16.

7: Finding the Activity of a Solute from the Activity of the Solvent

We have seen that the activity of any component of an equilibrium system contains information about the activities of all the other components. From
¯¯
¯ ¯
¯¯¯ ~
dμ = −S dT + V P + RT dln a 

and the Gibbs-Duhem equation, we can find a general relationship among the activities. Substituting dμA and dμB into the Gibbs-Duhem equation, we have, for a two-
component solution,

−S dT + V dP = nA dμA + nB dμB

¯¯
¯ ¯
¯¯¯ ~ ¯¯
¯ ¯
¯¯¯ ~
= nA (−S A dT + V AP + RT dln aA  ) + nB (−S B dT + V B P + RT dln aB  )

¯¯
¯ ¯¯
¯ ¯
¯¯¯ ¯
¯¯¯ ~ ~
= − (nA S A + nB S B ) dT + (nA V A + nB V B) dP + nA RT dln aA   + nB RT dln aB  

Since S = n and V , this simplifies to

¯¯
¯ ¯¯
¯ ¯
¯¯¯ ¯
¯¯¯
A
S A + nB S B = nA V A
+ nB V B

~ ~
0 = nA dln aA   + nB dln aB  

or, dividing by n A + nB ,
~ ~
0 = yA dln aA   + yB dln aB  

For simplicity, let us consider a system in which a non-volatile solute, A, is dissolved in a volatile solvent, B. Measuring the pressure of the system and applying the
equations that we developed in Section 16.1 for volatile component A to the volatile solvent, B, in the present system, we can determine the activity of the solvent, B. Let us
use mole fractions to measure concentrations and take pure liquid B at its equilibrium vapor pressure as the activity standard state for both liquid- and gas-phase B. When
⦁
B is in its standard state, we have xA = 0 , xB = 1 , and ¯
¯¯
V
¯
B
(g) = V
¯
¯¯¯

B
(g) . Then, since the solute is non-volatile, we can determine the activity of the solvent, B, from the
pressure of the system. We have

¯
¯¯¯
⦁
P
P ⎛V (g) 1 ⎞
~ B
ln[aB (P , yA , yB )] =  RT ln[ ]−∫ −  dP
⦁ ⦁ RT P ⎠
P P
B
⎝
B

Assuming that the integral makes a negligible contribution to the activity, we have
P
~ ~
aB (P , yA , yB ) = aB = yB γ B (P , yA , yB ) =
⦁
P
B

(solvent)
so that
P
γ B (P , yA , yB ) = γ B =
⦁
yB P
B

(solvent)
Since the gas-phase concentration of A is immeasurably small, we must determine its activity indirectly. Let the standard state for solute activity be the hypothetical pure
liquid, y = 1 , whose equilibrium vapor pressure is equal to the Henry’s law constant of solute A. (We can determine the solute’s activity without measuring its Henry’s
A

law constant.) We have

(H yp ℓ, ҝA ) + RT ln[aA (P , yA , yB )] 

o ~
μ (P , yA , yB ) = μ̃
A A

where
~ ~
aA (P , yA , yB ) = aA = yA γ A

(solute)
Since we are able to measure the activity of the solvent, we can determine the activity of the solute from the relationship 0 = y A
~ ~
dln aA   + yB dln aB   . Rearranging, we have

yB 1 − yA
~ ~ ~
dln aA   = − dln aB   = − ( ) dln aB  
yA yA

For two solutions in which the mole fractions of A are y and y , and in which the activities of A and B are a
A
~
,a
#

A
~
,a
~
, and a
~
, we have A B
#

A
#

~ yA
aA 1 − yA
~
ln = −∫ ( ) d ln aB
~# # yA
a y
A A

Graphically, the integral is the area under a plot of −(1 − y A )/yA versus ln a
~
  , from y
B
A
#
to y .
A

Typically, we are interested in solutions for which y ≪ y . In the limit as the solution becomes very dilute, the activity, mole fraction, and activity coefficient of the
A B

solvent, B, all approach unity: a

~
→ 1, y → 1 , and γ
B → 1 . The activity of the solute, A, approaches the mole fraction of A. As a matter of experience, the approach is
B B

asymptotic: as the mole fraction approaches zero, y → 0 , the solute activity coefficient approaches unity, γ → 1 , and does so asymptotically, so that a
A
~
→ y . For dilute A A A

solutions, ln a
~
→ −∞  and ln γ
A → 0  asymptotically. In consequence,
A

lim (dln γ A  )  = 0

y →0
A

Because the activity coefficient approaches a finite limit while the activity does not, we can express the solute’s activity most simply by finding the solute’s activity
coefficient. Since a
~
= y γ and a
A
~
= y γ
A A
= (1 − y ) γ
B
, we have
B B A B

~ ~
0 = yA dln aA   + yB dln aB  

= yA dln yA γ A   + yB dln yB γ B   = yA dln yA   + yA dln γ A   + yB dln yB   + yB dln γ B  

= (dyA + dyB ) + yA dln γ A   + yB dln γ B   = yA dln γ A   + yB dln γ B  

(Since y A + yB = 1 , we have dy A
+ dy
B
= 0 .) We can rearrange this to
1−y
A
dln γ A   = − ( ) dln γ B  
yA

As the solute concentration approaches zero, (1 − y A )/yA becomes arbitrarily large. However, since lim y
A
→0 (dln γ A  )  = 0 , it follows that

lim dln γ B    = 0
y A →0

We see that the solvent activity coefficient also approaches unity asymptotically as the solute concentration goes to zero. The solute activity coefficient at any y A > 0 is then
given by
yA yA
1 − yA
∫ d ln[γ A (yA )] ln[γ A (yA )] ∫ −( ) d ln[γ B (yA )]
0 0
yA
 Figure 6. Graphical representation of ln γ for the solute when γ of the solvent is known.
As sketched in Figure 6, the latter integral is the area under a graph of − (1 − y )/y versus ln[γ (y )] , between y = 0 and y . Since γ (y ) → 1 as y → 0 , this integral
A A B A A A A A A

must remain finite even though − (1 − y )/y → −∞ as y → 0 . This can occur, because lim
A A A
dln γ    = 0 , as we observe above. Nevertheless, the fact that the
y A →0 B

integrand is unbounded can limit the accuracy of the necessary integration. For accurate measurement of the solute activity coefficient, it is important to obtain solvent-
activity data at the lowest possible solute concentration.

The most desirable situation is to collect solvent-activity data down to solute concentrations at which the solvent activity coefficient, , becomes unity. If
#
γB γ B (y ) = 1
A

when the solute concentration is y , ln[γ (y )]  can be evaluated with y , rather than zero, as the lower limit of integration. In some cases, ln[γ (y )]  may be known
# # △
A A A A A A

from some other measurement at a particular concentration, y ; if so, we can find ln[γ (y )]  by carrying out the numerical integration between the limits y and y .
△
A A A
△
A A

If the measurement of γ cannot be extended to values of y at which γ (y ) = 1 , we must find an empirical function, call it f (y ) , that fits the experimental values of
B A B A A

ln[γ (y )]  , for the smallest values of y . (That is, the empirical function is f (y ) = ln [γ (y )]   .) The differential of f (y ) is then a mathematical model for dln[γ (y )] 
B A A A B A A B A

over the region of low solute concentrations. Letting y

A
#
be the smallest solute concentration for which the solvent activity can be determined, we can integrate, using the
function for dln[γ that we derive from this model, to estimate ln[γ . Uncertainty about the accuracy of the mathematical model becomes a significant source
#
B (yA )]  B (y )] 
A

of uncertainty in the calculated values of γ . A

Of course, if we can find an analytical function that provides a good mathematical model for all of the solvent-activity data, the differential of this function can be used in the
integral to evaluate ln[γ (y )]  over the entire range of the experimental data. If necessary, the evaluation of this integral can be accomplished using numerical methods.
A A

It is essential that any empirical function, f (y ) = ln [γ (y )]   , have the correct mathematical properties over the concentration range to which it is applied. If it is to be
A B A

used to extend the integration to y = 0 , f (y ) must satisfy f (0) = 0 and df (0) = 0 . This is a significant condition. For example, consider the approximation
A A

αB
f (yA ) = ln[γ B (yA )]  = cB y
A

This model gives

dln[γ A (yA )] 
α B −2
= − (1 − yA ) c αB y
B A
dyA

and
dln[γ A (yA )] 
lim   = 0
y A →0 dyA

requires α B > 2 .