DETERMINATION OF SCTIVITY COEFFICIENTS FRO31 TOTAL PRESSURE NEASUREMENTS

[Manuscript received March 12, 19531 Summary A method of least squares is described for calculating activity coefficients from results of total vapour pressure measurements.

I. INTRODUCTION The present paper describes a procedure, based on the method of least squares, for calculating activity coefficients for solutions from total vapour pressures. Although several methods are described in the literature (see, for example, Musil and Schramke 1 9 4 9 ~ ~ 19493 ; Redlich and Kister 1949), there is no systematic method capable of extracting the best results from given experimental data, with due allowance for vapour phase non-ideality. The author believes that activity coefficients derived from total pressure measurements can be comparable in accuracy with those derived from the more difficult partial pressure measurements, particularly for solutions of high relative volatility where the difficulty of obtaining a vapour sample in true equilibrium with the liquid is most marked. This belief is supported by the example worked below. A general discussion of the relative advantages of total and partial pressure measurements is given by Scatchard (1952). The activity coefficients and excess chemical potentials are given by the equations (Scatchard and Raymond 1938) :

Here pf and p are the changes in chemical potentials in excess of the : changes for an ideal solution, P is the total pressure, P, and P, are the vapour pressures of the pure components, cc and y are liquid and vapour mole fractions respectively, y, and y, are activity coefficients, V, and V, are liquid molar volumes, Pll, P2,, and PI, are second virial coefficients in the equations of state of the pure and mixed vapours, and 612=ZP12-Pll-P22. It follows from equations (1) that the total pressure is given by P =yip; +y2p;, ................................ ( 2 ) p;
=G'1

exp [(Vl-

p;=x2P,

.... ( 3 ) exp [(V,-P,,)(P-P,)/RT-PS12y,2iR.Tl. ( 4 ) ....

P 1 1 ) ( P - ~ 1 ) / ~ ~ - ~ ~ 1 2 ~ ~ / ~ ~ 1 ,

* Division
B

of Industrial Chemistry, C.S.I.R.O., Melbourne.

208

J. A. BARKER

The symbols p i and p',, introduced for brevity and defined by equations ( 3 ) and (4), represent the values the partial pressures would have if the liquid solution were ideal but the vapour phase remained non-ideal. It has been usual in the past to assume that the mixture virial coefficient p,, is the arithmetic mean of p,, and p,,, SO that a,, is zero. This assumption does not lead to serious error for solutions of non-polar components, but Scatchard and Ticknor (1952) showed that it could be a serious source of error for solutions containing associated liquids. I n describing the method we shall therefore not assume that a,, is zero. We assume that the concentration dependence of the excess free energy of mixing can be described by the equation (Scatchard 1949)

I b=c=. f =0, equation (5) has the form appropriate for a regular An equation of the form (5), with sufficient terms, can represent solution. measured values of G z with any desired accuracy. Equation (5) implies that t'he activity coefficients are given by

...

In y,=AT, +Bm, +Cn, f ...., In y,=A& +Bm, +Cn, .....

where
1, =xi, m, = -xt(1 -4x,), 1, =x:, m2= +xT(l -4x,),

..................

(6)

n, =xg(l -8x, +12x,2), . , =x,2(l-8x, +12x$), n

and

A =a/RT, =b/RT, C =GIRT. B

............

(8)

1 . DESCRIPTION METHOD 1 OF We determine the constants A, B, C by a process of successive approximations. To find a first approximation we assume that the solution behaves like f a regular solution, so that B and 0 are zero. I we neglect the corrections for vapour phase non-ideality, then A is given by

where P* is the pressure for the equimolar mixture, which can be estimated graphically. Using this value of A we can calculate approximate vapour concentrations which are sufficiently accurate to use in the small correction term P8,,y2/RT in equations (3) and (4). We now calculate pi, p', for the experimental liquid compositions. Using the first approximation for A, we calculate y, and y, by equations (6), P by equation (2), the pressure residuals R =P,,,t. --P,,1,., and the derivatives dP/dA, dP/dB, dP/clC, which are given by

ACTIVITY COEFFICIENTS FROM PRESSURE MEASUREMENTS

209

W e determine the changes 6A, 6B, SO in A, B, C which will most nearly reduce the pressure residuals to zero, by fitting by least squares to the equation (dP/dA)GA+(dP/dB)GB +(dP/dC)$C =R. That is, we solve the equations 8AX(dP/dA)2 +6BX(dP/dA) (dP/dB)+SCX(dP/dA)(dP/dC) =XR(dP/dA), SAX(dP/dA)(dP/dB) +6BX(dP/dB)2+SCX(dP/dB)(dP/dC) =XR(dP/dB), GAX(dP/dA)(dP/dC) +GBX(dP/dB)(dP/dC)+GCX(dPldC)2 =XR(dP/dC), . . . . . . . . . . . . . . . . . . . . (11) where the summations X are taken over all the experimental points. Adding these increments to the initial values of A, B, 0 we find improved approximations. Since the second derivatives of P with respect t o A, B, G are not zero, these are not necessarily the best values, and i t may be necessary to repeat the process until A, B, C do not change significantly. In practice, it is not necessary to recalculate the derivatives dP/dA etc. a t each stage. We illustrate the method using the data of Brown (1962) on the system benzene +.n-heptane a t 80 "0. The second virial coefficients used were those given by Brown ; in particular 6,, was assumed to be zero. In Table 1we give
TABLE1
SUCCESSIVE APPROXlXATIONS TO

. . . . . . . . (10)

A, B ,

AND
I

Pressure residuals,

experimental liquid compositions and total pressures, the pressure residuals found with first, second, and third (final) approximations to A, B, and 0, the. calculated vapour compositions, and finally the measured vapour cornpositiona (which were, of course, not used in the calculations). The agreement betweem the two sets of values of vapour compositions is very satisfactory.

210

J. A. BARKER

Our values of the constants a, b, c in the equation for G z , together with those determined by Brown from total pressures and vapour concentrations, are shown in Table 2. The values of G: found with the two sets of constants are
TABLE2
CONSTAXTS I N EQUATIOX BOR

G:

Fronl total pressure

i
1

I
253.9 62.5
I

18.3

From total pressure vapour composition and

254.3
i

64.3

1
I

12.3

shown in Table 3. Since the estimated experimental error in Brown's was 3 cal/mol, the two sets of values are in remarkable agreement.
TABLE3 C :

values

VALUES O F
I l

$1

---

0.2

1
l

0.3

0.4

0.5

, ,
0.6

0.7

0.8

1
I

0.9

G: (total pressure and vapour composition) ..

111. REFEREXCES
BROWN, (1952).-Aust. J. Sci. Res. A 5 : 530. I. MUSIL,A., and SCHRAMKE, (1949a).-Acta Phys. Austr. 3 : 137. E. MUSIL,A., and SCHRANKE, (1949b).-Acta Phys. A u s t ~ .3 : 309. E. REDLICE, . , and KISTER, T. ( I 9 4 9 ) . 4 . Amer. Chem. Soc. 71 : 505. O A. SCATCHARD, (1949).-Chem. Rev. 44 : 7 . G. SCATCHARD, (1952).-Ann. Rev. Phys. Chem. 3 : 269. G. SCATCHARD, and RAYMOND, L. (1938).-J. Amer. Chem. Soc. 6 0 : 1278. G., C. SCATCHARD, and TICKNOR, B. ( 1 9 5 2 ) . 4 . Amer. Chem. Soc. 7 4 : 3724. G., L.