Activity Coefficient

This paper is in a collection of
"Historic Publications In Electrochemistry"

which is part of
Electrochemical Science and Technology Information
Resource (ESTIR)
(http://electrochem.cwru.edu/estir/)
THE JOURNAL
OF THE
Chemical Society
VOL. XLIII
JANUARY-JUNE
1921
Editor:
ARTHUR B. LAMB.
Anociate Editor.:
G. P. BAXTBR, L. W. JONES,
W. C. BRAY, S. C. LIND,
C. A. BROWNB, W. LASH MILLBR,
WARD HART, J. F. NORRIS,
W. F. HII.LBBRAND, W. A. NoYES,
OBN JOHNSTON, H. W. WILBY.
BASTON. PA
asCIIICHBACB PRINTING COIIP.AJfY
1921
1112 GILBERT N. LEWIS AND MERLE RANDALL.
[CONTRIBUTION FROM THE CHEMICAL LABORATORY OF THE UNIVERSITY OF CALIFORNIA.]
THE ACTIVITY COEFFICIENT OF STRONG ELECTROLYTES,I
By GILBERT N. LEWIS AND MERLE RANDALL.
Received March 10, 1921.
It has long been recognized that the formulas which first sprang from
Arrhenius' theory of electrolytic dissociation, and which very satisfac-
torily accounted for the behavior of weak electrolytes, were by no means
adequate to account for the behavior of strong electrolytes. This was
particularly the case with the Ostwald dilution law,
(
A )2
AO .m
~
AO
(1)
where A and A are the equivalent conductivities at the molality2 m,
and at zero molality, and KA. must be a constant if two assumptions are
correct: (1) that the molality of each ion is independent of the concentra-
tion, and (2) that for each substance concerned the molality is proportional
to the "active mass" or, as we should now say, to the activity.
To those who have not examined this question closely it may be sur-
prising to learn how far K
A
is from being a constant. For potassium
chloride, at three concentrations, m = 0.001, 0.01 and 0.1, K
A
is
O. 046, a .148 and 0.528, respectively.
It will be seen that K
A
increases nearly 12-fold when the molality
changes lOa-fold, and while these figures might be changed materially
on account of possible errors in A 0, such uncertainties are in no way ade-
quate to account for the enormous departure from the Ostwald dilution
law. It must, therefore, be concluded for such strong electrolytes, either
that A (because of some variation in the mobility of the ions) does not
measure the ion concentration, or that the substances concerned are very
far from obeying the ideallaws of the dilute solution.
These very peculiar characteristics of strong electrolytes in aqueous
solution, together with the presumption that the constituents of an elec-
trolytic solution, even at low concentrations, would depart radically from
1 This paper is a summary of several chapters of our book on "Chemical Thermo-
dynamics." Although that work is approaching completion, it seems that the material
here considered may be of some immediate interest to the growing number of inves-
tigators who are engaged in this important field. It has been impossible in such a
summary to discuss in detail the various methods which we have employed, or to give
derivations of all of our equations. These details will be reserved for our more com-
plete publication.
2 The term molality is llsed to express the number of mols of solute per 1000 g. of
water.
ACTIVITy COEFFICIENTS OF STRONG ELECTROLYTES. 1113
the ideal solution, led Lewis
l
to propose a general study, by purely thermo-
dynamic methods, of the activities of such constituents. In 1912 he col-
lected
2
the meager data which were then available for such thermodynamic
treatment, and showed that in all cases the activity of the ions is very
appreciably less than the ion concentration as calculated from AIA0.
He employed in these calculations measurements of electromotive force,
of the solubility of salts in the presence of other salts, and of freezing
points; showing how freezing-point data could be exactly employed in
these thermodynamic calculations.
S:ince the publication of that paper an extensive literature on this sub-
ject has developed, and the data obtained not only confirm Lewis' conclu-
sions, but furnish the abundant material for the calculation of ionic activi-
ties which we are to employ.
Notation and Conventions.
It is our custom to denote the activity of a solute by a2. If this solute
is a substance like sodium chloride we may denote by a+ and a_, respec-
tively, the activities of cation and anion, while a2 is called the activity of
undissociated NaCl, or more simply, of NaCI. In the case of a binary
electrolyte like this, the exact thermodynamic equation of chemical
equilibrium takes the form,
a+a_ = K, (2)
a2
where at any given temperature K is an exact constant.
\Ve are using the term activity3 in the newer sense of "relative activity."
Thus, at infinite dilution, we make the activity of each ion of sodium
chloride equal to its molality, which, assuming complete dissociation, is
the molality of NaCI. Now in the complete absence of any reliable in-
formation as to the concentration of the undissociated salt, we shall find
it extremely convenient to choose our standard state of that substance,
so that the K in Equation 2 becomes unity.4 We thus define the activity
of NaCl as the product of the activities of its two ions,
a+a_ = a2. (3)
At finite concentration the two ions may not have the same activity,
and it is often expedient to consider the geometrical mean of the two
ion activities, which we may denote by a and define by the equation
a = (a+aJ
Y2
= a2'1. Ci)
1 Lewis, Proc. Am. Acad., 43, 259 (1907); Z. physik. Chern. 61, 129 (1907); 70,
212 (1909).
2 Lewis, THIS JOURNAL, 34, 1631 (1912).
3 See Lewis, ibid., 35, 1 (1913); Lewis and Randall, ibid., 43,233 (1921).
< Likewise it is thermodynamically convenient, in the case of strong electrolytes,
to take the activity of any assumed intermediate ion, such as BaCl +. as equal to the
product of the activities of TIa ,"+ and Cl-.
1114 GILBERT ~ LEWIS A ~ D MERLE RANDALL.
(0)
The mean activity of the ions a ' divided by the molality of the elec-
trolyte, gives a quantity which has been called the thermodynamic de-
gree of dissociation, since it may be used to replace the degree of dissocia-
tion as used in the older approximate formulas. This quantity has also
been called the activity coefficient,1 and, in order to avoid any implica-
tion as to the molecular species which may be present, this is the term we
shall ordinarily employ henceforth. It will be denoted by 'Y.
So far we have been considering the case of a binary electrolyte like
KCI or CUS04. When we treat the more complicated types, such as
K
2
S0
4
, K
4
Fe(CN)6 and La2(S04)3 our equations become a little more com-
plicated. If an electrolyte dissociates into v = v+ + v- ions according
to the equation X = v+X+ + v-X-, we write for equilibrium,
a+
v
+ a_
v
- = a2; a = (a2)1/v. (5)
If now we wish to define the activity coefficient 'Y, so that in dilute solu-
tions it may be regarded as a thermodynamic degree of dissociation, and
become equal to unity at infinite dilution, we must no longer write it
equal to a / m. In a solution of barium chloride in which the molality m
is very small, a+ = m, a- = 2m, and a = [(m)(2m)2]7i = 2% m.
By defining the activity coefficient by the equation 'Y = a .. /(2% m) it
becomes equal to unity at infinite dilution. In the case of lanthanum
sulfate, La2(S04)3, which gives 2 positive and 3 negative ions, the cor-
responding factor is (2
2
3
3
)'/s. In general, we shall define the activity
coefficient by the equation,
m(v+v+v_V_)l/v'
We might have made this derivation clearer if we had introduced the
individual activity coefficients of the several ions. If in a solution of a
chloride the stoichiometrical molality of the chloride present is designated
by m-, and if the activity of the chloride is a ~ the activity coefficient of
chloride ion is defined as 'Y_ = ajm_. Furthermore, if we define the
mean molality of the ions, as we have their mean activity, and write it
as m' it is readily seen that
m = m(v+v+v_v_)I/v, (7)
and in place of Equation Gwe may write for the mean activity coefficient
(which we might for consistency have called 'Y )
'Y = a/m = ('Y+v+'Y_v_)I/v. (i:\)
1 The term activity coefficient has been used in two senses, sometimes to mean the
ion activity divided by the assumed ion molality, and sometimes to express the ion
activity divided by the gross molality of the electrolyte. This latter usage, to which
we shall find it desirable to adhere in the thermodynamic work, is more expressly desig-
nated by Briinsted (THIS JOURNAL, 42, 761 (1920)), as the stoichiometricul activity
coefficient.
ACTIVITY COEFFICIENTS OF STRONG ELECTROLYTES. 1115
Having established these conventions we may turn to the practical
problem of determining the activity coefficients of various electrolytes.
Activities from the Vapor Pressure of the Solute.
Few electrolytes are sufficiently volatile to permit the determination
of their activities from their own vapor pressures. However, when this
is possible it furnishes a method of great simplicity.
Bates and Kirschman1 have made a careful study at 25
0
of the partial
pressures of hydrogen chloride, bromide and iodide over their aqueous
solutions, and their results are given in Table I, in which the first column
gives the molality, the second, fourth and sixth the partial pressures of
the halides, and the third, fifth and seventh a quantity which is propor-
tional to the activity coefficient. If a2 is the activity of one of the undis-
sedated halides, and a is the mean activity of its ions, we have defined
the activity coefficient as a /111 = a/"/111. Hence if we consider a2 as
proportional to the vapor pressure p, then pY'i /111 is a quantity proportional
to the activity coefficient and may be written as k-y. The pressures are
given in millimeters of mercury, and we shall so leave them, as we already
have an undetermined constant.
TABLE I.
Activity Coefficients of Hydrogen Halides at 25 o.
HCI. HBr. HI.
kay.
0.00398
0.00610
0.0101
0.0191
0.0363
.-----"""'------..
P(mm.).
0,00057
0.00182
0.0065
0.0295
0.132
p(mm.), m. p(mm.). k!y.
4,0 0.0182 0.0337
5.0 0.0530 0.0460
6,0 0,140 0,0624 0.00151 0,00648
7.0 0.348 0.0843 0.00370 0.00869
8.0 0,844 0.115 0.0089 0.0118
90 1.93 0.154 0.0226 0.0167
10,0 4,20 0.205 0.059 0.0243
11.0 0.151 0.0353
If these measurements could be carried to high dilutions, k could be de-
termined as the limit approached by k-y at infinite dilution. This, how-
ever, is experimentally impossible in the present case, and therefore we
must leave k as an undetermined constant until a later section.
Activities from Distribution Ratios.
A very similar method of calculating the activities is available when we
know the distribution of an electrolyte between water and some non-
ionizing solvent, in which the activity of the electrolyte is known as a
function of the concentration (or assumed proportional to it). To illus-
trate such a case we may use the measurements of Rothmund and Drucker
2
on the distribution of picric acid between water and benzene. Their re-
sults, expressed in mols per liter, are given in Table II. The first column
1 Bates and Kirschman. THIS JOURNAL, 41, 1991 (1919).
2 Rothmund and Drucker, Z. physik. Chem" 46, 827 (1903).
shows the concentration in water, and the second in benzene. Here
again, taking a2 as proportional to CB, and identifying Cw with m, we may
write k'Y = CBY:l;Cwo 'The values of this quantity, which are given in
the third column, we have extrapolated to infinite dilution, obtaining 15.4.
But at infinite dilution 'Y = 1, and therefore 1.5.4 is the value of k. Divid-
ing now by this factor, we obtain the values of the activity coefficient
given in the last column.
TABLE II.
Activity Coefficient of Picric Acid at 18.
C\V_ CB
kyo y.
0.0334 01772 12.6 0.82
0.0199 0.0700 13.3 0.86
0.0101 0.0199 14.0 0.91
000701 0.0101 14.3 0.93
0.00327 0.00225 H.5 0.94
0.00208 0000932 14.9 0.97
0 0 15.4 1.00
Activities from the Vapor Pressure of the Solvent.
The method of calculating a2, the activity of the solute, from aI, the ac-
tivity of the solvent, we have already fully discussed in a previous paper. I
It is applicable without modification to electrolytic solutions. In dilute
solutions the accuracy obtainable by this method is far inferior to that
of the electromotive force or freezing-point measurements which we are
about to discuss, but in concentrated solutions it will furnish a very satis-
factory means of determining the ratio of the activities of the solute be-
tween two concentrations.
From Bronsted's2 measurements of the vapor pressure of water over
sulfuric acid solutions at 20 and at 30, we have interpolated the values
given in Table III, where the first and second columns give the mol frac-
tions and molalities of H
2
S0
4
, and the third gives PI, the vapor
pressure of water from the solution, divided by pl
o
, the vapor pressure of
pure water.
TABLE III.
Vapor Pressure of Sulfuric Acid Solutions at 25 C.
IV
2
m. p,/p,o. NdN,. lO+logpdPIo. k,y.
0.02175 1.236 0.959 449 9.982 0.00348
0.04255 2.467 0.878 22.5 9.944 0.00350
0.0801 4.88 0716 11.4 9.856 0.00469
0.1110 6.83 0.545 8.01 9.744 0.00760
0.2014 14.02 0.201 3.96 9.320 0.0227
02742 20.94 0.068 2.65 8.851 0.0477
I Lewis and Randall, THIS JOURNAL, 43, 233 (1921). In a case of this sort where
there is a wide departure from the laws of a perfect solution, there is no advantage to
be gained by employing the method of Equation 2.J. of that paper.
2 BriinstecJ, Z. physik. Chem., 68, 693 (1910).
Assuming that al = pI!PI 0, we may employ Equation 21, of our pre-
ceding paper,
I
N!
f d log a2 == - - d log a!.
. N2
We plot the values in the fourth column against those of the fifth, and
find the difference between the two values of log a2 by obtaining the area
under the curve. We may thus obtain a quantity which is proportional
to (J2, and if we then take the cube root and divide by m we have a series
of quantities which are proportional to 'Y. Such values of k!'Y we give
in the last column, and later we shall have occasion to check these values
with those obtained by two other methods of determining the activity
coefficient.
TABLE IV.
Activity Coefficient of Hydrochloric Acid Solutions at 25.
m,
')'. a'2. m.
"I
a,.
0.0005 0.991 0.000000246 0.75 0.788 0.348
0.001 0.984 0.000000969 1 0.823 0.6715
0.002 0.971 0.00000377 2 1.032 4.17
0.005 0.947 0.0000228 3 135 16.4
0.01 0.924 0.0000855 4 184 54.2
0.02 0.894 0.000319 5 2.51 158
0.05 0.860 0.00185 6 3.40 417
0.1 0.814 0.00664 7 4.66 1065
0.2 0.783 0.0246 8 6.30 2540
0.3 0.768 0.0530 9 8.32 5600
0.4 0.763 0.0929 10 10.65 11350
0.5 0.762 0.145 16 43,2 478000
0.6 0.770 0,213
Activities from Electromotive Force.
The e. m. f. of the cell, H
2
' HCII AgCII Ag (or HgCI, Hg) at 25
0
may be
expressed by the formula
0.1183 log 'Y = EO - (E + 0.1183 log m), (9)
where the e. m. f. is E when the molality is m, and EO at such concentra-
tion that a = 1. The very careful measurements of Linhart! permit
the determination of EO by direct extrapolation, and his measurements,
together with those of Ellis,2 of Noyes and Ellis,3 and of Lewis, Brighton
and Sebastian,4 permit the exact calculation of 'Y over a wide range of con-
cen.tration. By a careful plot of all these values we have obtained the
fig,ures of Table IV. In addition to the activity coefficient we have given,
in the third and sixth columns, the values of a2 (the activity of HCI as
1 Linhart, THIS JOURNAL, 39, 2601 (19li); 41,1175 (H119).
2 Ellis, ibid., 38, 737 (1916).
3 Noyes and Ellis, ibid., 39, 2532 (1917).
Lewis, Brighton and Sebastian, ibid., 39,2245 (1917).
such). It is instructive to see the enormous variation in this quantity.
For example, a2 increases half a million times between molal and 16 molal.
The quantity 'Y, even between 1Mand 2M rises abGve 100%, a phenomenon
which appears in the case of many electrolytes. Hence, while in dilute
solutions it is useful to think of the activity coefficient as a thermodynamic
measure of the degree of dissociation, such an idea loses all value in these
concentrated solutions where it might indeed be misleading to speak of a
"degree of dissociation" of over 4000%.
Vle may now compare these values with those obtained in Table I.
Assuming that 'Y = 1.84 at 4 M, k
1
of that table becomes 0.0183. The
values of k1'Y divided by this constant give the following values of 'Y,
which, except for the last concentration, are in excellent agreement with
Table IV.
m 4
"t (1.84)
5
2.51
6
3.41
7
4.60
8
6.27
9
8.43
10
11.20
Activity Coefficient from Freezing-Point Data.
l'he van't Hoff factor, i, which is the ratio between the molal lowering
caused by the electrolyte and the theoretical molal lowering (assuming
no dissociation), is a quantity which has been much used in the interpre-
tation of dilute solutions of electrolytes. It has been customary to con-
sider (i - 1) for a binary electrolyte as the measure of the degree
of dissociation, but the value so obtained is not the thermodynamic de-
gree of dissociation, nor does it give any valuable information regarding
the thermodynamic properties of the solute. At best the factor, i, gives
the ratio of the lowering of the activity of the solvent to the lowering which
would be produced by a normal undissociated substance. To proceed
from the activity of the solvent to the activity of the solute involves a
process of integration which requires a knowledge not only of the freezing
point at the concentration in question, but of a series of freezing points
extending to low concentrations.
The purely thermodynamic treatment of freezing-point data, which
permits the exact calculation of activities has been almost entirely neg-
lected, perhaps because of its mathematical complexity. Neve1"theless,
it furnishes a method
l
which is so general and at the same time so precise
that it has become more important than all other methods of calculating
the activity coefficient.
In developing the thermodynamic treatment of freezing-point data,
we shall proceed in three stages. First, we shall consider the very dilute
solutions; then the more concentrated solutions, in which, however, we
1 In this paper we are going to employ, with some extensions, the method used by
Lewis (THIS JOURNAL, 34, 1631 (1912)).
ACTIVITy COEFFICm;NTS OF STRONG EI,ECTROLYTES. 1119
neglect the heat of dilution; and finally, the general and exact calculation
for solutions of any concentration. If A is the molal lowering of any un-
dissociated solute at infinite dilution, namely,l 1.858, we have for any
solute iJ;1 dilute solution the thermodynamic formula
dfJ
d In az = -
Am
(10)
(11)
where fJ is the freezing-point lowering at the molality m. Now for an
electrolyte which produces v ion molecules when completely dissociated,
we may divide this equation through by v and find
1 I 1/. d. dfJ
- d In az = ( In az = In a", = -
v vAm.
For such an electrolyte the molal lowering at infinite dilution is vA. We
may define a function j, which becomes equal to zero at infinite dilution.
by the equation
or
j = ~
vAm'
By combining this with Equations 8 and 11 we find
d In 'Y = -- dj - jd In m,
In 'Y = - j -- J: jd In m.
(12)
(13)
(14)
The valuation of the last term, which might be done by graphic integra-
tion, is in the present case far more simply treated by the use of a very
general principle found by Lewis and Linhart. z They discovered that
electrolytes, of all the various types which have been accurately investiga-
ted, approach as a limiting law in dilute solutions
(15)
and that for all the salts considered this law was obeyed within the limits
of experimental error below 0.01 M. At higher concentrations the values
of j always become smaller than the ones calculated from the formula, and
in some cases j passes through a maximum.
Having determined {3 and a for some one electrolyte we at once deter-
mine 'Y in any dilute solution of that electrolyte by combining Equations
14 and 15 and integrating, whence
In 'Y = - a + 1 {3m" = _ a + 1 j. (16)
a a
1 Lewis and Linhart, THIS JOURNAL, 41, 1951 (1919)
7 This value is based 00 the recalculation of the heat of fusion of ice by Lewis
CTHIS JOURNAL, 30,668 (1908) aod has become very ienerallyaccepted in place of
the old value of 1.85.
1120 GtLBE:R.'f ~ LEWIS AND MERLE RAKDALt.
By employing this equation Lewis and Linhart obtained a series of
values for 'Y for various electrolytes, a part of which we reproduce
1
in
Table V.
T B L l ~ V.
Activity Coefficients of very Dilute Solutions (at Any Temperature) at Several Values
of m.
a ~ 0.01 0.005 0.002 0.001 0.0005 0.0002 0.0001
KCI, NaCI.. 0.535 0.329 0,922 0,946 0.967 0,977 0.984 0,990 0.993
KNO
s
....... 0.565 0.427 0.916 0.943 0.965 0.976 0.984 0.990 0.994
KIOs, NaIO
s
0.500 0.417 0.882 0.915 0.946 0.961 0.972 0.982 0,988
K
2
S04 0.374 0.572 0.687 0.749 0.814 0.853 0.885 0.917 0.935
H2S04 0.417 0.970 0617 0.696 0.782 0.831 0.871 0.910 0.932
BaCh 0364 0.477 0.716 0.771 0.830 0.865 0.894 0.923 0939
CoCI
2
...... 0.362 0.441 0.731 0,784 0.840 0.873 0.900 0.927 0.943
MeS04...... 0.38 1.44 0.40(4)0.50 0.61 0.69 0.75 0.81 0.85
K
s
Fe(CNk.0.420 1.148 0.571 0.657 0.752 0808 0.853 0.897 0922
La(NOsk ... 0.420 1.148 0.571 0.657 0.752 0.808 0.853 0897 0.922
This investigation showed for the first time the general trend of the
activity coefficient in very dilute solutions. The table may be in error
in some details, for it is based upon measurements in which the smallest
inaccuracies cause a relatively large error. However, we may place re-
I We have added to this table values of 'Y for KNO
s
obtained from a very accurate
series of measurements by Adams, which was inadvertently omitted by Lewis and
Linhart. In examining the plot for KIOs and NaIO
s
we find that the individual points
are hardly sufficiently concordant to determine with any accuracy the value of a.
In fact within the limits of experimental error a is one-half, probably for all of the llni-
univalent salts given in the table, and also for other uni-univalent salts which we are
going to study by another method. The value of a previously given for these iodates,
namely, 0.442, was the only one falling below 1/2. We have preferred to recalculate
these values on the assumption that a = 0.50, which, together with the experimental
value of j at 0.01 M, makes ~ = 0.417. Also it may be noted that, looking over the
numerical calculations of Lewis and Linhart, we are led to one minor change, namely,
for 0.01 M KCI and NaCI, 'Y = 0.922 instead of 'Y = 0.925. Finally, in considering the
sulfates of Mg, Zn, Cd and Cu, which we have grouped together as MeS04, we have
altered the values given by Lewis and Linhart. (Average a = 0.31, ~ = 1.08, 'Y
(at 0.01 M) = 0.33.) Those values rested chiefly upon measurements made by Haus-
rath in 1902. These were undoubtedly the most accurate freezing-point data ever
obtained up to that time. But in the meantime important improvements in ther-
mometry have been made, and Hausrath's results in very dilute solutions must be re-
garded as uncertain. By a method which we cannot give here in detail, and which
involves the use of Horsch's measurements (THIS JOURNAL, 41, 1787 (1919)) on cells
with zinc electrodes in ZnS04 and in ZnCh, we have obtained new values for the ac-
tivity coefficients of ZnS04. By this method at 0.01 M, 'Y = 0.47. Giving equal weight
to the two independent determinations, we obtain as an average at 0.01 :M, " = 0.40,
and this value will be corroborated by another method which we shall describe later.
Moreover, this value is in better agreement with the results on MgSO. alone, as obtained
by Hall and Harkins (THIS JOURNAL, 38, 2658 (1916)), than the one used by Lewis
and Linhart. Assuming, therefore, this value of 'Y and taking the average j at 0.01 M
as 0.25, we obtain the values given in the Table.
ACTIVITy COEFFICIEKTS OF STROt\G 1121
liance upon the general results of this calculation, and also upon most of
the individual results.
When we compare the results of Table V with the values of A/A 0,
which have ordinarily been used as a measure of the degree of dissocia-
tion, we find a divergence between A/A 0 and the thermodynamic de-
gree of dissociation which, especially in salts of the higher types, is truly
astonishing.
1
For the sake of comparison we give in Table VI the values
of y and of A/A0 at 18
0
for a number of salts at 0.01 M.
TABLE VI.
Comparison of 'Y and A/A.
o
for several Salts at 0.01 M.
KCl. NaCl. KNO,. ](,SO.. BaCh. CdSO.. CuSO.. La(NO.) .
'Y 0.922 0.922 0.916 0.687 0.716 0.404 0.404 0.571
A/A
o
0.941 0.936 0.935 0.832 0.850 0.53 0.55 0.75
The assumption that, at 0.01 or 0.001 molal, A/A 0 can be taken as a
satisfactory measure of the thermodynamic degree of dissociation or ac-
tivity coefficient, is evidently very far from the truth, especially for salts
of a higher type than the uni-uni.valent. In the case of copper sulfate,
where this great divergence was first pointed out by Lewis and J..,acey,2
we find that even at so great a dilution as O. 001 M, A/A 0 is 0.80 while
'Y is 0.69.
On the other hand, when we compare 'Y with the "corrected degree of
dissociation" which Lewis
3
obtained by combining conductivity and
transference data we find, not perhaps an identity, but certainly a re-
markable parallelism between those values and the activity coefficient.
This corrected degree of dissociation is always a little lower for nitrates,
bromates, and other salts of the oxygen acids than for chlorides or bro-
mides. J...ikewise it is smaller for silver and thallium salts than for salts
of the alkali metals. These are precisely the conclusions which we are
going to reach concerning the activity coefficient from our inspection of
Table V and following tables.
Before turning our attention from the Lewis and Linhart equation, we
may point out some interesting corollaries which will be found very use-
ful in practical calculations. We note that within the limits of experi-
mental error a for a uni-univalent electrolyte is 1h. This is shown not
merely by Table V, but also by other data which will be mentioned later.
1 Another proof that any relation between 'Y and the values of A/A0, which are
uncorrected for changing mobility, must be a very rough one, is furniEhed by a con-
sideration of the change of these quantities with the temperature. In the very dilute
solutions which we are now discussing, the heat of dilution is always negligible, and the
values of 'Y must be independent of the temperature. But the values of A/A0 dimin-
ish markedly with increasing temperature. For several salts at 0.08 N, studied by
Noyes and Melcher (Carnegie Inst. Pub., 63, 1907)) this diminution between 18
0
and
100
0
ranged from 3% for NaCI to 30% for MgS0
4
2 Lewis and Lacey, THIS JOURNAL, 36,804 (1914).

! Lewis, ibid., 34, 1631 (1912).
11:.2:.2 GILBERT 1\. LEWIS AND MERLE RA!\DALL.
Therefore, at least as a close approximation to the truth, for uni-univalent
electrolytes
(17)
(19)
(:.20)
the classic equation of
or by Equation 16,
In I' = - 3!3m
J1
= - 3j, (18)
which furnishes an extremely simple method of calculating the activity
coefficient in dilute solution from the freezing point. We may put this
equation in a different form by expanding in series, whence
I' = 1 -- 3m
Yz
+ 1/
z
(3m
Y
:i)" + ... ,
and for very dilute solutions the last term disappears, giving
I' = 1 - 3m)1 = 1 - 3j.
It is interesting to compare this equation with
van't Hoff, which we may write
I' = i - 1 = 1 - 2j.
In other words, in any dilute solution I' obtained from the van't Hoff
equation differs from unity by only zfa as much as the true 1'. Thus for
0.001 M potassium chloride the experimental value of j is 0.0290, whence
from Equation 20, 'Y = 0.97fi. while by the van't Hoff equation it would
be 0.983.
1
Finally, it is to be noted that if Equation 20 is correct at high dilution.
'Y should be a linear function of m
l
/", or, in other words, if we plot the ac-
tivity coefficient against the square root of the molality, the curve at
small values of m should become a straight line, intersecting the I' axis at
unity.
Preliminary Treatment of Concentrated Solutions.-When we pass
to the more concentrated solutions there are two corrections necessary in
Equation 14. The first of these involves the heat of dilution and this
one we shall ignore for the moment. The other is due to factors such
as the change of the heat of fusion with the temperature, and these to-
gether lead to a new term, so that in place of Equation 14 we write"
d I d
"d I + 0.00057 iJ d
o
('_')1)
n I' = - J - J n m --- - 'IJ.
)I m
Changing for convenience to common logarithms, and integrating,
j Jm 0.00025 Jm iJ
10g'Y = - 2.303 - 0 d log m + --v- 0 m
diJ
. (:.22)
In evaluating the first of these definite integrals we may no longer em-
1 It is a curious coincidence that for uni-univalent salts this false value of 'Y is
not far from the value of AlAo. This coincidence has doubtless been in part responsible
for the previous neglect of the correct thermodynamic method of using freezing-point
data.
2 For the complete development of these equations we can only refer the reader to
the more extended publication which is in preparation.
ACTIVITY COEFFICIENTS OF S'fRO)!G ELEC'fROLYTES. 11:23
ploy the equation of Lewis and Linhart, but the integration may be
readily made graphically by plotting i against log tn, and determining
the area under the curve.
1
The second definite integral, which usually
gives a small correction term, may be obtained by a very rough plot of
{} 1m against iJ.
We shall illustrate the use of Equation 22 by means of the freezing-point
data on sodium chloride. Here we make use of the measurements up to
(J. 1:36 ]\I! made by Harkins and Roberts,2 those of Rodebush,3 between
molal and the eutectic of NaC1 2H
2
0, and a few intermediate points of
Jahn. 4 In Table VII we give the various terms which enter into Equa-
tion 22 and the values obtained therefrom for the activity coefficient.
These provisional values of the activity coefficient which we have thus
obtained when ignoring the heat of dilution are denoted by ,,/'. Antic-
ipating the work of the following paragraphs, we give in the last column
of this table the true values of the activity coefficient of sodium chloride
at 2;) 0, obtained by the general method which we now have to consider.
TABLE VII.
Provisional Values of the Activity Coefficient of Sodium Chloride.
j
fin 'd
fin,'}
-logy'. -y'. In.
2.303'
o J l o ~ In.
0.000125 - d ~ '"'(29S.
o In
0.01 0.0126 0.0227 00353 0.922 0.922
0.02 0.0169 0.0327 0.0496 0.892 0.892
0.05 0.0235 0.0513 0.0748 0.842 0842
0.1 00287 0.0694 0.0001 0.0980 0.798 0.798
0.2 0.0339 0.0913 0.0002 0.1250 0750 0.752
D.S 0.0413 0.1259 0.0008 01664 0.682 0.689
LO 0.0456 o.lS6\i 0.0014 0.2008 0.630 0.650
:?O 0.0317 0.1840 0.0029 0.2128 0.613 0661
:3.0 o 0135 01937 0.0046 0.2026 0627 0.704
4.0 -00065 01950 0.0064 0.1821 0.657 0.765
ii.O -00347 0.1909 0.0090 0.1472 0713 0.852
6.2 -0.0425 0.189.5 0.0095 0.1375 0.729 0.874
Final Calculation of the Activity Coefficient at a Given Temperature,
Neither the Heat of Dilution nor its Temperature Coefficient being
Neg1ected.-When the heat of dilution is neglected, the values of the
activity coefficient which we have called "/' are independent of the tempera-
ture. In concentrated solutions, where the heat of dilution becomes ap-
preciable, such values of the activity coefficient are approximately valid
only in the neighborhood of the freezing point. If we wish to calculate
1 It is best to determine graphically only the area between 0.01 M and the molality
in question, while the integral to 0.01 1Y1 is obtained from the equation of Lewis and
Linhart.
2 Harkins and Roberts, THIS JOURKAL, 38, 2676 (lQ16).
3 Rodebush, ibid., 40, 1204 (1918).
I Jahn, Z. physik. Chem., 50, 1293 (1905); 59, 3(907).
11:2-i GILBERT r-.;. LEWIS AND MERLE RANDALL.
(:23)
from the freezing-point data the true values of 'Y at 25
0
we may still use
Equation 22 if we add another term, namely if we write
55 51 Jm dx
log 'Y = log 'Y' - -'-
II 0 m
where
1'''-1''
x = - LIT" +
2.303RT"T'
(
c - CO) (1''' 1''' - 1" _ ~ log 1''') (24)
Pl Pl 2. 303RT"T' R 1'"
where 1''' is 298.1
0
A, 1" is the absolute temperature of the freezing point
at the given concentration, Ll is the partial molal heat content of the water
when that of pure water is taken as zero, and C
P1
and ~ are, respectively,
the partial molal heat capacity of water in the solution of given concen-
tration, and at infinite dilution.!
Employing this general equation we obtain the values of 'Y at 2;') 0
(298.1
0
A) already given in the last column of Table VII. It is evident
that if we neglect the effect of heat of dilution we make in this case an
error of 3% at 1 M and an error of 20% at 5.2 M.
Comparison of these Activity Coefficients with Those from Electro-
motive-Force Measurements.-\Ve are now in a position to make one
of the most remarkably satisfactory checks which has yet been afforded
by applied thermodynamics. The results of this arduous calculation
of the activity coefficient of sodium chloride at 25
0
, from the freezing-
point data, we may compare directly with the results of Allmand and
Polack,2 who studied the electromotive force at 25
0
of the cell, Na(amalg.) I
NaCl(aq) IHgCll Hg, using the amalgam electrodes of Lewis and Kraus.
3
Their measurements at various concentrations enabled them to calculate
values which are proportional to the activity coefficient. Multiplying
those values by a constant factor in order to give at 1 M the same value
TABLE VIII.
Activity Coefficients of Sodium Chloride at 25
0
from Electromotive Force.
m.
o 1
D.5
1
3.196
6.12 (saturated)
-y.
0.784
0.681
(0650)
0.726
1.013
1 For a discussion of these partial molal quantities see Lewis and Randall (THIS
JOURNAL, 43, 233 (1921). The values of 1.1 and CPl are obtained from the work of
Randall and Bisson (ibid., 42, 347 (1920).
2 Allmand and Polack, J. Chem. Soc., 115, 1020 (1919).
3 Lewis and Kraus, THIS JOURNAL, 32, 1459 (1910).
ACTIVITY COEFFICIENTS OF STRONG EI.EcTROLYTEs. 1125
as that which we have found in Table VII, we obtain the activity coeffi-
cients of Table VIIU
In order to show the relation between these values and our previous
ones we show in Fig. 1 the activity coefficient as a function of m'lo. The
T
~
I--.-
~
/
i
I
I
i
1.0
0.0
0.0 0.5 1.0 1.5 2.0
m'l" square root of molality.
Fig. 1.-Activity coefficient of sodium chloride.
2.5
continuous curve is from the data of Table VII and the circles represent
the values of Allmand and Polack. The average deviation of these points
from the curve is only a few tenths of a per cent., and when we consider
that, in the concentrated solutions, the use of the Land cp values intro-
duced a term which alone amounted to 20%, the agreement is most strik
ing.
The Activity Coefficient of Sulfuric Acid by Three Independent Methods.
Another case in which we have material for a very satisfactory compari-
son of the activity coefficients obtained by various methods is afforded by
aqueous solutions of sulfuric acid.
Freezing Points.-In the freezing-point method we may employ the
data which we have already considered in obtaining Table V, together with
those of Drucker,2 Roth and Knothe,3 and Pickering and Barnes.
4
We
proceed just as in the case of sodium chloride, except that in Equation
23, II = 3. The results are given in Table IX.
5
1 We have not utilized their r e ~ u t s at concentrations below 0.1 M where the ex-
perimental uncertainty proved to be large.
2 Drucker, Z. Elektrochem., 17, a98 (1911).
3 Roth and Knothe, published only in Landolt-Bornstein Roth Tables.
Pickering and Barnes, Trans. Roy. Soc. Canada, [II] 6, 3, 37, (1900).
For this whole calculation of the activity of sulfuric acid from freezing-point
data, we are indebted to Mr. T. F. Young. We have not checked his calcUlations, but
have much confidence in their accuracy.
1126 GILBER.T N. MSR.LE RANDALL.
TABI.:R IX.
The Activity Coefficients of Sulfuric Acid at 25
0
from Freezing Points.
....
i .

"('. ")'111.
3 0 ...
0.01 0.142 0.617 0.617
0.02 0.197 0.519 0.519
0.05 0.258
.
0.397 0.397
0.1 0.293 0.00407 0.316 0.313
0.2 0.320 0.00795 0.248 0.244
05 0.329 0.0152 0.182 0.176
1.0 0.280 00230 0.155 0.147
2.0 0.134 0.0360 0.155 0.143
4.0 -0.350 0.1202 0.268 0.203
5.0 -0.650 0.2320 0.413 0.242
In this table the first column gives the molality and the second the value
of l calculated from the best value of the freezing point. The fourth col-
umn gives "'(', which is the activity coefficient calculated from Equation
22, or, in other words, calculated without consideration of the partial
heat contents and heat capacities. The last column gives the results of
the complete calculation of the activity coefficient at 25.
The partial molal heat capacities and heat contents have been calcula-
ted, respeclively, from the papers of Birol1s
1
'andof Bronsted.
2
In this
case 6f sulfuric acid the correction caused by the il1troditction of the thermal
terms is opposite in sign to the corresponding correction for sodium chloride,
so that in this case "Y298 is less thal1 "'('. These thermal effects are so large
that, at 5 M, "'(298 is only about half of "'('. Thus, in the concentrated
'Y is .changing witP, temperature, and at 5. M it
nearly doubles in going from 25, to a temperature in the neighborhood
of the freezing point, which is -46.
Electromotive Force.-The cell, H
2
/ H
2
S0
4
I Hg
2
SO" Hg has been
thoroughly investigated at 25. Randall and Cushman,3 who have made
the most recent study of this cell, obtail1ed very reproducible results,
which, moreover, they showed to be in the main concordant with the
earlier results of Bronsted and of Edgar, in concentrated solutions, and
with those of Lewis and Lacey in the dilute solutions. Their results (to-
gether with one of Bronsted's at 14 M) are given in Table X, where the
first column gives the molality of the sulfuric acid, and the second gives
three. m. f. of the cell (hydrogen at one atmosphere). The last column
gives values of k
2
"'( obtained as follows.
1 Birons, J. Russ. Ph,s. Chem. Soc., 31, 100 (1899);. see Domke and Bein, Z.
anorg. Chem., 43, 125 (1904).
2 BrOnsted,Z. ph,"k. Clum., 68, 698 (1910).
3 Randall and Cushman, THIS JOTJRNAI., 40, 398(1918).
TABU X.
Electromotive Force of the Cell H
2, H
2
SO" Hg
2
SO" Hg at 25.
m. E. k,'Y.
0.0506 0.7544 621
0.505 0.6960 283
1.031 0.6751 239
3.637 0.6187 291
8.204 0.5506 761
14.00 0.4977 1757
We have the equation,
or. by Equation G,
E
RT
EO - - In a2
2F '
(25)
E
-_ EO 3RT l ( 4
73
)
-- 2F n I'm .
We may, for simplicity, collect the constants and write
3RT
E = -- 2F In (k21'111).
(26)
(27)
Thus each electromotive-force measurement yields a quantity which is
proportional to the activity coefficient and is here represented as k
2
'Y.
These are the values given in the last column of Table X. It is difficult
in this case to determine k
z
(which is the same as determining EO) by a
process of extrapolation, since it was shown by Lewis and Lacey! and by
Horsch
z
that, in solutions as dilute as 0.005 M, the use of the mercurous
sulfate electrode gives results which are not in agreement with other
thermodynamic methods. This is undoubtedly due to the fact that at such
concentrations the solubility of mercurous sulfate cannot be ignored.
However, by assuming the value of I' at one concentration from Table IX,
namely, 'Y = 0.397 at 0.05 M, we nnd k
2
= 1578, and thus obtain the
other values of I' which we are about to compare with those of Table IX.
Activity Coefficient from the Vapor Pressure, and Comparison of the
Three Series.-We must also consider another series of values, which
we have already obtained in Table III, namely, the values of k1l' ,obtained
from the vapor pressure of water over aqueous sulfuric acid. Here again,
by interpolation from Table IX, we nnd I' = 0.238 at 4.88 M. Hence,
k
1
= 0.001\)7, and we obtain another series for 1'.
The agreement between our tbree series is shown in Fig. 2, where we
have plotted the values of I' obtained by all three methods, against the
cube root
3
of the molality. The agreement between the several series
is not quite as complete as it was in our previous case of sodium chloride,
but is, nevertheless, an even more powerful connrmation of the adequacy
1 Lewis and Lacey, THIS JOURNAL, 36, 804 (1914).
2 Horsch, ibid.; 41,1787 (1919).
a Th;s is usually the most convenient {llQt w\len v '" .3.
3.0 0.5 1.0 1.5 2.0 2.5
mIlo, cube root of the molality.
Fig. 2.-Activity coefficient of aqueous sulfuric acid.
I
I
o Electromotive Force
- [J Freezing
Points
I
Q V"por Pressure
/
\
,
7
"
/
"
c.--
0.0
0.0
0.5
2.0
2.5
of our several methods. For here we are dealing with an exceptionally
large range of concentrations, and the values of 'I' obtained from the freez-
ing points depend chiefly upon the successful employment of the partial
molal heat content and heat capacity. To illustrate this point we give
in the figure a few values of '1" (points in parentheses) which show the re-
sults which would have been obtained if these thermal quantities had been
neglected.
From the plot we have chosen the final values of 'I' given in 'fable XI.
TABU XI.
Final Values of the Activity Coefficient of Sulfuric Acid at 25 o.
m. ')'. m ')'.
0.01 0.617 2 0.147
0.02 0.519 3 0.166
0.05 0.397 4 0.203
0.1 0.313
fi 0.242
0.2 0,244 10 0.660
0.5 0.178 15 1.26
1 0.150 20 2.22
ACTIVITY COEFFICIENTf' OF STRONG ELECTROLYTES. 1129
Methods of Evaluating Activity Coefficients when Data at High Dilu-
tions are Lacking.
It frequently happens, as we have seen, that a series of data permits
the calculation of k'Y, a quantity proportional to the activity coefficient,
without affording means for any very accurate extrapolation which
would give the value of k. If we have a series of values of k'Y for some
uni-univalent salt, extending down to about 0.01 M, all of the values
having about the same percentage accuracy, the extrapolation can be
made with a fair degree of precision. For example, we may plot the
values of k'Y against the square root of the molality. The curve through
the several points approaches a straight line at low concentrations, and
the intercept of this line upon the axis of k'Y is the value of k.
Unfortunately in most cases the values of k'Y are subject to greater
uncertainty, the higher the dilution, and if we should give full weight to
the experiments in the most dilute solutions the extrapolation would be
entirely misleading. Accurate measurements of electromotive force always
become increasingly difficult at high dilution, and in employing existing
data it is often necessary to ignore completely results obtained with the
most dilute solutions.
Perhaps the best method of treating cases of this kind consists in com-
paring the data obtained for the given electrolyte with the data for some
other electrolyte of similar type, which has been carefully investigated.
Here we need only to assume that for salts of a similar type the activity
coefficients approach one another in some regular manner as the dilution
is indefinitely increased. In order to illustrate the method we may em-
ploy the data obtained by Noyes and MacInnes,l and their associates, for
potassium and lithium chlorides, and potassium hydroxide.
These investigators, using amalgam electrodes of the Lewis and Kraus
type, exercised the greatest care, by the exclusion of oxygen and other-
wise, to avoid the complications which attend the use of these high1;y re-
active amalgams. They were thus able to obtain an unexpected degree
of accuracy and reproducibility. Nevertheless it appears to be impossi-
ble to prevent side reactions which are comparatively harmless in con-
centrated solutions, but which, in the measurements with dilute solutions,
produce far greater error than the authors supposed.
In re-examining the activity coefficients given by Noyes and MacInnes,
we must therefore recognize two sources of uncertainty.2 Their ac-
tivity coefficients may have to be multiplied by a constant factor, be-
cause of a change in the method of extrapolation to infinite dilution, and
1 Noyes and MacInnes, Tms JOURNAL, 42, 239 (1920).
2 A third source of uncertainty enters in the piecing together of several overlapping
series of measurements. This probgbly causes no difficulty in the case of LiCI and KCl.
but may in the case of KOH. .
1130
GILBERT :\. LEWIS AND MERLE RA.NDALI.
0.25 0.50 0.75
m
1h
, square root of molality.
Fig. 3.
I
I i
,
,
I ! i
i I
i
I I
W:
j
,
i I I
I
I
0
: I
I
I
C I
I
I
I
I
I--
--
--
i I
.......
I
095
0.90
0.0
1.00
it may be necessary to give less weight to the measurements III dilute
solution.
Potassium Chloride.-These questions are answered immediately by
comparison of potassium chloride with sodium chloride, for which we
have obtained accurate values of the activity coefficient. Table XII
contains in the second column the values of 'Y given by Noyes and Mac-
Innes. The next column gives the values of 'Y for sodium chloride fur-
nished by our previous tables. The next shows the ratio between thc
values of the second and third columns. Now according to our assump-
tion, this ratio should in some regular manner approach constancy as the
concentration diminishes.
In Fig. 3 we have plotted these ratios aganist m
Y2
, and the smooth
curve which we have drawn is obtained by ignoring the points at low con-
centrations and extrapola-
ting from the points at
high concentrations. The
ordinate at zero concentra-
tion by this extrapolation
is 0.935, and if we divide
any other ordinate by this
value we obtain the value
of 'Y at the corresponding
concentration. The values
so obtained are given in
1.0 the last column of Table
XII. They differ as a rule
by several per cent. from
the values of Noyes and MacInnes, but are completely corroborated by
freezing-point measurements as we shall show presently.l We have
treated in a similar manner their results for lithium chloride and the re-
sults will be given later in Table XIV.
Cadmium Chloride.-A very beautiful illustration of the inter-
pretation of the data for one salt, by comparison with those for another,
is furnished by the case of cadmium chloride, whose abnormal behavior
puts it in a class by itself. For this salt we have the measurements of
Horsch
2
which, upon examination, prove to be more accurate than the
1 We might also have used the measurements of Harned (THIS JOURNAL, 38,
1986 (1916, and of Getman (ibid., 42,1556 (1920, IOn concentration cells with liquid
junctions (with transference). This, however, requires a knowledge of the transference
number, which can be most accurately determined by these very measurements. It
may be noted that if we take the ordinarily accepted transference number of KCI the re-
sults given in Table XII are Dot in accord in dilute solution with the measurements of
MacInnes and Parker (ibid., 37, 1445 (l915)) on concentration cells with transference.
2 Horsch, THIS JOURNAL, 41, 1787 (1919).
author himself claimed. He measured the electromotive force of a cell
with cadmium amalgam and silver--silver chloride electrodes, with various
concentrations of cadmium chloride.
TABLE XII.
Activity Coefficient of Potassium Chloride at 25 from Electromotive Force.
'I-KCI.
m. N & M. "(-NaCI. Ratio. "(-KCI.
0.001 0.979 0977 1.002 0.977
0.005 0.923 0.946 0.976 0.946
0.01 0.890 0922 0.964 0.922
o 0:) 0 .823 0 .869 0 .947 0 .869
0.05 0.790 0.842 0.938 0.841
o 1 0 .745 0 .791' 0 934 0 .796
0.2 0.700 0.752 0.931 0.749
o.5 0 638 0 .689 0 .928 0 .682
1.0 0.593 O.6,"iO 0.912 0.634
Our methods here will be somewhat altered. Knowing that cadmium
chloride is a peculiar electrolyte, we may nevertheless in very dilute solu-
tions assume, as a first approximation, that it has the same activity coef-
ficient as barium chloride. The calculations we are about to make will
show to what extent this assumption is justified.
Using the same equation as for sulfuric acid (Equation 26), we might,
from each measurement of E, calculate a value of EO on the assumption
that 'Y is the same as we have found for barium chloride in Table V. These
values, which we may call EO', should be constant and equal to the true
EO if the measurements are correct, and if 'Y is really the same for cadmium
and barium chloride. The results of this calculation are shown in Table
XIII.
same
as
BaCh
Chloride at 25,
"(CdC!,).
0.809
0.697
0.652
0.568
0.219
0025
of 'Y for the two
Fortunately, the
TABLE XIII.
Electromotive Force of the Cell and the Activity of Cadmium
m. "(BaC!'). E. EO'.
0.000103 0.939 09060 0.5676
0.000109 0.938 0.9023 0.5661
0000114 0.937 0.9011 0.5665
0.000127 0.934 08978 05673
0000]53 0.930 0.8926 0.5691
1)000214 0.923 0.8803 0.,'5705
o.000336 0 .905 0 .8644 0 .5702
0.000366 0.901 0.8614 0.5703
I) 000479 0 .894 0 .8520 0 .5690
11 .000924 0868 0.8296 0.5727
0.00258 0.816 0.7958 0.576]
0.00352 0.796 0.7864 0.5777
0.0074 0.741 0.7630 0.5802
0.0995 0.6996
6.62 0.6220
It is evident that the assumption of equal values
salts is far from correct, except at very high dilution.
experimental results are of such accuracy that we find a range of concen-
tration, namely, between 0.000366 and 0.000153 M in which EO' is nearly
constant and we may take the true EO as O. ;)700. At lower concentra-
tions the experimental uncertainty, and the effect of the solubility of
silver chloride, become too great to make the data useful.
Having found the value of EO we may now obtain the activity coefn-
cients for cadmium chloride by Equation 26. These values are given,
for the more concentrated solutions, in the last column of the table. The
great divergence between these values and the corresponding ones for
barium chloride persists to concentrations as dilute as 0,0005 M. We
have seen other cases of slight disagreement between the activity coeffi-
cients of salts of the same valence type, but this is evidently a different
kind of phenomenon. The fact is that cadmium chloride cannot be re-
garded as a strong electrolyte. It possesses in less degree the charac-
teristics of the analogous substance, mercuric chloride, which shows almost
no ionization.
Activity Coefficient of Several Typical Electrolytes.
It will be convenient to have before us a table of the activity coefficients
of a few electrolytes of the several types, which we present in Table XIV.
The values for several of these electrolytes are merely a recapitulation
of previous tables.
TABLE XIV.
Activity Coefficients of Typical Electrolytes.
.70
1 20
G.70J
1
0.823
0.776
0,650
0,634
o 75
0.244 0.178 0.150
0.271
0.119
m = 0,01 0.02 0,05 0.1 0,2 0,5
HCI (25) , 0.924 0.S94 0.860 0.814 0.783 0.762
LiCI (25) 0,922 0.892 0.843 0,804 0.774 0.754
(25) , 0.922 0.S92 0.842 0.798 0.752 0.689
leCI (25) ,., 0,922 0,892 0,840 0.794 0,749 0.682
leOH (25) 092 0,89 0,84 0.80 0,7fi 0,73
0 916 0.878 0.806 0.732
0.902 0.857 0.783 0.723 0,655 0,526 0,396
leIO
a
, NaIO
a
0.882 0.840 0.765 0,692
BaClz , 0.716 0.655 0.568 0.501
CdClz (25) 0.532 0.44 0,30 0.219
K
2
SO,. , , 0.687 0.614 0.505 0.421
H
2
SOd25) 0.617 0.519 0.397 0.313
""""'" 0,571 0.491 0.391 0326
MgSO,. ' 0.404 0.321 0225 0.166
CdSO, 0.404 0,324 0,220 0.160
CuSO, 0.404 0,320 0.216 0,158 0.110 0.067
The values for potassium nitrate and iodate, sodium iodate, barium
chloride, potassium sulfate, lanthanum nitrate, magnesium, cadmium and
cupric sulfates, were obtained by Equation 22, from the freezing-point
data collected in the paper of Lewis and Linhart. They correspond to
the values of "(' which we obtained for sodium chloride. Not having been
corrected for the heat of dilution, they are strictly valid only in the neigh-
borhood of the freezing point. However, at these moderate concentra-
tions the corrections due to the Land c
p
terms are certainly small, and,
except perhaps for one or two of the most concentrated solutions, it seems
safe to use the given values as the activity coefficients at 25.
A similar calculation for potassium chloride may be compared directly
with the results of Table XII. The values found from the freezing-point
data are given in Table XV. It is evident that in the dilute solutions the
values are practically identical with those of the previous table, thus
fully corroborating our method of employing the data of Noyes and Mac-
Innes, which has been used also for lithium chloride and potassium hy-
droxide.
TABLE XV.
m 0.01 002 0.05 0.1 0.2 0.5
,' 0.922 0.892 0.839 0.791 0.737 0.660
In the more concentrated solutions a difference is to be expected owing
to our neglect of the heat of dilution in Table XV. This difference amounts
at (1 .5 M to a little over 3%. The corresponding difference between 'Y298
and 'Y' for sodium chloride was about 1%. In fact such data as we pos-
sess show that the heat of dilution is two or three times as great for potas-
sium as for sodium chloride.
\Ve include also approximate values for silver nitrate, since it repre-
sents an important type of uni-univalent salt. The freezing-point measure-
ments seem reliable down to about 0.01 M. The results at higher dilu-
tions hardly warrant a determination of a and f3 by the method of Lewis
and Linhart, but by taking the experimental value at 0.01 M and assum-
ing, as we have done before that a = 1/
2
, we find f3 = 0.343, and 'Y =
0,902 at 0.01 M. The remaining values given below are obtained from
Equation 22, using the data of Roth
1
in the more dilute, and of Raoult
2
in the more concentrated solutions.
TABLE XVI.
Activity Coefficients of Silver Nitrate.
m.. .. . 0 .01 0 .02 0 .05 0 . 1 0 .2 0 ,5 1. 0 2 5
j....... 0.0343 0.055 0.080 0.099 0.122 0.199 0.291 0.419 0.630
,' 0.902 0.857 0.783 0.723 0.655 0.526 0.396 0.280 0.141
The Activity Coefficient in Mixed Electrolytes, and the Solubility of
One Salt in the Presence of Others.
The behavior of electrolytes, when two or more are dissolved in the same
solution, is of both theoretical and practical importance. The quantita-
tive information which we possess chiefly consists in measurements of
the solubility of one salt in the presence of others. Lately a series of
1 Roth, Z. physik. Chem., 79, 599 0(12).
2 Raoult, ibid., 2, 488 (1888).
(28)
measurements of the electromotive force of cells, with mixed electrolytes,
has furnished a still more direct method of studying the free energy and
the activity in mixed salt solutions. The data obtained in these ways
permit us to make some interesting and useful generalizations regarding
activity coefficients in mixtures of salts at moderate concentrations. At
the higher concentrations we must, in this case, as in others, have recourse
to direct experiment in each particular case.
The activity coefficient of an electrolyte in a mixture may be defined
in strict conformity with our previous usage. If m+ is the (stoichio-
metrical) molality of the positive ion, and m- that of the negative, then in
general G+ /m+ = 1'+ is defined as the activity coefficient of the first
ion, and a- /m- = "(- is the activity coefficient of the second ion. If a
given electrolyte gives (stoichiometrically) v+ molecules of the positive
ions and v- molecules of the negative ion, where v+ + v- = v, then the
mean activity coefficient of the ions or, more briefly, the activity coeffi-
cient of the electrolyte, is defined as
(
v+ v_) Ilv a
"( = 1'+ y- = -.
m
I t will be understood that we are speaking of the gross molality of the
ions quite irrespective (in the case of strong electrolytes) of any assumed
degree of dissociation, or intermediate ions. Thus in a solution contain-
ing O. 1 M potassium chloride and O. 1 M barium chloride m+ = O. 1 for
K + and also for Ba++; while for Cl-, m- = 0.3. Thus in this mixture
we find for potassium chloride, m = (0.1 X O. :i))1; while for barium
chloride, m = [0.1 X (0. 3)2]Y:l.
The Activity Coefficient in Mixtures Calculated from Electromotive-
Force Data.-In the study of the conductivity of strong electrolytes
an empirical rule has proved extremely serviceable, namely, that in a
mixture of electrolytes with a common ion, the value of A/A 0 for each
electrolyte depends not upon the amount of that electrolyte but only
upon the total equivalent concentration. When it was realized that
A/A 0 does not measure the thermodynamic degree of dissociation, it be-
came of interest to inquire whether any similar rule would apply to the
activity coefficients.
The first experiments in this direction were made by Lewis and Sargent, I
who measured the potential of a gold electrode in a solution of potassium
chloride, containing potassium ferro- and ferricyanides, in amounts
negligible compared with the amount of potassium chloride present.
Their experiments, in which the concentration of the potassium chloride
varied from 0,5 to O,S M, and in which the ratio of ferro- to ferricyanide
was varied several fold, show that to 0.1 of a millivolt their results could
JLewis and Sargent, THIS ]OVRNAL, 31,33:; (1909); see also later work of Schoch
and Felsing, ibid., 38., 1928 (1916); and Linhart, ibid., 39, 615 (1917),
ACTIVITy COEFFICIENTS OF STRONG ELECTROLYTES. l1:j,)
be interpreted by assuming the degree of dissociation to depend solely
upon the concentration of potassium chloride. Moreover, they obtained
identical values when potassium bromide was substituted for potassium
chloride.
Identical conclusions were reached by Linhart,l who studied the mer-
cury electrode in a solution of perchloric acid containing relatively small
amounts of mercurous perchlorate. He found, under these conditions,
that the thermodynamic degree of dissociation of the mercurous per-
chlorate was independent of the amount of that salt and varied only
with the total concentration of electrolyte.
Similar results have been obtained with hydrochloric acid in the pres-
ence of other chlorides. The work of Loomis and Acree
2
, and the early
work of Harned,3 on cells with mixed electrolytes, involved liquid junc-
tions between different electrolytes. Therefore these cells are not so well
adapted to simple thermodynamic treatment as those we are about to
discuss. The second paper by Harned
4
contained measurements of a
cell with hydrogen and calomel electrodes, and O. 1 M hydrochloric acid,
with the addition of various amounts of potassium chloride to the elec-
trolyte, and more recently cells of this type have been studied by several
authors.
Loomis, Essex and Meacham
5
investigated such a cell in which the con-
centrations of hydrogen chloride and potassium chloride were varied in
such a way that the total molality was kept at 0.1 M. Their pur-
pose was to see whether the electromotive force of the cell could be
calculated upon the assumption that the degree of dissociation, or the
activity coefficient, of the hydrochloric acid is independent of the rela-
tive amounts of the two electrolytes, when the total concentration of the
electrolyte is fixed. They found in fact what appeared to be a slight
departure from this principle in the direction of a greater activity coeffi-
cient in the solutions containing larger percentages of hydrogen chloride.
The difference, however, was hardly outside the limits of experimental
error, and later measurements of a similar cell by Ming Chow
6
do not show
any variation in the activity coefficient with varying percentages of the
two electrolytes. It seems probable that there is a slight variation in
the direction suggested by Loomis, Essex and Meacham, but not more
than 10 or of the amount which might be deduced from their measure-
ments.
In order to test the limits of validity of this principle, we may turn to
1 Linhart, THIS JOURNAL, 38, 2356 (1916).
2 Loomis and Acree, Am. Chem. J., 46, 1332 (1911).
3 Harned, THIS JOURNAL, 37, 2460 (1915).
4 Harned, ibid., 38, 1986 (1916)
, Loomis, Essex and Meacham, ibid., 39, 1133 (1917).
6 Chow, ibid, 42, 488 (1920).
Harned's! latest very comprehensive investigation of this cell in which
the HCI is kept at O. 1 M; in the presence of KCI, NaCI and LiCI,
ranging from zero to several molal.
The electromotive force of his cell is given by the equation
E = EO -- R: In (a+a-) = EO - In aoi:' (29)
where Q+ is the activity of H+, a- is the activity of CI-, and a= is their
geometrical mean. Thus calculating from his measurements the values
of aoi:' and also calculating moi: from his given concentrations, we have
obtained values of 'Y at various concentrations of each of the added salts.
These we have interpolated to round concentrations and the results are
given in Table XVII.
TABLE XVII.
The Activity Coefficient of HCl (0.1 M) in the Presence of Other Chlorides.
Total m.
')'.
Pure HC!.
')'. ')'.
HCI in LiC!. HCI in NaC!.
')'.
HCI in KC!.
0.1 0.81 0.81 0.81 0.81
0.2 0.78 0.78 0.78 0.78
0.5 0.76 0.78 0.76 0.75
1.0 0.82 0.86 0.80 0.75
2.0 1.02 1.09 0.94 0.84
3.0 1.35 1.47 1.17 0.97
4.0 1.84 2.02 1.47 1.17
The first column shows the total molality, MeCI and HCI, the latter
always being 0.1 M; the second column reproduces from Table IV the
activity coefficient of pure aqueous HCI and the remaining columns show
the activity coefficient of HCI in the presence of the three chlorides.
It appears from these results that the behavior in concentrated solu-
tions cannot be quantitatively predicted from any simple rule. The
activity coefficient of the hydrochloric acid depends upon the specific
nature of the added chloride. It seems to be smaller in the presence of
a salt whose own activity coefficient in the pure state is small (d. Table
XIV), but this can hardly be universally true, for we observe that 'Y is
higher in the presence of lithium chloride than in the pure acid, although
the activity coefficient of lithium chloride itself is lower than that of
pure acid.
When we confine our attention to the dilute solutions we see that in
the presence of NaCl up to 0.5 M, and in the presence of the other
chlorides to at least 0.2 M, the principle which we are discussing
is valid. In other words, the activity coefficient is the same in any of
these solutions, in which the total ion concentration is constant, and does
not exceed 0.2 M.
Without doubt identical values of 'Y would have been obtained by
Harned at the same total concentrations if he had used a smaller concen-
I Harned, THIS JOURNAL, 42, 1808 (1920).
ACTIVITy COEFFICIB:\TS OF STRO:\G ELECTROLYTES. 1137
tration of hydrochloric acid throughout. In so far, therefore, as these
various measurements go, they support the general rule, for which, more-
over, we shall find abundant evidence from another source: in any dilute
solution of a mixture of strong electrolytes, of the same valence type, the ac-
tivity coefficient of each electrolyte iepends solely upon the total concentra-
tion.
It is unfortunate that the measurements which we have here discussed
have not been extended to include cases where the added salt is of a differ-
ent valence type like BaCb or MgS0
4
or La2(804)3. We shall show how
this deficiency is made up by a different kind of experiment, which will
permit a wider generalization. From that generalization it is possible
to predict in advance the electromotive force of such cells. We shall
now turn to the evidence which is furnished by the solubility of salts in
the presence of one another.
The Activity Coefficient in Mixed Electrolytes from Solubility
Measurements.I-When, at a given temperature, a solid salt is in
equilibrium with a solution, the activity of that salt in the solution is fixed.
It cannot be changed by any change in the nature of the solution, such as
would be produced by the solution of other electrolytes. Thus for the
salt whose solid phase is present, the value of a2 and also of a is con-
stant. Since by Equation 28 the activity coefficient is defined as 'Y = a/m,
we see that whatever happens (isothermally) to the solution, the activity
coefficient of the salt in question must remain inversely proportional to
the mean molality of its ions.
This principle will be of the greatest service in determining the activity
coefficient in mixtures, and in some cases we shall see that it furnishes a
very accurate means of obtaining the activity coefficients of pure salts
as well.
In order first to illustrate the method in the simplest case, let us con-
sider the solubility of a uni-univa1ent salt in the presence of another uni-
univalent salt with no common ions. We shall consider the solubility
, The experimental development of this subject, especially in the field of dilute
solution, is due in large measure to the work of A. A. Noyes and his collaborators.
beginning with an early paper by Noyes (Z. physik. Chem., 6, 241 (1890)), and cul-
minating in an exhanstive experimental and critical study by Noyes, Bray, Harkins,
and their assistants, appearing mainly in THIS JOURNAL, 33 (1911). Very recently,
Briinsted (ibid., 42, 761 (1920)) has published a brilliant paper on the interpretation of
solubility measurements, in which he mentions extensive investigations of the influence
of one salt upon the solubility of another, the greater part of which has not yet been
published. His theoretical treatment of this subject is of the utmost interest, and he
hab shown for the first time how cases with and without a common ion may be treated
by identical thermodynamic methods. We are unable here to follow his methods in
deetul, since he bases his work upon an empirical rule, which is inconsistent with the
facts that we have deduced. (It is equivalent to making a = ';' in Equation 15 for
all types of electrolytes.)
1]38 GILBER'(K. LEWIS AKD MERLE RANDALL.
KNO,. KCl. HC!. TINO,. BaCh. Tl,SO,. K,SO,.
0.01607 0.01607 0.01607 0.01607 0.01607 0.01607 0.01607
0.01716 0.01034 0.01779
0.00869 0.00866 0.00880 0.00898
0.01826 0.00590 0.00583 0.00624 000618 0.00677 o 01942
0.01961 0.00396 0.00383 0.00422 0.00416 0.00468 0.02137
000268 000253 0.00282
0.02313 0.02600
0.03072 0.03416
of thallous chloride in the presence of potassium nitrate, as given in the
second column of Table XVIII, which we take directly from the work of
Bray and Winninghoff.
1
TABLE XVIII.
Solubility of Thallous Chloride at 25
0
in the Presence of Other Salts.
Added salts,
equivalents
per liter, a
o
0.020
0.025
0.050
0.100
0.200
0.300
1.000
a It ill be noted that the unit in this table is the equivalent and not the mol.
Moreover, in this and in some subsequent tables the values given are expressed in
amount per liter and notin amount per 1000 g. of water. 'We have not taken the trouble
to recalculate since we are interested primarily in the dilute solutions, where the two
methods of expressing the composition are essentially identical.
In the simple case where we have no common ion, m is equal to the
solubility, and the activity coefficient is proportional to l/m. If then
we plot the reciprocal of the solubility against the total molality of the
solution, or, better, against the square root of the total molality, as we
have done in Fig. 4, we may find the proportionality factor by a simple
extrapolation. The points for potassium nitrate are represented in the
figure by circles which lie on a curve which is rapidly approaching :1
straight line. This tangent line, which we have obtained not alone from
the measurements upon potassium nitrate, but from the measurements
in the other salts, cuts the axis of zero concentration
2
at l/m = 70.:3.
If, therefore, we divide any of our values of l/m by this number, we ob-
tain at once the activity coefficient of thallous chloride in the correspond-
ing solution. This method of calculation and extrapolation gives at once
absolute values of the activity coefficient. The results rival in accuracy
those of the best instances which we have chosen for the corresponding
calculation from electromotive force and freezing-point data.
Before giving the figures so obtained we may illustrate the treatment
of the case where a common ion is present. Let us consider the solubility
of thallous chloride in the presence of potassium chloride. We find,
for example, that when m = 0.050 for KCI, m = 0.00590 for TICl. Thus
tn+ = 0.00590, tn- = 0.050 + 0.00590 = 0.05590, and tn = (0.00590 'X
0.05590) 72 = 0.0181. Operating similarly at the other concentrations,
1 Bray and Winninghoff, THIS JOURNAL, 33, 1663 (1911).
2 This method of extrapolating solubility data to obtain ion activities is equivalent
to the one used by Lewis in his paper, "The Activity of the Ions," THIS JOURNAL,
34, 1631 (1912).
AC'fIVITY COEFFICIENTS OF STRO!\'G ELECTROLYTES. 1] 30
we obtain the points indicated by squares in the figure. These points
approach more rapidly than those of the previous case to the same straight
line, which cuts the axis of zero concentration at 70.3. Our example thus
shows that in the mixture of total molality, 0.0559, the activity coefficient
of TICI is 1/(0.0181 X 70.3) = 0.784.
75.------,------,------,....-----,.----..,....-----,
------..._.
o TICI
o KN0
3
C KCI
x HCI
o TIN0
3
() Ba CI
2
Q TI
2
S0
4
o K
2
S0
4
70
I
U
"C
::
+<'l 65
'0
>,
.-::
3 60
o
S
::
S
<lJ
'J 55
'0
o
...
0-
50
...

0.6 0.5 0.1
40 '--__-J'--__--l -l. --l.. -.L>--..::..-_.....J
0.0 0.2 0.3 0.4
IJ.
I/
,. square root of the ionic strength.
Fig. 4.
Proceeding similarly with the data for the solubility in hydrochloric
acid and in thallous nitrate. we find other curves
l
which differ slightly
from those for potassium chloride and nitrate at higher concentrations,
but up to several hundredths molal the curves are all identical and lead
to the same extrapolated values of l/m, namely, 70.3. By reading
l The curves are drawn on so large a scale as to greatly exaggerate their differences.
The maximum difference at 0.05 M is only 1%.
1140 GILBERt N. LEWIS AND MERLE RANDALL.
from the various curves at round total molalities the values of 11m"" and
dividing by 70.3, we obtain the activity coefficient of thallous chloride,
in the various mixtures, as given in Table XIX.
TABLS XIX.
Activity Coefficient of Thallous Chloride at 25 o.
Total m. In KNO.. In KCl. In HC!. TINO.
0.001 0.970 0.970 0.970 0.970
0.002 0.962 0.962 0.962 0.962
0.005 0.950 0.950 0.950 0.950
0.01 0.909 0.909 0.909 0.909
0.02 0.872 0.871 0.871 0.869
0.05 0.809 0.797 0.798 0.784
0.1 0.742 0.715 0.718 0.686
0.2 0.676 0.613 0.630 0.546
Up to a concentration of about 0.02 M the rule stated in the preceding
section is completely corroborated, and we need not hesitate to assume
that in the dilute solutions the values of 'Y given in the table represent
also the activity coefficients of pure thallous chloride. Thus we may pre-
dict that if we could obtain a supersaturated solution of this salt at 0.02 M
its activity coefficient would be found to be 0.871. At values below
saturation (0.016 M) the values given in the table could be readily checked
either by the electromotive force or by the freezing-point method.
By comparing Table XIX with Tables V and XIV, we see that the ac-
tivity coefficient of thallous chloride is distinctly below that of sodium or
potassium chloride, which again corroborates a rule which we have men-
tioned before, and which was first observed in a study of the corrected
degree of dissociation, obtained by a combination of conductivity and
transference data.
Mixtures Containing Two or More Valence Types.-The remaining
cases in Table XVIII involve mixtures of thallous chloride with uni-
bivalent salts. Hence we are now forced to consider activity coefficients.
in mixtures of different valence types. We might guess the activity
coefficient of a given electrolyte to depend simply upon the total molal
concentration of electrolytes, or to depend upon the total equivalent con-
centration (which would be assuming that one molecule of a bivalent ion
has the effect of two molecules of a univalent ion). Neither of these two
guesses is correct, although the latter comes nearer the truth than the
former.
In attempting to solve this problem we have brought to light what ap-
pears to be a remarkably simple and precise generalization, which we shall
find to be in exact agreement with all of the experimental results which
are at present available. Before stating this new principle, we shall
introduce a new term, the ionic strength.
In any solution of strong electrolytes let us multiply the stoichiometrical
molality of each ion by the square of its valence (or charge). The sum
of these quantities, divided by two (since we have included both positive
and negative ions), we shall call the ionic strength, and designate by }.I.
Thus in pure solutions of potassium chloride, magnesium sulfate and
barium chloride, all at 0.01 M, we have, respectively, }.I = 0.01, }.I =
0.04, and }.I = [(4 X 0.01) + 0.02J/2 = 0.03. We may now state
our general principle: In dilute solutions the activity coefficient of a given
strong electrolyte is the same in all solutions of the same ionic strength.
In a solution which is 0.01 M in K
2
S0
4
the solubility of TlCl is
0.01779. We take half the sum of 0.01779 for 1'1+, 0.01779 for Cl-,
0.02000 for K+, and 4 X 0.01000 for S04--' hence }.I = 0.04779. Ob-
taining by this procedure the other values for K
2
S0
4
, TbS04 and BaC1
2
,
given in Table XVIII, we plot once more in Fig. 4, 11m: against the
square root of the ionic strength, }.I. We see that all the series give
curves which coincide, in dilute solutions, with one another and with the
curves obtained with the uni-univalent salts. As a further check it would
be desirable to have measurements of the solubility of thallous chloride in
salts like magnesium sulfate and lanthanum sulfate.
The Solubility of Uni-bivalent Salts.-We may next test the new prin-
ciple by a study of the influence of other salts upon the solubility of a
uni-bivalent salt. Here we shall utilize the data of Harkins and Win-
ninghoff
1
on the solubility of barium iodate in the presence of barium
nitrate and of potassium nitrate (represented, respectively, in Fig. 5
by circles and squares). Their results, together with the calculated
values of 11m: and of }.I, the ionic strength, are shown in Table XX.
1000
500
I---
---
~ ~
f----
, :e-
I---
--
r--- ~
--
-
Cl
~
f--
o
0.0 0.1 0.2 0.3
}.Io'h, cube root of the ionic strength
0.4 0.5
Fig. 5.
These values of 11m: are plotted against the cube root of the ionic
strength in Fig. 5. Our rule for mixtures evidently holds with great
1 Harkins and Winninghoff, THIS JOURNAL, 33, 1827 (1911).
0.00444
0.01274
0.0528
0.2008
776
690
477
39.';
0.000812
0.000913
0.001320
0.001595
0.002
0.01
0.05
0.2
precision. For practically all of the range shown in the figure the two
series fall on the same curve, which moreover proves to be a straight line.
TABLE XX.
Solubility of Barium Iodate in Barium Nitrate and in Potassium Nitrate.
m Ba(NO,),. m Ba(IO,),. I!m:. 1'. m KNO. m Ba(IO.),. I!m: 1'.
o 0.000790 797 0.00237
0.0005 0.000681 770 0.00354
0001 0.000606 751 0.00482
0.0025 0.000488 706 0.00896
0.01 0000337 597 0.0310
0.025 0.000307 472 0.0759
0.05 0.000283 396 0.1508
0.1 0.000279 317 0.3009
Throughout the range of concentration shown in Fig. 5 we may cer-
tainly assume that the value of 'Y is the same at a given ionic strength
as it would be for pure barium iodate, and can therefore be found from
the figure, if we divide the ord;nate at any concentration by the limiting
ordinate. Owing to the accuracy of the results, and especially to the in-
solubility of barium iodate, which renders necessary only a small extrapola-
tion, we may obtain the activity coefficients of barium iodate with much
certainty. In Table XXI values of 'Y so obtained are given at round
values of the molality, and compared with those of barium chloride (Tables
V and XIV). In accordance with the observation that we have previously
made, it is seen that in this case also the activity coefficient of a salt of
an oxygen acid is less than that of the corresponding halide.
TABLE XXI.
Activity Coefficients of Barium Iodate, at 25 o.
m. Ba(IO,)z. BaCk
0.001 0.834 0.865
0.002 0.790 0.830
0.005 0.714 0.771
0.Q1 0.639 0.716
002 0.549 0.655
Bi-bivalent Salts.-In the study of salts with two bivalent ions we have
little to add to our previous theoretical treatment, but we have material
for a still more severe test of our rule that the activity coefficient of any
electrolyte varies only with the total ionic strength of the solution.
Let us use for study the data of Harkins and Paine! on the solubility
of CaS04 in the presence of MgS0
4
, CUS04 and KN0
3
The solubilities
are given in the second column of Table XXII, the first column showing
the molality of the other salt which is added. The third column gives
the reciprocal of the mean molality, and it will be noted that m: in the
case of added potassium nitrate is simply the solubility of CaS04, while in
the case of the other two salts it is the geometrical mean of the total
calcium and total sulfate. The fourth column gives the ionic strength.
1 Harkins and Paine, THIS JOURNAL, 41, 1155 (1919).
ACTIVITY COEFFICIENTS OF STRONG ELECTROLYTES. 1U3
TABl.S XXII.
Solubility of Calcium Sulfate in the Presence of Other Salts at 25.
m.
(l!m",)
added m.
salt. CaSO. I!m",. 1'. -r(MgSO.). -r(MgSO.).
None 0.01535 65.15 0.0614 0.358 182
MgSO.
0.00502 0.01441 59.77 00777 0.327 183
0.010]2 0.01362 55.59 009.50 0.305 183
001528 0.01310 51.92 0.1135 0.286 181
CuSO.
001254 0.01360 .53.08 0.1046 0.298 179
00.5010 0.01239 35.97 0.250 0.200 180
0.1010 0.01242 26.67 0.454 0.162 166
0.2120 0.01329 12.92 0901 0.113 114
0.9771 0.01654 7.81 3.974
KN0
3
0.02766 0.01812 55.19 0.1001 0.300 1R4
0.0.5293 0.02019 49.53 0.1237 0.277 179
0.1038 0.02130 46.95 0.1890 0.231 204
\Ve already have activity coefficients for a substance very similar to
calcium sulfate, namely, magnesium sulfate. The best way, therefore, to
determine the proportionality factor which converts the values of 11m",
into values of "'I, would be to divide each value of 11m", by "'I for magnesium
sulfate at the same ionic strength, and then to find the limiting value of
this ratio at zero concentration (zero ionic strength). This is essentially
the method which we have illustrated before in determining the activity
coefficients of potassium chloride. We have, in Tables V and XIV, "'I
for MgS0
4
at several molalities, and therefore at the several values
of the ionic strength (which is four times the molality). By in-
terpolating we have thus obtained (Col. 3) the values of "'I (magnesium
sulfate), corresponding to the values of j.J. in our table, and the ratios
which we have just mentioned are given in the last column.
The results are very striking. Evidently, within the limits of experi-
mental error, the activity coefficient of calcium sulfate is identical with
that of magnesium sulfate up to an ionic strength of over 0.1. Taking
the limiting value of the ratio as 182, we find for calcium sulfate, in molal
copper sulfate, (j.J. = 4), 'Y = 7.81/182 = 0.043.
It may be that these results do not show superficially what an extra-
ordinary confirmation of our rule is furnished by these figures. Supposing
that we had measurements of the solubility of calcium sulfate in the pres-
ence of magnesium or copper sulfate, and were obliged to estimate from
these values its solubility in the presence of 0.05 M potassium nitrate.
Reversing our previous procedure (and using the method of approxima-
tions) we should calculate a value within 1% of that actually found, or
within the limits of expeIImental error. On the other hand, if we made a
similar calculation, assuming that it is not the ionic strength, but the
1144 GILBERt' N. LEWIS AND MERLE RANDALL.
equivalent concentration, which determines the activity coefficient, we
should make an error of over 20%. Or if we should assume neither of
these, but rather the molal concentration to be the governing factor, our
error would prove to be over 50%.
Uni-trivalent Salts.-It is evident that our rule for activity coefficients
in mixtures will be given a very severe test in systems involving trivalent
ions, for here, according to our rule, a certain concentration of trivalent
ion is nine times as effective as a univalent ion at the same molal concen-
tration. It is regrettable that no one has studied the influence of salts
containing polyvalent ions upon the activity coefficients of uni-univalent
electrolytes, although this could readily be done, either by adding a
lanthanum salt to the hydrochloric acid in the cell with hydrogen and
calomel electrodes, or by measuring the solubility of thallous chloride in
the presence of such a salt. On the other hand, Harkins and Pearce
l
have
measured the solubility of lanthanum iodate in the presence of lanthanum
nitrate, lanthanum ammonium nitrate, and sodium nitrate. Unfortu-
nately an error seems to have slipped into the first two of these series,
the source of which we are unable to discover, but the series with sodium
nitrate furnishes an excellent opportunity of testing our rule of mix-
tures.
'tABLE XXIII.
Solubility and Activity Coefficient of Lanthanum Iodate in Solutions of Sodium
Nitrate at 25.
(I!m",,)
NaNO,. La(IO,),.
I!m"".
p.. 'Y-La(NO,). 'Y-La(NO.); 'Y-La(IO.)o.
0 0.00103 426 0.00618 0.809 527 0.806
0.001 0.001043 421 0.00726 0.796 529 0.796
0.002 0.001056 415 0.00834 0.784 529 0.784
0.01 0.001150 382 0.0169 0.720 529 0.720
0.025 0.001309 335 0.0329 0.645 520 0.633
0.050 0.001492 294 0.0589 0.570 517 0.556
0.1 0.001748 251 0.1105 0.505 497 0.475
Once more, in Table XXIII, together with the molality of added salt
and the solubility of lanthanum iodate, we give the values of 11m"" cal-
culated therefrom. Lanthanum iodate and nitrate should closely re-
semble each other. We have already, in Tables V and XIV, obtained
values of 'Y for the nitrate, and from these we have interpolated the figures
corresponding to the several values of }J, (where for pure lanthanum
nitrate, }J, = 6m). Dividing 11m"" by 'Y of lanthanum nitrate, we obtain
the figures in the next to the last column, which afford once more a re-
markable confirmation of our rule, and show that for values of }J, up to
about 0.02, the activity coefficient of lanthanum iodate is the same as
that given by Lewis and Linhart for lanthanum nitrate, the average
1 Harkins and Pearce, THIS JOURNAL, 38, 2679 (1916).
deviation being hardly a O. 1%.1 In order to show the deviation at
higher concentration we have given in the last column of Table XXIII
the activity coefficient of lanthanum iodate in the mixture, obtained by
dividing each value of 1/m by 259.
The Activity Coefficients of Individual Ions.
In developing our equations we have made use of the activity coefficient
of the separate ions, and we have shown that, for a salt like potassium
chloride, the activity coefficient is the geometrical mean of the activity
coefficients, 'Y+ of potassium ion, and 'Y- of chloride ion. It remains for
us to consider whether these separate values can be experimentally de-
termined. This is a problem of much difficulty, and indeed we are far
from any complete solution at the present time.
At infinite dilution the activity coefficient is the same for all ions, and
equal to unity. As the concentration increases, we might expect that the
activity coefficients of two ions of similar type would remain approximately
the same, up to moderately concentrated solutions. We have seen in
the preceding section that in dilute solutions the mean activity coefficient
of an electrolyte is independent of the particular ions present in the solu-
tion, and this would hardly be true if it were not true also of the activity
coefficients of the individual ions.
MacInnes
2
has assumed that in two univalent salts with a common ion,
and at the same concentration, the common ion has the same activity
coefficient in both. This assumption, which should be confined to dilute
solutions, will be seen to be a corollary of a much wider generalization,
which the consideration of the preceding sections now permits us to make,
a generalization which will give us a very novel idea of the activities of
the separate ions in salts of a mixed valence type.
Hypothesis of the Independent Activity Coefficients of the 10ns.-In
studying the mean activity coefficients of the ions of numerous electro-
lytes it has been seen how advantageous it is to employ the quantity
which we have called the ionic strength, and the conclusion was reached
that, except in rather concentrated solutions, the activity coefficient of a
certain electrolyte is independent of the particular character of any
other strong electrolytes which may be present, but depends solely upon
the total ionic strength. Now, unless some peculiar compensation oper-
ates, this could hardly be true in general if it were not true of the individual
1 From considerations which we are about to make, we should expect the values
for the iodate to be slightly smaller than those for the nitrate, even at 0.01 iII. The
ahove (lata hardly suffice to warrant any exact conclusion in this regard.
:! :\'laeInnes, THIS ]OCRKAL, 41,1080 (lOla).
1146 GILBERT :\C. LEWIS AND MERLE RANDALL.
activity coefficients of the several ions, and we are thus led to the follow-
ing simple hypothesis: Tn dilute solution
l
the activity coefficient oj any
ion depends solely upon the total ionic strength oj the solution.
Accepting this hypothesis, it is possible to calculate the activity coef-
ficient of a salt when the activity coefficients of other salts are known.
For example, let us consider KCI, KIO
a
, BaCh and Ba(IO
s
)2' each at an
ionic strength of 0.01; in other words, we will consider the first pair at
0.01 M and the second pair at 0.0033 M. From Table V (interpolating in
the case of barium chloride), we find I'KCI = 0.922, I'KlO, = 0.882, and
I'BaCI, = 0.800. Let us calculate from these the value of I' for barium
. d - Woo 2 3 2 d d
10 ate. ntmg I'KCl = I'K+l'ct-; I'BaCI, = I'Ba ++I'ct-, an so on, an
taking the activity coefficient for each ion as the same in whichever salt
it appears, we obtain by simple algebra
(
I'KCI )4 (I'BaCh)3 (30)
'YKI03 = 'YBa(I03)2
and substituting the above values we find I'Ba(103h 0.754, while by in-
terpolating the experimental values of Table XXI we find 0.746. The
agreement is well within the limits of error.
. Unfortunately, the number of such chccks which are at present avail-
able is small, and our principle must therefore be regarded rather as a
prediction than as a summing up of accurate data now existing. We feel
confident, however, that this hypothesis will be verified by further exact
experimentation, and that even now we are safe to employ it, sometimes
in preference to experimental data, if these are of a low grade of accuracy.
Numerical Values of Ion-Activity Coefficients.-It is evident that if
we ascertain, or if we arbitrarily assume, the individual activity coeffi-
cient of some one ion, at a given value of the ionic strength, we can then
proceed to determine the values for other ions. Thus, at a given value of
f.L, if we had given the activity coefficient of sodium ion, we could combine
this with the known activity coefficient of sodium chloride to obtain the
value of chloride ion. Using then the known value for barium chloride,
we could obtain that of barium ion, and so on. MacInnes has suggested
that at each concentration the two ions of potassium chloride, "which
have nearly the same weight and mobility," be considered to have equa
I 'When we use the rather \'ague phrase "dilute solution" in a case like this we mean
that the principle as stated approaches complete validity as the dilution is indefinitely
increased. It becomes then a matter of experiment to determine to what concentra-
tions such a principle may be rell;arded as valid within certain limits of permissible
error, say 1%. With such interpretation it is our belief that this hypothesis is correct
over the same range as our previous rule of mixtures, namely, up to an ionic strength
of a few hundredths to a few tenths according to the nature of the ions. The degree
of departure in concentrated solutions doubtless depends upon numerous factors such
as the amount of hydration of the ion.
ACTIVITY COEFFICIENTS OF STRONG ELECTROLYTES, 1147
activity coefficients. This convention which, at least in the concentrated
solutions, must be regarded as purely arbitrary, we may adopt as well
as any other for the sake of obtaining a table of individual activity coeffi-
cients for the ions.
Thus when p. = 0.01 we take, both for K+ and Cl-, l' = 0,922 from
Table V. From the same table I'KIO. = 0,882, whence for 10
3
-, l' =
(0.882)2/0,922 = 0,845. For BaCb, when p. = 0.01 or m = 0.0033,
I'is 0.800. Whence, for Ba++, l' = (0.800)3/(0,922)2 = 0.602. Hence
the activities in the solution at this concentration are, for Ba++, a+ =
0.00333 X 0.602 = 0.00201, and for Cl-, a- = 0.00333 X 2 X 0.922 =
0.00614. Obviously this is a very radical departure from the idea that
the concentration, and hence approximately the activity, is half as great
for the barium as for the chloride ion.
By the method that we have just sketched, and by means of the pre-
vious tables, the values of Table XXIV have been obtained. These
values may be interpolated at intermediate concentrations, and thus we
find for MgSO. at 0.01 M, where p. = 0.04, 1'+ = 0.44, and 1'- =
0.38, whence l' = 0.41. This is to be compared with the value 0.40,
which we obtained from experimental measurements. The agreement
is better than we should expect, considering the uncertainty of some of
the experimental data.
TABU XXIV,
Activity Coefficients of Individual Ions at Various Values of the Ionic Strength.
p. ,= .. ' 0.001 0.002 0.005 0.01 0.02 0.05 0.1
E[+ .................. 0.98 0.97 0.95 0.92 0.90 0.88 0.84
OH- 0.98 0.97 0.95 0.92 0.89 0.85 0.81
CI", Br-, 1- 0.98 0.97 0.95 0.92 0.89 0.84 0.79
Li+ 0.98 0.97 0.95 0.92 0.89 0.85 0.81
0.98 0.97 0.95 0.92 0.89 0.84 0.80
K+, Rb+. Cs+ 0.98 0.97 0.95 0.92 0.89 0.84 0.79
Ag +. .. .. .. .. .. .. . .. 0.97 0 .16 0 .93 0 .90 0 .85 0 .80 0 .77
TI' 0.97 0.96 0.93 0.90 0.85 0.75 0.64
0.97 096 0.94 0.91 0.87 0.77 0.68
CIO.-, BrO.-. 10.- 0.95 0.93 0.89 0.85 0.79 0.70 0.61
1!e++
G
............. 0.78 0.74 0.66 0.60 0.53 0.43 0.34
SO.- - 0.7 0.'i1 0.63 056 0.47 0.35 0.26
.. 0.73 0.66 0.55 047 0.37 0.28 0.21
Under Me+ + we include Mg+ +, Ca + +, Sr+ +. Ba + +, Cu + +. Zn + +, Cd+ +,
except that for Cd + ; the value is not to be used for obtaining the activity coefficient
of a halide.
This table is prepared from fragmentary data and will need revision
when more data are available. 'Ve may predict, however, that when it
is so revised it will permit the accurate calculation of the activity coeffi-
cient of any strong electrolyte, except at one or two of the higher concen-
trations.
1148 GILBSRT N. LSWIS AND MSRLS RANDALL.
The Possible Determination of Ion-Activity Coefficients without any
Arbitrary Assumption.-Although the development as far as we have
carried it suffices for all ordinary thermodynamic calculations, it would be
of much theoretical interest if we could determine the actual activity
of an ion in a solution of any concentration. This indeed might be ac-
complished if we had any general method of calculating the potential
at a liquid junction. Such an attempt to estimate individual ion activi-
ties even in very concentrated solutions has been made by Hamed,1 but
he was obliged to make certain assumptions regarding the elimination of
liquid potentials which may be very far from valid.
Supposing that we consider the cell, H
2
Idil. HCII conc. HCII H2, the
electrotnotive force may be expressed by the equation
RT a+
E = -F In -, + E
L
(31)
a +
where the activity of hydrogen ion in the dilute solution is a+ and in
the concentrated a'+, while E
L
is the potential at the liquid junction.
Now, if a+ were very small, so that it could be taken as equal to the
molality, we could determine directly a'+ if we knew E
L
. This, unfor-
tunately, we do not know. The Nemst equation for the liquid potential
between the two concentrations of an electrolyte might be improved by
replacing concentration by activity, but even so the equation could
at best be valid only in very dilute solutions. Efforts to eliminate
such a liquid potential by interposing, between the two solutions, some
concentrated salt solution, like potassium chloride, have doubtless served
in many cases to reduce such a potential to a few millivolts. But by the
method so far proposed the elimination almost certainly has never been
complete, and the uncertainties increase with the concentration of the
solutions, between which the liquid potential is to be estimated.
In the case of a liquid junction between two different solutions of the
same concentration, conditions are quite similar.
2
The equation of Planck,S
as modified by Lewis and Sargent,
4
appears to give results in solutions of
moderate concentration which are approximately correct. Insofar as these
results go they are entirely consistent with our rule for the activity coef-
cient of the ions. Thus let us consider the cell Hg I HgClj HCl(O.1 M)I
KCI(O.1 M) [HgCl I Hg. Assuming that the activity of chloride ion
is the same in both solutions, the total e. m. f. of the cell must be equal
1 Harned, THIS JOURNAI., 42, 1808 (1920).
! On the other hand, the potential between two concentrations of the same elec-
trolyte is absolutely constant and reproducible, while that between two different elec-
trolytes depends upon the method of making the junction. and upon the time. See
Lewis, Brighton and Sebastian, THIS JOURNAI., 39, 2245 (1917).
I Planck, Ann. physik., 40, 561 (1890).
4 Lewis and Sargent. THIS JOURNAL, 31,363 (1909).
ACTIVITY CO:UFFICIaNTS OF STRONG :UL:UCTROLYTlS. 1149
to the potential between the liquids. The measured value at 18
0
is 0.0286
V,., while that calculated by the formula of Lewis and Sargent is also
0.0286. (The unmodified Planck equation gives 0.0271.) In more con-
centrated solutions, however, it is certain that such equations would
completely lose their validity. At the present time we must conclude
that the determination of the absolute activity of the ions is an interesting
problem, but one which is yet unsolved.
The Concentration of the Ions and the "True" Degree of Dissociation.
It will certainly be evident from the preceding that for all thermo-
dynamic calculations it is the activity of the ions and not their more or
less hypothetical concentration which is of value. Nevertheless, if the con-
centration of the ions, and the concentration of undissociated electrolyte,
are terms which have any exact significance, the determination of these
concentrations would be of much theoretical interest. Since our thermo-
dynamic methods do not give us these "true" concentrations we may con-
sider for a moment other methods which may have a bearing.
There is a group of properties which .give some evidence, though of a
somewhat conflicting character, upon this problem. The various partial
molal quantities which pertain to an electrolytic solute prove to be com-
pletely additive at infinite dilution; that is to say, they can be represented
as the sum of the two numbers, one characteristic of the cation, and the
other of the anion. How far this additive relation persists into the realm
of finite concentration, is an experimental problem which is far from
solved. Perhaps the best evidence that we have is that which concerns
the partial molal heat content.
The old law of the thermo-neutrality of salt solutions, which was a
powerful support of Arrhenius' theory of electrolytic dissociation, might
perhaps with equal force have been used against his assumption of
partial dissociation. If on mixing very dilute solutions of two salts,
let us say potassium chloride and sodium nitrate, we find no heat
evolved or absorbed, this is in accord with the additivity of heat contents
and is what would be expected from the theory of complete dissociation.
But if we obtain the same result at higher concentrations we might, by
this method of reasoning, be led to assume complete dissociation where
other criteria may lead us to assume a dissociation of only 50% or less.
The fact is that from the great mass of data obtained by Thomsen we may
state that, within the limits of error of his measurements, the partial molal
heat contents of all the strong electrolytes, up to a concentration of 0 . 5 M
or greater, seem to obey the additivity principle.!
1 Thus, for example, in the neutralization of any strong acid by a strong base. the
reaction may be regarded merely as H+ + OH- = H
2
0. and the heat of neutraliza-
tion is constant within the limits of his error. This is true of acids like Hel, HBr and
HNOI, and of bases like KOH. LiOH and Ba(OH)2 (whose dissociation at 0.5 M is
ordinarily assumed to be not over 70%). It is not true of acetic or phosphoric acid,
1150 CILI:l;R'l' LWIS AND 1r12RL RANDALL.
There are other properties which show the same kind of additivity.
Thus to quote A. A. Noyes,! "The optical activity and the color of salts
in solution * * * are additive with respect to the properties of
the constituent ions, even up to concentrations where a large propor-
tion of the salt is in the un-ionized state."
As Lewis
2
has remarked, "If we had no other criterion for the degree of
dissociation these facts would undoubtedly lead us to regard salts, up to
a concentration of normal or half-normal, as completely dissociated."
Indeed this conclusion had been drawn by Sutherland, I and more recently
this theory of complete dissociation has been adopted by Bjerrum, 4 Milner,
Ghosh
6
and numerous other authors who explain the properties of strong
electrolytes by the mutual influence of ions already separated to a con-
siderable distance. The diminution in equivalent conductivity and in
the thermodynamic degree of dissociation with increasing concentration
are both ascribed to the nearer approach of the charged particles rather
than to chemical union of the ions. Whether there is any essential differ-
ence between these two views is 8: question to which we shall revert pres
ently.
The Interpretation of Conductivity.-There remains only one method
of studying the concentration of the ions as distinct from their activities,
and this is the method of conductivity.
7
Unless our present theories of
nor even for sulfuric acid. which. with respect to its second dissociation, shows many
of the characteristics of a weak electrolyte.
Hamed (THIS JOURNAL, 42, 1808 (1920 has made an exact study of the partial
molal heat content of hydrochloric acid in 0.1 M solution when other chlorides are added.
He finds that this heat content changes by only 150 calories in the presence of 1.0 M
KCI. but by about 5000 cal. in the presence of 1.0 M LiCl.
I A. A. Noyes. address before International Congress of Arts and Sciencd, St.
Louis. Technology Quarterly, 17,293 (1904).
2 Lewis, Z. Physik. Chern. 70, 212 (1909).
a Sutherland, Phil. Mag., 3
1
161 (1902); 7, 1 (1906).
4 Bjerrum. Proc. VII Intern. Congo Appl. Chern. (London), 1909; Z. Elektrochern.
24,321 (1918); Medd. K. Vetenskapsakad. Nobelinst., 5, No. 16 (19Hl).
6 Milner. Phil. Mag. 25, 742 (1913).
6 Ghosh, J. Chem. Soc . 113, 149.627, 790 (1918). Since we are dealing primarily
with the thermodynamics of electrolytic solutions, we cannot discuss in full the methods
by which Ghosh is led from mechanical considerations to a remarkably simple equation
for electrolytic conductivity in aqueous and non-aqueous solutions. It is, howeTer.
our opinion that his work marks a new departure which will in time. and after lome
modification. lead to a satisfactory kinetic picture of the trend of conductivity and of
the activity coefficient. Readers of his papers will notice that the numerical factor
which occurs in his equation for the work of completely separating ions which are at
a given distance from one another, is identical with the factor by w1:ich we have multi
plied the molality of an electrolyte to obtain its ionic strength.
7 One other method has occasionally been attempted. namely, (0 t:ctennine the
concentration of ions by the speed of some reaction. In addition to the experimental
difficulties involved here, the method is probably unsound in principle. For any re-
ACTIVITY coEFFICIBNTS OF STRONG ELECTROLYTES. 11tH
the mechanism of electrical conduction are quite erroneous, there would
seem to be no objection to regarding the conductivity due to each ion as
the product of the mobility and the concentration of that ion. If then
we should adopt the assumption of Kohlrausch that the ion mobility is
independent of the concentration of the electrolyte, it would be neces-
sary to regard the conductivity as a direct measure of the ion concentra-
tion.
]ahn
1
was the first to express serious doubt as to the validity of the
Kohlrausch assumption of constant ion mobility. His conclusions were
somewhat discredited, because they were largely based upon the tacit
assumption that the quantities obtained by thermodynamic methods,
and which we now call the activities, must be equal to the ion concentra-
tions. Nevertheless, we are now certain that his contentions were in
the main correct.
After a careful scrutiny of existing data on conductivity and transfer-
ence numbers, Lewis
2
was able to show that certainly in many cases, and
presumably in all cases, the ion mobilities change with the concentra-
tion of electrolyte, and indeed by an amount which is some function of
the mobility of these ions at infinite dilution. He showed further that
by making an assumption, which is far more plausible than the one of
Kohlrausch, it is possible to determine, not an absolute degree of dissocia-
tion, but the ratio between the degrees of dissociation of any two salts.
We have had occasion to remark the striking parallelism between these
"corrected" degrees of dissociation, and the activity coefficients obtained
by thermodynamic means.
Since, however, these corrected degrees of dissociation are not abso-
lute but only relative, they permit us to assume either that many salts
are wholly dissociated, or are dissociated to an extent even less than that
given by the value of AlA0, according as we assume that. the mobility
of the ions diminishes, or increases, with increasing concentration.
What do we Mean by Degree of Dissociation?-Before discussing
further the significance of the conductivity values, it may be of interest
to view for a moment the logical implications of such a term as "degree of
dissociation."
Let us consider the equilibrium in the vapor phase, between diatomic
and monatomic iodine, and at such a temperature that on the average
each molecule of 1
2
, after it has been formed by combination of two
atoms, remains in the diatomic condition one minute before it redisso-
c i t t ~ s During this minute such a molecule will traverse several miles
action which is near equilibrium conditions, at least. it i ~ thermodynamically necessary
that the speed depend upon the activity, and not upon the concentration of the reacting
substances.
II Jahn, Z. physik. Chem., 33, 545 (1900).
:1 Lewis, THIS JOURNAl" 34, 1631 (1912).
1152 GILBSR'l' N. LBWIS AND MBRI.B RANDALL.
in a zigzag path, and after its dissociation each of its constituents will
traverse a similar path before it once more combines with another atom.
If we imagine an instantaneous photograph of such a gaseous mixture,
with such enormous magnifying power as to show us the molecules as
they actually exist at any instant, then by counting the single and double
molecules we should doubtless find the same degree of dissociation which
is actually determined by physico-chemical methods.
On the other hand, if we should choose a condition in which the dissocia-
tion and reassociation occurs 10
13
or 1014 times as frequently, the atoms
of the dissociated molecules would hardly emerge from one another's
sphere of influence before they would once more combine with each other,
or with new atoms. In such a case the time required, in the process of
dissociation, would be comparable with the total time during which the
atoms would remain free, and even our imaginary instantaneous photo-
graph would not suffice to tell us the degree of dissociation. For, first, it
would be necessary to know how far apart the constituent atoms of a
molecule must be, to warrant oUt" calling the molecule dissociated. But
such a decision would be arbitrary, and according to our choice of this
limiting distance, we should find one or another degree of dissociation.
Until a problem has been logically defined it cannot be experimentally
solved, and it seems evident in such a case as we are now considering
that, just as we should obtain different degrees of dissociation by different
choices of the limiting distance, so we should expect to find different de-
grees of dissociation when we come to interpret different experimental
methods.
Now it is generally agreed that ionic reactions are among the most rapid
of chemical processes, and it is in just such reactions that we should expect
to find difficulty in determining, either logically or experimentally a really
significant value of the degree of dissociation.
Conclusions.-On the whole, we must conclude that the degree of dis-
sociation and the concentration of the ions are quantities which we cannot
determine by existing methods, and which perhaps cannot be defined
without some degree of arbitrariness. The question is one which should
be left open, especially as its answer is of no immediate concern to those
who employ purely thermodynamic methods.
1
1 Since the above was written a paper has appeared by Harkins (Proc. Nat. A&ad.
Sci., 6, 601 (1920)), in which the author brings forward several arguments against the
assumption of the complete ionization of strong electrolytes. While we may subscribe
to Harkins' statements in the main, we cannot agree that his experimental work on
solubility demonstrates the presence of intermediate ions in strong uni-bivalent elec-
trolytes. Our treatment of his results in the preceding section has shown to what an
extraordinary degree the various types of strong electrolytes obey identical laws. These
calculations do not. of course, prove the theory of complete ionization, but they surely
cannot be used as an r ~ n t against that theory.
Whatever conclusions may ultimately be reached regarding the degree
of dissociation of strong electrolytes, there can be no doubt whatever that
the phenomenon of dissociation is a very different thing in strong and in
weak electrolytes. In accordance with the theory of valence and molecu-
lar structure advanced by Lewis
1
we may explain such a difference as fol-
lows.
If we have a uni-univalent electrolyte whose cation is M+ and whose
anion is X-, the molecule may be represented by the formula M:X:,
where the pairs of dots represent the valence electrons, or the electrons
of the outer shells.
2
The pair lying between the atomic kernels M and X
constitutes the chemical bond. In the weak electrolytes, like acetic acid
or mercuric chloride, this approximates to the typical bond of organic
chemistry, but as we pass to stronger electrolytes the kernel of the cation
draws away from this bonding pair until, in the limit, this electron pair
may be regarded as the property of the anion alone. Then the positive
ion, which is the kernel M, is held to the symmetrical anion :X:, only
by the fact that they are oppositely charged. When an electrolyte in a
strongly polar or electrophilic medium approximates to this condition it
may be classified as a strong electrolyte. Whether we should call such
an electrolyte completely dissociated is, as we have seen, a matter of
choice. In all probability the additivity, in dilute solutions, of certain
physical properties, such as the heat content, accompanies the practical
disappearance of the chemical bond.
If then we agree that a strong electrolyte is one which is completely
polar,S and that the ions are held to one another by a simple electrostatic
force which obeys Coalomb's law, it becomes merely a matter of termin-
ology to decide whether we shall say that a certain fraction of such an elec-
trolyte is dissociated, or, with Ghosh, that a certain fraction of the ions
are free, or outside the "sphere of mutual attraction."
Summary.
After discussing various methods of determining the exact values of
the activity coefficient (thermodynamic degree of dissociation), we have
shown that the activity coefficients of sodium chloride (0-6 M), calculated
from freezing points, agree with those calculated from electromotive
force within a few tenths of a per cent., notwithstanding the fact that one
term alone in the freezing-point equation, which involves the heat of
1 Lewis, THIS JOURNAL" 38, 762 (1916).
2 We have for simplicity represented the ions as unso1vated. The same remarks
apply equally, however, to the more complicated case.
I We use the word polar in the sense proposed by Bray and Branch (THIS JOURNAL
35, 1440 (11113)) and by Lewis (ibid., 35, 1448 (1913)). .
1154
ARTHUR B. LAMB AND GORTON R. FONDA.
dilution, affects some of the results by 20%. An even more extensive
test of the several methods is furnished in the case of sulfuric acid, where
satisfactory agreement was obtained over the range from 0 to 20 M (66%
H
2
S0
4
) by three different methods.
Calculations of activity coefficients in mixtures, especially from the
solubility of salts in the presence of other salts, lead to an extremely power-
ful and general rule based upon the new concept of ionic strength. Each
ion contributes to this ionic strength by an amount which is proportional
to its stoichiometrical molality multiplied by the square of its valence.
In accordance with this rule the activity coefficient of every strong elec-
trolyte (and the individual activity coefficient of every ion) is the same in
all dilute solutions of the same ionic strength. It therefore depends in
no way upon the number or the nature of the ions of which the solution
is composed.
The general problem of the dissociation of strong electrolytes is dis-
cussed from several standpoints.
BIlRltllLIlY, CALIF.
[CONTRmUTJON FROM THE HAVEMJJ:YER CHEMICAL LABORATORY OF NEW YORX UNI-
VERSITY, AND THE CHEMICAL LABORATORY OF HARVARD UNIVERSITY.]
THE HYDROLYSIS OF DICHLORO AND HEXA-AQUO CHROMIC
CHLORIDES.
By ARTHUR B. LAMB AND GORTON R. FONDA.
Received March 15, 1921.
In a paper! published in 1906, it was shown that the change of r ~ n
chromic chloride into the blue (violet) isomer actually takes place through
the intermediate formation of an hydrolysis product of the green chloride,
presumably CrCbOH. Niels Bjerrum,2 in an able and comprehensive
paper published during the same year, arrived independently at the same
conclusion, and presented a large amount of quantitative evidence based
on kinetic measurements which agreed with it in a striking fashion.
In the course of his investigation Bjerrum measured the degree of
hydrolysis of both the green and the blue isomers, obtaining as average
values of the hydrolysis constants, 3.8 X 10-6, for the green, and 0.89 X
10-
4
for the blue salt at 25, the latter hydrolysis taking place according
to the equation
Cr(H
2
0)6+++ + H
2
0 = Cr(H
2
0)60H++ + H+.
Since then Sand and Grammling,8 from measurements of the potential
1 Lamb, THIS JOURNAL, 28,1710(906).
2 Bjerrum, Kgl. D. Vid. Selsk. Skr., (7) 4. 1 (1906); also Z. physik. Chem., 59,
336, 581 (1907).
2 Sand and Grammling, Z. physik. Chtm., 62, 1,28 (1908).

Activity Coefficient

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Activity Coefficient

Uploaded by

Copyright:

Available Formats

This paper is in a collection of

"Historic Publications In Electrochemistry"

2 Lewis and Lacey, THIS JOURNAL, 36,804 (1914).

You might also like