaussian field expression, equation (3.4), in discretised form [with the under- standing that χ−1 in = (χin) −1]. Injecting this formula into equation (A.5) also gives the same expression as Chandler for the equilibrium solvation free-energy, equation (3.3), except the last logarithm-of- determinant term. A.3. Direct minimization in outer volume (reduced number of variables) Instead of performing the minimisation with the Lagrange multipliers, we can minimise the reduced functional which depends only on ρout: F [ρout] = 1 2 ∆ρT out(X−1)out∆ρ T out − (ρ 0 in) T(X−1)inter∆ρout + βvToutρout + C, (A.14) where C ≡ 1 2 (ρ0 in) T(X−1)inρ 0 in. Taking the derivative (X−1)out∆ρout − (X−1)Tinterρ 0 in + βvout = 0 (A.15) and ∆ρout = [(X−1)out] −1 [ (X−1)Tinterρ 0 in − βvout ] . (A.16) To see that this is the same as (A.13) we need to invert the matrix X. To do it, let us define Xin ≡ A, (X−1)in ≡W, Xinter ≡ B, (X−1)inter ≡ Y, Xout ≡ C, (X−1)out ≡ Z. 33005-10 Gaussian field and density functional theory By the definition of the inverse matrix we have( A B BT C ) · ( W Y YT Z ) = ( I 0 0 I ) , (A.17) where I is an identity matrix of appropriate size. We have the following equations: AW + BYT = I, AY + BZ = 0, BTW + CYT = 0, BTY + CZ = I. (A.18) Multiplying the first by A−1: W = A−1 − A−1BYT (A.19) and inserting this into the third equation: BTA−1 − BTA−1BYT + CYT = 0, (A.20) (C − BTA−1B)YT = −BTA−1. (A.21) From this we find YT YT = −(C − BTA−1B)−1BTA−1 (A.22) and Y Y = −(BTA−1)T[(C − BTA−1B)T]−1 = −A−1B(C − BTA−1B)−1. (A.23) (Here, we use A = AT, C = CT, which is true since X is symmetric). Now, from the last equation in (A.18) C−1BTY + Z = C−1, (A.24) Z = C−1 ( I − BTY ) . (A.25) Inserting here the expression of Y: Z = C−1 [ I + BTA−1B(C − BTA−1B)−1] . (A.26) We can further simplify this expression. We first express the identity matrix I as I = (C − BTA−1B) · (C − BTA−1B)−1. (A.27) Inserting this into the expression of Z we have Z = C−1(C − BTA−1B + BTA−1B)(C − BTA−1B)−1. (A.28) Cancelling BTA−1B and C−1C we get Z = (C − BTA−1B)−1. (A.29) Returning to the expression (A.16) ∆ρout = Z−1(YTρ0 in − βvout) = (C − BTA−1B) [ −(C − BTA−1B)−1BTA−1ρ0 in − βvout ] . (A.30) Opening the brackets ∆ρout = −BTA−1ρ0 in − Cβvout + BTA−1Bβvout = BTA−1(−ρ0 in + Bβvout) − Cβvout (A.31) or, returning to the original definitions: ∆ρout = XT inter (Xin) −1 (−ρ0 in + Xinterβvout) − Xoutβvout (A.32) which is the same as (A.13). This terminates the proof for the equilibrium density. The same equivalence can be proved for the equilibrium solvation free-energy. 33005-11 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis References 1. Hansen J.P., McDonald I.R., Theory of Simple Liquids, Academic Press, London, 1989. 2. Gray C.G., Gubbins K.E., Theory of Molecular Fluids. Volume I: Fundamentals, Oxford University Press, Oxford, 1984. 3. Gray C.G., Gubbins K.E., Joslin C.G., Theory of Molecular Fluids. Volume II: Applications, Oxford University Press, Oxford, 2011. 4. Di Caprio D., Stafiej J., Badiali J.P., J. Chem. Phys., 1998, 108, No. 20, 8572–8583, doi:10.1063/1.476286. 5. Di Caprio D., Stafiej J., Badiali J.P., Mol. Phys., 2003, 101, No. 16, 2545–2558, doi:10.1080/0026897031000154293. 6. Di Caprio D., Stafiej J., Badiali J.P., Mol. Phys., 2003, 101, No. 21, 3197–3202, doi:10.1080/00268970310001632318. 7. Di Caprio D., Badiali J.P., J. Phys. A: Math. Theor., 2008, 41, No. 12, 125401, doi:10.1088/1751-8113/41/12/125401. 8. Di Caprio D., Badiali J.P., Entropy, 2009, 11, No. 2, 238–248, doi:10.3390/e11020238. 9. Holovko M., di Caprio D., Kravtsiv I., Condens. Matter Phys., 2011, 14, No. 3, 33605, doi:10.5488/CMP.14.33605. 10. Di Caprio D., Stafiej J., Holovko M., Kravtsiv I., Mol. Phys., 2011, 109, No. 5, 695–708, doi:10.1080/00268976.2010.547524. 11. Kravtsiv I., Holovko M., di Caprio D., Mol. Phys., 2013, 111, No. 8, 1023–1041, doi:10.1080/00268976.2012.762615. 12. Kravtsiv I., Patsahan T., Holovko M., di Caprio D., J. Chem. Phys., 2015, 142, No. 19, 194708, doi:10.1063/1.4921242. 13. Chandler D., Andersen H., J. Chem. Phys., 1972, 57, 1930, doi:10.1063/1.1678513. 14. Hirata F., Rossky P.J., Chem. Phys. Lett., 1981, 83, 329, doi:10.1016/0009-2614(81)85474-7. 15. Hirata F., Pettitt B.M., Rossky P.J., J. Chem. Phys., 1982, 77, 509, doi:10.1063/1.443606. 16. Reddy G., Lawrence C.P., Skinner J.L., Yethiraj A., J. Chem. Phys., 2003, 119, 13012, doi:10.1063/1.1627326. 17. Blum L., Torruella A.J., J. Chem. Phys., 1972, 56, 303, doi:10.1063/1.1676864. 18. Blum L., J. Chem. Phys., 1972, 57, 1862, doi:10.1063/1.1678503. 19. Patey G.N., Mol. Phys., 1977, 34, 427, doi:10.1080/00268977700101821. 20. Carnie S.L., Patey G.N., Mol. Phys., 1982, 47, 1129, doi:10.1080/00268978200100822. 21. Fries P.H., Patey G.N., J. Chem. Phys., 1985, 82, 429, doi:10.1063/1.448764. 22. Richardi J., Fries P.H., Krienke H., J. Chem. Phys., 1998, 108, 4079, doi:10.1063/1.475805. 23. Richardi J., Millot C., Fries P.H., J. Chem. Phys., 1999, 110, 1138, doi:10.1063/1.478171. 24. Belloni L., Chikina I., Mol. Phys., 2014, 112, No. 9–10, 1246–1256, doi:10.1080/00268976.2014.885612. 25. Dyer K.M., Perkyns J.S., Pettitt B.M., J. Chem. Phys., 2007, 127, 194506, doi:10.1063/1.2785188. 26. Dyer K.M., Perkyns J.S., Stell G., Pettitt B.M., J. Chem. Phys., 2008, 129, 104512, doi:10.1063/1.2976580. 27. Evans R., Adv. Phys., 1979, 28, No. 2, 143, doi:10.1080/00018737900101365. 28. Evans R., In: Fundamentals of Inhomogeneous Fluids, Henderson D. (Ed.), Marcel Dekker, New York, 1992, 85–176. 29. Evans R., In: Lecture Notes, 3rd Warsaw School of Statistical Physics, Cichocki B., Napiórkowski M., Pi- asecki J. (Eds.), Warsaw University Press, Warsaw, 2010, 43–85. 30. Chandler D., McCoy J.D., Singer S.J., J. Chem. Phys., 1986, 85, No. 10, 5971, doi:10.1063/1.451510. 31. Chandler D., McCoy J.D., Singer S.J., J. Chem. Phys., 1986, 85, No. 10, 5977, doi:10.1063/1.451511. 32. Biben T., Hansen J.P., Rosenfeld Y., Phys. Rev. E, 1998, 57, R3727–R3730, doi:10.1103/PhysRevE.57.R3727. 33. Oleksy A., Hansen J.P., Mol. Phys., 2009, 107, No. 23–24, 2609–2624, doi:10.1080/00268970903469022. 34. Oleksy A., Hansen J.P., J. Chem. Phys., 2010, 132, No. 20, 204702, doi:10.1063/1.3428704. 35. Oleksy A., Hansen J.P., Mol. Phys., 2011, 109, No. 7–10, 1275–1288, doi:10.1080/00268976.2011.554903. 36. Ramirez R., Gebauer R., Mareschal M., Borgis D., Phys. Rev. E, 2002, 66, 031206, doi:10.1103/PhysRevE.66.031206. 37. Ramirez R., Borgis D., J. Phys. Chem. B, 2005, 109, 6754, doi:10.1021/jp045453v. 38. Ramirez R., Mareschal M., Borgis D., Chem. Phys., 2005, 319, 261, doi:10.1016/j.chemphys.2005.07.038. 39. Löwen H., J. Phys.: Condens. Matter, 2002, 14, No. 46, 11897–11905, doi:10.1088/0953-8984/14/46/301. 40. Rosenfeld Y., Phys. Rev. Lett., 1989, 63, No. 9, 980–983, doi:10.1103/PhysRevLett.63.980. 41. Kierlik E., Rosinberg M.L., Phys. Rev. A, 1990, 42, No. 6, 3382–3387, doi:10.1103/PhysRevA.42.3382. 42. Kierlik E., Rosinberg M.L., Phys. Rev. A, 1991, 44, No. 8, 5025–5037, doi:10.1103/PhysRevA.44.5025. 43. Roth R., Evans R., Lang A., Kahl G., J. Phys.: Condens. Matter, 2002, 14, No. 46, 12063, doi:10.1088/0953-8984/14/46/313. 33005-12 https://doi.org/10.1063/1.476286 https://doi.org/10.1080/0026897031000154293 https://doi.org/10.1080/00268970310001632318 https://doi.org/10.1088/1751-8113/41/12/125401 https://doi.org/10.3390/e11020238 https://doi.org/10.5488/CMP.14.33605 https://doi.org/10.1080/00268976.2010.547524 https://doi.org/10.1080/00268976.2012.762615 https://doi.org/10.1063/1.4921242 https://doi.org/10.1063/1.1678513 https://doi.org/10.1016/0009-2614(81)85474-7 https://doi.org/10.1063/1.443606 https://doi.org/10.1063/1.1627326 https://doi.org/10.1063/1.1676864 https://doi.org/10.1063/1.1678503 https://doi.org/10.1080/00268977700101821 https://doi.org/10.1080/00268978200100822 https://doi.org/10.1063/1.448764 https://doi.org/10.1063/1.475805 https://doi.org/10.1063/1.478171 https://doi.org/10.1080/00268976.2014.885612 https://doi.org/10.1063/1.2785188 https://doi.org/10.1063/1.2976580 https://doi.org/10.1080/00018737900101365 https://doi.org/10.1063/1.451510 https://doi.org/10.1063/1.451511 https://doi.org/10.1103/PhysRevE.57.R3727 https://doi.org/10.1080/00268970903469022 https://doi.org/10.1063/1.3428704 https://doi.org/10.1080/00268976.2011.554903 https://doi.org/10.1103/PhysRevE.66.031206 https://doi.org/10.1021/jp045453v https://doi.org/10.1016/j.chemphys.2005.07.038 https://doi.org/10.1088/0953-8984/14/46/301 https://doi.org/10.1103/PhysRevLett.63.980 https://doi.org/10.1103/PhysRevA.42.3382 https://doi.org/10.1103/PhysRevA.44.5025 https://doi.org/10.1088/0953-8984/14/46/313 Gaussian field and density functional theory 44. Yu Y.X., Wu J., J. Chem. Phys., 2002, 117, No. 22, 10156, doi:10.1063/1.1520530. 45. Roth R., J. Phys.: Condens. Matter, 2010, 22, 063102, doi:10.1088/0953-8984/22/6/063102. 46. Wu J., AIChE J., 2006, 52, No. 3, 1169–1193, doi:10.1002/aic.10713. 47. Wu J., Li Z., Ann. Rev. Phys. Chem., 2007, 58, 85–112, doi:10.1146/annurev.physchem.58.032806.104650. 48. Gendre L., Ramirez R., Borgis D., Chem. Phys. Lett., 2009, 474, 366, doi:10.1016/j.cplett.2009.04.077. 49. Zhao S., Ramirez R., Vuilleumier R., Borgis D., J. Chem. Phys., 2011, 134, 194102, doi:10.1063/1.3589142. 50. Borgis D., Gendre L., Ramirez R., J. Phys. Chem. B, 2012, 116, 2504, doi:10.1021/jp210817s. 51. Levesque M., Vuilleumier R., Borgis D., J. Chem. Phys., 2012, 137, No. 3, 034115, doi:10.1063/1.4734009. 52. Levesque M., Marry V., Rotenberg B., Jeanmairet G., Vuilleumier R., Borgis D., J. Chem. Phys., 2012, 137, No. 22, 224107, doi:10.1063/1.4769729. 53. Jeanmairet G., Levesque M., Borgis D., J. Chem. Phys., 2013, 139, No. 15, 154101, doi:10.1063/1.4824737. 54. Jeanmairet G., Levesque M., Vuilleumier R., Borgis D., J. Phys. Chem. Lett., 2013, 4, 619–624, doi:10.1021/jz301956b. 55. Jeanmairet G., Marry V., Levesque M., Rotenberg B., Borgis D., Mol. Phys., 2014, 112, No. 9–10, 1320–1329, doi:10.1080/00268976.2014.899647. 56. Sergiievskyi V.P., Jeanmairet G., Levesque M., Borgis D., J. Chem. Phys. Lett., 2014, 5, No. 11, 1935–1942, doi:10.1021/jz500428s. 57. Beglov D., Roux B., J. Phys. Chem. B, 1997, 101, 7821, doi:10.1021/jp971083h. 58. Kovalenko A., Hirata F., Chem. Phys. Lett., 1998, 290, 237, doi:10.1016/S0009-2614(98)00471-0. 59. Hirata F. (Ed.), Molecular Theory of Solvation, Kluwer Academic Publishers, Dordrecht, 2003. 60. Yoshida N., Imai T., Phongphanphanee S., Kovalenko A., Hirata F., J. Phys. Chem. B, 2009, 113, 873–886, doi:10.1021/jp807068k. 61. Sergiievskyi V.P., FedorovM.V., J. Chem. Theory Comput., 2012, 8, No. 6, 2062–2070, doi:10.1021/ct200815v. 62. Palmer D.S., Sergiievskyi V.P., Jensen F., Fedorov M.V., J. Chem. Phys., 2010, 133, No. 4, 044104, doi:10.1063/1.3458798. 63. Liu Y., Zhao S., Wu J.Z., J. Chem. Theory Comput., 2013, 9, No. 4, 1896–1908, doi:10.1021/ct3010936. 64. Fu J., Wu J.Z., Fluid Phase Equilib., 2016, 407, 304–313, doi:10.1016/j.fluid.2015.05.042. 65. Chandler D., Phys. Rev. E, 1993, 48, 2898, doi:10.1103/PhysRevE.48.2898. 66. Lum K., Chandler D., Weeks J.D., J. Phys. Chem. B, 1999, 103, 4570, doi:10.1021/jp984327m. 67. Rein ten Wolde P., Sun S.X., Chandler D., Phys. Rev. E, 2001, 65, 011201, doi:10.1103/PhysRevE.65.011201. 68. Varilly P., Patel A.J., Chandler D., J. Chem. Phys., 2011, 134, No. 7, 074109, doi:10.1063/1.3532939. 69. Zhu C., Byrd R.H., Lu P., Nocedal J., ACM Trans. Math. Software, 1997, 23, No. 4, 550–560, doi:10.1145/279232.279236. 70. Lazaridis T., J. Phys. Chem. B, 1998, 102, 3542–3550, doi:10.1021/jp972358w. 71. Sergiievskyi V., Jeanmairet G., Levesque M., Borgis D., J. Chem. Phys., 2015, 143, No. 18, 184116, doi:10.1063/1.4935065. 72. Jeanmairet G., Levesque M., Sergiievskyi V., Borgis D., J. Chem. Phys., 2015, 142, No. 15, 154112, doi:10.1063/1.4917485. 73. Rosenfeld Y., J. Chem. Phys., 1993, 98, No. 10, 8126–8148, doi:10.1063/1.464569. 74. Oettel M., J. Phys.: Condens. Matter, 2005, 17, No. 3, 429, doi:10.1088/0953-8984/17/3/003. 75. Tang Y., J. Chem. Phys., 2004, 121, 10605–10610, doi:10.1063/1.1810473. 33005-13 https://doi.org/10.1063/1.1520530 https://doi.org/10.1088/0953-8984/22/6/063102 https://doi.org/10.1002/aic.10713 https://doi.org/10.1146/annurev.physchem.58.032806.104650 https://doi.org/10.1016/j.cplett.2009.04.077 https://doi.org/10.1063/1.3589142 https://doi.org/10.1021/jp210817s https://doi.org/10.1063/1.4734009 https://doi.org/10.1063/1.4769729 https://doi.org/10.1063/1.4824737 https://doi.org/10.1021/jz301956b https://doi.org/10.1080/00268976.2014.899647 https://doi.org/10.1021/jz500428s https://doi.org/10.1021/jp971083h https://doi.org/10.1016/S0009-2614(98)00471-0 https://doi.org/10.1021/jp807068k https://doi.org/10.1021/ct200815v https://doi.org/10.1063/1.3458798 https://doi.org/10.1021/ct3010936 https://doi.org/10.1016/j.fluid.2015.05.042 https://doi.org/10.1103/PhysRevE.48.2898 https://doi.org/10.1021/jp984327m https://doi.org/10.1103/PhysRevE.65.011201 https://doi.org/10.1063/1.3532939 https://doi.org/10.1145/279232.279236 https://doi.org/10.1021/jp972358w https://doi.org/10.1063/1.4935065 https://doi.org/10.1063/1.4917485 https://doi.org/10.1063/1.464569 https://doi.org/10.1088/0953-8984/17/3/003 https://doi.org/10.1063/1.1810473 V. Sergiievskyi, M. Levesque, B. Rotenberg, D. Borgis Сольватацiя в атомних рiдинах: зв’язок мiж теорiєю гауссового поля i функцiоналом густини В. Сергiєвський1, M. Левек1, Б. Ротенберг2, Д. Боржiс1,3 1 Унiверситет Сорбонна, Унiверситет П’єра iМарiї Кюрi, Вища нормальна школа, Париж, Францiя 2 Унiверситет Сорбонна, Унiверситет П’єра iМарiї Кюрi, Париж, Францiя 3 Будинок моделювання, Унiверситет Парi-Сюд, Унiверситет Парi-Саклє, Жiф-сюр-Iветт, Францiя Для проблеми молекулярної сольватацiї, що формулюється як рiдина в зовнiшньому потенцiальному полi, створеному молекулами довiльної форми, якi розчиненi в розчиннику, ми приводимо зв’язок мiж теорiєю гауссового поля, виведеною Давидом Чандлером [Phys. Rev. E, 1993, 48, 2898] i класичною тео- рiєю функцiоналу густини (DFT). Ми показуємо, що результати Чандлера щодо сольватацiї твердого ко- ру довiльної форми можуть бути зрегенерованi або шляхом мiнiмiзацiї лiнеаризованого HNC функцiо- налу, використовуючи допомiжне поле множникiв Лагранжа для накладання умови зникаючої густини всерединi кору, або мiнiмiзацiєю цього функцiоналу напряму в областi зовнi кору, що є насправдi про- стiшою процедурою. Цi еквiвалентнi пiдходи порiвнюються з двома варiантами DFT, або в наближеннi HNC, або в наближеннi частково лiнеаризованого HNC, для сольватацiї розчиненої речовини iз взаємодi- єю Леннарда-Джонса зi зростаючим розмiром в розчиннику iз леннард-джонсiвською взаємодiєю.Щодо порiвняння з моделюванням методом Монте Карло, всi цi теорiї дають прийнятнi результати для нео- днорiдної структури розчинника, але є повнiстю поза дiапазоном для сольватацiйних вiльних енергiй. Це може бути поправлено в DFT за допомогою додавання твердосферної мiсткової поправки до функцiоналу HNC. Ключовi слова: статистична механiка, класичнi плини, 3-вимiрнi системи, теорiя функцiоналу густини, теорiя гауссового поля 33005-14 Introduction Density functional theory and HNC approximation Chandler's Gaussian field theory Linearised and partially-linearised HNC approximations and connection to Gaussian field theory Hard-sphere bridge correction Conclusion Connection between ``inner'' and ``outer'' DFT formulations, and Chandler's GFT Notation Lagrange multipliers minimization Direct minimization in outer volume (reduced number of variables)

Ähnliche Einträge