A. Three-layer model of an ion in electrolyte solution* 5 10 15 20 In Sec. II.B, we used a model of the ion which is not fully selfconsistent: on the one hand, the ion is considered a point charge with respect to the ion-solvent interactions (i.e., the cavity radius Rcav is set to zero), but on the other hand, the ion-ion interactions are modeled with a hard-sphere potential. For this reason, if one sets LD = ∞ in Eq. (39), instead of obtaining the Born energy of a charge of finite size, Eq. (11), one obtains the energy of a point charge, Eq. (10). The self-consistent model will assume that the ion is placed into a cavity (r < Rcav, subscript “cav”), inside of which = 0 and Q = 0; the respective Coulomb-Ampere law reads: 02cav ei δ(r ) , r < Rcav. (S1) Outside the cavity (r > Rcav), there is a medium of dielectric permittivity and quadrupolarizability Q, with free charges of density = -/LD2. However, these free charges cannot approach the ion to a distance smaller than R = R+ + R-, so that near the ion there is an ion-free, “insulating” layer (subscript “ins”) (S2) 2ins L2Q 4ins 0 , R > r > Rcav; The diffuse electric layer is situated at r > R (subscript “dif”): (S3) 2dif L2Q 4dif dif / L2D , r > R. 25 55 60 65 70 75 The general solution of Eqs. (S1)-(S3) (which fulfils the condition dif → 0 at r → ∞) is: cav 0 X / r ; exp r / LQ exp r / LQ A1 A A ; r r r exp(r / lQ ) exp(r / lD ) dif BD BQ . (S4) r r These equations generalize Eq. (37). There are 8 integration constant to determine: 0, X, A0, A1, A+, A-, BD and BQ. The three conditions at r = Rcav (the boundary cav|ins) are derived in Ref. [48]. The first one is the condition for continuity of the potential: (S5) cav (Rcav ) ins (Rcav ) . 80 ins A0 30 85 90 The second one is the Gauss law Dins, r ( Rcav ) Dcav, r ( Rcav ) S , 35 which involves the quadrupolar moment density and E through Eq. (2). In explicit form and after some simplifications, it reads Q d 2 Eins, r (S6) 0 Ecav, r r R 0 , Eins, r cav 3 dr 2 r R cav where Er = -d/dr. The last boundary condition reads48: Qins, rr ( Rcav ) Qcav, rr ( Rcav ) 0 , 40 (S7) or in explicit form, E dE (S8) Q ins, r ins, r 0. d r r r Rcav The conditions at r = R (the boundary ins|dif) are the same as those used in Sec. III.A: ins (R) dif ( R) ; dins / dr |r R ddif / dr |r R ; 45 50 d 2ins / dr 2 |r R d 2dif / dr 2 |r R ; d3ins / dr 3 |r R d3dif / dr 3 |r R . (S9) The final 8th equation is the electroneutrality condition (38). The solution of the linear system of equations (S5),(S6),(S8), (S9) and (38) for the integration constants in Eq. (S4) is trivial, but the result for 0 and the energy uel = ei0/2 is too complicated. Therefore, we will only discuss several important features of the solution (S4). The first one is that in the limit of Rcav → 0, it simplifies to Eq. (37), which justifies the non-classical boundary condition for finite potential, used in Sec. II (see also the discussion in Ref. [48]). In the limit of infinitely dilute solution (LD → 0), the result for uel following from Eq. (S4) simplifies to the Born energy (11) (while the simplified Eq. (37) yields (10)). At least for small ions, the final result stemming from Eq. (S4) differs insignificantly from its simplified variant (39) obtained under the assumption that Rcav = 0. With larger ions such as SO42, the deviations are more significant. This is demonstrated in Fig. 5, where the three layer model is compared to the simplified equation (41). The simplified model (the sum of Eq. (41) and (46)) always predicts slightly lower energy. In view of the complicated form of the exact expression 0 and the fact that we neglect various other effects which yield corrections of similar order of magnitude as the difference between the three-layer model and the simplified model in Sec. II.B, we believe that this complicated correction is meaningless when theory is compared to experimental data. Let us finally comment on the interesting fact that the same classical expression (36) will be obtained for the Debye-Hückel energy from the three-layer model (Eqs. (S1)-(S3) with LQ = 0) or the simplified model defined with Eqs. (34) and (5) with LQ = 0 (although the Born energies corresponding to the two models are different). This means that Born energy (i.e., the ion-solvent interaction energy) and Debye-Hückel (energy of the interaction ion-ionic atmosphere) are not coupled in non-quadrupolarizable solvents and there one can analyze both interaction energies separately. In the case of non-zero quadrupolarizability of the medium, this is not true anymore. Therefore, the diffuse layer and the Born potential distribution must be rigorously considered together. This coupling is rather weak, which allows us to use the approximate equation (48) where the electrostatic contribution u0,i + uel,i of the two interaction energies is assumed additive. Nevertheless, the fact that ion-solvent and ion-ion interactions are coupled (or in other words, the structure of the ionic atmosphere is interrelated with the structure of the dipole cloud around the ion) is yet another fundamental property of our general quadrupolar Coulomb-Ampere law (4), which has no analogue within the classical Poisson-Boltzmann equation. *A Referee of Ref. [48] contributed to a certain extent to the considered model. 95 B. Maple 17 code for the calculation of the integration constants in Eq. (37) 65 (paste the following code in the Maple worksheet and execute it with the “!!!” icon) 5 10 15 20 25 30 35 40 45 50 55 60 > restart; > #solutions for the potentials at r<R (phi1) and at r>R (phi2) - Eq (36): phi1:= A0+Ap*(exp(+r*sqrt(1/lQ^2+1/lD^2))1)/r-Am*(exp(-r*sqrt(1/lQ^2+1/lD^2))-1)/r; phi2:= BD*exp(-r/lD)/r+BQ*exp(-r/lQ)/r; #r-components of the field E: Er1:= -diff(phi1,r): Er2:= -diff(phi2,r): #charge density rho:=-epsilon*phi2/LD^2; #self potential phi0:= limit(phi1,r=0); > #relation between lD, lQ, LD and LQ - Eq (19): LD:= sqrt(lD^2+lQ^2); LQ:= sqrt(1/(lD^2+lQ^2))*lQ*lD; > #verification that phi1 and phi2 are solutions of the generalized Coulomb-Ampere law - Eq (5): de1:= 1/r^2*diff(r^2*Er1,r) - LQ^2/r^2*diff( r^2*diff(1/r^2*diff(r^2*Er1,r),r) ,r): simplify(%); de2:= 1/r^2*diff(r^2*Er2,r) - LQ^2/r^2*diff( r^2*diff(1/r^2*diff(r^2*Er2,r),r) ,r)1/epsilon*rho: simplify(%); > #boundary conditions - 4 equations for continuity and the condition for electroneutrality: bc1:= subs(r=R, phi1-phi2=0): bc2:= subs(r=R, diff(phi1-phi2,r)=0): bc3:= subs(r=R, diff(phi1-phi2,r,r)=0): bc4:= subs(r=R, diff(phi1-phi2,r,r,r)=0): eln:= ei+int(rho*4*Pi*r^2,r=R..infinity)=0 assuming R>0,lD>0,lQ>0: > #solution for A0, Ap, Am, BQ, BD: sol:= simplify(solve({bc1,bc2,bc3,bc4,eln},{A0, Ap, Am, BQ, BD})) assuming lD>0,lQ>0,lD>lQ,R>0: assign(sol); > #activity coefficient lngamma:= ei/2/T*simplify(phi0ei/4/Pi/epsilon/LQ) assuming lD>0,lQ>0,lD>lQ,R>0: #the result simplifies to Eq (38) > #limit for diluted solutions LQ:='LQ': LD:='LD': lD:= LQ/sqrt(1/2-1/2*sqrt(1-4*LQ^2/LD^2)); lQ:= LQ/sqrt(1/2+1/2*sqrt(1-4*LQ^2/LD^2)); #in first approximation, it yields the limiting Debye-Hueckel: convert(series(lngamma,LD=infinity,3),polyno m): lngammaDH:= simplify(%) assuming LQ>0; > #extended Debte-Huckel model as the limit at LQ = 0 70 75 80 85 90 95 lngammaDHe:= simplify(limit(lngamma,LQ=0,right)) assuming LD>0,R>0; > #parameter values with(ScientificConstants): T:= GetValue(Constant('k'))*(273.15+25): e:= GetValue(Constant('e')); ei:=e; epsilon:= 78.303*GetValue(Constant('epsilon[0]')); LD:= sqrt(epsilon*T/2/e^2/c); c:= GetValue(Constant('N[A]'))*1e3*cM; LQ:=2.1e-10; R:=2.35e-10; > #plot of the potential - Fig. 3 plot([ [1e10*r,1e3*evalf(subs(cM=0.01,phi2)),r=2.35 e-10..2e9],[1e10*r,1e3*evalf(subs(cM=0.01,phi1)),r=0 ..2.35e-10], [1e10*r,1e3*evalf(subs(cM=1,phi2)),r=2.35e10..2e9],[1e10*r,1e3*evalf(subs(cM=1,phi1)),r=0..2 .35e-10], [2.35,y,y=0..48] ],r=0..20,phi,color=[ ColorTools:-Color([0., 0.65, 0.85]),ColorTools:-Color([0.9, 0.1, 0.1]), ColorTools:-Color([0., 0.65, 0.85]),ColorTools:-Color([0.9, 0.1, 0.1]), black],thickness=[2,1,2,1,1],linestyle=[soli d$4,dot],axesfont=[Times,Roman],tickmarks=[4 ,5]); C. Summary of the initial parameters of the electrolytes used for the calculations 5 10 35 For the convenience of the reader, a summary of the parameters involved in the calculations in Sec. III.B is provided here. We used two different regression formulas for the experimental activity coefficients presented in Fig. 4-Fig. 6. For all aqueous electrolytes, the following equation is used15,24: 3 Z Z A k C n (S10) lg D is m n k isCm ; 1 B k isCm n 1 osm 1 A Cm 1 B Cm 0Cm 1Cm exp(1Cm ) kCm2 ; (S11) here A = 1.294 (kg/mol)1/2; B = 3.2 (kg/mol)1/2; 0 = 0.1058 kg/mol; 1 = 0.5578 kg/mol; 1 = 2 kg/mol; k = 0.083476 (kg/mol)2. The activity coefficient was calculated accordingly through Gibbs-Duhem relation. Table S I. Values of the parameters of the regression formula (S10) for the activity coefficient of the aqueous electrolytes, used in Fig. 4 and Fig. 5. salt 1 B 2 3 i Cmax Ref. (kg/mol)1/2 kg/mol (kg/mol)2 (kg/mol)3 mol/kg ii NaF 1.280 -0.0180 1 [24] KCl 1.295 7.0×10-5 3.60×10-3 -1.95×10-4 0 0 5 [24] Na2SO4 1.280 -0.06081 5.17×10-3 -1.3×10-4 4 [15] MgSO4 1.022 -2.78×10-4 1.74×10-3 0 3 [15] (i) Eq. (S10) is valid up to Cm = Cmax. (ii) All solutions are aqueous. 25 The parameters of the empirical (Cel) dependence (43) used for the curve-fitting of the experimental data for the static dielectric permittivity of salt solutions are summarized in Table S II. Table S II. Parameters of Eq. (43) for the dependence of the dielectric permittivity of electrolyte solutions on the molarity of the solution. salt 1 2 3 4 6 (l/mol)1/2 l/mol (l/mol)3/2 (l/mol)2 (l/mol)3 ii 30 i Cmax Ref. mol/l NaF 0 -5.84 0 -4.3 0 0.9 [20] KCl 0 -12.4 0 0.95 0 4 [20] Na2SO4 4.4 0 -28.1 12.5 0 1.6 [58] MgSO4 14.7 0 0 -23.4 7.6 1.1 [20] NaBr in CH3OH 0 -47 0 119 -144 0.5 [20] (i) Eq. (42) is valid up to Cel = Cmax. (ii) All solutions with the exception of NaBr in CH3OH are aqueous. The radii Rcav and R, as well as the partial molar volumes of the electrolytes used for the calculations in Sec. III.B are given in Table S III. Note that the dependence of the mass density of the Table S III. Cavity radii Rcav+ and Rcav- of the cation and the anion, minimal distances R of approach between two counterions and partial molar volumes Vel of the electrolytes at infinite dilution. All data are calculated by using the crystallographic radii and the single ion partial molar volumes selected by Marcus62. salt where AD is the Debye-Hückel coefficient in water (-0.5108 kg1/2/mol1/2 at 25˚), and the empirical parameters B and n are summarized in Table S I. For NaBr in methanol, we used Pitzer’s regression formula for the osmotic coefficient data from Ref. [20]: 15 20 40 methanol solution on Cel was neglected, due to the absence of data. For the relation between molality and molar concentration in methanol, we used Cm = Celm, where m was assumed equal to the mass density of pure methanol, 791.8 kg/m3. 45 i i Rcav+ = Rcav- = gV1/3(R+ + Lcav+) gV1/3(R- + Lcav-) R = R + + R- Vel Å Å Å l/mol NaF 1.90 1.48 2.35 2.4 KCl 2.30 2.02 3.19 26.8 Na2SO4 1.90 2.57 3.32 11.6 MgSO4 1.56 2.57 3.02 -7.2 NaBr in CH3OH ii ii 2.98 - 1.90 2.19 (i) Lcav+ = 0.69Å, Lcav- = 0 Å, gV = 1.39, cf. Ref. [48]. (ii) Cf. also supplementary material D. D. Size parameters of the ion cavity in methanol solutions – comparison with experimental solvation energies 5 10 In this section, we will use the data collected by Kelly et al. 63 to test our formula for the cavity radii, Eq. (44), and to defend our choice of the values of the parameters Lcav+, Lcav- and gV. The data of Kelly et al. 63 is insufficient to repeat the detailed optimization procedures used in Ref. [48] (data for only six simple ions in methanol is available). It nevertheless is enough to test our assumptions. 40 45 50 55 15 20 FIG. S 1. Comparison between the experimental solvation energies of various ions in methanol assembled by Kelly et al.63 and the theoretical equation (S12). We used the comparison to determine the solvation energy of H+ in methanol as the single fitting parameter. We use the data63 for the conventional solvation energies 0+ of the anions and 0+ of the cations. Our model (11) of the Born energy predicts for their dependence on the crystallographic radii Ri the formulae: 1 3LQ / Rcav ei2 1 1 Δ0+ Δ0,H+ ; 2 2 8πRcav 0 1 3LQ / Rcav 3LQ / Rcav Δ0 Δ0,H 25 (S12) where 0,H+ is the solvation energy of the conventional ion, H+. We will determine the value of 0,H+ from the data, by minimizing the dispersion Δ2 (Δ0,H ) 35 Δ i 0 i ,th (Δ0,H ; Ri ) Δ0i ,exp ( Ri ) 2 ; (S13) N f here 0i,th is the calculated solvation energy from Eqs. (S12) for an ion of crystallographic radius Ri (using Lcav+ = 0.69 Å, Lcav- = 0 and gV = 1.39 according to our assumption that the dielectric lengths and the packing factor of water and methanol are similar); 0i,exp is the respective experimental value from Ref. [63]; N is the number of data points (N = 6) and f is the number of free parameters (f = 1). Note that Eq. (S13) does not involve a term for the standard states in gas and in methanol solution 48, due to the choice of Kelly et al.63 for the standard state in the gas phase. The value of LQ in Eq. (S12) is unimportant for the final result since the 1/ term in Eqs. (S12) is negligible in comparison with 30 1 3LQ / Rcav ei2 1 1 , 2 8πRcav 0 1 3LQ / Rcav 3L2Q / Rcav 1/0 (as it is with water48). Nevertheless, we used the value LQ = 1.1 Å obtained in Sec. III. B for methanol. The optimization of yield for the best value of the solvation energy of H+ in methanol the value 0,H+ = -1052±12 kJ/mol. The value is by 4±12 kJ/mol higher than the hydration energy of H+, -1056 kJ/mol62, which is another manifestation of the fact that the first solvation shells of simple ions in water and in methanol are similar. Kelly et al.63 also obtained almost equal solvation energies of H+ in methanol and in water (0,H+ is by 12±10 kJ/mol higher in methanol; note that they preferred another value for the hydration energy of H+). From the TATB approximation, for the same difference the value 9 kJ/mol was obtained64. The comparison between Eq. (S12) and the experimental data is shown in FIG. S 1. The good agreement confirms the validity of the assumed values of Lcav+, Lcav- and gV. We can conclude that our model for the cavity size, Eq. (44), provides not only a good first approximation for the hydration energies but also allows for a fast determination of the conventional energies of accuracy comparable to the existing methods. E. List of symbols 5 10 15 20 25 30 35 40 45 50 Ccr critical concentration for potential oscilations, Ccr = kBT/2kise2LQ2 Cel electrolyte concentration (in units [m–3] or [M]) Ci ith ion concentration, Ci = iCel Cm electrolyte molality [mol/kg] D electric displacement field E electric field intensity e elementary charge ei ith ion charge, ei = Zie kis ratio between ionic strength and Cel, kis = (+Z+2+-Z-2)/2 kB Boltzmann constant Lcav dielectric length, Lcav+ = 0.69 Å and Lcav- = 0 LD Debye length, LD2 = kBT/2kise2Cel lD first characteristic length of the quadrupolar Debye-Ampere law, Eqs. (18)-(19) lIm imaginary part of lD at LD < 2LQ, Eqs. (26) LQ quadrupolar length, LQ2 = Q/3 lQ second characteristic length of the quadrupolar DebyeAmpere law, Eqs. (18)-(19) lRe real part of lD at LD < 2LQ, Eqs. (26) Mel electrolyte molecular mass Ms molar mass of the solvent P macroscopic polarization vector, P = QE Q macroscopic quadrupolarization tensor, Q=Q(E - U∙E/3) R distance of closest approach between counterions, R = R+ + RRi crystallographic or Pauling radius of the bare cation/anion Rcav,i cavity radius of the ith ion, Rcav,i = gV1/3(Ri + Lcav) r spherical radial coordinate T temperature u0 self-energy (Born energy) of the ion u0∞ Born energy u0 at infinite dilution uel full electric energy of the ion, u0 + kBTlnDH Vel partial molar volume of the electrolyte xi molar fraction of the ith ion Zi dimensionless charge of the ion P macroscopic polarizability of a medium, P = – 0 Q macroscopic quadrupolarizability of a medium ± bulk mean activity coefficient, (++-)ln±=+ln++-lnBorn self-salting-out (Born energy or ion-medium interaction) contribution to the activity coefficient; kBTlnBorn = u0 – u0∞ DH Debye-Hückel (ion-ion interaction) contribution to the activity coefficient i activity coefficient of the ith ion x contribution to the activity coefficient originating from the choice of standard concentration, Eq. (51) Dirac delta-function absolute dielectric permittivity of the electrolyte solution 0 absolute dielectric permittivity of vacuum ∞ absolute dielectric permittivity of pure solvent i stoichiometric number of the ith ion free charge density of the electrolyte solution 55 m mass density of the solution m0 mass density of pure water electrostatic potential 0 electrostatic potential at r = 0