T H E MECHANISM OF T H E COAGULATION OF SOLS BY ELECTROLYTES. I. FERRIC OXIDE SOL BY HARRY B. W E B E R
Two general theories of the mechanism of the coagulation of sols by electrolytes have been proposed : the adsorption theory of Freundlich and what may be termed the solubility theory of Duclaux and Pauli. The Adsorptzon Theory. The widely accepted adsorption theory’ assumes in the first instance that the particles of hydrophobic sols owe their charge to the preferential adsorption of ions from the intermicellar solution. Thus silver chloride formed in the presence of a slight excess of silver nitrate is positively charged because the particles adsorb silver ion more strongly than nitrate ion whereas silver chloride precipitated with a slight excess of sodium chloride is negatively charged because, in this case, the anion is the more strongly adsorbed. Similarly, hydrous ferric oxide thrown down in the presence of a slight excess of ferric chloride or hydrochloric acid is positively charged because of preferential adsorption of cations. The addition of suitable amounts of electrolytes to such colloidal systems causes coagulation as a result of preferential adsorption of the ions whose charge is opposite to that on the colloidal particles. This adsorption lowers the charge on the particles below a critical value and the colliding particles agglomerate into clumps sufficiently large to settle out. The ions whose preferential adsorption by the sol particles is responsible for their charge are called stabilizing ions while the added ions of opposite charge whose adsorption lowers the particle charge, are called precipitating ions. Since the ions of a precipitating electrolyte which have the same charge as the sol help to determine the critical concentration necessary for coagulation, the precipitation concentration or precipitation value of an electrolyte for a sol has been definedl as that concentration which results in sufficient adsorption of the precipitating ion to neutralize below a critical value, the combined adsorption of the original stabilizing ion and that added with the electrolyte. That the precipitating ions are carried down by the coagulated particles was demonstrated first by Linder and Picton3 and then by Whitney and Ober4 and has been confirmed repeatedly by a number of investigators. To account quantitatively for the wide variation in the precipitating power of electrolytes of varying valence, Freundlich assumed that equal amounts of precipitating ions are adsorbed from equimolar solutions and that equivalent amounts of ions of different valence are carried down a t the precipitation value. The evidence to support these assumptions was not con-
’ Freundlich: “Kapillarchemie,” 1 s t Ed.,345 (1909); 2nd Ed.,572 (1922).
* Weiser and Nicholas: 4
J. Phys. Chem , 25, 742 (1921). J. Chem. Soc., 67, 63 (1895). J. Am. Chem. SOC.,23, 842 (1902).
HARRY B. WEISER
2
elusive and it was demonstrated I O years ago in the author's laboratory' that they are not generally true. Freundlich2 has recently convinced himself that his original postulates are not in accord with the facts. To account for the variation from equivalent adsorption during precipitation of sols, it was assumed3 that the adsorption of equivalent amounts of the several ions was necessary to neutralize the charge below the critical value and that the observed variation from equivalence was due to varying adsorption of the several electrolytes by the agglomerating micelles. As will be pointed out in the experimental part of this paper, these assumptions are not borne out by the facts. Since the two ions of the precipitating electrolyte are not adsorbed in equivalent amounts, the adsorption of the precipitating ion must be in part by exchange. Thus Linder and Picton4 showed that chloride ion passes into solution by exchange when a ferric oxide sol prepared from ferric chloride is coagulated by potassium sulfate; with As&& sols, hydrogen ion enters the solution by exchange during coagulation.5 I n concluding this brief survey of the adsorption theory of the coagulation process, it may be said that the theory furnishes a satisfactory semi-quantitative picture of the behavior of sols in the presence of electrolytes, but gives little insight into what actually takes place in any given case. The Solubility Theory. The solubility theory of the mechanism of the coagulation process probably had its origin in a concept of Wyrouboff6 that the various dialyzed ferric oxide sols are basic salts or chlorides of "condensed" ferric hydroxides. This idea was further developed by Duclaux' and especially by Pauli and his pupilsJ8who consider the stability of sols from the point of view of solubility. The colloidal particles are assumed to be highly complex colloidal ions resulting from the ionization of complex electrolytes allied to the Werner compounds. Coagulation in terms of this theory is believed to consist essentially of a chemical change involving the precipitatation of a difficult soluble electrolyte. Ferric oxide sol formed by adding ammonium hydroxide to ferric chloride solution followed by dialysis, has been investigated extensively from this point of view. Pauli9 determined the chloride content of the sol by direct analysis and by a potentiometric method and measured the conductivity of the sol a t various dilutions. I t was found that not all of t'he chlorine could be detected potentiometrically as chloride ion. The possibility that a part of the chloride was adsorbed'o and so would not be subject to potentiometric measurement Weiser and Middleton: J. Phys. Chem., 24, 30, 630 (1920). Joachirnsohn, and Ettish: 2. physik. Chem., 141,249 (1919). a Weiser and Middleton: LOC.cit. J. Chem. Sac., 87, 1908 (1905). Whitney and Ober: J. Am. Chem. Soc., 23, 842 (1902). E Bull., 21, 137(1899). 7 J. Chim. phys., 5, 29 (1907); 7, 405 (1909). 8 Pauli and Matula: Kolloid-Z., 21, 49 (1917); Pauli and Walter: Kolloidchem. Beihefte, 17, 256 (1923); Pauli and Rogan: Kolloid-Z., 35, I31 (1924). B LOC.cit. 10Cf. Maffia: Kolloidchem. Beihefte, 3, 85 (1911). 1
* Freundlich,
COAGULATION OF SOLS BY ELECTROLYTES
3
was ruled out and instead, it was assumed that the sol was merely an incompletely dissociated electrolyte which yielded only a fraction of its chlorine as ion. I n support of the dissociation theory, Pauli showed that the addition of an electrolyte to the sol which is equally concentrated in chloride ions causes no displacement of the chloride ion concentration of the sol. Moreover, under certain conditions the cations of an added electrolyte with a common anion may also decrease in concentration, a phenomenon which is attributed to a driving back of the disso9iation by the common ion of the ferric oxide sol. As Freundlichl points outrneither of these arguments is conclusive. For if the chloride ion is in adsorption equilibrium, it is only natural that the equilibrium should be maintained if the chloride ion concentration remains unchanged. Furthermore, the fact that the anions in the intermicellar liquid are sufficiently free to drive back the dissociation of an added salt with a common anion is likewise in entire accord with the adsorption equilibrium which exists in the sol. Finally, Pauli and Matula2 emphasize that the behavior of sols as regards conductivity cannot be interpreted simply from the point of view of the dissociation theory. Thus a mixture of ferric oxide sol and alkali chloride containing equivalent amounts of chloride ion exhibits a conductivity higher than the arithmetic mean of the conductivity of the components. This is probably due to higher mobility of the colloidal particle owing to an increase in charge by adsorption of a portion of the added cations. In this connection Lottermose? found the specific conductance of ferric oxide sols to be higher than that of the ultrafiltrates, the differences being regarded as the true conductivity of the micelles. If the micelles P are considered to be complex electrolytes, the equivalent conductivity a t infinite dilution may be calculated from the equation, AP, = xoooKp/Kcl, where K signifies specific conductance. The mobility of the micelles was found to rise abnormally with purified sols containing but small amounts of chlorine. This fact indicates that the micelles are really adsorption complexes, the abnormality being due to the displacing of the adsorption and hydrolysis equilibria by dilution. The general forrnula of a ferric oxide sol is written by Pauli [xFe(OH)3. yFeOCl.FeO]+ C1-. The addition of sufficient electrolyte such as potassium sulfate, to the sol causes coagulation and the amount of sulfate dragged down is the same as the amount of chloride in the supernatant solution after coagulation. The precipitation is attributed to the formation of a complex insoluble sulfate in accord with the equation [ X F ~ ( O H ) ~ . ~ F ~ O C ~ . F ~ O ] C ~ K Z S O= ~ [ X F ~ ( O H ) ~ . ~ F ~ O C ~ .2FKC1. ~ O ]Actually, ~ S ~ ~ of course, the alleged insoluble sulfate is not thrown down until a critical concentration of sulfate is added. This anomalous behavior as compared with that in the precipitation of simple insoluble compounds calls for an explanation. “The
+
+
l“Kapillarchemie,”621 (1922). Kolloid-Z., 21, 498 (1917). a Z.Elektrochemie, 30, 391 (1924).
+
4
HARRY B. WEISER
addition of sulfate and similar acting salts to the sol, says Pauli and Walter,’ “causes precipitation owing to the formation of insoluble compounds. The anomalous behavior as compared to simple electrolytes, namely that an amount of sulfate almost equivalent to the entire chlorine content of the sol is necessary for flocculation, is explained by a peculiar equilibrium bet ween the complex bound and the ionic chlorine as well as by the peptizing action of the undischarged complex. The coagulated sol has the formula of a complex double salt, a chlorine-poor and sulfate-rich chlor-sulfate. The ratio of chlorine to sulfate is in the first instance a function of particle size. I t increases with particle size since the analytically determined maximum cxchange of chlorine (with excess sulfate) decreases with the growth of the particles.” The coagulating action of an alkali chloride is attributed to a driving back of the dissociation of the salt and a similar action is assumed to account for the precipitation with alkali nitrate. It will be noted that Pauli’s interpretation of the coagulating action as a chemical precipitation process, involves the use of the phrases “peculiar equilibrium between the complex bound and the ionic chlorine” and “peptizing action of the undischarged complex.” I t would be interesting to know just how Pauli visualizes the “peculiar equilibrium” relations and the “peptizing” action referred to. The general point of view of Pauli is accepted by Wintgen, Rabinowitsch, and others. Considering the micelles to be ordinary ions, Wintgen’ determines the amount of colloidal substance deposited by one Faraday of electricity and designates this the electrochemical equivalent or “equivalent aggregate” of the colloid. The number of electrochemical equivalents of colloid per liter is called the zorma1i:y of the colloid, N . This value is obtained by applying Kohlrausch’s law, 1000K, = ri(u v). K,, the specific conductance of the micelle is estimated from the conductance of the colloidal system before and after ultrafiltration; u the mobility of the colloidal particle is gotten from U-tube measurements; u the mobility of the anion is known, and N is calculated. The similarity between colloidal behavior and ordinary ionic reactions in solution is indicated. Thus colloidal chromic oxide is regarded as an amphoteric electrolyte which reacts with either acids or bases to give salts.3 I t is pointed out that equi-normal colloidal solutions containing oppositely charged particles should mutually precipitate each other. Whatever merits Wintgen’s formulation of the constitution of colloids and the mechanism of the coagulation processes may have, it should be pointed out that his work is vitiated by a methodical error. Laingl showed that the fraction of the current carried by any charged body whether ion, colloid, wall, or bubble is equal to the ratio of its actual conductivity to the total conductivity of the system. That is fraction of current = clfl/u,
+
Kolloidchem. Beihefte, 17, 291 (1923). Z. hysik. Chem., 103, 250 (1922).Wintgen and Bilta: 107, 403 i Wintgen: kolloid-Z., 40,300 (1926). Lowentgal: 109, 378 (1924); a Wintgen and Weisbecker: Z. physik. Chem., 135, 182 (1928). J. Phys. Chem., 28, 673 (1924).
Kintgen and
COAGULATION OF SOLS BY ELECTROLYTES
5
where ci is the concentration and fl is the conductivity of unit concentration and p is the sum total of all such cf terms for all constituents present. The bodily movement nl differs from the above by a factor ni which is the number of units to one electrical charge. Thus bodily movement = nl = clmlfl/p. l\lcBain’ points out that Wintgen and his pupils neglected this factor m. They should have divided their mobilities by ml to find the conductivity. A further error was made in interpreting U-tube experiments on cataphoresis. Both Lash Miller and G. N. Lewis showed that the law of conservation of matter requires that the “moving boundary” method must give results which are identical with those derived from the Hittorf method of quantitative analysis. These errors and the neglect of the Donnan equilibrium makes Wintgen’s numerical values for the charges on the particles something like ten times too large. If one insists on regarding ferric oxide sols merely as electrolytes with complex ions it must be emphasized that there is a fundamental difference between sols and non-colloidal complex electrolytes such as potassium ferrocyanide, the cobalt amines, the complex platinous salts, etc. formulated by Werner. There is also a distinct difference between a colloidal ferric oxide and such colloidal electrolytes as the soaps and Congo red in that the latter contains ionic micelles made up of groups of ions of definite composition and carrying one charge for each equivalent of the ion, whereas the micelle of the former has no definite composition and may carry hundreds or thousands of equivalents for each free charge. In conclusion, there seems no justification for assuming the absence of adsorption in a system such as colloidal ferric oxide where an adsorption equilibrium between minute solid particles of hydrous ferric oxide and the surrounding solution, must certainly exist. I t was believed that further light could be thrown on the structure of colloidal systems and the mechanism of coagulation by following the change in composition on adding a precipitating electrolyte stepwise to a sol. Since the precipitating ion is taken up in exchange with chloride in ferric oxide sol prepared from ferric chloride solution, Rabinowitsch and Kargan* followed the change in the chloride ion concentration potentiometrically on adding sulfate and other ions stepwise t o a measured portion of the sol. The procedure was as follows: X 2 0 cc portion of sol in which was some suspended calomel, was placed in a beaker containing mercury thus making one half of a calomel concentration cell. The other half was a saturated calomel electrode. The difference in potential was found with a potentiometer at the outset and shortly after each addition of a small amount of electrolyte. From the potential measurements, the concentration of chloride ion a t each step was calculated using the Nernst equation. Typical curves obtained by Rabinowitsch Colloid Symposium Monograph, 4, 14(1926). Chem., 133,203 (1928).
* Z. physik.
6
HARRY B. WEISER
for the titration of two different sols with N/z sodium sulfate are given in Fig. I . The straight line represents the concentration of sulfate added while the irregular curves represent the changing concentration of chloride. I t will be noted not only that the curves are irregular but that they run above the sulfate line indicating that considerably more chloride is displaced than there is sulfate added. Indeed special attention is called to the observation that
FIG.I
Titration curves of Fe203sols with K 2 S 0 4(from data by Rahinowitsch and Kargan) 0.05 cc S 1 2 Xa2S04displaced more than 4.6 times the equivalent of chloride in one sol. There is no apparent reason for this behavior and Kabinowitsch’s explanation of it is not satisfactory. Moreover, as will be discussed later on, Rabinowitsch and Kargan either do not give all the facts or there is some error in their calculations of chloride concentrations from the observed potentials. In any event, the general method seemed to offer a fruitful method of approach t o the problem of what takes place when an electrolyte is added to a sol. The results of investigations along this line are reported in the next section. Experimental Preparations of Sols. Three ferric oxide sols were used in the course of this investigation. The first was prepared by dissolving roo grams of pure ferric chloride in approximately 500 cc of water, filtering, and adding dilute ammonia just short of precipitation. The concentrated sol was diluted to 4 liters and dialyzed in four Neidle dialyzers using cellophane bags which held approximately one liter. The dialysis proceeded with a continuous change of water for two days when the chlorine content was found to be approximately
7
COAGULATION O F SOLS BY ELECTROLYTES
0.015 normal. The sol was set aside in a pyrex flask for four months before using. The second sol was prepared b y the same procedure outlined above except that the ageing following the dialysis was accomplished by heating for a week on the steam hot-plate. Loss by evaporating was prevented by attaching a reflux condenser to the mouth of the flask. This sol was used within a month after preparation. The third sol was formed by dissolving 7 5 grams of ferric chloride in 500 cc of water and adding slowly to 3500 cc of boiling water. After dialysis a t 90' for two days, the sol was set aside for three months before using. The sols were analyzed for the FerOa content and the chlorine content as follows: The ferric oxide in 2 5 cc portions of the sols was coagulated by ammonia and the precipitate was collected, washed, ignited and weighed in the usual manner. The chlorine content of 2 5 cc portions of sol was obtained by first adding a small excess of silver nitrate followed by adding nitric acid and heating. This process dissolved the oxide leaving a precipitate of silver chloride. After diluting, the silver salt was collected in a Gooch crucible, dried and weighed. The analyses are given in Table I.
TABLE I Analysis of Ferric Oxide Sols Sol number
I 11 I11
Fe20a grams per liter
c1
mols per liter
6.92
0.0158 o ,0166
4.08
0.0047
6.46
Method of Tztration. Preliminary experiments disclosed quite promptly that satisfactory data could not be obtained by adding the electrolyte t o the sol stepwise and measuring the change in chloride ion potentiometrically after a few minutes. The reason is obvious when one considers that it takes time to establish equilibrium a t the calomel electrode. Indeed it is always recommended that the newly prepared standard calomel half element be set aside at least a day and preferably several days before use. Accordingly the procedure adopted by Rabinowitsch and Kargan was discarded as unsatisfactory and the following method adopted: In general, the method consists in mixing separate portions of sol with gradually increasing amount of electrolyte and making the observations on the separate samples. In the sol was suspended a small amount of freshly precipitated calomel. This was prepared by the interaction of pure mercurous nitrate and hydrochloric acid followed by thorough washing with the aid of the centrifuge in the presence of a small amount of mercury, first with water and then with the sol. To secure rapid uniform mixing of the sol with electrolyte 2 0 cc of the former was placed in the outer compartment of an all-glass mixing apparatus and a definite amount of electrdyte diluted to 5 cc in the inner compartment. After mixing, the mixture was transferred to a small bottle which was placed in the thermostat
8
HARRY B . WEISER
a t 25’ and shaken a t intervals for a period of two days in order to saturate with calomel. I t was then transferred to an electrode vessel like that shown in Fig. 2 . This was prepared by rounding out the bottom of a 30 cc weighing bottle to prevent its breaking, followed by sealing in a small platinum wire. After placing mercury in the bottom of the vessel, the sol-electrolyte mixture was poured in and allowed to stand in the thermostat 2 4 hours before makingthe potentiometric measurements. Thus an interval of three days elapsed between the mixing of sol with electrolyte and the determination of the chloride ion content. Potentiometric -Weasztrements. The potentiometric measurements were made by means of a N / r o calomel electrode prepared in the usual way from highly purified potassium chloride, calomel, and mercury. Contact between the standard electrode vessel and the “unknown” was made directly by the aid of a glass tube filled with the K / I O KCl and drawn down to a fine capillary to minimize diffusion. A type K Leeds and Sorthrop potentiosot meter was used in conjunction with a Hartman and Braun moving coil galvanometer sensitive to less than 0 . I millivolt. Calculations. From the observed potential the molar concentration of chloride as potassium chloride was obtained by the use of the modified Kernst equation substituting activities for concentrations. At 25’the activity I coefficient of W I OKCl is 0.749’ and the equation takes FIG.z the form T = 0.0591 log 0.0794/01. The calculated value of CY is converted into molar concentration by dividing it by the corresponding activity coefficient as read off from a graph prepared from data given by Lewis and Randall.2 I n this connection, attention should be called to an apparent error in the calculations of Rabinowitsch and Kargan. For example, a potential of 0.1589volts a t 19’ using a saturated calomel electrode as a standard mas calculated to give a chloride ion concentration of 0.00794 mol per liter. The activity coefficient of such a solution is 0.933 and the activity is thus 0.00741. Substituting this value in the Nernst formula at 19’ and solving for the activity of the chloride in saturated KCI at 19’, one gets: 0.1589 = 0.0579 log
a K C l Sat.
0.0074
from which, a K C 1 Sat. = 4.I I . Now the difference between the potential of the saturated and N / I O calomel electrode a t 19’ is approximately 0.0874 volt.3 To give this potential 1 Lewis and Randall: “Thermodynamics,” 362 (1923). 2“Thermodynamics,” 344 (1923). Clark: “The Determination,of Hydrogen Ions,” zoo (1927). J
COAGULATION O F SOLS BY ELECTROLYTES
9
the activity of the chloride in the saturated half element is given by the equation 0.880 = 0.0579 log
a K C l Sat. ~
9.0794
from which C L K C ~sat = 2.63. Since the value for a which Rabinowitsch and Kargan must have used is more than 60 per cent higher than the above, their concentrations calculated from the observed potentials would appear to be erroneous. I n this connection Rabinowitsch and Kargan report the same value for the total chlorine in the sol by gravimetric analysis as silver chloride after dissolving the precipitate in nitric acid, and by potentiometric analysis of the supernatant solution after adding a precipitating electrolyte. Since ferric oxide sol prepared according to the procedure employed by Rabinowitsch and Kargan contains chlorine that is not displaced by a coagulating electrolyte, the two methods should not yield the same results. That they did is most likely the result of some compensating error in the potentiometric procedure.
Observations on Ferric Oxide Sol I Tztration with K2S04. After a number of preliminary experiments which led to the procedure described above, the following observations were made of the change in chloride concentration on adding varying amounts of N/2 j K2S04to the sol. Above the precipitation value of the salt, the suspended calomel was carried down completely by the agglomerating oxide. I n order to make certain that the supernatant solution was saturated with HgzClz a small amount of calomel paste was added after the coagulation had taken place. The results of the titration are given in Table I1 and shown graphically in Fig. 3. TABLE 11 Titration of Fe203Sol I with KZSO4 Cc of N/z5 K 2 S O I added to 20 cc of sol Total volume 25 cc.
A
volts
a Cl-
x
IC11
lC11 displaced
[Cl] equivalent to [SO,] added
103
X 10s
2.82 3.08
2.93
0
0
0.j
0.0857 0.0834
3.20
0.27
I .o
0.0800
0.80 I .60
0.0768 0.0746
2.5 2 .o
0.0725
4.34 4.71
0.0707
5 .os
3.5
0.0686 0.0674 0.0616
5.48
5.75
3.68 4.20 4.57 4.96 5.34 5.81 6.10
0.75
1.5 2 .o
3,52 4 .oo
7.11
7 .60
0.0608
7.43 7.91 8.22
7.95 8.47 8.82
0.0
4.0
4.5 4.75 5 .oo 2.5
of N / I o
0.0592 0.0582
x
103
I .27
I
.64
2.03
2.41 2.88 3.17 4.67 5.02
5.54 5 .89
x
2
103
.40
3.20
4.00 4.80
5.60 6.40 7.20
7.60 8 .oo 10.00
IO
HARRY B . WEISER
Referring to the data and the accompanying curve, it is obvious that the chloride displaced during the titration follows a regular course. Particularly significant is the fact that the amount of chloride displaced is always appreciably less than the amount of sulfate added, contrary to the observation of Rabinowitsch and Kargan. Since the sol was prepared by essentially the
FIG.3 Titration curve of Fe203Sol I with K2SOa
same procedure and was similar in oxide and chloride content to the one used by them, it seems probable that their experimental method led to erroneous observations. Attention has already been called to possible errors in their calculations. I t is of interest to note that the curve for chloride displaced followsa nearly linear course until the precipitation concentration is approached when it increases rapidly. Above the precipitation concentration, indicated by a vertical line cutting the curve, the chloride displaced follows the usual adsorption type of curve. The precipitation concentration 4.5 cc of K ( ’ 2 j KzSO, in z j cc was the smallest amount of electrolyte which would just cause complete coagulation in three days. At the precipitation concentration, the supernatant solution gave no test for sulfate with barium chloride even after several days. Since the “sulfate added” comes quite close to the “chloride” curve a t the precipitation value, it is clear that the amount of sulfate adsorbed is approximately equivalent to the chloride in the supernatant solution, as Pauli has demonstrated. Since all the added sulfate is taken up at the precipitation value, it is reasonable to conclude that all of it is taken up below the precipitation value. It is of particular importance to note that at the precipitation value as well
COAGULATION OF SOLS BY ELECTROLYTES
I1
as a t all points above and below the precipitation value, the chloride equivalent to the sulfate added is derived in part from the chloride originally present in the intermicellar solution and in part is displaced from the micelles. This behavior will be considered in detail in a later section. The marked increase in the chloride displaced as the precipitation concentration is approached is undoubtedly the result of partial agglomeration of
FIG.4 Titration curve FegOa Sol I with potassium citrate
particles with the consequent decrease in specific surface as the charge approaches the critical value below which complete flocculation takes place. Finally, the form of the upper portion of the “chloride” curve represents the adsorption of sulfate by precipitated ferric oxide in concentrations above the precipitation value, since the sulfate taken up is approximately equivalent to the chloride in solution. Attention has been called to this behavior in earlier communications.’ Titration with X3CsH30;. Experiments with X/z5 potassium citrate were carried out by the same procedure used for potassium sulfate. The data are recorded in Table I1 and shown graphically in Fig. 4. I t is evident that the displacement of chloride by citrate follows a course strikingly similar to that for sulfate. The only difference is the earlier upward turn in the curve and the lower precipitation concentration of the salt as indicated in the figure. The maximum amount of chloride displaced above the precipitation value is almost identical in the two cases. Titration with K$e(C-V)s. Since K3Fe(C?J)swhich is the salt of a strong acid might be expected to give largely trivalent anions, it was of interest to compare the behavior of the sol during the stepwise addition of this salt Weiser: J. Phys. Chem., 25, 299 (1921); Weiser and Middleton: 24, 30, 630 (1923).
HARRY B. WEISER
I2
TABLE I11 Titration of FeaOs Sol I with Potassium Citrate [CI] equivalent CC O f N / z ~KjCliH3CI T ci c1IC11 IC11 added to 20 cc of sol. volts x 103 X 103 disnlaced to ICoHjO; j:
Total volume 2 5 cc. 0.0
0.0857
0.j
0
2.82
,0823
3.22
.o
0.080;
1.5
0,0777
.o
0.0;58
3.45 3.85 4.16 4.44 4.75 j 36 5 75 6.19 6 66 7.49 7.9'
I
2
2 . j
3 .o 3.5 4.0 4.25 4.50 4.00'
7.50'
o.oj40 0.0723 0.0692 0.0674
0.0655 0.0636 0.0606 o .oj92
2.93 3.36 3.61 4 .O-l
103
added X
0.0
0.0
0.13 0.68
0.80
1.11
2.40
I
.60
3.20 4.00
4.38 4.68
I
45
I
.75
,j . O I
2.08
4.80
5.69 6.10
j .60 6.40 6.80
8.OI
2.76 3.17 3.65 4.10 j .08
8.48
5.5;
10.91
6.58 7.10
103
.20
8 .oo
Sol repeptized, with negative charge.
with that of citrate which doubtless gives some trivalent anions but divalent and monovalant anions as well. The titration data are given in Table IT.-and the curve in Fig. j . The potentiometric measurements for chloride are very satisfactory so long as practically all of the added salt is adsorbed, but the measurements are uncertain as soon as a n appreciable amount of ferricyanide remains in the supernatant solution. The curve for displacement of chloride by ferricyanide is characterized by the earlier upward bend and the distinctly lower precipitation value a s TABLE
IT
Titration of Fe203Sol I with K3Fc(CN)B Cc of N / 2 5 K3Fe(CS)6 T added t o 2 0 cc of sol. volts Total volume 2 5 cc.
1.5
0.0857 0.0828 0.0801 0,0777
2 .o
0.0742
2 . 5
0,0717
0.0
0.5 I
.o
3 .o 3.25 3 .so 4.00 5 .oo
0.0693 o ,0674 0.0655 0.0630
Uncertain
CI
x
C1103
2.82 3.17 3 .so 3.85 4.41 4.86 5.34 5.75 6.19 6.82
vi1
X lo3
2.93 3.31 3.66
UI x 103
displaced
Ci] equivalent t o [Fe(CS)6] added X103
0.0
0.0
0.38 0 73
0.80 I .60
4.04
I .I1
2
4.65 5.13 j .66 6 . IO 6.58 7.27
I .72
3.20
2.20
4.00
2.73 3.17 3.65 4.34
4.80
40
j . 2 0
5.60 6.40
COAGCLATION O F SOLS BY ELECTROLYTES
I3
compared with both sulfate and citrate. Since all the ferricyanide and sulfate are carried down at the respective precipitation values and since the precipitation value in equivalents of the former is appreciably lower than that of the later, it is obvious that Freundlich's assumption that equivalent amounts of all ions are carried down a t the precipitation value cannot be right. My later assumption that the adsorption of equivalent amounts is
CC. N/25
& Fe (CN),
Added
FIG.5 Titrrttion curve of F e 2 0 3Sol I with K3Fe(CN)b
necessary to lower the charge to the coagulation point, likewise isnot generally true. The total chloride in the intermicellar liquid near the precipitation value and after coagulation, appears somewhat greater than the ferricyanide added. This is because coagulation with consequent decrease in specific surface starts a t very low concentrations and so the displaced chloride is relatively greater for ferricyanide than for other salts a t the same concentration. At the same time, it should be noted that there is appreciably less chloride displaced and total chloride in the supernatant solution a t the lower precipitation value of K3Fe(CS)6than a t the higher value for K2S04. Titratzon wzth Kh'Oa. Since univalent anions in general precipitate positive sols only in relatively high concentration, it was necessary to use a normal solution of KN03 to effect the coagulation within 5 cc of electrolyte in 2 5 cc of mixture. I n the interest of accuracy N/IO solutions were used for lower concentrations. The titration data given in Table V are plotted in Fig. 6. The form of the curve is distinctly different from that for salts with multivalent precipitating ions. The amount of chloride displaced for the same concentration is very much less for the univalent ion. Moreover the
I4
HARRY B. WEISER
C c . N KNO,
Added
FIG.6 Titration curve of FerOa Sol I with KKOs
amount of chloride displaced goes up fairly rapidly a t relatively low concentrations but soon attains a practically constant value or even decreases slightly. The latter observation may be the result of a salt error introduced with relatively high concentrations of KKOS. I n any event, the maximum amount of chloride displaced is appreciably less for nitrate than for the multivalent ions. The variation in the chloride displacing power of equivalent amounts of potassium salts of mono-, di-, and trivalent anions is clearly indicated in the composite diagram Fig. 7 .
FIG.7 Tit,ration curves of FepOsSol I with different types of electrolytes
COAGULATION OF SOLS BY ELECTROLYTES
15
TABLE V Titration of Fez03Sol I with KNO, Cc of K X 0 3 added to 2 0 cc of Sol. Total volume 2 5 cc.
7
volts
I
0.0857 0.0812
2
0.0785
0
of N / I O of N / I O 3 of S / I O 5 of X / I O I of N 2 of N 3 of N 4 of N 5 of x
a C1-
x
103
0.0718
2.82 3.37 3.73 3.92 4.27 4.66 4.81 4.84
0.0722
4.76
0.0772
0.0750
o ,0728 0.0720
ICll X
103
[Cl] [Cl] equivalent displaced to [No3]added
x
103
x
102
2.93 3.53 3.91 4.12
0.0
0.60 0.98 1.19
4.0 8 .o I 2 .o
4.49 4.92 5.08
1.53 1.99
40.0
2 .I5
80 .o
5.11
2.18
120.0
5.03
2 . IO
0.0
20.0
160.0 200.0
Experiments with F e m c Oxide Sol I1 Sol I1 was prepared in much the same way as Sol I but was aged by heating instead of by long standing. The results with the former merely confirm and extend those with the latter. Titration with KzSU4. The data for the titration with potassium sulfate is given in Table VI and shown graphically in Fig. 8. As compared with Sol I, it will be noted that the curve starts to bend upward a t a lower concentration and the precipitation value is somewhat lower. As was to be expected the general form of the curve is the same as with Sol I.
TABLE VI Titration of Fez03 Sol I1 with X2SOd Cc of K2S04added to 20 cc. of Sol. Total volume 25 cc.
a
volts
2 . j
0.0910 0.0880 0.0843 0.0804 0.0778 0.0748
3.0
0.0727
3.5 4.0
0.0700
0.0
0.5 I
.o
1.5 2 .O
4.25
4.50 5 .oo
6.25
0.0675 o ,0647 o ,0626 o ,0605 0.0597
a C1-
X io3
[Cll
X
103
2.29 2.58
2.37 2.69
2.98 3.48 3 .E3 4.31 4.67 5.19 5.73 6.38 6.93 7.52
3.11 3.64 4.02 4.54 4.93 5.49
7.76
6.07 6.68 7.40 8.04 8.31
IC11 [Cl] equivalent displaced to [SO,] added
x
103
X
103
0.0
0.0
0.32
0.80 I .60 2.40 3.20 4.00 4.80 j .60 6.40 6.80
0.74 I .27
.65 I7 2.56 3.12 3.70 4.31 5.03 5.67 5.94 I
2 .
j .20
8 .oo 10.00
16
HARRY B. WEISER
I n order to test the accuracy of the experimental procedure for determining chloride potentiometrically, an analysis was made of the supernatant solution after precipitation. This was done by precipitating 40 cc of sol with I O cc of Niz5 Ei~S04and analyzing an aliquot part of the supernatant solution for chloride by titrating with silver nitrate. The silver nitrate was standardized against S / j o silver chloride and the titrations were made to the same endpoint. The results given in Table VI1 show the potentiometric procedure to be quite accurate.
Cc. N/Z5
K,SO,
Added
FIG.8 Titration curve of Fe203Sol I1 with K2S04
TABLE VI1 Analysis of Supernatant Solution for Chloride after Precipitation Fe203Sol I1 Substances mixed 0.001934 N A g S O I [CI] x 1 0 3 Fe203Sol cc
K/25 K z S 0 4
to titrate z j cc cc
CC
volumetric method
potentiometric method
8.10 8.04
40
IO
10.48
40
IO
10.qj
8.16 8 .os
Fez03 Sol 40
N/z j KzCrOl
40
10
10.80 10.80
8.40 8.40
IO
Titration with K2C204. The data and corresponding curves with K2C204 as precipitating electrolyte are given in Table S’III and Fig. 9, respectively. The upper portion of the curve with K2C204runs slightly above that with K2S04 and the precipitation concentration of the former is slightly lower than the latter. As in the case of sulfate, all of the oxalate is adsorbed a t the precipitation value.
COAGULATION O F SOLS BY ELECTROLYTES
I7
FIG.g Titration curve of Fez03 Sol I1 with KzC104
TABLE VI11 Titration of Fez03Sol I1 with KzCz04 Cc of N/25 K Z C 2 0 4 c added to 20 cc of Sol. volts Total volume 2 5 cc. 0.0
0.5
I .o
1.5 2
.o
2.5
3 .o 3.5 4.0 4.25 4' 50 5 .oo 6.25
0.0910 0.0875 0.0840 0.0815 0.0778 0.0758 0.0730
0.0697 0.0660 0.0643 0.0623 0.0595 0.0583
a C1-
x
102
2.29 2.66 3 .OI 3.32 3.83 4.15 4.62 5.25 6.07 6.48 7 .OI j .81 8.22
IC11 X
103
2.37 2.77 3.14 3.47 4.02 4.36 4.87 5.56 6.45 6.90 7.47 8.36 8.82
[CI] displaced
x
103
[Cl] equivalent to [ C 2 0 r added ] X 103
0.0
0.0
0.40 0,77 1.10
0.80 I .60 2.40
I .65
3.20
1.99
4.00 4.80 5.60 6.40 6.80
2 .SO
3.19 4.08 4.53 5.10
7.20
5.99 6.45
8 .oo 10.00
Titration with KzCrOa. The data using K2Cr04as precipitating electrolyte and the corresponding curve are given in Table IX and Fig. IO. The curve parallels that obtained with K2C2O4 almost throughout the entire range and the observed precipitation concentration lies between the values for oxalate and sulfate. A11 of the chromate is carried down a t the precipitation value. Satisfactory potentiometric measurements were not obtained with appreciable concentrations of K2Cr04in the supernatant solution above
r8
HARRY B. 'U'EISER
FIG. I O Titration curve of Fe203 Sol I1 with KICrO,
the precipitation value. The point for 5 cc X 2 5 K2Cr04 with 2 0 cc of sol was obtained by volumetric analysis of the supernatant solution with the results as given in the second part of Table T'II. The value 8.40 X IO-^ mols per liter corresponds with 8.36 X IO-^ obtained potentiometrically with oxalate as precipitating electrolyte.
CC.
N/25 K, Fe (CN), A d d e d
FIG. 1 1 Titration curve of Fe209Sol I1 with KaFe(CN)a
19
COAGULATION O F SOLS BY ELECTROLYTES
TABLE IX Titrations of FezOs Sol I1 with K2Cr04 Cc of N/25 K*CrOl added to 20 cc. Sol. Total volume 2 5 cc. 0.0 0.5
I
.o
1.5 2
.o
2.5 3 .o 3.5 4.0 4.25 5 .oo
T
volts
0.0910 0.0875 0.0845 0.0812 0.0175 0,0746
a
x
c1103
0.0686 0.0658 0.0643
2.29 2.66 2.95 3.35 3.87 4,34 4.71 5.49 6.12 6.48
-
-
0.0725
[C11
x
108
[ClI displaced
x
103
[Cl] e uivalent to [Cr8,] added X
103
2.37
0.0
0.0
2.77
0.40 0.71 1.13
0.80 I .60 2.40 3.20 4.00 4.80 5.60 6.40 6.80 8 .oo
3.08 3.50 4 .ai 4.55 4,97 j .81 6.50 6.90 8.40~
I .70
2.18 2.60 3.44 4.13 4.53 6.03
Determined analytically, (See Table VII).
Titration with K3Fe( C,V)6. The observations with K3Fe(CN)Gas precipitating electrolyte merely confirm those obtained with Sol I. The data are given in Table X and shown graphically in Fig. I I .
TABLE X Titrations of Fe203Sol I1 with K3Fe(CN)e [CI] equivalent a c1IC11 IC11 X 103 X 103 displaced t o [Fe(C”)s] x 1 0 3 added X 1 0 3
Cc of N/25 K3Fe(CN)s T added to 20 cc. of Sol. volts Total volume 2 5 cc. 0.0 0.0910 0.5 0.0870 I
.o
1
.s
2
.o
2 . 5
3 .o 3.25 3.50 4.00 5 .oo
0.0842 0.0815 0.0778 0.0750 0.0712
0.0692 o ,0662 0.0637
Uncertain
2 .29 2.68 2.99 3.32 3.83 4.27 4.96 5.36 6.02 6.65
2.37 2.79 3.12 3.47 4.02
0.0
0.0
0.42 0.75
0.80 I .60 2.40 3.20 4 .oo 4.80 5.20 5.60
1.10
45 ’. 25 40
.65 .13 2.87
5.68 6.40 7.09
3.31 4.03 4.72
I
2
6.40
8.00
The results of the observations on Sol I1 are collected in Fig. 12. It is of interest to note the marked similarity in the behavior of the salts with divalent anions. The precipitation concentrations are quite close together and the chloride concentrations of the respective supernatant solutions are almost identical. Since the adsorption of the multivalent anions observed is complete a t the precipitation value, Freundlich’s assumption that equivalent amounts of ions are adsorbed a t the precipitation concentration holds almost
HARRY B. WEISER
20
I
I
cc.
N/25 € l e c t r o t y t e
Added
FIG.IZ Titration curves of FeXO, Sol I1 with different types of electrolytes
quantitatively for the three ions under consideration. But when one turns to trivalent ferricyanide, one encounters a distinctly lower equivalent precipitation value and correspondingly lower adsorption at the precipitation value, confirming the observation with Sol I.
Observations with Ferric Oxide Sol I11 Sol I11 prepared by hydrolysis of FeC13 in hot water and dialyzed in the hot was much freer from chloride than either Sol I or 11. Moreover, since no ammonia was added, the sol contained simply the hydrolysis products of FeCL, hydrous FeaOaand HCl. That its behavior on titrating with K&04
$
LYY
2
*
4
e
n 147
0 4
X or+'=:
u
0
Cc.
N/50 K2S0, Added
FIG.13 Titration curves of Fe203 Sol 111 with KXSOa
-
COAGULATION O F SOLS BY ELECTROLYTES
21
is similar in all respects to that observed with less pure sols obtained by somewhat different procedures, is shown by the data given in Table XI and plotted in Fig. 13. TABLE XI Titration of FelOI Sol 111 with K 8 0 4 Cc of 3 / 5 0 K2S04 a added to 20 cc. of Sol. volts Total volume 2 5 cc.
a
c1-
x
103
[Cll X io3
[Cl] [Cl] equivalent displaced to [SO,] added X io3 X io3 0.0 0 .o
0.0
0.1220 0.1154
0.69 0.91
0.70
0.5
0.93
0.23
I .o
0.1097
1.11
1.14
0.44
0.40 0.80
.51 .71
0.81
I .20
.47 .66
1.5
0.1025
2 .O
0,0994
I I
2.5
0.0926
2.15
I I
.60
I .OI
I
2.23
1.53
2 .oo
3.0
0.0925
2.16
2.24
1.54
2.40
4.0 5 .o
0 0924
2.17
2.25
1.55
3.20
0.0922
2.19
2.27
1.57
4.00
Discussion of Results A theory of the mechanism of the coagulation of colloidal ferric oxide must explain the following facts reported in the preceding section: I. At the precipitation value of potassium salts of multivalent ions the chloride in the supernatant solution is equivalent to or but little greater than the amount of added electrolyte. 2. Only a part of the chloride that is found in the supernatant solutions after precipitation, can be detected potentiometrically in the original sol before adding electrolyte. Not all of the chloride in the micelles is displaced by a large excess of precipitating electrolyte. 3 , The chloride measured potentiometrically following the stepwise addition of electrolyte, consists of the chloride in the sol originally, together with an additional amount that is displaced when the added anion is taken up. 4. The multivalent ions investigated are taken up practically completely by the sol particles in concentrations up to and including the precipitation concentration. The chloride displaced so that it can be detected potentiometrically, is less than half the amount equivalent to the multivalent ion taken up. 5 . The amount of chloride displaced follows nearly a linear course a t the outset of the stepwise addition of multivalent ions but becomes proportionately greater as the precipitation concentration is approached, 6. The chloride displacement curves for multivalent ions of varying valence follow an almost identical course until the precipitation concentration is approached when there is a marked divergence for ions of different valence. 7. The three salts of divalent anions exhibit a strikingly similar behavior as regards the entire course of the chloride displacement curves and the precipitating power.
22
HARRY B. WEISER
8. The trivalent ferricyanide coagulates at a distinctly lower concentration than the divalent ions and the chloride displaced at the precipitation value is proportionately less with the former than with the latter. 9. The chloride displacement curve with potassium salts of univalent ions such as nitrate, follows a course distinctly lower than that for the multivalent ions. The solubility theory of Pauli would attempt to explain all of these facts on the basis of metathetical reactions with the formation of insoluble salts of the added electrolytes. Even if one disregards the objections to the theory which have been referred to already, the Pauli mechanism appears inadequate to account for the observations. A few cases may be mentioned. First of all, it seems altogether improbable that the alleged complex salts of such widely varying anions as sulfate, oxalate, and chromate should all possess practically the same solubility and so precipitate at the same concentration of added electrolyte. ?Tor is there any reason for supposing that the solubility of the alleged complex trivalent ferricyanide and citrate should be appreciably less than that of the divalent complexes. Moreover, it is difficult to explain on the basis of solubility relationships, why potassium oxalate and potassium ferricyanide, say, which must be assumed to form salts of widely varying solubility, should give chloride displacement curves which follow an almost identical course until the precipitation concentration of the trivalent ion is approached. Finally, I do not think that the time has yet arrived when anything can be gained by applying the term salt to a complex formed when a positively charged particle of variable composition takes up a given negative ion in varying amounts depending upon the conditions of preparation of the particle. An adsorption mechanism which will now be outlined appears to offer a more rational interpretation of the facts: When ferric chloride is hydrolyzed there are formed positively charged micelles which vary in size, composition, and charge, depending on the conditions of the hydrolysis. I n general, the colloidal particle exclusive of the outer layer has a composition represented by some point in the three-component diagram, Fe203-HC1-Hz0which may be represented symbolically [xFez03.yHC1.zH20]. Thls indicates the observed fact that there is chloride within the micelle which is not displaced even by high concentrations of precipitating electrolytes. It also indicates what is well known, that the actual composition of the particle is determined by its method of formation and the subsequent history. Assume for the sake of simplicity that the hydrolysis has taken place to the point where the electrolyte active in sol formation is HCl and not FeCL Xow the finely divided solid particles of xFe2O3.yHCl.zHz0exhibit such a marked tendency to adsorb hydrogen ion as compared with chloride that the sols prepared as described above are almost neutral having a pH value of 6 to 6. j whereas an appreciable concentration of chloride ion is observed potentiometrically. If n represents the number of hydrogen ions adsorbed by the micelle and y equals the chloride concentration of the sol that can be measured potentiometrically, the composition of the sol may be formulated as follows:
COAGULATION OF SOLS BY ELECTROLYTES
23
+
[xFe203.yHC1.zH20]Hi.n-qC1qC1-. This is represented diagrammatically in Fig. 14in which the inner layer of the double layer surrounding the particle is composed of the adsorbed hydrogen ions. The outer portion of the double layer is a diffuse layer' consisting of an equivalent amount of chloride ions as shown. Some of the chloride ions because of their relatively higher kinetic energy exert sufficient osmotic repulsive force against the electrical attraction CC
c 1-
Cf
c 1FIG.14 Diagrammatic representation of the structure of the micelle in a ferric oxide sol.
of the Hf ion layer, to influence the calomel electrode and are therefore detected potentiometrically. Such ions are represented in the diagram beyond the dotted circle. When potassium sulfate is added to the sol, the more strongly adsorbed divalent sulfate ion forces itself into the double layer closer to the adsorbed hydrogen ion and displaces chloride as indicated diagrammatically in Fig. I 5 . Potentiometric analysis enables one to determine the chloride which has been displaced. The difference between the chloride in the sol originally and that after the addition of sulfate, in other words, the displaced chloride is not equivalent to the adsorbed sulfate since a part of the sulfate which enters the layer corresponds to chloride measurable potentiometrically in the original sol. The adsorbed sulfate lowers the charge on the particle in the following way: A sulfate ion possesses the same average kinetic energy as a chloride ion but it possesses double the charge. Accordingly, if one assumes for the moment G o u ~ :J. Phys., (4) 9, 457
(1910).
HARRY B. WEIS'ER
24
that the valence only determines the adsorbability, the divalent particles in the outer layer would be drawn closer to the inner layer and the thickness of the double layer would be decreased. Since the potential difference between two layers of opposite sign with constant charge density is directly proportional to the distance between them, it follows that the reduced thickness of the layer will be accompanied by a decrease in charge on the particle.
c 1CT
c 1-
Cf
CI'
c1-
FIQ.15 Diagrammatic representation of the structure of the micelle in a ferric oxide sol after adding some K1S04 Since the three divalent ions sulfate, oxalate, and chromate exhibit such a strikingly similar behavior in displacing chloride and in reducing the charge to the critical coagulation value, it follows that in the case of these three ions the valence is the most important factor determining the adsorbability. The behavior of trivalent ions such as ferricyanide would follow from what has been said. The ions having the same average kinetic energy but with three charges will be drawn closer to the inner layer than the divalent ions and the further reduction in thickness of the double layer manifests itself in a lower precipitation value. The chloride ion displaced is obviously less since less ferricyanide needs t o be adsorbed to lower the charge on the particles to the critical coagulation value, The adsorption of both the divalent and trivalent ions is sufficiently great that the amount added to cause coagulation is practically completely adsorbed. The adsorption a t this concentration is not completely irreversible
COAGULATION OF SOLS BY ELECTROLYTES
25
however, since shaking the precipitated gel with water results in partial repeptization of the sol owing to some of the precipitating ions breaking away from the close adsorption layer. The increase in the chloride displaced for a given increment in the multivalent ion added in the region of the precipitation value, is the result of agglomeration and partial coalescence of the colloidal particles into micelles having a lower specific surface for a given mass. The behavior of nitrate ion as compared with multivalent ions might be inferred from what has gone before. A much higher concentration of the univalent ions is necessary to produce the same effect as the multivalent ions since the latter are attracted so much more strongly toward the adsorbed hydrogen ion and so would be expected to effect the necessary lowering of charge in much lower concentration. The smaller amount of chloride displaced by nitrate as compared with the multivalent ions is not surprising in view of the fact that chloride is quite as effective as nitrate in lowering the charge. There is some chloride displaced because the higher concentration of nitrate causes some of the latter to enter the double layer and force out an equivalent amount of chloride.
Summary The following is a brief summary of the results of this investigation. A procedure is described for accurate potentiometric determination I. of the change in chloride concentration on adding electrolytes stepwise to hydrous oxide sols containing a slight excess of hydrochloric acid or ferric chloride as stabilizing electrolyte. 2. Only a part of the chloride that is found in the supernatant solution after coagulation can be detected potentiometrically in the original sol before idding electrolyte. The chloride measured potentiometrically following the stepwise addition of electrolyte, consists of the chloride in the sol originally together with an additional amount that is displaced when the added anion is taken up. 3 . Titration curves are given which show the increase in chloride ion concentration on adding potassium sulfate, chromate, oxalate, ferricyanide, and nitrate stepwise to different ferric oxide sols. 4. The multivalent anions are taken up practically completely by the sol particles in concentrations up to and including the precipitation concentration. The chloride displaced so that it can be detected potentiometrically is less than half the amount equivalent to the multivalent ion taken up. At the precipitation value, the chloride in the supernatant solution is equivalent to or but little greater than the amount of multivalent ion added. 5 . The titration curves follow a nearly linear course a t the outset of the stepwise addition of multivalent anions but the amount of chloride displaced for a given increment of precipitating ion is relatively greater as the precipitation concentration is approached. Above the precipitation concentration, the curves take the form of an adsorption isotherm.
HARRY B. WEISER
26
6 . For simple multivalent ions of the same valence, the titration curves are strikingly similar. For multivalent ions of varying valence there is a marked divergence. Thus trivalent ferricyanide coagulates a t a distinctly lower concentration than divalent sulfate and the chloride displaced a t the precipitation value is proportionately less with the former than with the latter. The titration curve with a univalent precipitating ion such as nitrate follows a course distinctly lower than that for the multivalent ions. 7 . I n ferric oxide sols such as those under consideration the composition of the micelle exclusive of the outer layer is given by some point in the three component diagram Fe2O3-HC1-H20which may be represented symbolically as xFe,O,.yHCl.eHIO. This indicates the observed facts that there is chloride within the micelle which is not displaced by electrolytes and that the composition varies with the conditions of preparation and the subsequent history of the sample. 8. The outer capsule of the micelle which largely determines its colloidal properties consists of an ionic double layer. The inner portion is adsorbed hydrogen or ferric ions; the outer portion is a diffuse layer consisting of chloride ions most of which are held by the electrical attraction of the adsorbed positive layer while others because of relatively higher kinetic energy exert sufficient osmotic repulsive force against the inner layer, to influence the calomel electrode and are therefore detected potentiometrically. 9. An adsorption mechanism is outlined to account for the change in composition and nature of the double layer which results in a decrease in charge on the micelle when electrolytes are added to the sol. At the same time, the proposed mechanism accounts for the form of the chloride displacement curve with different electrolytes. IO. The mechanism outlined accounts for the observed fact that relatively less of a trivalent ion must be adsorbed than of a divalent ion in order to lower the charge on a particle to the critical coagulation potential. 11. The relative merits of the proposed adsorption mechanism of the coagulation process and the solubility theory of Pauli are discussed. The Rice Institute, Houston, Texas.