Alfred w.

%rancisl

Mobil Oil Coroorotion

Paulsboro, N e w Jersey
and Norman O. Smith
Fordham University
N e w York, N e w York 10458

I
I

Ternmy Liquid Systems

O v e r the past thirty years an extensive

body of data on ternary liquid systems has been published. A survey of these data and, in particular, the
considerable experimental work of one of us (AWF)
with such systems have revealed certain features of
phase diagrams found only in monographs or research
papers but which are not general knowledge. The
frequency with which some of the features have heen
observed suggests that there are guiding principles to be
followed, and that textbooks are frequently in error in
certain matters. I n most cases apparent exceptions to
these principles have been found to be the result of
known or probable impurities. This paper is an attempt
to summarize what generalizations have been observed
and to draw attention to some examples of novel and
fascinating behavior which add interest to a classical
subject. It is hoped, at the same time, that it will
help to discourage the drawing of unrealistic phase
diagrams for instructional purposes. Except where
stated, it is understood that ternary systems are being
discussed, so that the presence of a fourth substance
(especially water), even in small amounts as an impurity, is excluded.
Two Coexisting Phases

The most familiar phase behavior is that in which

two of the three liquid components are immiscible or
partially miscible, but each is miscible with the thud.
Addition of a sufficient quantity of the latter increases
the mutual solubility of the other two to the point where
complete miscibility is attained, as shown in Figure 1
(1) for the system acetic acid-isopropyl ether-water a t
23-24C. The binodal curve or "bite," XPY, separating the one-phase and two-phase areas, is the locus of
the ends of the tie lines. At P, the plait point, the tie
lines have shrunk to a point. As applied to isothermal
extraction for an unusually favorable situation, B is the
original solvent for the solute A, and C is the extracting
solvent. The extract lies on Y P and the residual solution (raffinate) on XP. For a more common, hut less
favorable, situation the meanings of B and C are reversed. If the two-phase area is small then it is
usually also shallow. Othmer and coworkers (1)
observed that the plait point is located at a flatter portion of the binodal curve. While the curve is generally
convex, a concavity can occur, as in the system watersulfuric acid-diethyl ether ($) at 25'C, where chemical
interaction between components takes place to some
Presented before the Division of Chemical Education of the
American Chemical Society, Minneapolis, Minnesota, April 1969.
Present Address: Apt. 12A-3, Redfield Village, Metuchen,
New Jersey 08840.

extent. (Such interaction does not destroy the ternary

character of the system as long as the composition of all
phases a t thermodynamic equilibrium can still he
expressed in terms of the original three components.
I n general, a quadrangle is needed to represent the
isothermal behavior of truly reciprocal ternary systems
but, with the example mentioned, the composition
range studied is confined to a triangle, thus making the
latter an adequate means of representation.) It is
clear, however, that a concavity cannot occur at the
plait point for then the tie lines immediately adjacent to
it would have to cross the one-phase area and so result
in a contradiction. A symmetrical hinodal curve, with
a plait point near the apex of the triangle is rare, but
observed in the system water-acetone-carbon disulfide
at 20-21C (5) when compositions are plotted in
weight per cent. This system also possesses an isopycn i e a phenomenon to he discussed below.
The tie lines, when produced, intersect the side of the
triangle on which the binodal curve is based only when
that side is also extended, as shown in Figure 1. Alleged examples of tie lines intersecting the unextended
side have, in several cases tested, been shown to be the
result of impurities (4).Similarly, a tangent to the curve
at the plait point does not intersect the unextended
side. Alternatively, if straight lines be drawn from
the plait point to all three corners of the triangle, only
one of these fails to cross the corresponding two-liquid
region. Meeting this restriction which, so far a3 is
known, has no thermodynamic proof, requires considerable care in the construction of figures with two or
three plait points. Such systems will be discussed
below. I n these the reverse relation also probably
holds, viz., only one plait point can be reached by a
straight line from any one corner without crossing its
ACETIC ACIDO

Figure 1.
23-24'C.

The system acetic ocid(A1-iropropyl other(B1-water(C1 at

Volume 46, Number 12, December 1969

81 5

own hinodal curve. This correspondence of components (corners) with plait points suggests that more
than three plait points are impossible. The slope of the
tie lines usually changes steadily with increase in the
quantity of distributed component (A). This is to be
expected, at least for small additions of A, from the
distribution law. In many systems the direction of
change of slope reverses as the amount of distributed
component increases (6); in a few systems, e.g., waterpropionic acid-benzene a t 40C (6) there are at least
two reversals. (It is worth observing, however, that
changing from weight per cent to mole per cent composition coordinates not only changes the slopes of the
tie lines but may even change their sign.)
I n many two-phase areas there is, at a given temperature, one tie line connecting phases of equal density
(an isopycnic). Total compositions anywhere on such
a line have the same overall density as that of the
phases. It may he noted that isopycnics are to he
avoided in extraction processes, since settling of the
layers is prevented. (An interesting lecture demonstration can he arranged by adding traces of colored
substances to an isopycnic system in order to make
the two phases readily distinguishable (7).) The
system water-o-toluidine-propionic acid at 20C (7)
has two isopycnics such that the addition of propionic
acid to the immiscible water-o-toluidinc system causes
the water-rich and o-toluidine-rich layers to change
places twice before complete homogenization is attained.
By analogy with isopycnics there are also iso-opticsunique tie lines which join phase compositions that have
the same refractive index for a given wave length of
light. Such two-phase systems are usually colored.
There may he a set of such iso-optics adjacent to each
other, all giving differently colored pairs of layers.
Figure 2 (8)shows a binodal area in the system benzenenitromethane-n-heptane at 25C to be covered with
"colored" tie lines. Thus the complete tie line pattern
can he observed colorimetrically. The colors are also
temperature-sensitive.
The above description applies, in general, to systems
which show two or even three separate binodal areas.
I n Figure 3 (9) for the system carbon dioxide(1)methyl phthalate-n-heptane at 21-26'C under pressure,
which is typical of many systems, the addition of the
third component to the other two exerts an homogeniz-

!BON DIOXIDE

METHYL PHTHALATE

wHEPTANE

Figure 3. Tho ryrtem carbon dioxide(1)-methyl phfhdote-n-heptone a t

21-26'C lunder prcsrurel.

ing effect if not too much is added; continued addition,

however, can cause two liquid phases to reappear if the
composition of the original mixture lies in a suitable
range. Three separate hinodal curves are also common
if one is not limited to ambient pressures. To show the
phenomenon at atmospheric pressure, however, components must be chosen so that the three binary
critical solution temperatures are near together-a rare
combination. The system ethylene glycol-nitromethane-lanryl alcohol at 29C (10) is representative
(see Fig. 10a, to he discussed later).
When a single two-phase region extends from one
binary system to another, either as a result of the merging of two bites or, more commonly, from the enlargement of one bite until it reaches the other side of the
triangle, a hand results as in Figure 4 for the system
furfural-water-ethyl acetate at 25'C (11). This situation is quite common. The short side of the band is
often slightly concave. The slope of the tie lines
crossing the band, although usually changing steadily
as the triangle is crossed, can somewhat rarely reverse
the direction of change, as in the system water-hydrogen
bromide-n-heptane at 25C (If?).
There are also "island curves," hinodal areas completely surrounded by a one-phase area. These curves
are smooth in their entirety, although occasionally
drawn erroneously with discontinuities in slope. The
three components, taken in pairs, may or may not he

BENZENE

FURFURAL

h
NITROMETHANE

n-HEKANE

Figure 2. The system benzene-nitmmefhone-n-heptane a t 25'C rhowing

rtruclurol o%rr: a, light blue; b, indigo; c, purple; d, lilac; e, magenta;
f, brick red; g, amber; h, amber yellow; i, yellow.

8 16

Journol o f Chemicol Education

Figure 4. The ryrtem furfurol-voter-ethyl acetate a t 25'C.

line is on bopycnic.

The dashed

miscible. The island area will have two plait points.

The phenomenon is often associated with the reversibIe
formation of a soluble product of chemical reaction
between two of the components. I n the system wateracetone-phenol between 66' and 92'C (IS) there is an
island curve, but complete miscibility in all three
binaries. The same is true for propionic acid-hydrogen
bromide-n-heptane a t 0C (14). Here the island is
almost triangular and nearly fills the entire diagram.
I n the system hydrogen chloride-water-dioxane at 20C
under pressure (14) there is an island near the dioxane
corner of the diagram and a small bite on the opposite
side near the hydrogen chloride corner as in Figure 6.
The tie lines shown in t,his figure, and those in T'g
4 1 ures
8, 9, 10, and 11 are speculative, but have been included
for emphasis. I t seems improbable that more than one
island curve can exist in a genuinely ternary system:
although two islands have been reported in the system
aniline acetic acid-isooctane at 51C (15) it would appear that the formation of aniline acetate is not rapid or
reversible enough to maintain the true ternary quality
of the system.
Three Coexisting Phases

At constant temperature and pressure three coexisting

phases (three liquids, two liquids and one solid, one
liquid and two solids, etc.) are represented by the
corners of a scalene "internal triangle." Only the
first two possibilities are pertinent to this discussion.

Figure 5. The ryrtem hydrogen chloride.woter-dioxane a t 20'C.

The coexistence of three liquid phases is illustrated

by Figure 6 for the system carbon dioxide@-waterphenol a t 25-C (under pressure) (9),simplified for the
present purpose by omitting the equilibria involving
solid phenol. (Inclusion of the latter would have
required a region for liquid saturated with solid phenol
which would have interrupted, and reduced the size of,
the miscibility gap based on the carbon dioxide-phenol
side of the triangle.) Similar three-liquid regions are
evident in Figures l o b and l l b described below.
The manner in which the borders of two areas, each
denoting two-phase equilibria, make contact with a
three-phase area is governed by the Schreinemakers
rule (16) according to which the tangents of ac and be
s t c in Figure 6, for example, must both lie within or
both lie without the three-phase regions. The same must
he true at the other corners of the internal triangle.
This rule is valid regardless of the state of aggregation
of the various phases. (Thus it applies to the well
known isothermal solubility curves in ternary systems
of two salts with a commou ion and water at the point
representing simultaneous saturation of liquid with two
solids.) A proof of the rule is given by Schreinemakers
(16) and by Williamson and Morey (17). Figure 7
illustrates an erroneous disposition of the isothermal
solubility curvecfor two salts (B and C) with a common
ion and water (A). The reader may wish also to examine Figures 8, lob, and l l b for conformity with the
Schreinemakers rule.
Examples in which the three coexistiug phases are

Figure 7. System of two ralh with o common ion and wafer showing
ermneour dirporition of isothermal solubility curves.

POTASSIUM CARBONATE

Figure 6. The system carbon dioxide-water-phenol a t 25'C (under prel?ha equilibria inrdving d i d phenol ore omiwed.

sure).

Figure 8. The ryrtem potosium carbonate-water-methml at 25'C.

represents the hydrate 2KzCOs .3Hz0.

Volume 46, Number 12, December 1969

817

two liquids and one solid are common, and arise, for
example, when the components are two miscible or
partially miscible liquids, and a solid which is soluble in
both of them as in the system aniline-isooctane-naphthalene at 20C (18). Systems in these categories also
arise in connection with the process of salting-out, as
with potassium carbonate-water-methanol a t 25%,
shown schematically in Figure 8 (19).
The manner in which a change in temperature can
cause two-phase regions to interact with each other has
been a matter of speculation for some time. It is only
recently that sufficient experimental data have been
available to provide examples in support of the speculation. Three such types of interaction may now be dis-

NlTR BENZENE

Figure 9. o, b, r The system nitrobenzene-methonoI.i~oocttne a t 15.0.'

14.1 and 1 4.0C, respectively.

'

818

Journal of Chemicol Education

tinguished: (1) external contact of two binodal

curves rut a "col," (2) external contact of two binodal
curves a t a "subcol," and (3) eruption of one twophase area from within another.
The first of these modes arises when the two binodal
areas first make contact at their plait points. This is
illustrated in Figure 9a,b,c representing schematically
the relations of the system nitrobenzene-methanolisooctane at 15.0, 14.1 and 14.0C (to). Points PI
and P, are the plait points. The name col or saddle
point, found at 14.1C, derives from the appearance of
the solid model obtained when a temperature axis is
erected on the composition triangle. Careful experimental study has shown that a t the col temperature the
two curves which have just made contact, Figure 9b,
do not have a common tangent as sometimes postulated,
but appear as two nearly straight lines intersecting a t a
finite angle. Thereafter, the borders of the resulting
band are highly concave but have no cusps, Figure
9c.
The second mode of contact arises when two binodal
areas make contact involving only one of the two plait
points. Figures 10a and lob show schematically the
sequence of events in the system nitromethane-ethylene glycol-lauryl alcohol at temperatures just above
and just below the subcol temperature, 28.5"C, respectively (10). The subcol tie line ab, Figure 10a, splits
to form the internal triangle a'af'b of Figure lob. It is
to be noted that one of the contacting curves-the one
the plait point of which is not involved in the coutact-

Figure 10. a, b. The system nitmmethme-ethylene glycol-lauryl alcohol

l u a above and /"st below the subcol temperoture, 28.5'C, respestiveiy.
The subcol tie line is ab.

and water are partially miscible and that the addition

of sufficient acid produces homogeneity, just as the
addition of A in Figure 1 makes the partially miscible
system B-C homogeneous. When the water is replaced
by salt solution the triangular diagram is still adequate
for expressing all compositions if the salt solution,
alcohol, and acid are mixed in proportions such as to
give a single liquid phase. When, however, the proportions are such as to give more than one phase (the
only situation of interest in heterogeneous equilibrium)
it is not, in general, possible t,o describe all the phase
compositions in terms only of the three variables salt
solution, alcohol, and acid, and therefore not possible to
represent them by a point within the triangle--unless
the salt/water ratio is the same in every phase as it is in
the original solution represented by the corner of the

Figure 1 1 . a, b. T h e system nitroethane-ethylene glycol-lauryl alcohol

lust above a n d lust below 16..SC. The equilibria involving d i d loury1
olcohol ore omitted.

Figwe 12. The 3ystem toluene-97% acetic acid-n-heptone ot 2SPC

ghoxing phore boundory cunrer.

must be concave at the point where the merger occurs,

in order not to contravene the Schreinemakers rule
given above.
The t,hird
kind of behavior is shown in F i ~ u r e sl l a
~-~
and l l b which illustrate schematically the system
nitroethane-ethylene glycol-lauryl alcohol (10) a t temperatures just above and just below the temperature at
which the new binodal area erupts from within the old,
also a subcol temperature, 16.5"C. (The phase relationships are actually slightly more complex than
shown, because of the presence of solid lauryl alcohol,
m.p. 22.7'C. The equilibria involving this solid have
been omit,ted from the diagram for simplicity, but the
omission does not affect the liquid-liquid equilibria
features under consideration.)

triangle. The latter event is unlikely since the salt

would tend to enter the water-rich layer to a greater
extent than the alcohol-rich layer.
In spite of the inability to represent even two-phase
equilibria, one can still plot "phase boundary curves"
on triangular diagrams in such cases. These are
curves resembling XPY of Figure 1. Total compositions lying outside of the "bite" area give a single phase
of the same composition; total compositions lying within
the area give more than one phase, but the compositions
of these phases cannot be accurately represented since
such phases are not ternary. As they cannot be
plotted they do not necessarily lie on the phase boundary curve and, strictly, tie lines cannot be drawn across
the area. The curve is not, therefore, a binodal curve,
and the comments made earlier concerning the shape of
binodal curves are not applicable to it. Nevertheless,
it is possible to plot a plait point under these circumstances, since it, too, lies on the edge of an area of
homogeneity.
Phase boundary curves frequently consist of reverse
curves with concavities, as shown by Figure 12 for the
system toluene-aqueous acetic acid-n-heptane at 25'C
(81). When anhydrous acetic acid is used, the three
components, acetic acid, toluene, and n-heptane, are
miscible in all proportions, so the phase diagrams consist of an empty triangle. When 98% acetic acid
(in water) is used, however, partial miscibility develops
between the aqueous acid and the heptane. When
97% acetic acid is used partial miscibility also develops

NlTR ETHANE

~~

Presence of a Fourth Substance

The presence of small amounts of a fourth substance

can exert a profound influence on the phase behavior of
the other t,hree. The location of plait points, for
example, is very sensitive to impurities. The fourt,h
substance is sometimes deliberately introduced in
appreciable quant,it,iesby, for example, replacing one of
the component,^ of a ternary system by a solution of the
fourth substance in that component. The system is
now quaternary (21), but certain features of ternary
systems may be retained under these circumstances.
Consider, for example, the use of an aqueous salt
solution in place of pure water in the ternary system
wat,er-alcohol-acid. Let us suppose that the alcohol

Volume 46, Number 12, December 1969

819

between the aqueous acid and the toluene, a seen in

Figure 12. By regarding water as an impurity in this
svstem one can readily
" auureciate the drastic influence
on liquid-liquid equilibria of the presence of soluble
impurities.
.A

Literature Cited
( 1 ) O T R ~ RD.
. F.. WBITE. R. E., A N D T R U ~ ~E..
~ R
Ind.
, Enp. Chem..
33, 1241 (1941).
( 2 ) Trrrr, A,. Bull. Soc. C h m . Fmnce, 13,583 (1946).
E.,, Suomen. Kem., 13B,13 (1940). See ~ 1 8 0Ref. (22). p. 845.
( 3 ) LEIKO~A
( 4 ) F n ~ ~ o xA.
s , W., "Liquid-Liquid Equilibriums." Intemoiencs (division
of John Wiiey &Sons, Ine.). New York. 1963, P. 36.
( 5 ) S M I T ~J., C., STIBOLT,
V. C.. A N D D A Y . R. W.. Ind. Ens. Cham., 43,
.n,,

(6)
(7)
(8)
(9)
(10)

,.r".,.

,.,,K,,

BIANCANI.
M.. A N D DEFILIPPO,
D.. G ~ u .Chim. I t d . . 88, 1202 (1958).
F n * ~ c ~ A.
s , W.. Ind. Eno. Cham.. 45, 2789 (1953).
F n ~ ~ mA.
s , W.. J . Phys. Chcm.. 56, 510 (1952).
F R A N CA.
~ .W.. J . Phya. Cham.. 58, 1099 (1954).
FRANCIS,
A. W., J . Phya. Chem., 60, 20 (1956).

820

1 lournal of Chemical Education

(11) TRIMBLF.
F.. A N D DUNLOP.
A. P.. Ind. EW. chsm.. ~ n a l ~. d . 12,
. 721
(1940).
( 1 2 ) F m m r a . A.
A N D KING.
W. H., "chemistry of P
~ nydro- ~
oarbone."
.a=.
- "A*(Editor: Rnooxs. B. T.. etol.) ~ e i n h d dN, ~ ~Wo r k vol.
,
I,

.,..,-,
".

w..

r.",..

( 1 3 ) S o n n e m ~ u ~ x e aF.
s . A. H.. Z.Phyaik. Chem.. 33, 79 (1900).
s ,W.. J . Phya. Cham.. 6 2 , 579 (1958).
(14) F n ~ ~ c lA.
(15) R e f . ( 4 ) p. 88. 9. Z~unAvbev.E. F.. A N D V o ~ n o v .A. A,. Imesl.
Vysshikh Uchob. Zovdenii Khim. i Khim. Tekhnol 3.427 (1960).
(16) S c ~ = m ~ n r ~ F.
m ~A.s .H., "Die Heterogsnen Gleiohgewiohte Vom
Standpunkte Der Phssenlehre," (Editor: B ~ a m o sRoozsaoow.
H . W.) Braunschweig, Vol. 111, Part 1. 1913, pp. 71-74, 260-270.
(17) W r ~ ~ ~ ~ r sEo. rD..
i , h m Monzr, G. W.. J . Am. Chcm. Soc.. 40, 49

,.-.",.
,<",a,

(18) A x o e ~ ~ s c u
E... A N D ZINC*, 8.. Bull. Slcl. Sci. Acod. Roumoinc. 24,

\."-",.

**a ,>o*.,,
7-"

(19) FRANKFOATER,
G . B., ANDFRART,
F. C., J.Phya. Chem., 17,402 (1913).
(201 FRANOIS.
A. W.. J . Am. Chcm. Sor.. 7 6 , 393 (1954).
(21) For detailed studies of quaternary liquid systems as such, see, for
example. Bnmonm, A. V., H U N T ~ RT,. G., A N D NASH.A. W.. J .
, F., W*Lsn, T. J., A N D
Phys. Chem.. 44, 683 (1940) and P n o m o ~C.
D E S A A.
~ , M., Ind. Eno. Chsm., 42, 1210 (1950).
(22) FRANCIS.
A. W.."Tern&w Systems Separating into T w o Liquid Layers,"
Solubilities of Inorganic and Organio Compounda. Supplement to
3rd ed.. (Editor*: SELDELL,A , . A N D LINKE,W. P.) Vsn Noatr*.~.d.
N e w York. 1952, 8. 1077.