^^^^^HH^IHVM^^^I Q D I UC-NRLF ; 1 & 1 $B 35 200 EXCHANGE H Ferric Oxide Hydrosol Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in the Faculty of Pure Science, Columbia University. By ALEXANDER FRIEDEN, B.S.,M.S..M.A. NEW YORK CITY 1922 Ferric Oxide Hydrosol Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in the Faculty of Pure Science, Columbia University. By ALEXANDER FRIEDEN, B.S.,M.S.,M.A. NEW YORK CITY 1922 ACKNOWLEDGMENT The author wishes to express his sincere appreciation for the constant advice and guidance of Professor A. W. Thomas, at whose suggestion this research was undertaken. KXCMAMCHC HISTORICAL The history of ferric oxide hydrosol may be divided into three periods: (1) 1827 to 1855, (2) 1855 to 1909, and (3) 1909 to the present time. The reports by Berzelius, 1 Maus, 2 Souberain, 3 Rose, 4 Schon- bein, 5 and Schorer, 6 during the first period, showed that they regarded colloidal ferric oxide as a definite chemical compound of iron. The second period starts with the discovery of Pean de St. Gilles 7 that when ferric acetate solution was heated on a water bath for four or five hours, the red solution became turbid by reflected, while clear to transmitted light. A trace of alkali or sulfuric acid precipitated all of it in the form of a red brown residue, insoluble in concentrated, but soluble in dilute acid. The acid precipitate when dried on a porus plate, turned dark brown. In this condition it was still readily soluble in water but it lost this property after being thoroughly dried. Scherer Kastner 8 prepared a similar colloidal ferric oxide by plunging test tubes containing dilute solutions of ferric nitrate into boiling water, and Debray 9 obtained a similar hydrosol upon heating dilute solutions of ferric chloride. Wiedemann 10 showed that any solution of a ferric salt contains some colloidal oxide in it ; and recently Wagner 11 showed that a 1-400 equivalent normal solution of ferric chloride hydrolyzes completely to colloidal ferric oxide and hydrochloric acid in 40 minutes at 25. Collodial ferric oxide was regarded by these early scientific workers as a modification of ferric hydroxide, but all the early attempts to determine the nature and composition of the sub- stance were doomed to failure because in all cases the substance formed contained not only the products but all the reagents in- volved, as well. This difficulty was removed by the introduction of the process of dialysis by Thomas Graham. 12 Thomas Graham prepared his ferric oxide "solution" by disolving freshly precipitated ferric hydroxide in a solution of ferric chloride. This solution, containing) from 4 to 5 percent of solid matter, was dialyzed through a parchment membrane for 11 days against distilled water. Analysis of it showed that the sol consisted of one equivalent of hydrochloric acid to 30.J) equivalents of ferric oxide, the hydrochloric acid being regarded as an impurity by Graham. (In this connection it is interesting to note that Berzelius 1 prepared colloidal ferric oxide by the method that Graham used, but called it "basic ferric chloride".) 3 That an iron oxide hydrosol prepared from ferric chloride con- tains chloride was also shown by Magnier de la Source 13 and by Wyrouboff. 14 Hantz and Desch 15 dialyzed Graham's sol to the absence of chloride iron as tested by the addition of silver nitrate to the so! and then, upon quantitative analysis of the sol found that it con- tained a great deal of chloride. The analysis showed that there was one molecule of ferric chloride to -every 18 molecules of ferric hydroxide. They explained the failure to get the qualitative test for chloride ion as due to the formation of a "chlorine complex". Ruer 16 made the same observations, but he attributed the failure of silver nitrate to show the presence of chloride to a "protective" action of the sol, an "explanation" which unfortunately is fre- quently used in colloidal chemistry. Linder and Picton 17 obtained results which agreed with those of Hantz and D-esch and concluded that, "since on dialysis of a solution of ferric hydroxide and ferric chloride for sixty-one days, hydrochloric acid could still be detected; and on another occasion, decided traces of chlorine could be detected in the dialysis product its-elf after 210 days, the substance present here is a hydroxychloride and not a hydrate associated with ferric chloride or free hydrochloric acid." Studying the composition of the substances formed by heat- ing dilute solutions of ferric chloride, Krecke 18 concluded that dep-ending on the concentration of the salt and on the temperature, the solution would undergo the following changes : first, the iron oxide of Graham will be formed, then that of Pean de St. Gilles, then the oxychloride, and finally, ordinary ferric oxide. The third period in the history of ferric oxide hydrosol began with the enunciation of the "Complex Theory" of colloids by P. P. von Weimarn. 19 He and others applied only to the common metals but it has been recently shown by Beans and Eastlack 20 that it applies to the nobk metals as well. In connection with ferric oxide hydrosol, this idea was first taken up by Malfitano. 21 Reasoning from the freezing point depression, he concluded that because of the low depression much lower than would be observed if the chloride were in the ionic condition the complex must exist in. the form somewhat as follows: [(FeO 3 H 3 )n FeCl 2 ]Cl, [(FeO 3 H 3 )nFeCl]Cl 2 , { [(FeO 3 H 3 )nFeO 2 H 2 Cl]mFe } C1 3 Neidle 22 attempted to determine the purity, or, rather, the amount of electrolyte present in a sol by obtaining a number of precipitation values with sulfuric acid, and, plotting, these values against the hydrochloric acid contents as determined analytically, he extrapolated the maximum purity of that particular sol. The assumptions were made 1 that an equivalent of 'sulfuric acid displaces an equivalent of hydrochloric and that this is con- tinued on extrapolation. He concluded from such results that there is a series of "oxychlorides," that when a colloid is dialyzed to a certain point, a gel will be formed and a sol of less purity will remain. This will, on further dialysis, precipitate another gel with the formation of another sol, etc., thus under- going a series of changes. He considers that the first colloid has a ratio of 21 Fe-3 to Cl, and cites as further evidence the fact that a solution of ferrous sulfate oxidized in air in presence of metallic platinum forms a precipitate at about that ratio. The idea of the formation of a chain of compounds is not new. Von Weimarn 23 uses a stepladder as illustration of the behavior of aluminum chloride upon dialysis. Aluminum chloride is at the highest rung and aluminum hydroxide at the lowest, each rung from top downward representing a different substance con- taining less free energy and being less soluble than the preceding one. Neidle disagrees with Nicolardot 24 who believes that there are two oxychlorides, one having the ratio in equivalents of 6 Fe to 1 Cl, the other of 125 to 1. The first no longer gives a test for ferric ion, the second shows absence of chloride ion. According to Nicolardot, any particular sol is composed of a mixture of these two oxychlorides. Pauli and Matula 25 consider ferric oxide hydrosol to be composed of a complex according to the Complex Theory, and picture it as xFe(OH) 3 yFe+++3yCl . They found their prep- arations to be neutral in reaction and to have a chloride ion concentration less than the total chlorine present ; also that the concentration of the chloride ion increases with dilution. CRITICISM The question of the nature of ferric oxide hydrosol is far from settled. Nicolardot's assertion that there are two oxy- chlorides, one when the test for ferric ion is absent and the other on the absence of chloride ion, is questionable since the test for ferric ion by ammonium thiocyanate is shown at much higher purities than those recorded by him. Magnier de la Source 13 stated that the sols which he analyzed were perfectly clear and "gave the ferric test with potassium ferrocyanide." Well dialyzed sols, we found, do not give this test, precipitation of ferric oxide gel taking place without evi- dence of blue color. Magnier de la Source could not have had sols of highest purity. The efforts of Linder and Picton were valuable in that they emphasized the fact that chloride "impurity" was an essential part of the sol, but their quantitative data concerning the compo- 5 sition of the sols are without value since they could not have ob- tained more than a fraction of the chloride present by the pro- cedure which they employed. The earlier work of Krecke and Pean de St. Gilles show how faulty Malfitano's method was, and therefore his results are not conclusive. Neidle considered his dialysis complete when addition of ammonium thiocyanate to the sol gave no color, after the precipi- tate settled or was centrifuged. Since ferric oxide gel carries down ferric chloride with it, his failure to obtain the color reaction does not mean that all the ferric ion in excess of that absorbed had been dialyzed away. Furthermore, in his deter- mination of chloride, the sol was heated after concentrated nitric acid had been added to it, to accomplish solution of the gel. We have found that it requires a high temperature and prolonged heating to effect this and hence there is the danger of loss of chlorine. His precipitation values are entirely relative. An amount of sulfuric acid which will not completely precipitate a sol in, say, five minutes, will do so upon longer standing. It is doubtful if the method of extrapolation is valid. METHODS Preparation of the Hydro sols The following methods were used to prepare ferric oxide hydrosols : (1) About 14 M ammonium hydroxide solution was delivered drop by drop from a burette into a solution of 3 M ferric chloride, which was continually and vigorously agitated by a motor stirrer. The addition of ammonium hydroxide solu- tion was continued until the resultant precipitate no longer readily dispersed. (2) Ammonium hydroxide was added as above until a permanent precipitate was just formed. l /z M ferric chloride solution was then added, and the mixture stirred until The precipi- tate dispersed. (3) To a hydrosol prepared by method (1) ferric chloride solution was added until the entire sol precipitated, dispersing the resulting precipitate in distilled water. (4) 50 cc. of molar hydrochloric acid were added to a freshly prepared and washed precipitate of ferric hydroxide prepared 'from 250 cc. 3 M FeCl 3 . The mixture was allowed to stand until the precipitate was peptized. The hydrosols prepared by the first two methods were blood red in color and perfectly clear to reflected and transmitted light. Those prepared by the third method were clear to transmitted, but slightly turbid to reflected light, while the sols prepared by the last method were decidedly turbid to reflected, though clear to transmitted light. Dialysis Cups* of very fine unglazed porcelain were first tried as dialyzers. It was found, however, that these were permeable to ferric ions for only a short time. After a few days of dialysis, there was but a slight amount of ferric ion in the diffusate, though the hydrosol in the cup contained a large amount of ferric chloride. Evidently, the membrane of ferric oxide formed within the walls of the cup is impermeable to ferric ion. Unglazed porcelain cups could not, therefore, be used at this stage of dialysis. In subsequent dialysis experiments, where all of the unadsorbed ferric chloride had already left the solution, and further dialysis was merely a slow hydrolysis, porcelain cups of smaller size and thinner walls were employed. These were permeable to chloride and hydrogen ions. For the preliminary dialysis, collodion bags were -employed. These bags were prepared in a two liter Florence flask and were changed as soon as a coating of ferric oxide was formed on the walls of the membrane, in order to speed up the process of dialysis. Some of these sols were dialyzed at room temperature and others at 60. The diffusate was changed every twenty-four hours, distilled water being used throughout. The preliminary dialysis was considered complete when the diffusate of twenty- four hours, usually one liter in volume, acidified and evaporated to 10 cc., gave no color upon addition of 5 cc. of molar ammonium thiocyanate. Since this is a very delicate test for the ferric ion, its concentration in the hydrosol is negligible at this point. (The ratio of the total ferric oxide in grams to the total ferric chloride in grams F^Os/FeCls was about 10 at this point.) The test for the ferric ion as given by ammonium thiocyanate applied to the sol directly was negative long before this ratio was reached. The length of time required for the completion of dialysis varies from two to five months, depending upon the concentration of the sol and the temperature at which the dialysis is performed ; the more dilute the sol is, the quicker is the hydrolysis process, and the easier for the dialyzable hydrolytic substances to leave the solution. Dilution of the dialyzing hydrosol was avoided by having the level of the water in the outer vessel much lower than that of the hydrosol inside. ANALYSES Iron. Iron was determined by titration with potassium perman- * Obtained from Coors Porcelain Company, Golden, Colorado 7 ganate, the colloidal solution having been evaporated with sul- furic acid to fumes of sulfur trioxide and reduced by the Jones reductor. Chlorine. Since hydrogen ions peptize the colloidal particks, due to their combination with hydroxyl ions that may be associated with the complexes as a result of the hydrolysis of the absorbed ferric chloride, nitric acid dissolves the sol with great dificulty. The first effect of the acid is to make the hydrosol more stable, but on the addition of a considerable amount, precipitation of hydrous ferric oxide takes place. This precipitate redissolves only on prolonged contact with the acid. This caused unexpected difficulties in the analyses of the sols for the determination of chlorine. Aifter considerable experimentation, the following method was adopted for the dissolution of the precipitated hydrosol and determination of the chloride concentration, To a definite volume of the hydrosol, nitric acid was added to make the final concentration about three molar. The covered beaker w r as allowed to stand in the dark until all the ferric oxide had dissolved, which usually required a week or ten days. Ap- proximately 0.1 molar silver nitrate was then added in excess of the amount necessary to precipitate all the chloride. A slight variation in the procedure was to add the measured amount of silver nitrate before the addition of nitric acid. Separate ex- periments, however, on pure-, potassium chloride showed that the procedures could be used ^iterchangeably, and that there was no danger of loss of chloride" by oxidation. The chloride con- centration was determined either gravimetrically or by the method of Volhard. EXPERIMENTAL At the beginning of this investigation, measurements of the conductivity of ferric oxide hydrosol during dialysis were made with the hope of getting quantitative indication of an end point in the purification. The measurements showed, as anticipated, that the conductivity gradually decreases as dialysis proceeds, but after a certain time the hydrosol showed a conductivity lower than that of the distilled water against which it was being dialyzed. The conductivity of ferric oxkle hydrosol has been reported by several investigators 20 , 23 , 2G , but the results differ widely due to the variable quantities of peptizing electrolyte present and are obviously of no value. Due to the failure of the conductivity method, freezing point depression was tried, and since this also failed to serve our pur- pose (see later), the observation of beginning of precipitation was adopted as end point in dialysis. The incipience of precipitation does not by any means signify the end of the hydrosol, since there- after precipitation proceeds gradually, the system assuming a turbid appearance which increases with continued dialysis until finally the entire sol becomes a gel. The ammonium salts formed in the preparation of the sol disappear within a comparatively short time after the beginning of dialysis. Thus, for sol No. 9, no traces of ammonium salts were detectable after two weeks dialysis. When dialyzed for four weeks, this sol contained: Fe 2 O 3 , 4.6695 gm./L* and FeCl 3 , 0.2664 gm./L* showing a molar ratio Fe 2 O 3 :FeCl 3 =11.7. A part of this hydrosol was removed from the large col- lodion bag and placed in a small bag in order that the incipience of precipitation might be obs-erved more easily. The level of the water outside was kept lower than that of the sol inside the sack in order to present dilution. Upon the first appearance of a precipitate, analysis of the hydrosol showed : Fe 2 O 3 , 7.0713 gm./L, FeCl 3 , 0.3282 gm./\L giving a molar Fe 2 O 3 : FeCl 3 =21.8. Two and one half liters of sol No. 10 were dialyzed for four weeks at room temperature against a volume of one liter of distilled water, the outside water being changed several times each day for the first week and once a day thereafter. Samples were withdrawn for analysis at intervals as noted in Table I until precipitation began, after which the analyses were made less frequently, and dialysis was continued until almost all of the sol was converted to a gel. This took place after ten weeks of continued dialysis. TABLE I Number of Sample Time of Ferric Oxide Dialysis (Fe 2 O 3 ) Molar 7& r ?, londe Ratio of (FeCl s ) Fe 2 3 /FeCl 3 lOa 24 days 9.310& gm/L 0.7143 gm./L 13.2 lOb 27 4.2923 0.2754 ' 15.9 lOc 32 3.6443 0.2150 ' 17.7 lOd 40 3.3555 0.1720 ' 20.3 lOe(ppfion) 47 3.2653 a. 1539 ' 21.5 lOf 52 3.4588 0.1080 ' 23.6 10g 60 * 1.6704 0.06039 ' 28.1 10h*(finaJ) 73 ' 1.4347 0.04965 ' 29.3 Precipitate 47.9 * In all tabulations throughout this paper, the chlorine found by analysis was calculated to ferric chloride, the ferric chl .ride computed to ferric oxide which was subtracted from the total ferric oxide (from total iron found by analysis), giving the ferric oxide values us?d in the data. * In the interval between lOa and lOb, the sol became considerably more dilute due to an accident in the collodion bag. Precipitation began after 47 days of dial- ysis. lOh was almost entirely in the gel state. It was dried in an oven for six hours at 110 deg, for three hours at 160 deg., powdered, washed until several consecutive wa=,r- ings gave the same turbidity with silver nitrate when viewed in test tubes, and analyzed. The results of the analysi's are shown under the heading "precipitate" in the above table. 9 Sol No. 15 was dialyzed at 60 for ten weeks. At the end of this time, the twenty-four hour diffusate when evaporated to 10 cc. volume gave no test for ferric ion. Five hundred cc. of this sol were diluted to about 3 liters (No. 15a) and a series of five collodion bags of 500 cc. each was allowed to dialyze. TABLE II Number of Sample Time of Dialysis Ferric Oxide (Fe 2 3 ) Ferric Chloride (FeCl 3 ) Molar Ratio of Fe 2 3 /FeCl 3 15 15a I5b 15c 15d ISe 1 18 days 21 " 30 " 40 ' 81017 gm./L 6.1521 ' 1.1961 " 1.1960 " 1.4164 " 1.5773 ' 0.8397 gm./L 0.3330 ' 0.05927 " 0.05924 " 0.05552 " 0.04896 " 9.8 '18.9 20.5 20.5 25.8 32.7 No. 15 represents the analysis of the original impure sol at the time the preliminary dialysis was stopped. No. 15a represents the original sol dialyzed until precipitation began ; it became somewhat diluted during this process. The diluted sols showed a slight precipitation after 18 days of dialysis, but no turbidity could be detected in the supernantant solution. However, with continued dialysis, precipitation increased, and after forty days of dialysis a large part of the sol in 15e had turned to a gel. This would seem to indicate that dilution has no marked effect on the point at which precipitation begins. To verify this, sol No. 13 which had been dialyzed at 60 for 8 weeks, was mad-e up to various dilutions and these allowed to dialyze in 500 cc. collodion bags to first appearance of precipitate : TABLE III Number of Sample Dilution Ferric Oxide Ferric Chloride (Fe 2 O 8 ) (Fed,) Molar Ratio of Fe 2 3 /FeCl 3 13 4.1838 gm./L 0.3637 gm./L 11.7 13b 1.87 2.3144 ' 0.1118 ' 21.8 13c 3.04 1.3711 " 0.06504 " 21.4 13d 7.75 tt.5396 " 0.02657 " 20.6 13e 13.80 0.3019 " 0.01485 " 20.6 Sol. No. 17, after having been allowed to dialyze at about 60 for two months, was then subjected to the same treatment as No. 13. TABLE IV Number of Sample Dilution Ferric Oxide (Fe 2 3 ) T- n 1 -j Molar Ferr c Chloride Ratio of (FeCl 3 ) Fe 2 O 3 /FeCl 3 V 17 17a I7b 17c 17d 17e 2.19 3.24 4.44 8.08 8.87 5.5385 2.5305 1.7102 1.2471 0.6833 0.6245 gm./L 0.4326 gm./L 0.1275 ' 0.08270 " 0.06097 " 0.02894 " 0.02693 " 12.9 20.1 21.0 20.7 23.9 23.5 10 In the last two a considerable amount of gel had formed. It was deemed probable that the gradual precipitation, which sets in after the initial appearance of precipitate, might be due to excessive hydrolysis and dialysis near the walls of the collodion bags. If so, this would affect the limiting ratio. To test it, a series of sols was dialyzed in small unglazed porcelain cups. The solutions were stirred throughout the entire period by a current of nitrogen and were kept at a temperature of 50-60, and were analyzed when a turbidity became perceptible. The following results were obtained : TABLE V Number of Sample Dilution Time of Dialysis Ferric Oxide (Fe 2 3 ) Molar Ferr 'c Chloride Ratio of (FeCl 3 ) Fe 2 3 /FeCl 3 23 23a 23b 23c 1.55 2.14 2.66 7 days 2 " 5.4244 3.5111 2.5388 2.0409 gm. / L 0.3864 gm./L 0.1723 " 0.1204 " 0.09497" 14.3 20.7 21.4 21.8 Hydrosol No. 21, after three months dialysis, was divided into portions which were diluted with distilled water to different extents, and these were dialyzed in porcelain cups at 60 (being stirred by a current of nitrogen) until incipience of precipitation with the following results : TABLE V Number of Sample Dilution Ferric Oxide Ferric Chloride (Fe 2 O 3 ) (FeCl 3 ) Molar Ratio of Fe 2 3 /FeCl 3 21 5.867 gm./L 0.3743 gm./L 15.9 21a None 4.9186 0.2453 ' 20.4 21b 1.33 3.6492 0.1829 " 20.3 21c 2.00 2.3434 0.1160 ' 20.6 21d 2.66 1.8119 0.09340 " 19.7 21e 4.00 1.4292 0.06690 " 21.6 Following the incipience of precipitation, gradual flocculation was observed in all cases. Analyses of samples of such sols showed that their molar ratios of ferric oxide to ferric chloride gradually increased as dialysis continued to final complete precipi- tation. These analyses were not very accurate because it was impossible to get the hydrosol entirely free from precipitated particles that were held in suspension. Centrifuging at about 1,000 times gravity for the purpose of removing these suspended particles frequently resuite.l in the breaking out of the entire dis- persed phase from dispersion in the form of a fairly continuous jelly phase. This is very significant in that it reveals the jelly-like nature of this hydrosol. Due to the inaccuracies of the analyses the results are not reported, but it is of interest to recall that 11 Duclaux 27 claimed that ferric oxide hydrosol could be dialysed to a limiting value of 170 Fe 2 O 3 .l FeQ 3 . Since the dispersed phase of hydrosols, prepared as just described, migrate to the cathode when subjected to the action of an -electrical current, the particles are naturally said to be positively charged due to the ferric chloride of the complex, the ferric ions thereof remaining in contact with the ferric oxide while the chloride ions are located in the water phase directly bathing the particles. Since like-charged bodies repel one another, the electrical charges of the particles are supposed to overcome the mutual attractive forces of the particles and this is the commonly accepted explanation for the stability of inorganic collodial particles. According to this explanation, the limiting ratio of ferric chloride to ferric oxide should increase with increased concen- tration of the particles, since the mutual attractive force varies inversely as some power of the distance between the particles. The more closely the particles are packed, the greater should be the charge required to keep them repelling one another. Examination of the data reveals no such tendency. In fact. in a few cases a concentrated sol showed a lower limiting ratio than a more dilute one. Apparently the electrical charge is not the predominating factor for the stability. The fact that the limiting ratio, i.e., the point corresponding to incipience of pre- cipitation, is always nearly the same* indicates that no matter what the concentration of the sol is, one mole of ferric chloride is required to keep about 21 moles of ferric oxide dispersed in the colloidal condition. Any amount of ferric chloride in excess of this ratio might be regarded as impurity. The stability of the ferric oxide hydrosol must then be due not to the electrical charge of the particles but to the solution forces (solubility) of the adsorbed ferric chloride. The high solution forces of the ferric chloride molecules pull the ferric oxide particles with which they are combined by secondary valence or "adsorption forces" into semisolution. Upon removal of the ferric chloride by hydrolysis the insoluble particles of ferric oxide having lost their "solution-link", precipitate. According to the "solution-link" hypothesis this hydrosol should be soluble in any liquid in which ferric chloride dissolves. It was found that dilution of the hydrosol with alchol ad infinitum had no effect. Addition of ether to this alcosol did not precipitate it either, provided too large an excess was not added. An iron oxide hydrosol stabilized by ferric sulfate should be precipitated by alcohol according to the hypothesis. Such a sol *The slipht deviations from the vahie of 21 may be ascribed as due to errors in the determinations of the end point of dialysis by means of the observation of the begin- ning of precipitation. 12 was prepared resembling the Pean de St. Gilles' sol in appearance. Addition of alcohol precipitated it instantly. Hydrogen Ion Concentration. Pauli and Matula 25 attempted to measure the hydrogen ion concentration of ferric oxide hydrosols. Good results were ob- tained by them when using sols which had "aged" for six months or which were heated for a few hours at 80, this being equivalent to aging, i.e., hydrolyizng excess ferric chloride. Their measure- ments had to be made quickly and their platinum electrodes were saturated with hydrogen before coming in contact with the hydro- nrutral, i.-e., a C H +of the order of 10- 7 . Measurements of hydrogen ion concentration would be iM- possible in the presence of ferric ion for obvious reason. Using well dialyzed sols in which the concentration of ferric ions was supposedly nil we found noi evidence of a disturbing reducing potential but could not get what is considered to be an absolutely satisfactory equilibrium reading due to the deposition of ferric oxide gel upon the plantinizecl electrode. However, taking the mean of a series of readings which were not widely divergent, a C H + of 10~ 4 - 9 was indicated. This was the same for a series of our "pure" hydrosols of varying concentration and consequently the hydrogen ion concentration does not appear to depend upon the concentration of the dispersed phase, at least over the range which we studied. We would say that our "pure" ferric oxide hydrosols showed a C H + = 10~ 5 . An acid reaction is to be expected since upon dialysis of sols from which the free ferric chloride has been removed, only hydrogen and chloride ions are found in the clifFusate across the collodion membrane. Consequently the ferric chloride of the dispersed phase is in equilibrium with the ions of hydrochloric acid in the dispersion medium, which in the case of our "pure" sols is of the order of M / 100, 000 (or less, since part of the acidity may be due to carbonic acid.) Behavior upon Freezing. As previously mentioned, freezing point depression was tried as a qauntitative measure for following purification but, it was found that a well dialysed sol gives a depression of the freezing- point within the range of experimental error of measuring the same by means of the Beckmann thermometer. Consequently any molecular weight figures based upon such determinations, as given by several investigators 28 , 29 , are valueless as has been previously suggested, 20 - 30 Pean de St. Gilles was inclined to regard his hydrosols as true solutions because they froze in a "normal" manner. In 1889, 13 Ljubawin 31 froze among otlier colloidal substances, ferric oxide hydrosol and found that on continued cooling particles of ferric oxide concentrated in the center while the periferous layers of the ice became colorless. Upon melting, all of the iron oxide par- ticles redispersed. Lottermoser 32 found that only sols which are deficient in electrolyte will precipitate out on freezing. Sols rich in electrolyte are not affected even on continued freezing. Our experiments showed that when a pure ferric oxide hydrosol is only partially frozen, ice crystals are formed which, upon melting, leave the sol as homogeneous as it was before the operation. But when cooling is continued until freezing is com- plete, the sol on melting will undergo some precipitation. The longer the sol has been cooled, the greater will be the amount of gel formed. The precipitate is in the form of short, amorphous, shiny, needle-like particles. When the sol, placed in a test-tube, is allowed to remain in contact with the ice-salt mixture for some length of time, the entire solution turns to a dark red solid mass. Upon continued cooling, separation of the water begins as a layer of colorless ice near the walls of the test-tube, and such layers continue inward until in the center there are deposited the red brown particles above described. These particles are ar- ranged in a string-like formation throughout the height of the tube. Upon melting the ice they do not redisperse. The water obtained from melting the ice shows a barely perceptible test for chloride ion and none for ferric ion. The gel particles are prac- tically insoluble in dilute nitric acid, but readily soluble in con- centrated acid. Analysis of them showed that about 80 percent of the ferric chloride of the original hydrosol particles was re- tained in this gel. This behavior is quite different from that of Bredig gold hydrosols upon freezing, since Beans and Beaver 33 find that all of the stabilizing electrolyte is removed from the gold particles through the congelation. These observations strengthen the conclusion reached from the limiting ratio of the hydrosol that the stability of those hydrosols is due to the solution forces of the adsorbed ferric chloride rather than to the electrical charge of the particles. The ferric chloride in the congelation gel from the pure sols must be dispersed throughout the compact slid mass, for although there is sufficient ferric chloride present to redisperse the par- ticles, at least partially, such does not take place, but when an impure sol is frozen, i.e., one to which some excess ferric chloride had been added, the gel particles redisperse upon melting. In the latter case the gel cannot be so massive as in the case of the gel from a pure sol and consequently the pull of the water on the adsorbed ferric chloride is able to disintegrate the particles and redisperse them. The appearance, upon freezing, of an im- pure sol is different from that of a pure sol. The preliminary ice 14 mush has a lighter color and upon complete congelation the gel particles deposited in the center are somewhat larger in appear- ^ance and less distinct individually. In a recent article, Gutbier and Flury 34 reported that in the case of selenium oxide sols, the extend of the reversibility after freezing depends upon the amount of peptizing electrolyte originally present. A sufficiently well dialysed selenium oxide sol is entirely irreversible after freezing; while the deposit from an impure sol readily redissolves upon melting. Relationship Between Graham's Hydrosol and the So^Called "Metairon" Hydrosol of Pean de St. Gilles. Water of Hydration. The hydrosol that Pean de St. Gilles prepared, by heating and boiling solutions of the acetate, differed slightly in properties from Graham's in that it was not so clear and that a precipitate formed on continued heating which was insoluble in concentrated acids but soluble in dilute acids and water. Graham, in analogy to the two modifications of tin oxide sol, called it the "metairon" oxide hydrosol. This appellation is now common and it is generally admitted that a study of this modification is greatly needed. In the course of this investigation, it was found that this conception of two modifications of ferric oxide hydrosol is not justifiable, the main difference between the two being water of hydration of the particles. When ferric chloride or hydrochloric acid is added to the ordinary Graham hydrosol, an ochre colored precipitate is formed which dries on a porous plate to a chocolate colored mass. This dry mass, as well as the original wet precipitate, is readily soluble in water, giving rise to a hydrosol identical in all its properties to the "metairon" modification. On redispersion of the powder an water, the resulting sol is slightly turbid, because during the precipitation some dehydration of the colloidal particles took place. The more thoroughly the precipitate is dried before its redispersion, the more turbid is the solution formed. After dry- ing in an oven at 110, it forms an unstable dispersion of yellow particles in water which settle out after standing a few days. In alcohol, with which Graham's hydrosol is miscible in all pro- portions, this wet precipitate forms a coarse dispersion of yellow particles similar to that formed by dispersing the over-dried precipitate in water. When precipitated ferric hydroxide washed free from ammonium salts is treated with a solution containing hydrochloric acid or ferric chloride and is allowed to stand for a few months, the hydrosol formed is more turbid than the "metairon" sol. This, again, may be attributed to a lower hydration or to a larger size of the particles. 15 But if this were an effect of the size of the particles, we should expect the limiting values of ferric chloride to be higher than that established above, that is, more electrolyte would be required to keep the larger particles dispersed. The following experiment was made to decide this question. A hydrosol was prepared by addition of ammonium hyd- roxide to ferric chloride solution until the resultant precipitate no longer readily peptized. Ferric chloride was then added to precipitate the dispersed phase and the gel so obtained was dried on a porous plate. This residue was then dispersed in water resulting in a sol slightly turbid to reflected light. This was dialyzed in collodion sacks at room temperature for three months when it showed upon analysis a molar ratio of Fe 2 O 3 /FeCl 3 =18.0. A portion of this dialyzed sol was then dialyzed in a porcelain cup at about 60 with continuous stirring by a current of nitrogen. It was analyzed (a) after five days of dialysis and then again (b) after nine days. In (a) incipience of precipita- tion was not evident while in (b) a decided precipitation had started. The ratios of Fe 2 O 3 /FeCl 3 were (a) 19.7 and (b) 25.9. Another hydrosol prepared by peptizing ferric hydroxide gel with a small amount of hydrochloric acid was dialyzed in porcelain as described above. This initially turbid sol increased in turbidity as dialysis was continued so that the beginning of precipitation could not be determined accurately. When the end point was presumed to have been reached, it showed upon anal- ysis a ratio of Fe 2 O 3 /FeCl 3 =22.6. It is thus seen that the limiting value for these "meta-iron oxide," or Pean de St. Gilles' sols is of the same magnitude as that of the Graham sol. This would indicate that the turbidity of this sol is due to dehydration rather than to the presence of larger ferric oxide aggregates. It has been suggested before that the difference between the two types of ferric oxide sols is one of hydration 35 . The Graham sol was suggested to be FeoO 3 .3HoO and the Pean de St. Gilles, Fe 2 O 3 .H 2 O. Since, according to Einstein 36 , the viscosity of a colloidal dis- persion is expressed by the formula N*=N (1-j-kf), where N* =viscosity of the system, Af=viscosity of the dispersion medium, and f=the ratio of total volume of the dispersed phase over the total volume of the system, which means as Wo. Ostwald 37 and others have shown that the viscosity of a colloidal solution in- creases with the amount of the dipersion medium taken up by the dispersed phase, the Pean de St. Gilles sol should have a lower viscosity than the Graham sol of the same concentration if our supposition is correct. To determine this, lOcc. portions of a Graham hydrosol were treated with varying amounts of 2N ferric chloride until the tur- 16 bidity which first appeared gradually changed to coarse bright yellow dispersion and the viscosities were measured. The measurements were made by means of an Ostwald vis- cosimeter, in a constant temperature bath at 25 0.1. The following tabluation shows the results. The figures signify the time in seconds for outflow : (1) Sol No. 15 10 cc. 81 sec. Distilled Water 71 sec. (2) (3) (4) (5) (6) (7) + 0.1 cc. FeCl 3 73 sec. 10 cc. +2 cc.FeC! 3 -80 4- 0.2 " 73 + 0.3 " 74 4- 0.5 ' " 75 + 1.0 " " 78 + 2.0 " " 99 10 cc. 2N FeCl 3 137 Number 1 was perfectly clear, numbers 2, 3, 4, and 5 were clear to transmitted, but turbid to reflected light. They resembled the Pean de St. Gill-es sol in all respects. The turbidity gradu- aly increased with increasing amounts of ferric chloride. Number 6 was decidedly brown while No. 7 was yellow. After standing for three hours, the viscosities were again measured and found to be unchanged. From these results it is seen that the addition of ferric chloride first decreases the viscosity and then increases it. The decrease in viscosity indicates a diminution in size of the parti- cles which could only have been caused by a loss of water of hydration by the dispersed phase, i.e., dehydration. The increase in viscosity observed upon the addition of larger amounts of ferric chloride is due to the coalescence of the particles prelim- inary to precipitation. As to the mechanism of the dehydration of the sol by ferric chloride, we can only venture to say that it is possibly due to the high hydration of the electrolyte added, thus causing a partial dehydration of the dispersed Fe 2 O 3 FeCl 3 H 2 O phase, or in view of the fact that this sol will migrate in an electrical fi-eld showing that the adsorbed and peptizing electrolyte is ionized, even though the degree be extremely small, then the Donnan effect of the added ferric chloride would result in a decrease in swelling (hydration) of the dispersed phase as in the case of the addition of hydrochloric acid or a neutral salt to gelatin jelly swollen in a solution of hydrochloric acid. Both suggested mech- anisms may operate at the same time. NEGATIVE IRON OXIDE HYDROSOL Linder and Picton in 1892 and Coehn in 1898 38 have shown that ferric oxide hydrosol is charged positively. H. W. Fisher 39 prepared a negatively charged ferric oxide hydrosol by runnings two thirds normal ferric chloride solution into five normal sodium hydroxide solutions to which a consid- erable amount of glycerol had been added. It is by no means 17 certain that a negative iron oxide hydrosol was thus prepared, and Fisher himself admits the possibility of the formation of a compound with the glycerol which would account for its anodic migration, analogous to the action of tartrate in Fehling's solution. Powis 40 prepared a negative sol by slow addition of lOOcc. of .01V ferric chloride to loOcc. .OlV sodium hydroxide solution with constant stirring. The sol was reddish brown, clear, and remained for several weeks without precipitating. On dialysis, it precipitated in a few hours. He prepared a similar sol by using a positive sol instead of the ferric chloride used above and concluded from this experiment that the hydrosol ion is adsorbed and that the negative charge is due to it. H. B. Kruyt and J. Van de Speck 41 find that upon adding from 1.55 to 2.8 millimoles of sodium hydroxide to a ferric oxide hydrosol containing 0.75 gm. of ferric oxide and 0.064 gm. of chlorine per kilogram and allowing the mixture to stand for three hours, complete precipitation takes place and in some cases immediately. The sol is not precipitated when sodium hydroxide of the concentration of 3.99 to 27.9 millimoles is added, although it becomes turbid. When more than 30.3 millimoles is added, complete precipitation again ensues. They call this zone of no precipitation the "tolerance" zone of ferric oxid-e sol and con- sider the substance within that zone to consist of a negatively charged ferric oxide hydrosol. In this laboratory, Powis's experiments were duplicated, using ferric chloride. Clear reddish brown solutions were ob- tained which were precipitated by the addition of sodium sul- fate. However, none of the sols prepared in the course of these experiments was stable. A gradual settling began after a couple of days and all the dispersed phase settled out within a week or two. With well dialyzed positive sols, the "negative sol" was -obtained only when extremely high dilutions were used. The dispersions formed were stable for only a few hours and in no case remained dispersed longer than a day. The addition of sodium sulfate caused immediate precipitation of ferric oxide. Using .01V ferric chloride and various concentrations of sodium hydroxide, it was found that concentrations above .OlV .sodium hydroxide resulted in the immediate formation of sus- pensions which settled out within a few hours. Increasing the concentrations of ferric chloride likewise resulted in suspensions which s-ettled out within a short time. These dispersions could be thrown down by centrifuging immediately after their preparation at 1,000 times gravity only when they were more or less concentrated. For instance, when a considerable amount (lOOcc.) of .01 N ferric chloride was added 18 to .01 A 7 " sodium hydroxide, ferric hydroxide was thrown out on centrifuging quite easily; but when smaller amounts of ferric chloride (up to 50cc) were added, centrifuging had no effect on them. On standing, for a day at most, they would begin to settle out. These samples from which ferric oxide gel could not be thrown down by centrifuging immediately after their preparation, when allowed to stand for a few hours or a day, depending upon the dilution, were easily thrown out on centrifuging. Powis believes these solutions to consists of a negative ferric oxide hyrosol due to the adsorption of hydroxyl ions by the ferric oxide particles. However, it is conceivable that it may be a negative hydrosol stabilized by a ferrate ion or that it may be a negative suspension. Since none of the conditions considered necessary for the preparation of a ferrate is satisfied in these experiments, a fer- rate is presumably not formed, although its formation would fit the behavior of this sol remarkably well. It is our opinion that the solution formed in these cases are negative suspensions. There is no reason why settling of ferric oxide should begin when the solutions are allowed to remain in stoppered "Non-Sol" bottles. SUMMARY 1. It has been shown that coloidal ferric oxide remains stable as long as the particles contain one mole of ferric chloride to approximately 21 moles of ferric oxide. Beyond this point gradual precipitation begins with the formation of hydrosols of lower ferric chloride and lower ferric oxide content. 2. Evidence has been offered o show that ferric oxide hydro- sol is not merely a complex consisting of ferric oxide and ferric chloride but that water is a third essential part of the colloidal particle ; and that the stability of ferric oxide hydrosol, stabil- ized by ferric chloride, is due to the solution forces of the ab- sorbed ferric chloride in the dispersed medium rather than to the mutual repulsive forces of the particles presumed to reside in their electrical charges of like sign. 3. The so-called "metairon" sol of Pean St. Gilles has been shown to be merely a les hydrated form of Graham's sol. 4. The hydrogen ion concentration of pure ferric oxide hydrosol has been shown to be approximately 10~ 4 - 9 . 5. Freezing of pure ferric oxide hydrosols causes irreversible coagulation, a small part of the peptizing ferric chloride being split off and the greater part remaining in the gel. 6. The reversibility of the sign of the charge of ferric oxide 19 hydrosol to form negative sols has been investigated and -evidence offered which shows that if a negative sol is actually formed it it exceedingly unstable and probably is merely a suspension of ferric oxide gel. BIBLIOGRAHY 1 Berzelius: Lehrbuch der Chemie 3, 555 (1845) 2 Maus: Ann. d. Phys. u. Chem. //, 75 (1827) 3 Souberain: Ann. Chim. et Phys. 44, 325 (1830) 4 Rose: Ann. d, Phys. u. Chem. 24, 301 (1832) 5 Schonbein: ibid. 39, 141 (1836) 6 Scherer: ibid 44, 453 (1838) 7 Pean de St. Gilles: J prakt. Chem. (1) 66, 137 (1855) 8 Kastner: Ann. Chim. et Phys. (3) 57, 231 (1859) 9 Debray: Compt. rend. 59, 174 (1864) 10 Wiedemann: Ann. d. Phys. u Chem. (3) 5, 45' (1878) 11 Wagner; Kolloid Z. 14, 149 (1914) 12 Graham: Phil. Trans. 161, 183 (1861) 13 Mjagnier de la Source: Compt. rend. 90, 1352 (1880) 14 Wyrouborf: Ann. Chim Phys. 7, 449 (1905) 15 Hantz and Desch: Ann: Chem. 323, 38 (1902) 16 Ruer: Z. anorg. allgem. Chem. 43, 85 (1905) 17 Linder and Picton: J. Chem. Soc. 87, 1919 (1905) 18 Krecke: J. prakt. Chem. (2) 3, 286 (1871) 19 von Weimarn: Zur Lehre der Zustande der Materie /, 60 (1914) published by Theodor Steinkopff, Leipzig. 20 Beans and Eastback: J. A. C. S. 37, 2667 (1915) 21 Malfitano: Compt. rend. /?9, 1221 (1904); 140, 1245 (1905); 141, 660, 680 (1905); 143, 172, 1141 (1906); Z. physik. Chem. 68, 232 (1909) 22 Neidle: J. A. C. S. 39, 2334 (1917) 23 von Weimarn: Zur Lehre der Zustande /, 46 (1914) 24 Nicolardot: Ann. Chim. Phys. 6, 334 (1905) Comp. rend. 140, 310 (1905) 25 Pauli and Matula: Kolloid Z. 21, 49 (1917) 26 Duclaux: Compt. rend. 140, 1468 (1905), Kolloid Z. ?, 126 (1908), Goodwin and Graver; Phys. Rev. 9, 251 (1896), //, 193 (1900), "Dnmanski: Z. physik Chem. 60, 553 (1907) 27 Duclaux: Compt. rend. 143, 296 (1906) 28 Kraftt: Ber 32. 1608 (1899) Dumanski: Kolloid Z. 8, 232 (1911) 29 Gladstone and Hibbert: Phil. Mag. 28, 38 (1889) 30 Bruni and Pappada: Atti R. Accad. Lincei (5) 9, 334 (1900) 31 Ljubawin: Z. physik. Chem. 4, 486 (1889) 32 Lottermoser: Ber. 41, 3976 (1908) 33 D. J. Beaver: Dissertation. Columbia University, New York, 1921 34 Gtitbier and Flury: Kolloid Z. 29, 161 (1921) 35 Roscoe and Schorlemmer: 2, 998, D. Appleton and Co., New York .^6 Hafcchek: Koiioid Z. 7, 301 (1910); S, 34 (1911) 37 Wo Ostwald: Trans. Faraday Soc. 9, 34 (1913) 38 Linder and Pictoa: Trans. Chem. Soc. 61, 160 (1892) Coehn Z. Elektrochem. 4, 63 (1898) 39 Fisher: Biochem. Z. 27, 311 (1910) 40 Po^is: J. Chem. Soc. 707, 818 (1916) 41 Kruyt and van de Spek: Kolloid Z. 25, 1 (1919) VITA Alexander Frieden was born October 5, 1895 and attended the public schools of Norfolk, Virginia. He entered the Uni- versity of Virginia in September, 1915 and graduated in June, 1919, receiving the degrees of B.S. and M.S. Since September, 1919, he has been a graduate student in Chemistry under the Fac- ulty of Pure Science, Columbia University. In Jun-e, 1920, he received the degree of Master of Arts from Columbia University. During the war he was a member of the Chemical War- fare Service, U. S. A. Since September 1920 he has occupied the position of Assistant in Chemistry at Columbia University. THIS BOOK IS DUE ON THE LAST DATE STAMPED BELOW AN INITIAL FINE OF 25 CENTS WILL BE ASSESSED FOR FAILURE TO RETURN THIS BOOK ON THE DATE DUE. THE PENALTY WILL INCREASE TO SO CENTS ON THE FOURTH DAY AND TO $1.OO ON THE SEVENTH DAY OVERDUE. 8 1935 1944 .'58 LD 21-100m-8,'34 Binder Gaylord Bros. Makers Syracuse, N. Y. Ml. JAM 21.1908 __ 5,3 is UNIVERSITY OF CAUFORNIA LIBRARY it 1&3 -" : v I "