You are on page 1of 27

2

Ion Solvation and Resolvation


TOSHIYUKI OSAKAI Department of Chemistry, Kobe University, Kobe, Japan
KUNIYOSHI EBINA Division of Sciences for Natural Environment, Kobe
University, Kobe, Japan

I. INTRODUCTION

Much attention has been directed since olden times towards ion solvation, which is a key
concept for understanding various chemical processes with electrolyte solutions. In 1920, a
theoretical equation of ion solvation energy (G0s ) was first proposed by Born [1], who
considered the ion as a hard sphere of a given radius (r) immersed in a continuous medium
of constant permittivity (), and then defined G0s as the electrostatic energy for charging
the ion up to ze (z, the charge number of the ion; e, the elementary charge):
 
N z2 e2 1
G0s ðBornÞ ¼  A 1 ð1Þ
80 r 
where NA is the Avogadro constant, and 0 the permittivity of vacuum. Since the term ‘‘1’’
in the parentheses of the RHS of Eq. (1) is 1/[the relative permittivity of vacuum (¼ 1, by
definition)], G0s (Born) can also be regarded as the electrostatic energy required for the
transfer of the ion from vacuum to the medium having the permittivity of . Accordingly,
0;O!W
the Gibbs free energy of ion transfer (Gtr ), i.e., the resolvation energy for the
transfer of an ion from one medium (e.g., organic solvent, O) to another (e.g., water,
W) can be expressed as
 
N A z2 e2 1 1
Gtr0;O!W
ðBornÞ ¼   ð2Þ
80 r O W
where O and W are the relative permittivities of O and W, respectively.
The Born equation thus derived is based on very simple assumptions that the ion is a
sphere and that the solvents are homogeneous dielectrics. In practice, however, ions have
certain chemical characters, and solvents consist of molecules of given sizes, which show
various chemical properties. In the simple Born model, such chemical properties of ions as
well as solvents are not taken into account. Such defects of the simple Born model have
been well known for at least 60 years and some attempts have been made to modify this
model. On the other hand, there has been another approach that focuses on short-range
interactions of an ion with solvent molecules.
The purpose of this chapter is to discuss mainly recent development in the theory of
ion solvation energy. Because we allowed much space for introducing our new, non-
23

Copyright © 2001 Marcel Dekker, Inc.


24 Osakai and Ebina

Bornian theory of the Gibbs free energy of ion resolvation, a vast amount of work dealing
with ion solvation and resolvation could not fully be reviewed. The reader is referred to
some books [2–5] and review articles [6–10] for a general survey.

II. MODIFICATIONS OF THE BORN MODEL

The defects of the Born equation [1] may be put into two categories; one is for large ions
and the other for small ions. The defect for large ions is concerned with the fact that the
Born equation does not take account of the energy of the formation of a cavity for
introducing an ion into the solvent. This energy, which corresponds to the so-called
‘‘solvophobic’’ interaction, is more noticeable for large ions. Consequently, a modification
has generally been made by adding a nonelectrostatic term, G0s ðneÞ, to the Born equa-
tion:

G0s ¼ G0s ðelÞ þ G0s ðneÞ ð3Þ

where G0s ðel) is the electrostatic term which is given by the Born equation (or its mod-
ifications). As described by the previous authors [9], the division shown in Eq. (3) is
somewhat arbitrary, because different effects may overlap. However, such a division is
useful for evaluating the individual effects in a heuristic way.
The evaluation of the nonelectrostatic term is usually empirical and multifarious.
Based on a skillful argument, however, Volkov and coworkers [9,11,12] proposed that the
nonelectrostatic term of the ion transfer energy, Gtr 0;O!W
(ne), should be expressed by a
semiempirical equation called the Uhlig equation [13], which is given by

Gtr
0;O!W
ðneÞ ¼ 4NA r2 O;W ð4Þ

when the surface tension at the boundary of organic solvent (O) with air is smaller than
that of water (W) with air. In this formula, the difference in the energies for forming a
surface around an ion in O and in W is to be evaluated simply by using the interfacial
tension (O;W ) between the two phases. The Uhlig formula is shown to be valid for O;W >
10 mN m1 and r > 0:2 nm [9].
The other defect of the Born equation is for relatively small ions and is concerned
with the change in the solvent structure around an ion when it is inserted in the solvent;
this change is caused by the high electric field formed by the ion. Many attempts have been
made since long ago to mitigate this defect by modifying the Born equation in various
ways: (1) by correcting the ionic radii (for example, see Ref. 14), (2) by taking account of
the effect of dielectric saturation (i.e., the lowering of the permittivity of solvents adjacent
to an ion due to the high electric field) [9,11,12,15–17], or (3) by doing both (1) and (2) [18].
In modification (1), a constant value (r) is added to the crystal ionic radius (r), and
the value of r þ r is used in place of r in Eq. (1). By using the values of r þ r (dependent
on the ionic valence) determined empirically, fairly good agreement was obtained between
experimental and theoretical values for the ion hydration energy. An example [14] is
shown in Fig. 1. The solid lines in the figure represent the plots obtained by using the
crystal ionic radii for r in the Born equation. It should be noted that the plots for the
cations showed a curved line inconsistent with the Born equation [See Eq. (1) again].
However, adding a constant rð¼ 0:85 Å) to the crystal ionic radius for all the cations
could make the curved line straight, as shown by a dashed line. But, the physical meaning

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 25

FIG. 1 Modification of the Born equation by correcting ionic radii. Solid curves for crystal radii,
dotted for corrected radii (for cations, r ¼ 0:85 Å; for anions, r ¼ 0:1 Å). (From Ref. 14.
Copyright 1939 the American Institute of Physics.)

of r still remains obscure. Such a modification seems unable to come to the essential
solution to this problem.
The modification by method 2 is more acceptable. Although several types of mod-
ifications have been reported, Abraham and Liszi [15] proposed one of the simplest and
well-known modifications. Figure 2(b) shows the proposed ‘‘one-layer’’ model. In this
model, an ion of radius r and charge ze is surrounded by a local solvent layer of thickness
(b  r) and dielectric constant 1 , immersed in the bulk solvent of dielectric constant b .
The thickness (b  r) of the solvent layer is taken as the solvent radius, and its dielectric
constant 1 is supposed to become considerably lower than that of the bulk solvent owing
to dielectric saturation. The electrostatic term of the ion solvation energy is then given by
    
N z2 e2 1 1 1 1 1
G0s ðelÞ ¼ A 1  þ 1 ð5Þ
80 r 1 r b b b

On the assumption that 2 ¼ 2, the theoretical values of the ion solvation energy were
shown to agree well with the experimental values for univalent cations and anions in
various solvents (e.g., 1,1- and 1,2-dichloroethane, tetrahydrofuran, 1,2-dimethoxyethane,
ammonia, acetone, acetonitrile, nitromethane, 1-propanol, ethanol, methanol, and water).
Abraham et al. [16,17] proposed an extended model in which the local solvent layer was
further divided into two layers of different dielectric constants. The nonlocal electrostatic
theory [9,11,12] was also presented, in which the permittivity of a medium was assumed to
change continuously with the electric field around an ion. Combined with the above-
mentioned Uhlig formula, it was successfully employed to elucidate the ion transfer energy
at the nitrobenzene–water and 1,2-dichloroethane–water interfaces.
It should be noted that in the modified Born equation of Eq. (5), the value of 1 ð¼ 2Þ
is extremely small compared with that of b ð¼ 78:54 for water). In the Born equation as
well as its modifications, the reciprocal of the permittivity (1=) crucially determines the
magnitude of the ion solvation energy. As b > 10 (i.e., 1=b < 0:1) for most polar sol-

Copyright © 2001 Marcel Dekker, Inc.


26 Osakai and Ebina

FIG. 2 (a) The Born model [1]. (b) The one-layer model proposed by Abraham and Liszi [15].
(From Ref. 10. Copyright the Japan Society for Analytical Chemistry.)

vents, we see how great the modification is when we set 1 ¼ 2 (i.e., 1=1 ¼ 0:5). It seems
far beyond the ‘‘modification.’’
What does the assumption ‘‘1 ¼ 2’’ really mean? Since the permittivity reflects the
extent of the rotational fluctuation of molecular dipoles of the solvent, the low permittivity
means that the solvent molecules are strongly restricted in their movements in the vicinity
of the ion which has the high electric field at the surface. This suggests that the short-range
ion–solvent interactions, including such chemical interactions as hydrogen bonds, should
play the most significant role in the ion solvation energy. As mentioned above, the Born
equation is essentially based on the long-range ion–solvent interaction. Ironically enough,
however, its modifications insist on the importance of the short-range interactions rather
than the long-range ones.

III. EMPIRICAL APPROACHES TO SHORT-RANGE INTERACTIONS

In the Born equation, the ion–solvent interaction energy is determined only by one phy-
sical parameter of the solvent, i.e., the dielectric constant. However, since actual ion–
solvent interactions include ‘‘specific’’ interactions such as the charge-transfer interaction
or hydrogen bonds, it is natural to think that the Born equation should be insufficient. It is
well known that the difference in the behavior of an ion in different solvents is not often
elucidated in terms of the dielectric constant.
Since around 1950, in studies of solvent effects for organic reactions, empirical
solvent parameters have been used; these parameters represent the capabilities of solvents
for the solute–solvent interactions (especially Lewis acid–base interactions). Though the
solute–solvent interactions should depend on the solute as well as on the solvent, the
empirical solvent parameters are considered to be irrelevant to solutes; in other words,
the use of only these parameters enables us to evaluate the solvation energies. Strictly

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 27

speaking, this is not correct. In practice, however, good correlations of the empirical
parameters with the solvent effects have been seen in a number of systems (for reviews,
see Refs. 19 and 20).
The best-known solvent parameters are the donor number [21] and acceptor number
[22] proposed by Gutmann and coworkers. The donor number (DN) for a donor solvent D
is defined as the positive value of the enthalpy difference H (kcal mol1 ) for the reaction
of D with an acceptor-halide SbCl5 (D þ SbCl5 ! D  SbCl5 ) in an inert medium such as
1,2-dichloroethane. DN is a fair measure for the donor properties of a solvent. The
correlations of DN with the solvation energies are known to be good particularly for
solvation of cations. A typical example [19] is shown in Fig. 3.
The acceptor number (AN) for a solvent A is determined by comparing the electron-
pair accepting ability of the solvent with that of SbCl5 , from the oxygen atom of triethyl-
phosphine oxide Et3 PO in 1,2-dichloroethane [22]. To put it concretely, the 31 P-NMR
chemical shift of Et3 PO in solvent A, ðEt3 PO  AÞ, which would reflect the electron
density around the 31 P nucleus, was measured, and then its relative value against the
chemical shift of Et3 PO  SbCl5 in 1,2-dichloroethane was defined as AN:
ðEt3 PO  AÞ
AN ¼  100 ð6Þ
ðEt3 PO  SbCl5 Þ

AN is known to show good correlations with the solvation energies of anions. Also, AN
has good correlations with other solvent parameters defined in different reaction systems,
e.g., Grunwald and Winstein’s Y-value [24], Kosower’s Z-value [25], Dimroth and
Reichardt’s ET -value [26,27], etc.
Taft et al. [28] proposed the solvatochromic parameters,  , , and , which describe
three solvent’s abilities, respectively, to stabilize a charge or a dipole by virtue of its
dielectric effect, to donate a proton (or accept an electron pair), and to accept a proton
(or donate an electron pair). It was shown that the AN for nonprotonic solvents correlates

FIG. 3 Correlation between the Gibbs free energy of transfer of Kþ from CH3 CN to a solvent and
DN. The free energies of transfer were determined by EMF measurements [23] based on the assump-
tion of negligible liquid junction in a cell: KðHgÞj0:01 M KClO4 (solvent) k0:1 M Et4 NPic
ðCH3 CNÞk0:01 M KClO4 ðCH3 CNÞjKðHgÞ. NM: nitromethane; TMS: tetramethylene sulfonate;
PDC: propanediol-1,2-carbonate; DMF: dimethylformamide; NMP; N-methyl-2-pyrrolidone,
DMA: dimethylacetamide; DMSO: dimethyl sulfoxide; HMPA: hexamethylphosphoramide.
(From Ref. 19. Copyright 1978 Plenum Press, New York.)

Copyright © 2001 Marcel Dekker, Inc.


28 Osakai and Ebina

well with  and for protonic solvents with a linear combination of  and . It was
therefore concluded that AN was, in fact, a combined measure of solvent polarity/polar-
izability and hydrogen bond donor ability. It was also shown that DN was linear to for
oxygen bases and RCN nitrogen bases, though the correlation broke down for pyridine. In
Taft et al.’s studies, the macroscopically defined solvent parameters were analyzed at a
molecular level; therefore their conclusions are full of interesting suggestions for the
understanding of solute–solvent interactions.

IV. A VOLTAMMETRIC STUDY WITH POLYANIONS

The above-mentioned solvent parameters are the solvent-side measures for describing the
solute–solvent interactions. Accordingly, they are useful for predicting the effects observed
when the solvent is changed, but inadequate when the solute is changed. On the other
hand, the Born equation includes the parameters for both the solvent and solute (ion), i.e.,
the dielectric constant () as a solvent parameter and the ionic charge and radius (z and r)
as ion parameters. This is an advantage of the Born equation and therefore the reason why
the Born equation has been widely used for many years notwithstanding a lot of criticisms.
However, in theoretical studies with the Born equation or its modifications, the r-depen-
dence of the solvation energy was generally examined for the ions with the common z-
value; the dependence on the z-value has not been elucidated successfully.
Previously, Osakai and coworkers [29–31] employed ion-transfer voltammetry to
determine the standard ion-transfer potentials (W O
) of heteropoly- and isopolyoxome-
0

talate anions (in short, polyanions) at the nitrobenzene (NB)/W and 1,2-dichloroethane
(1,2-DCE)/W interfaces; W 0
O
is directly related to the transfer energy by
0;O!W
Gtr
W
O
¼ 
0
ð7Þ
zF
The values of W O
as well as Gtr
0 0;O!W
, obtained for the NB/W interface, are summar-
ized in Table 1. As seen, the value of WO
for the polyanions, so their transfer energies as
0

well, depend only on the ionic size and charge. Thus there was no indication that the ion-
transfer potential (or the energy) was significantly influenced by such ‘‘specific’’ solvation
as due to electron localization in a polyanion. This seems to support the hypothesis that
the surface charge of such a polyanion is nonlocalized. If we assume the polyanion to be a
hard sphere with a uniform surface electric field strength (E), then E is a function only of z
and r:
ze
E¼ ð8Þ
40 r2
In order to clarify the role of the surface field strength, we have plotted the values of W O

against E [see Fig. 4(A) [29]]. Surprisingly, all the data lie on a straight line despite the
large differences in sizes, charges, and even structures, as seen in the figure. A similar linear
plot was also obtained for the polyanion transfer at the 1,2-DCE/W interface [30].
Because polyanions are generally quite large, a contribution from the solvophobic
interaction (i.e., the cavity formation energy) to its solvation energy should not be
neglected. Accordingly, from the observed W O
-values, the contribution was subtracted
0

with the help of the Uhlig equation [Eq. (4)]. Even after this subtraction, the linear
relationship still manifested itself in the resultant ‘‘electrostatic’’ term, W O
ðelÞ
0

¼ ½Gtr 0;O!W
ðelÞ=zF, as shown in Fig. 4(B) [33]. This excellent correlation strongly

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 29

TABLE 1 Values of W 0 0;O!W


O
and Gtr of Polyanions for the NB/W System [29] and Their
Values of r and E

r Ea WO

0
G0;O!W
tr
No. Polyanion (nm) (10 V m1 )
10
(V) (kJ mol1 )

1 ; -[XM12 O40 4 0.56b 1:84 0:067  0:003 25.9


(X ¼ Si; Ge; M ¼ Mo; WÞ
2 ; -½XM12 O40 3 0.56b 1:38 0:248  0:004 71.8
(X ¼ P; As; M ¼ Mo; WÞ
3 -½X2 Mo18 O62 6 0.648c 2:06 0:005  0:000 2.9
ðX ¼ P; AsÞ
4 -½S2 Mo18 O62 4 0.648c 1:37 0.269 103.8
5 ½S2 VMo17 O62 5 0:648c 1:72 0.085 41.0
6 ½P2 Mo18 O61 4 0.644c 1:39 0.239 92.2
(containing P2 O47 Þ
7 ½Mo6 O19 2 0.437c 1:51 0.164 31.6
8 ½VMo5 O19 3 0.437c 2:26 0:119 34:4
9 -½Mo8 O26 4 0.485c 2:45 0:137 52:9
a
Calculated from Eq. (8).
b
Literature value [32].
c
Evaluated using the simple relation r ðnmÞ ¼ 0:164n1=3 (with n the number of oxygen atoms in the polyanion)
[30].
Source: From Ref. 33.

FIG. 4 Plots of (A) WO


and (B) O
(el) of polyanions at the NB/W interface against E. (From
0 W

Ref. 33. Copyright 1996 Elsevier Science B.V., Amsterdam.)

Copyright © 2001 Marcel Dekker, Inc.


30 Osakai and Ebina

suggests that the surface field strength E should play an important role in the resolvation
energies of polyanions.
Then, what would the linear E-dependence of W O
mean? If we suppose that the
0

Gtr0;O!W
for the polyanions is given by the Born equation, W O
ðelÞ should be propor-
0

tional to z=r [cf. Eqs. (2) and (7)]; this dependence would come from the overall inte-
gration of a function of the electric field. In actuality, however, W 0
O
ðel) depends on the
2
value of E at the surface, which is proportional to z=r . This suggests that the short-
range ion–solvent interactions, rather than the Born-type long-range electrostatic inter-
action, play a major role in the z-dependence of the resolvation energy of polyanions.
Thus it is unsuitable to refer to W O
ðelÞ as the ‘‘electrostatic’’ term. Instead, the term
0

will be referred to as the charge (z)-dependent term, being written as W O


ðz-dep) [cf.
0

Eq. (26)].

V. A MODEL HAMILTONIAN APPROACH TO ION–SOLVENT INTERACTIONS

The authors [33] have elucidated the linear dependence of W O


ðz-depÞ on E for the
0

polyanions by a quantum chemical consideration. A model Hamiltonian approach to


the charge transfer (CT) interaction between a polyanion and solvents has been made
on the basis of the Mulliken’s CT complex theory [34].
Let us consider a case in which an ion (donor, Dz ) and a solvent (acceptor, A) form a
CT complex. The ground state energy WN (see Fig. 5) can be obtained as a solution of the
secular equation:
 
 W0  X W01  S01 X 

 ¼0 ð9Þ
 W01  S01 X W1  X 
where W0 is the energy in a no-bond state, j 0 i ¼ j ðDz    AÞi, in which Dz and A come
into contact without changing their electronic configurations; W1 is the energy in a dative
state, j 1 i ¼ j ðDzþ1  A Þi, corresponding to the transfer of an electron from Dz to A;
and W01 and S01 are defined by W01 ¼ h 0 jHj 1 i (H being the Hamiltonian operator) and
S01 ¼ h 0 j 1 i, respectively. The CT interaction energy W, i.e., the resonance energy in
the ground state is given by
W ¼ W0  WN ð10Þ

FIG. 5 The charge transfer between an ion (donor, Dz ) and a solvent (acceptor, A).

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 31

In the previous paper [33], the effect of the ionic charge z on W was investigated
for the CT interaction between a polyanion and water. The effect will come out through
W1 and W01 .
By using a simple model in which an electron in the surface atom of a polyanion is
transferred to a solvent molecule, the effect of z on W through W1 is considered as
follows: if the ionic radius r is much larger than the CT distance d and the radius of the
surface atom rO , the z-dependence of W1 can be expressed in terms of the surface field
strength E of the polyanion as
 
ed
W1 W~ 1 þ E ð11Þ
r
where the E-independent term will be approximated as
e2
W~ 1 ¼ W0 þ Ip0  Ea ðAÞ  ð12Þ
40 r d
with Ip0 being the ionization potential of the surface atom, Ea ðAÞ the electron affinity of the
solvent, and r the relative permittivity of the solvent in the first solvation shell (being
assumed to be equal to the optical permittivity, op ¼ 1:8 for water). The last term has
been introduced as a correction factor for electrostatic effects of the liquid medium.
Next, the effect of z on W through the transition matrix element W01 is considered as
follows: for rigorous determination of W01 , all electrons in the system should be treated.
However, for the sake of simplicity, we devote our attention only to the transferring elec-
tron; the other electrons would be regarded as forming the ‘‘effective’’ potential Veff ðxÞ for
the transferring electron (x the coordinate of the electron given from the ion center). This
enables us to reduce the many-body problem to a one-body problem:
D   E
W01 HOMO D
H one-body  LUMO
A ð13Þ

where HOMO
D is the highest occupied molecular orbital (HOMO) of Dz , LUMO
A the lowest
unoccupied molecular orbital (LUMO) of A, and H one-body the one-body Hamiltonian
being given by
h2
H one-body ¼  r2 þ Veff ðxÞ ð14Þ
82 m
In order to give a concrete expression for Veff ðxÞ, a simple model has been assumed
(see Fig. 6). In this model, it is postulated that an electron in the HOMO (essentially the 2p
orbital) of the surface oxygen atom (Osurg ) of the polyanion partially transfers to the
LUMO of water (the 4a1 molecular orbital). Then Veff ðxÞ can be expressed as
Veff ðxÞ ¼ Vcharge-electron ðxÞ þ Vcharge-hole þ Vps ðxÞ ð15Þ
where Vcharge-electron ðxÞ and Vcharge-hole represent potential energies ascribed to the Coulomb
interactions between the ionic charge and the transferring electron and between the ionic
charge and a positive hole formed on the Osurf atom, respectively, and where Vps ðxÞ is the
pseudopotential, i.e., the effective potential formed by the nucleus of the Osurf atom and
its electrons other than the transferring electron. Although it is difficult to obtain an
exact expression for Vps ðxÞ, its appropriate expression only in the overlapping region
(shown in D in Fig. 6) will be necessary for evaluating the matrix element with Eq. (13).
As seen in Fig. 6, the overlapping region located around the place where the electron
density of the 2p (Osurf ) orbital is the highest. The pertinent potential around this place

Copyright © 2001 Marcel Dekker, Inc.


32 Osakai and Ebina

FIG. 6 Quantum chemical model of the CT complex of a polyanion and water. (From Ref. 33.
Copyright 1996 Elsevier Science B.V., Amsterdam.)

should be used; the Coulomb potential has been employed here, on assuming that a
positive charge of 2 is located at the center of the Osurf atom. Thus, the corresponding
energy h HOMO
D jVps ðxÞj LUMO
A i can be calculated approximately.
Finally, under the condition that r  rO , the z-dependence of W01 can be expressed
as a function of E, as in the case of W1 [Eq. (11)]:
W01 ¼ W~ 01 þ eE ð16Þ
with
D E
 ¼ rO  D ¼ rO  HOMO
D j LUMO
A ð17Þ
*   +
 2  D   E
 h  HOMO 
W~ 01 ¼ HOMO
 2 r  A2 LUMO
þ Vps ðxÞ LUMO ð18Þ
D
 8 m  D A

From the above analysis, it is found that both W1 and W01 are linear functions of the
surface electric field strength E, so that W obtained from Eq. (9) is also a function of E.
The important point is that the z-dependence of the matrix element can be expressed only
through E. By actually solving the secular equation (9), it has been found that the E-
dependence of W is well represented by a quadratic equation:
W ¼ 0 þ 1 E þ 2 E 2 ð19Þ
It should be noted, however, that this equation represents the energy for a one-to-one ion–
solvent interaction.
When the ion is much larger than the solvent, it can be assumed that the number (N)
of solvent molecules adjacent to the ion is proportional to the surface area of the ion: N ¼
4r2 (where is the number of solvent molecules per unit surface area of the ion).
Accordingly, the contribution of the CT interaction to the ion solvation energy G0s is
given by

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 33

G0s ðCTÞ ¼ N  W ¼ 0 þ 1 E þ 2 E 2 ð20Þ


with n ¼ 4r n ðn ¼ 1; 2; 3Þ; note that n is proportional to r . The contribution of
2 2

the CT interaction to the ion-transfer energy Gtr0;O!W


is then given by
Gtr
0;O!W
ðCTÞ
G0;W
s ðCTÞ  G0;O
s ðCTÞ ¼  0 þ  1 E þ  2 E
2
ð21Þ
with  n ¼ W n  n . Here, the superscripts ‘‘W’’ and ‘‘O’’ refer to water and organic
O

phases.
Accordingly, the contribution of the CT interaction to the ion-transfer potential
[being given by Eq. (7)] can be expressed as
1
O
ðCTÞ ¼ 0 E
W 0
þ 1 þ 2 E ð22Þ
with n ¼ ð W W O O
n  n Þ=0 which depends neither on z nor r. In view of the necessity of
knowing an approximate value of the coefficient 2 of the linear term in Eq. (22), we have
made a daring assumption that Gtr 0;O!W
(CT) consists of only the contribution from
water, namely, Gtr
0;O!W
ðCTÞ Gs ðCTÞ. This leads to the following approximation:
0;W

W 1
W
O
ðCTÞ ¼ 0 E
0
þ 1W þ 2W E ð23Þ
where
nW ¼ W W
n =0 ðn ¼ 1; 2; 3Þ ð24Þ
Thus the coefficient 2W
can be evaluated from and
W
W
2 . The approximate value of W
2
ð¼ 9:4  10 molecules m ) is obtained from the density of bulk water by assuming that
18

the thickness of the first solvation shell equals the diameter (0.28 nm) of a water molecule,
22
whereas the value of W 2 has been obtained to be 2:025  10 ðeVÞ V2 m2 from the
estimation of the E-dependence of W. Finally, the coefficient 2 has been evaluated
W

to be 3:5  1011 m. This value is close to the experimental value of 2:85  1011 m, which
comes from the slope of the W O
ðelÞ vs. E plot in Fig. 4(B). From this result, the authors
0

[33] have concluded that the CT interaction plays a major role in the observed linear
dependence of W O
ðelÞ or O
ðz-depÞ on E, though this conclusion seems to be over-
0 W 0

stated to some extent.

VI. SELECTIVE HYDRATION OF IONS IN ORGANIC SOLVENTS

In almost all theoretical studies of Gtr 0;O!W


, it is postulated or tacitly understood that
when an ion is transferred across the O/W interface, it strips off solvated molecules
completely, and hence the crystal ionic radius is usually employed for the calculation of
0;O!W
Gtr . Although Abraham and Liszi [17], in considering the transfer between mutually
saturated solvents, were aware of the effects of hydration of ions in organic solvents in
which water is quite soluble (e.g., 1-octanol, 1-pentanol, and methylisobutyl ketone), they
concluded that in solvents such as NB and1,2-DCE, the solubility of water is rather small
and most ions in the water-saturated solvent exist as unhydrated entities. However, even a
water-immiscible organic solvent such as NB dissolves a considerable amount of water
(e.g., ca. 170 mM H2 O in NB). In such a medium, hydrophilic ions such as Liþ , Naþ ,
Ca2þ , Ba2þ , Cl , and Br are selectively solvated by water. This phenomenon has become
apparent since at least 1968 by solvent extraction studies with the Karl–Fischer method
[35–45]. Rais et al. [35] and Iwachido and coworkers [36–39] determined hydration num-
bers, i.e., the number of coextracted water molecules, for alkali and alkaline earth metal

Copyright © 2001 Marcel Dekker, Inc.


34 Osakai and Ebina

ions in NB. Some inorganic anions (Cl , Br ,, I , SCN , ClO 


4 , and NO3 ) have also been
found to coextract water into NB [40–42]. Similar phenomena have further been observed
in solvents other than NB (e.g., 1,1- and 1,2-dichloroethane, 4-methyl-2-pentanone,
chloroform, NB–benzene mixtures) [42–45]. If these observations are valid, we should
reconsider the previous theories of Gtr 0;O!W
in which the hydration of ions in O is not
properly taken into account. From this point of view, Osakai et al. [46] reaffirmed the
coextraction of water into NB for a variety of common ions and determined their accurate
hydration numbers in NB.
In the previous study [46], various kinds of cations and anions (listed in Table 2)
were extracted from water to NB using several extractants: tetraphenylborate (TPB ) and
dipicrylaminate (DPA ) for the cations; n-Bu4 Nþ , n-Pen4 Nþ , n-Hep4 Nþ , and tris(1,10-
phenanthroline)iron(II) ([Fe(phen)3 2þ Þ for the anions. The increase in the water concen-
tration in the NB phase, ½H2 O, with extraction of an ion was evaluated as a function of
equilibrium concentration of the ion in NB. A typical example, obtained in the DPA

TABLE 2 Numbers (n) of Coextracted Water Molecules in NB and Radii (rh ) of Hydrated Ions
at 25 C

Cation TPB system DPA system av rh (nm) ra (nm)

Liþ 6.3 5.7 (4.2[35]; 5.5[36]) 6:0  0:4 0.351 0.073


Naþ 4.0 3.6 (3.6[35]; 3.5[36]) 3:8  0:3 0.307 0.116
Kþ 1.0 (1.0[35]; 1.3[36]) 1.0 0.220 0.152
Rbþ 0.7 (0[35]; 0.7[36]) 0.7 0.212 0.166
Csþ 0.4 (0[35]; 0.7[36]) 0.4 0.206 0.181
Ca2þ 15 12 (13[36]) 14  2 0.467 0.114
Ba2þ 10 11 (9.4[36]) 11  1 0.435 0.149
Me4 Nþ 0 0 0.279
Et4 Nþ 0 0 00 0.337
n-Bu4 Nþ 0 ( 0[42]) 0 0.413
Ph4 Asþ 0 0 0.426

n-Bu4 Nþ n-Pen4 Nþ n-Hep4 Nþ ½FeðphenÞ3 2þ rh ra


Anion system system system system av (nm) (nm)

Cl 4.0 (3.3[43]) 4.0 0.322 0.167


Br 2.1 (1.8[43]) 2.1 2.0 —b (5.5[41]) 2:1  0:1 0.276 0.182
I 0.9 0.8 1.0 1.0c (2.1[41]) 0:9  0:1 0.248 0.206
SCN 1.1 1.1 1.0 1.1c (1.9[41]) 1:1  0:1 0.260 0.213
ClO4 0.3 0.2 0.2 0.1c (0.64[41]) 0:2  0:1 0.244 0.236
NO 3 ( 1:4½42Þ 1.7c 1.7 0.267 0.189
TPB 0ð 0[42]) 0 0 0c 0 0.421
a
Crystal (or bare) ionic radii (see Table 3).
b
No reliable value could be obtained due to extremely low extractability (the distribution ratio D < 0:03).
c
Corrected for the contribution (n ¼ 0:3) from ½FeðphenÞ3 2þ .
Source: From Ref. 46.

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 35

system, is shown in Fig. 7. The hydrophobic organic cations (Me4 Nþ , Et4 Nþ , and TPAsþ )
showed no increase in water concentration, indicating that these cations as well as DPA
had no ability to transport water to NB. In contrast, hydrophilic metal ions showed a
distinguishing increase. The slope of each linear plot should correspond to the number (n)
of water molecules coextracted to NB with a metal ion, which are listed in Table 2 together
with those determined in other systems. As seen in the table, for both cations and anions,
the n values determined were only slightly dependent on the nature of extractant (i.e.,
coion). This implies that ion-pair formation in NB is of less significance or that the ions are
almost dissociated. Thus accurate numbers of water molecules, being certainly associated
with individual ions, could be determined.
Based on these measurements, a new model of the transfer of hydrophilic ions across
the O/W interface was proposed (see Fig. 8). In this model, the hydrophilic ion transfers
from W to O with some water molecules associated with the ion. A typical example in Fig.
8 shows that a sodium ion transfers across the NB/W interface with four water molecules.
0;O!W
In theoretical treatment of Gtr of such a hydrophilic ion, therefore, the transferring
species should be regarded as the ‘‘hydrated’’ ion. In accordance, the radii (rh Þ of hydrated
ions in NB were estimated from the hydration numbers (n) and crystal ionic radii (r) by
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
3 n
rh ¼ þ r3 ð25Þ
4 d
where d is the density of water in the hydration shell (here, the value of d is assumed to be
the same as that of bulk water, i.e., 3:33  1028 molecule m3 ). The estimated values of rh
are also shown in Table 2.
As seen in Table 2, the order of the magnitude of rh for alkali metal ions is the
reverse of that of the magnitude of r. This means that a more hydrophilic ion has a larger
rh . However, this fact does contradict the expectation from Bornian electrostatic theories.
As can be seen in the Born equation [Eq. (2)], it is expected that the larger the radius an ion
has, the more positive the Gtr 0;O!W
value the ion has, that is, the more hydrophobic it

FIG. 7 Plots of the increase of water concentration in NB (½H2 OÞ with extraction of cations with
DPA against the equilibrium cation concentration in NB. Each value in the parentheses shows the
number (n) of coextracted water molecules per ion. (From Ref. 46. Copyright 1997 American
Chemical Society.)

Copyright © 2001 Marcel Dekker, Inc.


36 Osakai and Ebina

FIG. 8 Proposed model of the transfer of a hydrophilic ion across the O/W interface. The illustra-
tion shows the transfer of Naþ from W to NB as a typical example. (From Ref. 46. Copyright 1997
American Chemical Society.)

becomes. This would show that Liþ , as an example, is more hydrophobic than Csþ , but
this is of course not the case. Thus, if hydrated radii of ions are used, Bornian electrostatic
solvation models are invalid.
Then we made a new approach that recognizes short-range interactions of a
hydrated ion with solvents. By this approach, Gtr 0;O!W
for hydrophilic ions could be
elucidated on the basis of the proposed model in Fig. 8 (see Section VII). A similar model
was also employed by Sánchez et al. [47] to elucidate Gtr 0;O!W
for some hydrophilic ions,
but using a Bornian electrostatic theory.
Recently, the abilities of primary to tertiary alkylammonium ions with Me, Et, and
n-Bu groups to transport water to NB have been studied [48]. As the result of careful
consideration of the ion-pair formation, it has been shown that the hydration numbers (nh )
of the ammonium ions in NB, being little affected by the alkyl chain length, are simply
dependent on the class of the ammonium ion: nh ¼ 1:64, 1.04, 0.66, respectively, for the
primary, secondary, and tertiary ammonium ions.

VII. NON-BORNIAN THEORY OF THE RESOLVATION ENERGY OF IONS


A. Theoretical Model
Based on the above-mentioned experimental and theoretical findings, we have proposed a
new, non-Bornian theory of Gtr 0;O!W
of ions [49]. In this theory, it is assumed that a
hydrophilic ion which is selectively hydrated in organic solvent transfers across the O/W
interface as the hydrated ion, and a hydrophobic ion which is not hydrated in the O phase
transfers as a bare ion, as depicted in Fig. 9.
In a conventional manner we divide Gtr 0;O!W
into two terms:
Gtr
0;O!W
¼ Gtr
0;O!W
ðz-indepÞ þ Gtr
0;O!W
ðz-depÞ ð26Þ
0;O!W
where Gtr ðz-indepÞ is the charge-independent term corresponding to the solvopho-
bic interaction or the energy of the formation of a cavity in solvents, and Gtr
0;O!W
ðz-depÞ
is the charge-dependent term, which has so far been considered mainly as describing

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 37

FIG. 9 Proposed model of the transfer of (a) a hydrophilic ion and (b) a hydrophobic ion across the
O/W interface. (From Ref. 49. Copyright 1998 American Chemical Society.)

electrostatic (i.e., long-range) ion–solvent interaction, but is here treated as describing


specific (short-range) interactions.
For the evaluation of Gtr 0;O!W
ðz-indepÞ, we employed the Uhlig formula [Eq. (4)].
This formula may be applicable to the NB/W interface (O;W ¼ 25:2 mN m1 [50]) and to
all ions studied (see Table 3), which include the ‘‘hydrated’’ ions with a hydrated radius
rh ð¼ rÞ larger than 0.20 nm as well as the ‘‘unhydrated’’ ions of r > 0:23 nm.
The charge-dependent term Gtr 0;O!W
ðz-dep) in Eq. (26) is here assumed to be gov-
erned by short-range ion–solvent interactions. The long-range interactions of an ion with
the solvents in the second and further solvation shells are ignored in the present theory.
Because the short-range (i.e., chemical) interactions recognize some overlap of the electron
orbitals of the ion and the solvent molecule in its immediate vicinity, they should be
explained on quantum chemical considerations. In ab initio molecular orbital studies
[58], the self-consistent-field (SCF) energy of the ion–molecule interaction, USCF , is parti-
tioned into several terms including Coulomb (COU), polarization (POL), charge-transfer
(CT), and exchange (EX) terms:
USCF ¼ UCOU þ UPOL þ UCT þ UEX ð27Þ
The first two terms UCOU and UPOL correspond to the empirical energy of ion–dipole
and ion–induced dipole interactions, being given by
UCOU ¼ Ehcos i ð28Þ
and
1
UPOL ¼  E 2 ð29Þ
2
where  and  are the dipole moment and electronic polarizability of the solvent molecule,
respectively, is the angle between the dipole axis and the line connecting the point dipole
and point charge, h i indicates the ensemble average, and E is the ‘‘effective’’ electrical field
strength, which can be approximated by the surface field strength of the ion given by Eq.
(8), when r is much larger than the radius of the dipole (i.e., solvent molecule).

Copyright © 2001 Marcel Dekker, Inc.


38 Osakai and Ebina

TABLE 3 Charge Numbers and Radii of Ions and Their Hydration Numbers and Hydrated
Radii in NB at 25 C

Ion z ra (nm) nb rh c (nm)

Hydrated cations
Liþ 1 0.073 6.0 0.351
Naþ 1 0.116 3.8 0.307
Kþ 1 0.152 1.0 0.220
Rbþ 1 0.166 0.7 0.212
Csþ 1 0.181 0.4 0.206
Ca2þ 2 0.114 14 0.467
Ba2þ 2 0.149 11 0.435
Nonhydrated cations
Me4 Nþ 1 0.279d 0
Et4 Nþ 1 0.337d 0
n-Pr4 Nþ 1 0.379d 0e
n-Bu4 Nþ 1 0.413d 0
Ph4 Asþ ðTPAsþ Þ 1 0.416f 0
½NiðbpyÞ3 2þ 2 0.527g 0e
½NiðphenÞ3 2þ 2 0.544g 0e
½FeðphenÞ3 2þ 2 0.541g e
0 (0.3)
Hydrated anions
Cl 1 0.167 4.0 0.322
Br 1 0.182 2.1 0.276
I 1 0.206 0.9 0.248
SCN 1 0.213h 1.1 0.260
NO 3 1 0.189i 1.7 0.267
Nonhydrated anions
ClO 4 1 0.236i 0e (0.2)
IO4 1 0.249i 0e
2,4-dinitrophenol 1 0.315j 0e
2,4,6-trinitrophenol 1 0.332j 0e
Ph4 B ðTPB Þ 1 0.421f 0
Polyanions
; -½XM12 O40 4 4 0.56k 0e
; -½XM12 O40 3 3 0.56k 0e
-½X2 Mo18 O62 6 6 0.648k 0e
-½S2 Mo18 O62 4 4 0.648k 0e
½S2 VMo17 O62 5 5 0.648k 0e
½P2 Mo18 O61 4 4 0.644k 0e
½Mo6 O19 2 2 0.437k 0e
½VMo5 O19 3 3 0.437k 0e
-½Mo8 O26 4 4 0.485k 0e

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 39

The third term, UCT , in Eq. (27) is due to the partial electron transfer between an ion
and solvents in its immediate vicinity. The model Hamiltonian approach [33], described in
Section V, has shown that UCT ð¼ W in Ref. 33) per primary solvent molecule, for an
ion such as the polyanion, can also be expressed as a function of E, approximately a
quadratic equation:

UCT ¼ 0  1 E  2 E 2 ð30Þ

The coefficients 0 , 1 , and 2 (denoted as 0 , 1 , and 2 in Ref. 33) are influenced by


various molecular properties of the solvent and an ion, including their electron-donating
or accepting abilities. Hence, these coefficients are specific to the ion. Nevertheless, they
may be considered as common to a family of ions such as the polyanions whose surface
atoms, directly interacting with solvents, are oxygens. This is the case for ‘‘hydrated’’
cations or anions whose surfaces are composed of some water molecules that interact
with outer water molecules in the W phase or with organic solvents in the O phase.
The remaining term, UEX , in Eq. (27) represents the nonclassical repulsion term,
being inherently independent of E. By summarizing the above argument, we conclude that
the short-range interaction energy, USR ð¼ USCF Þ, can be given by a quadratic function of
E:

USR ¼ A  BE  CE 2 ð31Þ

with

A ¼ 0 þ UEX ð32Þ
B ¼ Ehcos i þ 1 ð33Þ

C ¼ þ 2 ð34Þ
2
The validity of Eq. (31) has been confirmed by an approach from redox potentials of
heteropoly oxometalate anions as well [59].
In Eq. (31) we should note that USR represents the interaction energy per primary
solvent molecule. Then the number of solvent molecules that can be interact directly with
an ion in phase S ð¼ O or W) is denoted as N S . Accordingly, the contribution of the short-
range interactions to the solvation energy G0;vac!S
tr of the ion phase S (i.e., transfer
energy from vacuum) is given from Eq. (31) as

Footnotes to Table 3
a
Shannon’s crystal ionic radii [51], except where otherwise noted.
b
Ref. 46.
c
Estimated by Eq. (25).
d
van der Waals radii calculated from the partial molar ionic volumes [52].
e
Assumed.
f
Ref. 6.
g
van der Waals radii calculated from the partial molar ionic volumes [53,54].
h
Ref. 55.
i
Thermochemical radii [56].
j
From van der Waals volumes calculated from atomic increments [57].
k
See Table 1.
Source: From Ref. 49.

Copyright © 2001 Marcel Dekker, Inc.


40 Osakai and Ebina

G0;vac!S
tr ðSRÞ ¼ N S USR ¼ AS N S  BS N S E  C S N S E 2 ð35Þ
where the superscript S of A, B, and C represents the S phase. Because the transfer energy
Gtr0;O!W
is the difference between G0;S
s ’s in O and W, the contribution of short-range
interactions to Gtr0;O!W
can be expressed as
Gtr
0;O!W
ðSRÞ ¼ 1  2 E  3 E 2 ð36Þ
with
1 ¼ A W N W  A O N O ð37Þ
2 ¼ BW N W  BO N O ð38Þ
3 ¼ C W N W  C O N O ð39Þ
In this manner Gtr
0;O!W
(SR) is given approximately by a quadratic equation of E. The
coefficients 1 , 2 , and 3 are related to various molecular properties of the ion and
solvents [see Eqs. (32)–(34) and Eqs. (37)–(39)]. To estimate theoretically these coefficients
using an appropriate model may not be impossible, but rather difficult at the present stage,
because there is insufficient information on the above-mentioned molecular properties (in
particular, the charge-transfer properties). Consequently, coefficients 1 , 2 , and 3 in Eq.
(36) have been determined empirically, as shown below.

B. Experimental Data
In the present analyses [49], 34 ions are classified into five groups: (1) hydrated cations,
(2) nonhydrated cations, (3) hydrated anions, (4) nonhydrated anions, and (5) polyanions.
Here, the term ‘‘hydrated’’ or ‘‘nonhydrated’’ means that the ion is associated with some
water molecules in the O phase or not, respectively.

1. Ionic Radii
All ions studied are assumed to be spherical, and radii (r) of the bare ions are listed in
Table 3. Unless otherwise noted, authorized values of crystal ionic radii or van der Waals
radii are adopted [51–57].

2. Hydration Numbers of Ions in NB


The numbers (n) of coextracted water molecules shown in Table 3 are from Table 2 [46]. In
the case of n-Pr4 Nþ , metal complex cations, larger anions of r > 0:23 nm, and polyanions,
it is assumed that n ¼ 0. Although the n value of ½FeðphenÞ3 2þ or ClO 4 was reported to be
as small as 0.3 or 0.2 [46], these ions have been classified as ‘‘nonhydrated’’ (i.e., n ¼ 0) so
that comparatively better results may be obtained.
In Table 3 are also shown the hydrated radii (rh ) which are evaluated with n and r by
Eq. (25). A good correlation of rh with the Stokes radius [60] (rs ) has been observed for
hydrated cations (alkali and alkaline earth metal ions) [46]:
rh ðnmÞ ¼ 1:310 rs ðnmÞ þ 0:055 ðR ¼ 0:997Þ ð40Þ
In this equation and the following remarks, R shows the correlation coefficient.
For the prediction of Gtr
0;O!W
of a hydrophilic ion whose n value is unknown, it is
desirable that the n value can be obtained with r and z only. Since the interactions of a
small ion with water molecules in the hydration shell may be mostly electrostatic, the

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 41

dependence of n on (z=r) has been investigated for hydrophilic cations and anions. As a
result, it has been found that the plots of n against (z=r) lie on a quadratic curve:
For cations:
n ¼ 0:566 þ 0:0091ðz=rÞ þ 0:04745ðz=rÞ2 ðR ¼ 0:966Þ ð41Þ
For anions:
n ¼ 5:844 þ 3:775ðz=rÞ þ 0:5661ðz=rÞ2 ðR ¼ 0:930Þ ð42Þ
where r is in nanometers. Equations (41) and (42) are applicable in the range of ðz=rÞ > 5
and ðz=rÞ < 4, respectively. Using these equations, we can predict, though somewhat
roughly, the n value of a hydrophilic ion. Regarding the cations, however, their n values
could be obtained more accurately from the Stokes radii by using the relations of Eqs. (25)
and (40).

3. Gibbs Energies of Transfer of Ions


In 1977 Koryta et al. [61] reported the Gtr 0;O!W
values of several common ions at the
NB/W interface, which were calculated from the extraction data using an extra thermo-
dynamic assumption. Afterward, a newly developed electrochemical technique (so-called
ion-transfer voltammetry) with a polarizable O/W interface was employed to determine
G0trO!W for a variety of ions [33,62–71]. In Table 4 the reliable values of Gtr 0;O!W

are compiled. Regarding the ions whose Gtr 0;O!W


values are available for both electro-
chemical and extraction measurements, the electrochemical data, which seem to be more
accurate, have been chosen preferentially. For several ions, somewhat different Gtr0;O!W

values from electrochemical measurements have been reported, as also seen in the data-
base provided by Girault on a website [72]. In this study, however, we have carefully
chosen reliable values for the respective ions, which were determined under well-defined
conditions (reference electrodes, solution compositions, etc.).

C. Data Analyses
1. Calculation of Cavity Formation Energies
Using the Uhlig formula [Eq. (4)] the values of Gtr 0;O!W
(z-indep) have been obtained
from ionic radii, as shown in Table 4. For the hydrated cations and anions, their hydrated
radii (rh Þ have been employed for r in Eq. (4). Table 4 also shows the values of Gtr0;O!W
(z-
dep) which have been obtained by subtracting Gtr 0;O!W
(z-indep) from the total Gtr0;O!W

[see Eq. (26)].

2. Regression Analyses for Hydrated Ions


Unless noted otherwise, it has been assumed that the number of the solvent molecules N S
(S ¼ O or W) that can interact directly with a hydrated ion in phase S is equal to the
hydration number:
NO ¼ NW ¼ n ð43Þ
For Ca2þ and Ba2þ , whose n values are larger than 10, however, it is thought that
some hydrated water molecules not only in the first hydration shell but also in the second
hydration shell are cotransferred into NB. Accordingly, it can be supposed that some
water molecules in the first hydration shell (i.e., in the vicinity of the ion) are covered
with the second hydration shell, so that they cannot be associated with outer solvent

Copyright © 2001 Marcel Dekker, Inc.


42 Osakai and Ebina

TABLE 4 Standard Gibbs Energies of Transfer of Ions from NB to W and Their Charge-
Independent and Charge-Dependent Components at 25 C

G0;O!W
tr G0;O!W
tr (z-indep)a G0;O!W
tr (z-dep)b
Ion (kJ mol1 ) (kJ mol1 ) (kJ mol1 )

Hydrated cations
Liþ 38:2c;d 23.6 61:8
Naþ 34:2c;d 17.9 52:1
Kþ 23:5c;d 9.2 32:7
Rbþ 19:4c;d 8.6 28:0
Csþ 15:4c;d 8.1 23:5
Ca2þ 67:3d;e 41.6 108:9
Ba2þ 61:8d;e 36.0 97:8
Nonhydrated cations
Me4 Nþ 3:4c;d 14.8 18:2
Et4 Nþ 5.3f 21.7 16:4
n-Pr4 Nþ 16:4g 27.4 11:0
n-Bu4 Nþ 26:5h 32.5 6:0
Ph4 Asþ ðTPAsþ Þ 35:9c;d 34.6 1.3
½NiðbpyÞ3 2þ 30:5i 53.0 22:5
½NiðphenÞ3 2þ 41:3i 56.4 15:1
½FeðphenÞ3 2þ 44:0j 55.8 11:8
Hydrated anions
Cl 38:2k 19.8 58:0
Br 27:8l 14.6 42:4
I 18:4l 11.7 30:1
SCN 15:8l 12.9 28:7
NO 3 25:2l 13.5 38:7
Nonhydrated anions
ClO 4 7:9l 10.6 18:5
IO4 6m 11.8 17:8
2,4-dinitrophenol 5:7n 18.9 24:6
2,4,6-trinitrophenol 6:7n 21.0 14:3
Ph4 B ðTPB Þ 35:9d 33.8 2.1
Polyanions
; -½XM12 O40 4 25:9o 59.8 33:9
; -½XM12 O40 3 71.8o 59.8 12.0
-½X2 Mo18 O62 6 2:9o 80.1 77:2
-½S2 Mo18 O62 4 103:8o 80.1 23.7
½S2 VMo17 O62 5 41:0o 80.1 39:1
½P2 Mo18 O61 4 92:2o 79.1 13.1
½Mo6 O19 2 31:6o 36.4 4:8
½VMo5 O19 3 34:4o 36:4 70:8
-½Mo8 O26 4 52:9o 44.9 97:8

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 43

molecules (W or NB). Such an effect is here called ‘‘shielding.’’ In these cases, N S cannot
be equated simply to n and should be reduced to some extent. It has been found that for
the above two cations, the subtraction of 4 from the net n values may lead to the best result
in the following regression analysis.
If we suppose that G0trO!W (z-dep) for the hydrated ions is ruled out by short-range
interactions, a combination of Eqs. (36)–(39) and Eq. (43) yields
G0trO!W ðz-depÞ=n ¼ A  BE  CE 2 ð44Þ
with A ¼ A  A , B ¼ B  B , and C ¼ C  C . In Fig. 10 the values of
W O W O W O

Gtr0;O!W
(z-dep)=n for the hydrated cations (*) and anions (*) are plotted against E,
being calculated by Eq. (8) with r ¼ rh . Note that the plots of Ca2þ and Ba2þ have been
compensated for the ‘‘shielding’’ effect: the symbols () in Fig. 10 represent the plots with
the net values of n ð¼ 14 or 11 for Ca2þ or Ba2þ , respectively).
As seen in Fig. 10, in accordance with Eq. (44), Gtr
0;O!W
(z-dep) per hydrated water
becomes progressively greater as E is enhanced. Since the contribution from the interac-
tion in W to Gtr 0;O!W
(z-dep) is probably more significant than that from the interaction
in NB, the dependences shown in Fig. 10 seem to suggest that the hydrogen bonds, which
are formed around a hydrated ion in W and which must be broken in its transfer to NB,
are strengthened by the surface field of the hydrated ion.
The solid lines in Fig. 10 represent the quadratic curves obtained in regression
analyses.
For hydrated cations:
0;O!W
Gtr ðz-depÞ=n ¼ 8:90 þ 4:506E  4:853E 2 ðR ¼ 0:931Þ ð45Þ
For hydrated anions:
Gtr
0;O!W
ðz-depÞ=n ¼ 50:17  52:51E  19:33E 2 ðR ¼ 0:999Þ ð46Þ
1 1
where Gtr
0;O!W
(z-dep) is in kJ mol 10
and E is in 10 V m .

3. Regression Analyses for Nonhydrated Ions


As for the nonhydrated ions including the polyanions, it has been assumed that N S is
proportional to the surface area of the ion:

Footnotes to Table 4
a
Evaluated using Eq. (4) (note that r ¼ rh for the hydrated ions).
b
G0;O!W
tr (z-depÞ ¼ Gtr0;O!W
 Gtr
0;O!W
 (z-indep).
c
Ref. 61.
d
Determined using extraction data; the other values were determined by electrochemical measurements.
e
Ref. 62.
f
Ref. 63. The value is revised by employing Me4 Nþ as an internal reference in the place of n-Bu4 Nþ .
g
Unpublished data obtained by means of the former electrochemical technique [64].
h
Ref. 65.
i
Ref. 66.
j
Ref. 67.
k
Ref. 68.
l
Ref. 69.
m
Ref. 70.
n
Ref. 71.
o
Calculated from the standard ion-transfer potentials compiled previously [33].
Source: From Ref. 49.

Copyright © 2001 Marcel Dekker, Inc.


44 Osakai and Ebina

FIG. 10 Plots of G0;O!W


tr (z-depÞ=n against E (with r ¼ rh ) for hydrated cations (*) and anions
(*). Note that the n values for the plots of Ca2þ and Ba2þ have been corrected for the ‘‘shielding’’
effect (see text) by subtracting 4 from their net values of n; () represents the plots with the net
values. Solid lines show the regression curves [Eqs. (45) and (46)]. (From Ref. 49. Copyright 1998
American Chemical Society.)

N S ¼ 4r2 S ð47Þ
where is the number of solvent molecules per unit surface area of the ion in phase S
S

ð¼ O or W). We have also assumed that G0;O!Wtr (z-dep) is ruled only by short-range
interactions in the same manner as the hydrated ions and then have obtained from Eqs.
(36)–(39) and Eq. (47):
0;O!W
Gtr ðz-depÞ=ð4r2 Þ ¼ A 0  B 0 E  C 0 E 2 ð48Þ
0 0 0
with A ¼ A  A , B ¼ B  B , and C ¼ C  C .
W W O O W W O O W W O O

In accordance with Eq. (48), the values of Gtr 0;O!W


(z-dep)/(4r2 ) have been plotted
against E in Fig. 11. The respective plots for the three ion groups have been found to lie on
a single quadratic curve.
For nonhydrated cations:
Gtr
0;O!W
ðz-depÞ=ð4r2 Þ ¼ 28:19  43:08E þ 9:582E 2 ðR ¼ 0:987Þ ð49Þ
For nonhydrated anions:
Gtr
0;O!W
ðz-depÞ=ð4r2 Þ ¼ 33:04 þ 47:53E þ 9:518E 2 ðR ¼ 0:967Þ ð50Þ
For polyanions:
Gtr
0;O!W
ðz-depÞ=ð4r2 Þ ¼ 13:40  7:685E  11:08E 2 ðR ¼ 0:989Þ ð51Þ
1 1
where Gtr
0;O!W
(z-dep) 10
is in kJ mol , r in nm, and E in 10 V m . It is noteworthy that
the plots of the polyanions with a wide variety of charges (z ¼ 2 to 6) lie on a single
curve.

4. Total Gibbs Energies of Transfer of Ions


The values of Gtr 0;O!W
have been calculated using the Uhlig formula [Eq. (4)] for the
charge-independent part and the empirical equations [Eqs. (45), (46), (49), (50), and (51)]
for the charge-dependent part. Table 5 gives the calculated values of Gtr
0;O!W
for 34 ions,
and the values are compared in Fig. 12 with the observed values shown in Table 4. As seen

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 45

FIG. 11 Plots of G0trO!W (z-depÞ=ð4r2 Þ against E for nonhydrated cations (~) and anions (~)
and polyanions (&). Solid lines show the regression curves [Eqs. (49), (50), and (51)]. (From Ref. 49.
Copyright 1998 American Chemical Society.)

FIG. 12 Comparison of the calculated and observed values of G0trO!W for 34 ions: (*) hydrated
cations, (*) hydrated anions, (~) nonhydrated cations, (~) nonhydrated anions, and (&) poly-
anions. (From Ref. 49. Copyright 1998 American Chemical Society.)

Copyright © 2001 Marcel Dekker, Inc.


46 Osakai and Ebina

TABLE 5 Calculations of Standard Gibbs Energies of Transfer of Ions from NB to W (25 C)

Ea G0;O!W
tr (z-dep)b G0;O!W
tr (calc)c
Ion (10 V m1 )
10
(kJ mol1 ) (kJ mol1 )

Hydrated cations
Liþ 1.17 (27.0)d 61:5 37:9 ðþ0:3Þe
Naþ 1.53 (10.7) 50:9 33:0 ðþ1:2Þ
Kþ 2.97 (6.23) 38:3 29:0 ð5:5Þ
Rbþ 3.19 (5.23) 30:7 22:1 ð2:7Þ
Csþ 3.38 (4.39) 19:6 11:5 ðþ3:9Þ
Ca2þ 1.32 (22.2) 114:1 72:5 ð5:2)
Ba2þ 1.52 (13.0) 93:1 57:1 ðþ4:7Þ
Nonhydrated cations
Me4 Nþ 1.85 18:3 3:5 ð0:1Þ
Et4 Nþ 1.27 15:7 5.9 ðþ0:6Þ
n-Pr4 Nþ 1.00 9:7 17.7 ðþ1:3Þ
n-Bu4 Nþ 0.84 2:9 29.6 ðþ3:1Þ
Ph4 Asþ ðTPAsþ Þ 0.79 0.1 34.7 ð1:2Þ
½NiðbpyÞ3 2þ 1.04 21:6 31.4 (þ0:9Þ
½NiðphenÞ3 2þ 0.97 17:3 39.1 (2:2)
½FeðphenÞ3 2þ 0.98 18:1 37.7 (6:3)
Hydrated anions
Cl 1:39 ð5:16Þd 64:1 44:3 ð0:3Þ
Br 1:89 ð4:35Þ 45:9 31:3 ðþ1:5Þ
I 2:35 ð3:39Þ 31:8 20:1 ðþ0:2Þ
SCN 2:13 ð3:17Þ 30:9 18:0 ð0:1Þ
NO 3 2:03 ð4:03) 42:7 29:2 ð1:3Þ
Nonhydrated anions
ClO 4 2:59 18:4 7:7 ðþ1:4Þ
IO4 2:32 20:3 8:9 ð2:4Þ
2,4-dinitrophenol 1:45 19:8 0:9 ðþ4:8Þ
2,4,6-trinitrophenol 1:31 17:7 3.3 ð3:4Þ
Ph4 B ðTPB Þ 0:81 1.6 35.4 ð0:5Þ
Polyanions
; -½XM12 O40 4 1:84 38:8 21.0 (4:9)
; -½XM12 O40 3 1:38 11.7 71.5 ð0:3Þ
-½X2 Mo18 O62 6 2:06 93:3 13:3 ð16:2Þ
-½S2 Mo18 O62 4 1:37 16.3 96.4 ð7:4Þ
½S2 VMo17 O62 5 1:71 31:6 48.5 ðþ7:5Þ
½P2 Mo18 O61 4 1:39 14.1 93.2 ðþ1:0Þ
½Mo6 O19 2 1:51 0:5 35.9 ðþ4:3Þ
½VMo5 O19 3 2:26 62:2 25:7 ðþ8:7Þ
-½Mo8 O26 4 2:45 101:1 56:3 (3:4)
a
Evaluated from Eq. (8). For the hydrated ions, their hydrated radii were employed for the value of r.
b
Calculated using Eqs. (45), (49), (46), (50), and (51) for hydrated cations, nonhydrated cations, hydrated anions,
nonhydrated anions, and polyanions, respectively.
c
Obtained by adding the calculated values of G0;O!W tr (z-dep) to the values of Gtr
0;O!W
(z-indep) in Table 4.
d
The values in parentheses are surface field strengths of the bare ions.
e
The values in parentheses show the deviations from the observed values.
Source: From Ref. 49.

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 47

from the figure, satisfactory agreement has been observed (R ¼ 0:994; standard deviation,
SD ¼ 4:63 kJ mol1 ). Also, for the W
O
being related to Gtr
0 0;O!W
by Eq. (7), the corre-
lation of the calculated and observed values has been found to be very good (R ¼ 0:995;
SD ¼ 0:023 V).

D. Predictions
Based on the proposed theory, the values of G0;O!W tr (and W O
) have been predicted for
0

hydrophilic ions (NH4 , Mg , Sr , Ni , Fe ; Fe , ClO3 , BrO


þ 2þ 2þ 2þ 2þ 3þ    
3 , CN , IO3 , HCO3 ,
 
NO2 , OH ), whose Gtr 0;O!W
values have been reported, but their hydration numbers are
unknown. In the predictions, the first decision on whether the ion is hydrated in NB is
made by means of ðz=rÞ; as described above, if ðz=rÞ > 5 (r in nm) for a cation or ðz=rÞ <
4 for an anion (except polyanions), the ion should be hydrated in NB to a certain extent.
For these hydrated ions, the hydration number and the hydrated radius are evaluated.
Then the charge-independent and -dependent terms of Gtr 0;O!W
are calculated using the
Uhlig formula and the empirical equation, respectively. Finally, the summation of the two
terms yields the total value of Gtr 0;O!W
.
In the calculation of Gtr0;O!W
(z-dep) for highly hydrated cations (Mg2þ , Sr2þ , Ni2þ ,
Fe2þ , Fe3þ ) with n > 10, the values of n for use in Eq. (45) have been corrected for the
‘‘shielding’’ effect, on the assumption that the corrected value of n is proportional to the
surface area of the hydrated ion:
ncorr ¼ f ð4r2 Þ ð52Þ
The factor f has been determined to be 42.05 by a regression analysis, so that the ncorr
values of Ca2þ and Ba2þ may be approximated to n  4 (see above).
Although the results of the predictions (see Table 5 in Ref. 49) are not shown here, it
has been found that the proposed theory is promising not only for an understanding of
0;O!W
Gtr but also for its prediction.

E. Discussion
In the proposed theory, short-range ion–solvent interaction energies are formulated as Eq.
(31). the coefficients A, B, and C in the quadratic equation are related to coefficients 1 ,
2 , and 3 in the formula for Gtr 0;O!W
[Eq. (36)]. The above-mentioned regression
analyses have clearly shown that these coefficients are common to a family of ions. This
suggests that the surface chemical constitution of an ion should play a major role in its
short-range interactions with solvents. For hydrated ions, their surfaces are composed of
water molecules, which may prevent the inner ion from interacting directly with outer
solvent molecules. Accordingly, the quantum chemical properties of the inner ion hardly
influence direct interactions of the hydrated ion with solvents. The inner ion may function
only as the origin of the surface electrical field of the hydrated ion, which should affect the
short-range interaction energy, as shown in Eq. (31). As for the nonhydrated cations
studied, their surfaces are mainly composed of C–H atomic groups, while the five non-
hydrated anions, being forced to be classified into one group, have miscellaneous surfaces.
As also described above, the surfaces of the polyanions are composed of oxygens without
exception. Thus Eq. (31) turns out to be quite promising for an understanding of
0;O!W
Gtr for ions with the common surface.
The coefficients in Eq. (44) or (48) coming from those in Eq. (31) have been deter-
mined empirically in the present study. As shown above, these coefficients are related to

Copyright © 2001 Marcel Dekker, Inc.


48 Osakai and Ebina

some molecular properties of the ion and the solvent. However, it seems premature to
discuss these relations quantitatively, because the coefficients would be heavily affected by
the experimental errors in Gtr 0;O!W
. It is desired to collect accurate and reliable values of
Gtr0;O!W
for many other kinds of ions.
In the present non-Bornian theory it should be noted again that the long-range
electrostatic interactions of an ion with solvents in the second and further solvation shells
are ignored. However, the electrostatic energies should contribute to a considerable extent
to the solvation energies of ions in each phase. Nevertheless, the proposed equations for
Gtr0;O!W
, in which short-range interactions are only considered, have been found to be
fitted very well to the experimental data. This may be understood by recognizing that
Gtr0;O!W
is the difference in the solvation energy between the two phases. It is most
probable that the long-range electrostatic solvation energies in the respective phases are
for the most part canceled in Gtr 0;O!W
. This will result in the predominant contribution of
0;O!W
short-range interactions to Gtr . Thus, our daring ignoring of the long-range electro-
static energies have given a better account of Gtr 0;O!W
than Bornian electrostatic models.

REFERENCES

1. M. Born. Z. Phys. 1: 45 (1920).


2. Y. Marcus, Ion Solvation, Wiley, Chichester, 1985.
3. Y. Marcus, Ion Properties, Marcel Dekker, New York, 1997.
4. R. R. Dogonadze, E. Kálmán, A. A. Kornyshev, and J. Ulstrup (eds.), The Chemical Physics of
Solvation: Part A, Theory of Solvation, Elsevier, Amsterdam, 1985.
5. P. Politzer and J. S. Murray (eds.), Quantitative Treatments of Solute/Solvent Interactions,
Elsevier, Amsterdam, 1994.
6. Y. Marcus. Rev. Anal. Chem. 5:53 (1980).
7. Y. Marcus. Pure Appl. Chem. 62:899 (1990).
8. Y. Marcus, in Liquid–Liquid Interfaces: Theory and Methods (A. G. Volkov and D. W.
Deamer, eds.), CRC Press, Boca Raton, 1996, pp. 39–61.
9. V. S. Markin and A. G. Volkov. Electrochim. Acta 34:93 (1989).
10. T. Osakai and K. Ebina. Bunseki 1998:589 (1998) (in Japanese).
11. A. A. Kornyshev and A. G. Volkov. J. Electroanal. Chem. 180:363 (1984).
12. A. A. Kornyshev, in Ref. 4, pp. 77–118.
13. H. H. Uhlig. J. Phys. Chem. 41:1215 (1937).
14. W. M. Latimer, K. S. Pitzer, and C. M. Slansky. J. Chem. Phys. 7:108 (1939).
15. M. H. Abraham and J. Liszi. J. Chem. Soc. Faraday Trans. 1 74:1604; 2858 (1978).
16. M. H. Abraham, J. Liszi, and L. Mészáros. J. Chem. Phys. 70:2491 (1979).
17. M. H. Abraham and J. Liszi. J. Inorg. Nucl. Chem. 43:143 (1981).
18. Y. Marcus. J. Chem. Soc. Faraday Trans. 87:2995 (1991).
19. V. Gutmann, The Donor–Acceptor Approach to Molecular Interactions, Plenum Press, New
York, 1978.
20. K. Burger, Solvation, Ionic and Complex Formation Reactions in Non-Aqueous Solvents:
Experimental Methods for Their Investigation, Akadémiai Kiadó, Budapest, 1983.
21. V. Gutmann and E. Wychera. Inorg. Nucl. Chem. Lett. 2:257 (1966).
22. U. Mayer, V. Gutmann, and W. Gerger. Monatsch. Chem. 106:1235 (1975).
23. D. A. Owensby, A. J. Parker, and J. W. Diggle. J. Am. Chem. Soc. 96:2682 (1974).
24. E. Grunwald and S. Winstein. J. Am. Chem. Soc. 70:846 (1948).
25. E. M. Kosower. An Introduction to Physical Organic Chemistry, Wiley, New York, 1968, p. 259.
26. K. Dimroth, C. Reichardt, T. Siepmann, and F. Bohlmann. Liebigs Ann. Chem. 661:1 (1963).
27. C. Reichardt. Angew. Chem. Int. Ed. Engl. 4:29 (1965).

Copyright © 2001 Marcel Dekker, Inc.


Ion Solvation and Resolvation 49

28. R. W. Taft, N. J. Pienta, M. J. Kamlet, and E. M. Arnett. J. Org. Chem. 46:661 (1981).
29. T. Osakai, H. Katano, K. Maeda, S. Himeno, and A. Saito. Bull. Chem. Soc. Jpn. 66:1111
(1993) and references therein.
30. T. Osakai, S. Himeno, A. Saito, K. Maeda, and H. Katano. J. Electroanal. Chem. 360:299
(1993).
31. T. Osakai, S. Himeno, and A. Saito. Bunseki Kagaku 43:1 (1994) (in Japanese with English
abstract).
32. T. Kurucsev, A. M. Sargeson, and B. O. West. J. Phys. Chem. 61:1567 (1957).
33. T. Osakai and K. Ebina. J. Electroanal. Chem. 412:1 (1996).
34. R. S. Mulliken. J. Am. Chem. Soc. 74:811 (1952).
35. J. Rais, M. Kyrs̆, and M. Pivon̆ková. J. Inorg. Nucl. Chem. 30:611 (1968).
36. S. Motomizu, K. Tôei, and T. Iwachido. Bull. Chem. Soc. Jpn. 42:1006 (1969).
37. M. Kawasaki, K. Tôei, and T. Iwachido. Chem. Lett. 417 (1972).
38. T. Iwachido, M. Minami, A. Sadakane, and K. Tôei. Chem. Lett. 1511 (1977).
39. T. Iwachido, M. Minami, M. Kimura, A. Sadakane, M. Kawasaki, and K. Tôei. Bull. Chem.
Soc. Jpn. 53:703 (1980).
40. Y. Yamamoto, T. Tarumoto, and T. Tarui. Chem. Lett. 459 (1972).
41. Y. Yamamoto, T. Tarumoto, and T. Tarui. Bull. Chem. Soc. Jpn. 46:1466 (1973).
42. T. Kenjo and R. M. Diamond. J. Phys. Chem. 76:2454 (1972).
43. T. Kenjo and R. M. Diamond. J. Inorg. Nucl. Chem. 36:183 (1974).
44. S. Kusakabe, M. Shinoda, and K. Kusafuka. Bull. Chem. Soc. Jpn. 62:333 (1989).
45. S. Kusakabe and M. Arai. Bull. Chem. Soc. Jpn. 69:581 (1996).
46. T. Osakai, A. Ogata, and K. Ebina. J. Phys. Chem. B. 101:8341 (1997).
47. C. Sánchez, E. Leiva, S. A. Dassie, and A. M. Baruzzi. Bull. Chem. Soc. Jpn. 71:549 (1998).
48. A. Ogata, Y. Tsujino, and T. Osakai. Phys. Chem. Chem. Phys. 2:247 (2000).
49. T. Osakai and K. Ebina. J. Phys. Chem. B 102:5691 (1998).
50. T. Kakiuchi, M. Nakanishi, and M. Senda. Bull. Chem. Soc. Jpn. 61:1845 (1988).
51. R. D. Shannon. Acta Cryst. A32:751 (1976).
52. E. J. King. J. Phys. Chem. 74:4590 (1970).
53. H. Yokoyama, K. Shinozaki, S. Hattori, F. Miyazaki. and M. Goto. J. Mol. Lig. 65/66:357
(1995).
54. H. Yokoyama, K. Shinozaki, S. Hattori, and F. Miyazaki. Bull. Chem. Soc. Jpn. 70:2357
(1997).
55. H. D. B. Jenkins and K. P. Thakur. J. Chem. Educ. 56:576 (1979).
56. Y. Marcus, in Ref. 2, pp. 46–47; for IO 4 , see A. F. Kapustinskii. Quat. Rev. 10:283 (1956).
57. J. T. Edward. J. Chem. Educ. 47:261 (1970).
58. A. Karpfen and P. Schuster, in Ref. 4, pp. 265–312.
59. T. Osakai, K. Maeda, K. Ebina, H. Hayamizu, M. Hoshino, K. Muto, and S. Himeno. Bull.
Chem. Soc. Jpn. 70:2473 (1997).
60. E. R. Nightingale, Jr. J Phys. Chem. 63:1381 (1959).
61. J. Koryta, P. Vanysek, and M. Br̆ezina. J. Electroanal. Chem. 75:211 (1977).
62. V. Marec̆ek and Z. Samec. Anal. Chim. Acta 151:265 (1983).
63. T. Osakai, T. Kakutani, Y. Nishiwaki, and M. Senda. Bunseki Kagaku 32:E81 (1983).
64. T. Kakutani, T. Osakai, and M. Senda. Bull. Chem. Soc. Jpn. 56:991 (1983).
65. Z. Samec, V. Marec̆ek, and D. Homolka. Faraday Discuss. Chem. Soc. 77:197 (1984).
66. D. Homolka and H. Wendt. Ber. Bunsen-Ges. Phys. Chem. 89:1075 (1985).
67. T. Kakutani, Y. Nishiwaki, and M. Senda. Bunseki Kagaku 33:E175 (1984).
68. B. Hundhammer and S. Wilke. J. Electroanal. Chem. 266:133 (1989).
69. T. Osakai and K. Muto. Anal. Sci. 14:157 (1998).
70. S. Kihara, M. Suzuki, K. Maeda, K. Ogura, and M. Matsui. J. Electroanal. Chem. 210:147
(1986).
71. T. Ohkouchi, T. Kakutani, and M. Senda. Bioelectrochem. Bioenerg. 25:81 (1991).
72. http://dcwww.epfl.ch/cgi-bin/LE/DB/InterrDB.pl.

Copyright © 2001 Marcel Dekker, Inc.

You might also like