305 EXCHANGE The Anomalous Osmose of Solutions of Electrolytes with Collodion Membranes A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN THE UNIVERSITY OF MICHIGAN By Dwight Clark Carpenter 1921 The Anomalous Osmose of Solutions of Electrolytes with Collodion Membranes A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN THE UNIVERSITY OF MICHIGAN By Dwight Clark Carpenter 1921 CONTENTS. I. History 5 II. Purpose of Investigation 15 III. Relationship between Osmose and Electrical Properties of Membrane.. 16 Preparation of Membrane 16 Construction and Assembly of Cell 18 Osmose of Solutions of Chlorides of Different Metals 21 Osmose of Solutions of Potassium Salts of Inorganic Acids 21 Osmose of Solutions of Potassium Salts of Organic Acids 23 Osmose of Solutions of Hydrochloric Acid and of Sodium Hydroxide 24 Measurement of Cell Potential 24 Sign of Membrane Charge 26 General Discussion and Conclusions 27 IV. Relationship between Membrane Pore Size, Osmose and Rate of Salt Diffusion through the Membrane 29 Permeability of Membranes 30 Measurement of Membrane Pore Size 32 Osmose through Membranes of Different Degrees of Permeability 33 Summary 37 Diffusion of Solute into Water Compartment during Osmose 41 Summary 44 V. Effect of Stirring Solutions during Osmose 46 Apparatus 47 Construction and Assembly of Rocking Cell 47 Thermostat and Rocking Machine 49 Method of Setting Up Cell for Experiment 49 Osmose Results in the Rocking Cell 51 Passage of Salt through Membrane during Osmose in the Rocking Cell. . . 52 Summary 54 VI. General Summary ! 56 The author wishes to acknowledge his indebtedness and gratitude to Professor Floyd E. Bartell, under whose direction this research was carried out, in sincere appreciation of ex- cellent advice, kindly encouragement, and many favors throughout the course of the work. THE ANOMALOUS OSMOSE OF SOLUTIONS OF ELECTROLYTES WITH COLLODION MEMBRANES BY DWIGHT CLARK CARPENTER I. HISTORY The phenomenon of osmosis, or the unequal rate of passage of two liquids through a membrane which separates them, was discovered by the Abbe Nollet 1 in 1748. After filling a vessel with alcohol, closing the orifice with bladder, and sub- merging in pure water, he observed that the bladder became distended, thereby showing that water had passed through the membrane more rapidly than alcohol. This observation attracted little attention of scientists for over half a century, and was evidently forgotten until Sommering, 2 experimenting with a pig's bladder, made a similar discovery. He found that when a pig's bladder, filled with an alcohol-water solution was suspended in air, the alcohol became more concentrated. When the experiment was repeated substituting a rubber bag for the bladder, the alcohol became more dilute. This appears to have been the first recorded observation of anomalous osmose. The above effects, the opposite of each other, established the important fact that the nature of the membrane material itself was an influencing factor. The first quantitative experiments on osmosis were carried out by Dutrochet 3 and Vierordt 4 between the years 1826 and 1848. They both found that when a salt solution was sep- arated from water by a membrane of pig's bladder, the water passed through the membrane more rapidly than the salt solution, resulting in a hydrostatic pressure. As this pressure 1 Xollet: Memoires de 1'Academy Roy des Sciences, 1748, 57-104. 2 Sommering : Pogg. Ann., 28, 17 (18l'4). 3 Dutrochet: Ann. Chem. Phys., 35, 37, 48, 49, 51, 69. 4 Vierordt: Pogg. Ann., 73, 519 (1848). was the result of osmosis, it was termed osmotic pressure. It was early recognized that the experimentally determined pressure was a resultant of the movement of both the solution and the water. Dutrochet gave us our nomenclature of these two oppositely moving liquids. The flow inward toward the more concentrated side, he called the endosmotic current, and the outward flow, the exosmotic current. The terms osmose and osmosis are now used to denote the process as a whole. In 1827 Dutrochet 5 announced an electrical theory to explain osmosis. He believed that the two sides of the membrane developed different "degrees of electricity," but that this difference could not be detected with a galvanometer. The work of Dutrochet and Vierordt showed that the rate of passage of pure water through the membrane depended not only on the salt used, but also on the concentration of the salt solution. From his later experiments with porous inorganic membranes, Dutrochet concluded that osmosis was also dependent on the nature of the membrane used. A number of explanations to account for this phenomenon were brought forward by various investigators. Poisson 6 believed that capillarity was the determining factor in osmosis. Briicke 7 considered it due to relative "attraction" of the mem- brane for the two liquids. Jolly 8 advanced a theory of hydro- diffusion, in which he claimed that the exosmotic current was replaced by the endosmotic current of water, which was characteristic of water and independent of the concentration . His final conclusion was that, in a given time, the amount of diffused substance was dependent on membrane area, the density of the solution, and the attraction of the separated substances for the membrane and for each other. Almost simultaneously, Liebig 9 gave reasons for believing that osmose was due to the ability of the membrane to absorb the sep- arated liquids. This directed the trend of investigation toward 5 Dutrochet: Ann. Chem. Phys., 35, 393 (1827). 6 Poisson: Ibid., 35, 98 (1827). 7 Briicke: Pogg. Ann., 58, 27 (1843). 8 Jolly: Ibid., 78, 261 (1849). 9 Liebig: Ann. Chem. Phys., (3) 25, 367 (1849). the study of different membranes and their function in the osmotic process. Thomas Graham 10 in 1855 published much data on osmosis with both organic and inorganic membranes. He advanced the theory that an alteration of the membrane was an indis- pensible condition to the maintenance of the "osmotic force." He thought one side of the membrane was always acid and the opposite side alkaline, and that the direction of the endos- motic current was from the acidic to the basic side ; or when the osmose of acids and bases were tested, the direction of flow was alway toward the side of lesser acidity in the former case, and toward the more basic side in the latter. In the develop- ment of this generalization, Graham did not include the results he obtained with porous earthenware membranes, for the reason that the osmotic effects observed with these membranes were usually opposite to the effects obtained with organic membranes. He had no satisfactory explanation to account for this difference. Later, influenced by his own work on dialysis and by that of I/Hermite 11 on selective or preferential solubility of two liquids in a separating membrane, Graham came to the same conclusions as those of Liebig. M. Trabue 12 prepared a membrane from a non-setting glue treated with tannic acid, which was the first artificial septum, permeable to water but impermeable to a crystalloid. He also prepared a number of precipitation membranes of different permeability. Pfeffer 13 devised the method of form- ing precipitation membranes within the walls of porous earthen- ware, thereby forming cells which were capable of withstanding great pressures. With these cells Pfeffer performed his classic experiments on the osmotic pressure of sugar solutions. His work proved that the osmotic pressure varied as the sugar concentration varied. Many investigators have since con- firmed Pfeffer's measurements. 10 Graham: Ann. Chem. Phys., (3) 45, 17 (1855). 11 L'Hermite: Ibid., (3) 43, 420 (1855). 12 Traube: Archiv. Anat. Phys. Und Wissensch. Medizin., 1867, 87. 13 Pfeffer: Osmotische Undersuchungen, Leipzig, 1877. 8 Up to this time Van't Hoff 14 had been studying gases and chemical equilibrium in solution with a view of throwing light on the question of chemical affinity. He recognized the close analogy between gases and dilute solution. This analogy between the behavior of such widely differing materials, led him to turn his attention to Pfeffer's data on the osmotic pressure of dilute solutions and to compare osmotic pressure with gas pressure. From Pfeffer's data he showed that the osmotic pressure of a dilute solution was equal to the pressure which the dissolved substance would exert if present in the gaseous state and present in the same volume as that occupied by the solution. This generalization has been extended to electrolytes by taking into account the increased number of solute particles produced by the dissociation of the electrolyte in solution. Van't Hoff was not concerned with the mechanism of osmosis, but simply in the quantitative relationships existing between gaseous and osmotic pressures. The majority of investigators in the field of osmosis, since the time of Van't Hoff, have directed their attention to the cause of osmosis and membrane permeability. L'Hermite 11 offered the first explanation of membrane permeability. His theory was that of selective or preferential solution of the two separated liquids in the membrane. Traube 12 believed that semipermeable membranes acted like atomic sieves, permitting molecules of a certain size to pass through but preventing the passage of larger particles. However it has been demonstrated by Bigelow and Bartell, 15 later by Bartell, 16 and again by Bigelow and Robinson, 17 and also by Tinker, 18 that osmotic phenomena were obtainable with membranes the pore diameters of which were much larger than molecular dimensions. H. B. Armstrong 19 advanced the theory that some kind of chemical association between the membrane pores 14 Van't Hoff: Zeit. Phys. Chem., 1, 481 (1887). 15 Bigelow and Bartell: J. Am. Chem. Soc., 31, 1194 (1909). 16 Bartell: J. Phys. Chem., 15, 659 (1911). 17 Bigelow and Robinson, Ibid., 22, 99, 153 (1918). 18 Tinker: Proc. Roy. Soc., 92A, 357 (1916). 19 Armstrong: Ibid., 81B, 94 (1909). and water took place which prevented the passage of any hydrolated molecules of solute but had no effect on unhydro- lated substancs. The most generally accepted explanation has been that of L/Hermite, whose view has been supported by a large number of experiments described by Nernst, 20 Kahlenberg, 21 Flusin, 22 and others. According to Findley, 23 theories accounting for osmosis have generally fallen into two classes. One, the kinetic in- terpretation, considers that osmotic pressure is due to bom- bardment of the semipermeable membrane by the imprisoned solute molecules, analogous to the kinetic explanation of gaseous pressure. The other view is that the osmotic pressure is the hydrostatic pressure produced by the passage of solvent into the solution. This latter explanation has been the most useful in the experimental study of osmosis, and defines the osmotic pressure as the hydrostatic pressure produced by the entrance of solvent into the solution. In negative osmose, the reverse is the case. The mechanism of the osmotic process has been explained in many ways. In 1891 Jager 24 proposed the surface tension theory. Although the theory has since been subjected to numerous modifications, in its simplest form, it stated that the osmotic pressure was proportional to the difference in surface tension between the solution and the pure solvent. Several years after this, Callender 25 proposed the vapor pressure hy- pothesis, according to which the membrane capillaries were regarded as not wetted by liquids, but rather that they acted as vapor pressure sieves. Perrin 26 and Girard 27 considered that osmosis was controlled chiefly by electrostatic phenomena. It must be borne in mind that Van't Hoff's generalization 20 Nernst: Zeit. Phys. Chem., 6, 35 (1890). 21 Kahlenberg: J. Phys. Chem., 10, 141 (1906). 22 Flusin: Ann. Chem. Phys., (8) 13, 480 (1908). 23 Findley: Osmotic Pressure, p. 99 (1919). 24 Jager: Werner, Ber., 100, 245, 493 (1891). 25 Callender: Proc. Roy. Soc., (A) 80, 466 (1908); Proc. Roy. Inst., 19, 485 (1911). 26 Perrin: Compt. rend., 136, 1388 (1903). "Girard: Ibid., 146, 927 (1908); 150, 1444 (1910); 153, 401 (1911). 10 was developed from Pfeffer's data on the osmosis of dilute sugar solutions, and was applied rigidly only to such dilute non-electrolytes, and only in the case of perfectly semiper- meable membranes. Such membranes are seldom realized in actual practice, and in the vital processes of living organisms, the group of cells which seem to act as osmotic membranes, are more or less permeable to the solute. The tendency of elec- trolytes to produce osmotic pressures of serious non-conformity to the gas laws has often been detected when refined measure- ments were made. Lord Berkeley and E. G. J. Hartley 28 found abnormal osmotic pressures with solutions of calcium, strontium and potassium ferrocyanides, and to explain these anomalies they assumed that the salts existed as ionized double molecules. H. N . Morse 29 found that membranes which proved satisfactory for finding the osmotic pressure of sugar solution, failed to give the quantitative results expected with alkali chlorides. Moreover after standing for some time with these chloride solutions in them, the cells were greatly impaired for use again with sugar. A cell returned very nearly to its original condition after soaking in water for several months. Fouard 30 and others have noticed this same lack of agreement between the experimental and calculated osmotic pressures of solution of electrolytes. This anomalous behavior of solutions of electrolytes has more often been noticed by the biologist and physiologist than by the chemist. The striking behavior of salt solutions with cells and tissues in the presence of acid or alkali, has been a problem, difficult of explanation and has been studied by Loeb, 31 Osterhout, 32 Lillie, 33 Girard, 34 and others. The swelling of 28 Berkeley and Hartley: Phil. Trans., (A) 209, 177, 319 (1908). 29 Morse: Osmotic Pressures of Aqueous Solutions, Carnegie Institution of Washington, p. 211-217 (1914). 30 Fouard: Bull. Soc. Chem., (4) 11, 249-261 (1912). 31 Loeb: Science, 37, 428 (1913). 32 Osterhout: Biol. Chem., 19, 493, 561 (1914). 33 Lillie: Am. Jour. Physiol., 194, (1911). 34 Girard: Compt. rend., 148, 1047, 1186 (1909); 151, 99 (1910); 153, 946 (1911); J. Phys. Path. Gen., 13, 359 (1911); Compt. rend., 155,308 (1912); 156, 1401 (1913); 159, 376 (1914); 167, 351 (1918); 168, 1335 (1919); 169, 92 (1919) 11 muscular tissue and of typical gels has been studied by M. Fischer, 35 Lloyd, 36 and others. Girard studied the osmotic pressures of electrolytes with porous CrCl 3 , gelatin, frog skin and other membranes. He found that the osmotic pressure of electrolytes varied greatly with their nature, as well as with their concentration. He obtained different results with differ- ent membranes, and in some cases recorded even negative osmose. Bartell 37 also observed both positive and negative effects at various concentrations of acetates, chlorides, nitrates and sulfates when porcelain membranes were used. In seeking an explanation of this type of anomalous behavior, Girard enunciated his electrostatic theory. He considered the osmosis of electrolytes to be essentially due to electrical influences and the osmotic process to be dependent upon the same general causes as electro-osmose, namely to the sign of the electrically charged, movable, liquid layer adjacent to the oppositely charged walls of the membrane capillaries, and to the potential difference existing between the two faces of the membrane. He considered the charge on the capillary wall to be due to a small excess of hydrogen or hydroxyl ions, the movable liquid layer assuming an equal but opposite charge. Later work convinced him that the membrane charge could be altered by ions other than hydrogen and hydroxyl. Girard found that the contact potential between two solutions may be raised or lowered, or that the orientation of the potential might even be reversed by the intercalation of a membrane. Other examples of such potential differences exhibited by membranes are found in the work of Loeb, 38 Brunnings, 39 Lillie 40 and Beutner 41 . Thus the membrane seems to play an important part in de- termining the nature of the osmotic effect and of the electrical state of the cell system. This indicates a close relationship be- 35 Fischer: Oedema and Nephritis, 1915. 36 Lloyd: Proc. Roy. Soc., 89B, 277 (1916); Biochem. J., 14, 147 (1920). 37 Bartell: J. Am. Chem. Soc., 36, 646 (1914). 38 Loeb: Science, 34, 884 (1911). 39 Brunnings: Pfluger's Archiv., 48, 241 (1903); 117, 409 (1907). 40 Lillie: Loc. cit. 41 Beutner: J. Phys. Chem., 17, 344 (1913). 12 tween electro-osmose and ordinary osmose. The most striking similarity between the two phenomena is found in the reversal of flow of liquid and in the resulting change of the cell potential due to the effect of acid, base and polyvalent ions. The re- lationship existing between the hydrostatic pressure and the H. M. F. of the electro-osmotic cell system has been stated by Wiedemann, 42 briefly as follows, for a given material, the dif- ference in hydrostatic pressure maintained between the two sides of a porous diaphragm is proportional to the applied potential. Wiedemann 43 stated that Munch obtained reversal of flow to the anode with dilute solutions of both neutral and acidified K 2 CrO 4 . Perrin 26 from his work on electro-osmose came to the conclusion that the acidity or alkalinity of the solution was one of the important factors. With a membrane of porous chromic chloride, acid solutions flowed to the anode while alkalies flowed to the cathode. He also found with cotton wool as a diaphragm that both acids and alkalies flowed to the cathode, and that as the concentration of an acid was increased, the flow decreased, but did not reverse in di- rection. Larguier des Bancels 44 found that diaphragms of wool, silk and cotton cloth were electro-negative against water. The negative charge of each was increased by alkali and no reversals were observed in acid solutions except with silk in .01 M. HC1. Barrett and Harris 45 found no reversal with agar or parchment membranes, but record the acid- alkali reversal with gelatin which was similar to silk. Engel- mann 46 found similar variations in electro-osmose with porous clay, frog's skin, pig's bladder, cat's lung and potato. It is manifestly evident that membranes which are protein sub- stances, more or less amphoteric in general character, appear to show the acid-alkali reversal with greater regularity than other membrane substances. Perrin's laws of contact elec- 42 Wiedemann: Pogg. Ann., 87, 321 (1852); 99, 177 (1856). 43 Wiedemann: Elektricitat., 2, 153 (1883). 44 Larguier des Bancels: Compt. rend., 138, 898 (1904). 45 Barrett and Harris: Zeit. Elekt., 18, 221 (1912). 46 Engelmann: Arch. Neerland, 9, 332 (1874). 13 trification 47 appear to have been confirmed by an unique ex- periment known as the Bose-Guillaume phenomenon. 48 If two wires, one of which is coated with porous material such as gelatin, be placed in a solution and the coated wire given a sudden twist, a momentary E. M. F. is produced, which is detectable with a ballistic galvanometer. This phenomenon may be interpreted as an enforced osmotic effect, in which liquid is momentarily squeezed out of the pores of the wire coating, and since the liquid is charged by contact oppositely to the coating itself, a very brief separation of charges occurs. This gives rise to a potential difference. Girard was convinced from his work that the sign of charge on the membrane regulated the amount of solute diffusing through into the solvent. The correctness of many of Girard 's conclusions are rendered somewhat doubtful by the careful studies of Hamburger. 49 Hamburger first showed that par- titioning off two solutions by a non-protein membrane such as collodion had no effect whatever on the contact potential of the two solutions. This was found to be the case with fairly con- centrated solutions of HC1 and Ce(NO 3 ) 3 , materials furnish- ing ions already known to be highly effective in altering the charges on the majority of negative colloids. From this work it seems reasonable to conclude that extensive selective ionic adsorption, at least for collodion, is not probable. Ham- burger further arrived at the conclusion that the magnitude of charge and the electrical orientation of the membrane had very little influence on the diffusion rate of salts. He also believed that the osmotic flow was in the direction demanded by the calculated diffusion potential of the salt, and also that the magnitude of osmose depended largely on the specific effect of the salt itself. It has been observed by Bayliss 50 and McClendon 51 and 47 Perrin: Compt. rend., 147, 55 (1908). 48 Bose-Guillaume : Compt. rend., 147, 53 (1908). 49 Hamburger: Zeit. Physik. Chem., 92, 385 (1917). 50 Bayliss: Proc. Roy. Soc., 84B, 245 (1911). 51 McClendon: Physical Chemistry of Vital Phenomena, p. 113 (1917). 14 often in this laboratory that the potential across a membrane boundary increased rapidly to a maximum value and receeded again practically to zero, especially with the more concentrated solutions. With thicker membranes such as porcelain, the change was slower, but with thin membranes the effect was very rapid. Bayliss found the maxium value approximated that calculated from Nernst's formula. The rapid fall of the potential has been ascribed to swelling of the membrane and consequent enlargement of the pores. That this view is in error, is evident from the identical effect encountered with porcelain, the pore size of which, it must be admitted is con- stant. Very recently Loeb 52 working with collodion membranes has made potential measurements of osmotic systems of so- called gelatin chloride solutions containing HC1 and finds experimentally the same potential value as calculated from Nernst's formula on the basis of the hydrogen ion concen- tration; again demonstrating the B. M. F. measured to be concentration cell and contact potential. Bartell and Hocker 53 working with porcelain, and Bartell and Madison 54 with gold beaters skin membranes came to the conclusion that osmosis of electrolytes was largely controlled by the same factors as those which are active in electro-osmose, namely a fall in potential along a membrane pore which in- fluences the direction of migration of the electrically charged water layer within the capillary. They considered the source of this potential as due to general diffusion of solute, relative ionic migration velocities of the salt used, and selective ionic adsorption, the usual resultant being a combined effect of the three factors. The usefulness of collodion as a membrane material was recognized in 1855 by Kick 55 who used it in his studies on diffu- sion. The ease with which dialyzing membranes can be pre- 62 Loeb: J. Gen. Physiology, 3, 557 (1921). "Bartell and Hocker: J. Am. Chem. Soc., 38, 1029, 1038 (1916). 64 Bartell and Madison: J. Phys. Chem., 24, 444, 593 (1920). 55 Kick: Pogg. Ann., 94, 59 (1855). 15 pared from' collodion has resulted in its use by many biol- ogists. In 1907, Bigelow and Gemberling 56 found that Poiseuille's law for the flow of liquids through capillaries applied to col- lodion as also to several other osmotic membranes. Mathews 57 made a number of qualitative experiments with collodion as an osmotic membrane, and concluded that it was not a truly semipermeable membrane. He believed that the direction and magnitude of osmose was largely a question of solubility. More recently Loeb 58 has prepared sacs of collodion with which he has studied the influence of electrolytes on the electrifica- tion of the walls of the sacs and also has studied the rate of diffusion of water through them. He found that the rate of diffusion of water was influenced not only by the concentra- tion, but also by the electrical forces. In nearly all of his ex- periments the collodion membrane was impregnated with gela- tin and in this respect his membranes differed from those used in this investigation. II. PURPOSE OF INVESTIGATION. The work undertaken in this investigation may be classed under three heads. 1. The determination of the relationship between the osmose of a number of electrolytes and the electrical proper- ties of the membrane system when collodion membranes are used. 2. The determination of the relationship between mem- brane pore size, osmose, and the rate of salt diffusion through the membrane. 3. The determination of the effect of continuous stirring of the solutions on the rate of osmose and also on the rate of salt diffusion. 56 Bigelow and Gemberling: J. Am. Chem. Soc., 29, 1576 (1907). 57 Mathews: J. Phys. Chem., 14, 281 (1910). 58 Loeb: Loc. cit. 16 III. RELATIONSHIP BETWEEN OSMOSE AND ELECTRI- CAL PROPERTIES OF MEMBRANE Preparation of Membrane The preparation of collodion sacs for dialysis has been undertaken by many investigators, among whom Novy 59 seems to have been the first to have developed a satisfactory and simple technique that yields sacs similar in properties. Loeb 60 has used collodion membranes which were formed on the inside of a 50 cc Erlenmeyer flask, then subsequently removed after drying and wetting with water. It was found by us that this method yielded sacs of varying permeability and thick- ness, also that different portions of the same sac varied con- siderably in these respects. Preliminary work with these sac membranes revealed the fact that to prevent the partial collapse of the sac, when the filled sac with its manometer tube was lowered into a dish of water, the experiments must be started under an hy- drostatic head, which in itself is undesirable for the experi- ment. Even under an hydrostatic head of 20-40 mm, small wrinkles in the sac were not eliminated, and one to five min- utes were necessary before the sac was filled sufficiently to reg- ister osmose on the manometer. Results obtained by us showed conclusively that a sac the shape of which was not maintained rigidly would stretch badly, and in the case of materials giving a strong osmose, one such sac was found to have stretched so that the initial sac volume was increased by 8 cc which represents approximately 15% change in total volume and an even greater percent error in osmose readings. The first problem then was to prepare uniform membranes of any required permeability and thickness, and of such shape that they could be held rigidly in place in the osmotic cell. The method of Bigelow and Gemberling 56 for casting uni- 69 Novy: J. Am. Chem. Soc., 29, 1578 (1907). 60 Loeb: J. Gen. Physiology., 1, 717 (1919); 2, 87, 173 (1919); 2, 387, 659, 673 (1920). 17 form flat membranes on a mercury surface was found to yield excellent results. To obtain membranes with similar properties it was found necessary to control the whole casting and drying process within a closed box. This was done in order to avoid uneven air currents and also to keep the tem- perature constant. The following procedure was found to give membranes of any desired permeability in sheets 12 x 18 inches and of uniform thickness. 100 cc of 3 percent soluble gun cotton, dissolved in a mixture of 75 percent ether and 25 percent alco- hol, was diluted with an equal volume of solvent to reduce its viscosity. This solution was poured evenly over clean mercury in a shallow pan placed within the casting box and allowed to evaporate slowly. At 25 C the casting time was about four hours for a membrane of medium permeability. A shorter drying period yielded more permeable, and a longer drying period or higher temperatures, less permeable mem- branes. The permeability was also affected by varying the proportion of water in the alcohol-ether diluent. The perfectly clear, colorless membrane was cut away from the sides of the casting dish with a razor blade, and turned the other side up for 15 to 20 minutes. The air was then changed in the casting box and the membrane turned over several times at short inter- vals. Very permeable membranes at this juncture began to ap- pear opalescent, while the less permeable did not alter in ap- pearance. The membranes were then immersed in water to dissolve out the remaining ether and alcohol. Bigelow and Gemberling have shown that the permeability of membranes prepared in a somewhat similar manner, in- creased gradually for about a month, after which the change was slight. With a view to shortening the curing period, the effect of soaking the membrane in varying concentrations of alcohol and water, as suggested by Brown 61 was tried. Such curing weakened the membranes so materially that this method was abandoned, and curing in distilled water which 61 Brown: Biochem. J., 9, 320, 591 (1915). 18 was changed daily for thirty days was adopted as the most satisfactory procedure. All of the membranes were then cut with a cork borer, so as to fit the osmose cell, and stored under water until used. The marginal inch from the large cast sheet, by which the latter had been handled during the drying and curing periods was discarded. Construction and Assembly of Osmose Cell The double-compartment osmose cell of the type used by Bartell and Hocker, 53 and later by Bartell and Madison, 54 was used and has proved satisfactory. This cell possesses certain desirable points, such as ease of detection of leakage, elimination of evaporation, elimination of alteration of hy- drostatic pressure due to temperature fluctuation, negligible capillary corrections, comparative simplicity and general ease of operation. As the membrane used in this research possessed neither the rigidity of porcelain, which could be wired into place, nor ,. .... the flexibility of gold-beaters skin, which could be waxed into position, and further- more, had to be kept moist constantly, a different type of holder was of necessity de- vised. The cell used in this work (Fig. 1) was constructed of two glass T-tubes of 15 mm diameter, each of which was fitted into a rubber stopper. These were fitted in turn into a threaded brass collar which carried on one end a circular brass plate. The stopper was held in the collar by a washer secured by a threaded sleeve which could be screwed onto the collar. The half-cells were bolted to- gether through the circular brass plates. Paraffined rubber stoppers were used to close the ends of the cell and also as connectors for the osmometer tubes. The assembled cell was made fast to a substantial support. In assembling the cell, the glass T-tube was adjusted 19 so that the surface of the rubber stopper on which the mem- brane was to be placed, projected slightly beyond the glass so that when the two halves of the cell were bolted together, the glass T-tube would practically touch the membrane. This definitely established the effective membrane area. In order to exclude the possibility of loss of liquid from either com- partment of the cell by leakage around the edge of the mem- brane, the faces of the rubber stoppers were coated with three layers of rubber cement, the first two layers thoroughly dried with the long axis of the cell in a vertical position, and the third coat dried to the "gummy" stage. The membrane which had previously been cut out with a cork borer, was removed by forceps from the water under which it had been stored. It was then dried superficially with filter paper, placed in position on the prepared stopper, and the other half of the cell was bolted on. The bolts were taken up slowly and uniformly so as to avoid buckling of the membrane. The cell was then filled with water and left for at least twenty -four hours before being used. The volume of each compartment was approximately 20 cc. The osmometer tubes were heavy-walled glass tubing of 2.5 mm internal diameter and were calibrated in the usual manner with mercury. The tubes used in the first part of this work agreed in specific volume to within about five per- cent, and were of fairly uniform diameter throughout. A given pair of tubes was always used with the same cell. In setting up a cell for use, the water was first emptied out and the respective compartments rinsed twice with the solutions they were to contain. The compartments were then filled, and the osmometer tubes, previously half filled, were adjusted so that the liquid columns were equal. The rubber stoppers were then immediately waxed into place. In this work, readings of the hydrostatic pressures which de- veloped were taken at two hour intervals over a twelve hour period. At the close of the experiment the cell was emptied, washed with distilled water, and then filled with the water. In order to wash the membrane capillaries free from electrolyte, the cell was put under an hydrostatic head for eight to ten 20 hours. In this way one membrane could be used a number of times, and still give results comparable with a fresh membrane. Results with the two-compartment cell, when the tempera- ture was accurately controlled, were readily duplicable to within one or two percent. What was actually obtained in this work was data showing the rate of flow of solutions through the membrane. In some cases the equilibrium pressure of the solution (expressed in terms of hydrostatic pressure) was determined; that is, readings were recorded when the rate of flow of liquid through the membrane in one direction was just balanced by the rate of flow in the other direction. Such readings will be designated as maxi- mum osmose values. It will be appreciated that the compari- son of the rates of flow of different solutions is by no means an exact way of comparing the absolute osmotic activity of the solutions. It does, however, give us a fairly accurate indi- cation of the order of maximum equilibrium pressures obtain- able with these solutions. Results were obtained (1) with chlorides of metals having different valencies, (2) with potassium salts having inorganic acid radicals of different valencies, (3) with potassium salts having organic acid radicals of different valencies and (4) with hydrochloric acid and with potassium hydroxide. The results obtained are shown in Table I. In this table readings obtained at the end of the two hour, six hour, and twelve hour periods are given. A medium porous membrane (pore diameter = 0.8 to 0.9 micron) was used. The salts were usually recrystallized twice from a good grade of distilled water. The HC1 was redistilled as "constant boiling" from C. P. acid. The KOH was a Kahlbaum product and was freed from carbonate by careful precipitation with Ba(OH) 2 . There was a slight trace of barium left in the solution. Stock solutions of the electrolytes were made up with conductivity water, analyzed, and diluted to the desired concentration. Conductivity water was used in one compartment of the cell in all experiments. The hydrostatic pressure was recorded in millimeters. The temperature was 25 C == 0.5 C. 21 4OO Osmose of Solutions of Chlorides of Different Metals Chlorides of monovalent metals gave a repressed effect at . 1 M con centr ation . This effect persisted throughout the entire 12 hour period. Solutions of chlorides of bivalent and trivalent metals at low concentrations (0.01 to 0.1M) tended to give either a low positive or neg- ative osmose, while the osmose was strongly posi- tive at higher concentrations The general shape of the curve representing initial os- mose rate, (i. e., osmose in mm at the end of a two hour period), plotted against the logarithm of molar con- centration, is the same in all cases as the maximum os- mose curve (12 hour period) plotted in the same manner. Fig. 2. CHLORIDES 12 HOUR OSMOSE MOLAR CONCENTRATION Fig. 2 These curves are shown in The data appear in Table I. Osmose of Solutions of Potassium Salts of Inorganic Acids These experiments were carried out to ascertain the effect of the valence of the anion on osmose. It was found that an increase in valence of anion (with salts which do not hydrolyze greatly) resulted in an increase of the initial osmose rate and also in a higher maximum osmose value. In concentrated solutions, readily hydrolyzable salts of inorganic acids (i. e., carbonate, phosphate, etc.) when com- pared with slightly hydrolyzable salts of inorganic acids of equal anion valence, were found to give an abnormally high osmose. In dilute solutions, the readily hydrolyzable salts 22 M|| W r^ -S W CO *^ *ts ^ H 'hr oi CD .J * # * # * * TtiOOr^tC^HC^ai^CC^OGCOLCiC I rHrHT^COO-sfHLCOiOOCOOOlCrH | iO CO O CQ (N O O t^ O O ^ O Cl^-t^-TFCC O2-O iO i trH(MCCCDCVJr-H. rH CQ CO i-H i HlOCOOC^ IT-IT ICNCO CC C^ rH rH CO LC I I T-H 4 solution) eliminate gassing and are excellent, but cannot be used in certain combinations such as ferrocyanides. Although a great deal of work was done with collodion membranes with a modification of Barratt and Harris' apparatus, and also by Briggs' method, 62 the cataphoresis method was found to be more satisfactory, since it was dependable and at the same time more rapid. A quantity of dried membrane was finely pulverized, after which a fairly stable water suspension of it was made. This suspension was examined under the ultra microscope, and the direction of migration of the collodion particles under an applied potential noted. The direction of motion of finely divided collodion par- ticles in water was found to be slowly to the anode which indicated that collodion bears a small negative electrical charge with respect to water. In all four dilutions of each of the fifteen electrolytes tried, the collodion was likewise neg- 62 Briggs: Jour. Phys. Chem., 22, 256 (1918). 27 4- -f- | - * f t A - f ~? ! ] - B ative. It also appears that no amount of acid up to and in- cluding 10M HC1 will cause it to become charged positively, nor will any concentration of salts of trivalent cation, such as aluminum, produce this effect. This appears to be an extreme case of an "unprecipitable" suspension. These results con- firm those of Loeb and also those of Gyemant, 63 which were obtained by the electro-osmose method. In this respect collodion is different from any other membrane we have used. General Discussion and Conclusions It will be seen from the data already presented that the membrane charge and the orientation of the cell systems for chlorides of Na, Ca, Mg and Al are given according to B of Fig. 5. The diagram represents one pore through the mem- brane. The solution side is understood to be above the membrane and the water be- neath. The arrow on the left, pointing up toward the solu- tion, represents the normal osmotic tendency while the one on the right represents the imposed electrical effect. Apply- ing this idea we see that the exosmotic tendency op- poses the normal osmotic tendency and we should expect a low degree of positive osmose (Na, Al) or even a negative osmose, which is actually realized with the case of Ca and Mg. At high concentrations, the normal osmotic tendency overcomes the imposed exosmotic effect, and positive osmose results. The condition with HC1 (also with KC1) is shown in A of Fig. 5, the orientation being the opposite of that in the above case. The osmotic tendencies should operate in the same direction and a high positive osmose should result. While in dilute solutions of HC1, positive osmose is obtained which is of greater magnitude than with Na, Ca, Mg, and Al, these latter substances in concentrated solutions give an osmose 63 Gyemant: Kolloid Zeit., 28, 103 (1921). 28 which materially exceeds that with HC1. It might be argued, though admittedly not at all conclusively, that except in the case of NaCl, this is due to the greater number of ionic units fur- nished by the dissociation of the more complex salts. The peculiar case of KOH with its reversal of orientation with change of concentration, is an admirable illustration of the applicability of the electrical theory. In dilute solutions, the condition of the system is such that a strong positive ten- dency should result. (A, of Fig. 5.) The results are in accord with this prediction. In a more concentrated solution, how- ever, the orientation changes and conditions are as in B, giv- ing a weak positive effect; with still higher concentrations, a strong negative effect results. All the potassium salts studied gave orientations of the cell system corresponding to A, Fig. 5, which would lead us to infer that abnormally positive osmose should result. This has been shown to be true (with the exception of K 2 C2O4) so far as initial osmose rate is concerned. The maximum osmose, however, shows that some factor, of which time is an important element, plays a part in reducing the initial high osmose rate to a considerably lower value. Summary and Conclusions 1. A method for preparing flat collodion membranes of uniform thickness and permeability is described. Methods for controlling the permeability to any desired value are given. 2. A non-leaking osmotic cell of two T-shaped glass compartments of equal volume, held together by a machined brass coupling which also holds the membrane firmly between the two compartments, has been developed and used success- fully. 3. The osmose values obtained with thirteen different salts, an acid and a base (chosen to show the relative influence of cation and anion valence, hydrolyzability of salt, and weak and strong acid radical) are given for 0.001, 0.01, 0.1, and 1M concentrations. 29 4. The maximum potential of each of these osmotic systems was measured and recorded. 5. The sign of the membrane charge has been determined by cataphoresis, using finely ground membrane material in suspension. 6. Consideration of the data obtained shows that the initial osmose rate of practically all the salts examined bears a definite relationship to the electrical properties of the mem- brane system. 7. The anomalous effects obtained with collodion are very similar to those obtained with membranes of porcelain, of gold beater's skin, of calf's bladder, and of parchment paper. The maximal and minimal values obtained with these different membranes do not come at exactly the same concentrations, but when consideration is given to the exact condition of the electrical orientations of the different membrane systems, the results are closely comparable. 8. It has been shown that anomalous effects are related somewhat to the time factor. For example the data for the osmose of potassium carbonate at the end of a two hour period, when plotted against log of concentration of potassium car- bonate gave no N shaped curve while the N shape was pro- nounced for the curve obtained at the end of a 12 hour osmose period. This fact makes it appear probable that the process of diffusion is in some way responsible for the repressing ef- fects noted at the intermediate concentrations. IV. RELATIONSHIP BETWEEN MEMBRANE PORE SIZE, OSMOSE AND RATE OF SALT DIFFUSION THROUGH THE MEMBRANE In the foregoing section of the thesis, it was shown that the direction and magnitude of initial osmose of solutions of a large number of salts, also of acid and alkali, with collodion membranes were in accord with the anomalous osmose theory '30 as previously outlined by Bartell and co-workers. 53 ' 54 ' 64 In the section above referred to, attention was called to the change in shape of the osmose rate curve with time. A so-called straight line curve was obtained when the initial osmose values (two hour osmose readings) were plotted against the concentration, while a peculiar N shaped curve was obtained when the maxi- mum osmose values (i. e., obtained after 12 hours or more) were plotted against concentration. The latter exhibited a minimum point at about 0.1 M concentration. From the fact that the straight line initial osmose rate- concentration curve developed into the N shaped curve with time as the only apparent influencing factor, it was in- ferred that diffusion of electrolyte was the active factor di- rectly responsible for the decided alteration of osmose rate through this concentration range. In order to test this idea, collodion membranes of different degrees of permeability were prepared in the manner described in the previous paper. The number of capillary holes through a given area of membrane will be shown to be nearly the same in all cases. The actual diameter of the capillaries was regu- lated at the time of setting and by the amount of' evaporation of solvent which occurred before soaking in water. Permeability of the Membranes The permeability of the membranes used in this work was determined by the hydrostatic method of Bigelow and Gemberling. 56 A known and constant hydrostatic pressure was applied to water in contact with the membrane and the rate of passage of water through the membrane accurately noted on a calibrated capillary tube by means of a cathetometer. The cells were sealed with wax around the rubber stoppers, so that leakage was eliminated. They were then immersed in a thermostat. An idea of the reproducibility of the col- lodion membranes and the constancy of the effective membrane areas of the various cells may be gathered from the close agree- 64 Bartell and Sims: J. Am. Chem. Soc., 44, 289 (1922). 31 TABLE III Membrane Permeability (Very Permeable Membrane) Membrane no. Pressure applied, mm HaO Water passed, cu. mm Time, sec. Cu. mm X 10 ~ 4 water passing per sec. per sq. cm diaph. area 1 2 3 4 385 385 385 385 18.48 19.50 23.16 18.90 337 355 421 344 319 319.3 320 319.5 Average 319.45 Diaphragm Area = 171.2 sq. mm. Temperature = 20 C TABLE IV Relative Membrane Permeability Temperature = 20 C Least permeable membrane Pressure applied, mm H2O Cu. mm X 10 ~ 4 water passing per sec. per sq. cm. dia. area Cu. mm X 10 ~ 4 water passing per sec. per sq. cm. area per mm pressure Relative permeability 485 385 285 185 54.76 42.93 31.84 21.37 0.1129 0.1115 0.1117 0.1155 - Ave. 0.1129 1 Medium permeable membrane 492 109.32 0.2221 392 87.82 0.2240 292 64.84 0.2220 192 43.29 0.2255 0.2234 1.978 Very permeable membrane 485 403.10 0.8311 385 319.45 0.8297 285 236.20 0.8288 190 158.08 0.8320 0.8304 7.355 32 ment of the results. Absolute results involving correction for viscosity of water at 20 C were not attempted, as compara- tive data at the one temperature were all that was desired to establish relative permeability. The preceding table was compiled from average values obtained in the same manner as shown in the foregoing table. From Poiseuille's Law for the flow of liquids through capil- KwPD 4 T lary tubes, we have the expression Q = . Q represents L the quantity of liquid passing through n, number of capillaries of diameter D and length L, in the time T, and under pressure P. K is a constant, the value of which is dependent upon the liquid used, the temperature, etc. It is assumed that the length of the capillary pore in the various membranes is a function of the thickness of the membrane. Under the conditions of the ex- periment, K, P, and T may be combined into a single constant kriD* k and the equation rewritten Q = -^ . If we assume for JL/ the time being that n is a constant for all three membranes, we are able to compare the relative pore diameters of the membranes. The actual thicknesses of the three membranes were, re- spectively, 0.08 mm, 0.15 mm, and 0.31 mm. The relative diameters of the pores of the three mem- branes, calculated as above, give the values 1 : 1.38: 2.31. Measurement of Membrane Pore Size The membrane pore size was measured directly by the method, based on Jurin's Law, as used by Bigelow and Bartell. 15 This gives us information relative to the assumption made con- cerning the number of pores in each of the membranes used. In case the relative pore diameters as measured by the two independent methods agree, we are justified in the previous assumption that the number of pores in a given area of mem- brane is the same for all three membranes. The membrane was supported in a holder by a fine-mesh 33 wire screen. Owing to the nature of the membrane, which was somewhat elastic, and which stretched under the pressure necessary to force water out of the pores, the values here reported must be considered as representing pores which have been slightly stretched. Results obtained follow: TABLE V Membrane Pore Size K. Cms pressure per sq. cm. Pore diameter, microns Relative Pore Diameter Jurin's Law Poiseuille's Law Least permeable membrane Medium permeable membrane Very permeable membrane 4.23 3.17 1.76 0.701 0.934 1.681 1.00 1.33 2.39 1.00 1.38 2.31 The above agreement between the values for relative pore diameter, as determined by the two different methods, indi- cates the validity of the assumption that the number of pores in a given cross-sectional area of the various membranes is very nearly the same, at least within 3 to 4%. We shall throughout this paper, distinguish between these three mem- branes by referring to them as (a) least permeable, (b) medium permeable, and (c) very permeable. Osmose through Membranes of Different Degrees of Permeability The validity of the hypothesis that diffusion was respon- sible for the change in shape of the osmose rate-concentration curve, may be tested by comparing the osmose through the membranes of different permeabilities. The membrane of least permeability should show an initial rate-concentration curve which approaches more nearly a straight line than do the curves obtained with the other membranes of greater permeability. The membrane of medium permeability should show less of the straight line effect, and should rapidly develop 34 the N type of curve, while the very permeable membrane should probably show no straight line effect at all, but should give instead the N shaped curve from the beginning. In' all of the work de- scribed in this paper, very carefully selected osmometer tubes of 2.5 mm internal bore were used. These were selected from among over four hundred tubes and the volumes of given lengths of them varied not more than one percent. This uniform- ity was considered satisfac- tory. The duplicability of osmose experiments with these refinements and with good temperature control was close to one percent. Inasmuch as it was de- sired to obtain results when membranes of different per- meabilities were used, both Fig. 6 Development of N shaped Osmose Curve with representative potas- as influenced by time in case of KC1. . . , . *:< Least permeable membrane. smm salts wlth amons of dlf- Time in hours ferent valencies and with salts of bivalent and triva- lent metals, the following were used: KC1, CaCl 2 , A1C1 3 , K 2 SO 4 , and KaFeCeNe. In addition, sucrose was also used in order to ascertain how the osmose of a substance supposed to give a normal osmose rate was affected by permeability. One concen- tration in addition to those used in the previous work was in- troduced, namely, 0.004 M. The recession in the N shaped curve started to develop with potassium salts at about 0.004 M concentration. In place of 1 M concentration 0.5 M was sub- stituted, the osmose in this region of concentration was of the positive type, and proved to be of no special import in this in- 35 vestigation. The hydrostatic pressure was read at thirty min- ute intervals for the first two hours, after which two hour read- ings were generally made. Close watch, however, was kept on the experiments as they neared the maximum, so that the ex- treme pressure developed would be known. The accompany- ing Table VI for KC1 contains data such as was obtained for all solutions investigated. In order to economize space, only values read at the end of 2 hr., 12 hr., and at maximum 100 .01 .1 MOLAR CONCENTRATION Fig. 7 Development of N shaped Osmose Curve as influenced by time in case of K 2 SO 4 . Least permeable membrane. Time in hours .001 .01 .1 MOLAR COA/CA/T/?AT/0A/ Fig. 8 Development of N shaped Osmose Curve as influenced by time in case of K 3 Fe(CN)i. Least permeable membrane. Time in hours osmose will be given. The development of the N shaped curve with time is shown for potassium salts in Figures 6, 7, and 8, the time being expressed in hours. The effect of in- creasing the membrane pore diameter on the initial osmose rate-concentration curve is shown in Figures 9, 10, and 11. 36 500 MOLAR CONCENTRATION Fig. 9 200 Fig. 10 Initial Osmose Rate (2 Hrs) of Potassium Initial Osmose Rate (2 Hrs) of Potassium Salts. Salts. With least permeable membrane With medium permeable membrane .001 .01 MOLAR CONCENTRATION Fig. 11 Initial Osmose Rate (2 Hrs) of Potassium Initial Osmose Rate (2 Hrs) of K 3 Fe(CN) 6 Salts. With membranes of different With very permeable membrane permeabilities 37 Figure 12 shows the initial (2 hr.) osmose rate-concentration curve for KsFeCeNe with all three membranes. TABLE VI Osmose of KC1 (Least Permeable Membrane) Osmose period, hrs. 0.001 M 0.004M 0.01 M 0.1 M 0.5M 0.5 2 4 4 7 18 1 4 8.5 10 12 32 1.5 6 12 15 16.5 44.5 2 8 16 19 21 56 4 13 30 33 34 74 6 17.5 42 42.2 40 86 8 21.5 50 49.5 44 93 10 25 55 54.5 46.5 98 12 28.5 59 57 48 98 14 28.5 62 49.5 16 59 18 67 20 59 22 67 Summary The following generalizations may be drawn from the fore- going data: 1. The least permeable membranes, i. e., the ones with pore diameters less than 0.7 microns, during the first five hours showed a tendency to give only positive osmose with all the salts used with the exception of CaCl2, which gave a negative osmose. At no time during the first few hours did the N shaped curve become pronounced. (K 2 SO4 solutions gave a slight N shape with this membrane.) These facts indicate that the phenomenon of negative osmose is not entirely de- pendent upon the specific salt used, neither is it dependent entirely upon concentration, but its appearance is dependent also, and to a large degree, upon the diameter of the membrane pores. It seems probable that negative osmose does not oc- cur until the diameter of the membrane pore reaches a certain limiting value, which is definite for a given salt solution. In this work we have noted that when negative tendencies pre- 38 a p 8 ^ 'S V rV ;! 2 rt to | I 1 .3 i I tO rH Oi CO rH CO iH tO C^ 1> rH CO O rH 00 d tO rH CO Tj 1 ^" !Q ^ to g oi <>q o oo os tO | 00 O CO o 1 cq co C5 g to co 1 1 8 00 t> CO CO CO 1 oq rn co cp 1 C5 T 1 T1 g co co cq o o co tO 00 C ; rj s e r-T II to O ^ O CO l> !> t^ O O GJ O rH oo cq tO 1>- O CO iO CO tO rH OO CO Oi 00 I T 1 rH rH 00 jqCjqoqcOO T^H T^H " ^^ ^i O5 | rH b- Oq tO CO rH C^ ^d^ *^H rH C^J to co cq o co co CO tO CO r^ O O CO rH C^q rH iO CO CO OO rH O5 >O cq i co to t^ cq to oq co Oi 0^ o cq ^ CO | TtH CO CO Oq Cl CO t cq | to CO oq rH tO 03 CO Cq O rH rH CO O V 3 rrt # -#!~s W i ** Q oj O r^, C^ rj ^2 PH tj Very perme membrane * 8E 3 Oi c T^ 1 IT- t N tO 2 1 CO CO CO O O5 ^ !> CO to CO O CO tO tO to a tO TJH T I I tO 00 i.O a 1>- rH 10 rH to N w EB 1 cq ^t 1 oo O5 I>- rH rH rH to tO tO X tO 1 ^5 co ^ co CO tO rH C^l >0 rH rH 1 zr. rH T^ O CO ^ to o O CM OS CO 1 rH cq r^H tO ,0 10 >Q O ^ ^g v I ^ ^ O D SSI 39 I I g o ti I o 53 '. PH 3 8 1 GO t^ O to Oi to O CO T i 10 O Gi O CO CO tO CO CO t^ 00 CO O T^ LO T^ oo oo cq cq o ga O .s io i ja Cq 00 CO CO CO CO CO CO 00 T^ ^ 00 i 10 LO oo cq cq 10 i O^ CO t^ ^"^ ^^ *-O TI (M o co T-I eq Tt< rH CO T-I iO I>- O O tO O LO T-I 00 CO O5 C5 1 T-H T 1 T 1 rH | O^ ^^ 1^ CO a T-H 5B ^^,00^0 O CO C5 cq co cq o co cq ci T-H THH to p o | T-H T-I CO T-H , ^Cq 1 S-9 . CO O CO CO O O ^ co oq ^h cq c^ CO CO CO TH CO CO 8 OH o S to & iO tO tO cr? o I>(M OS CO CQ Tfi -43 <. 2 ^ C1 >FeC 6 N 6 . For concentrated solutions the order for the least porous membrane was Cl > /// SO 4 >FeC 6 N 6 . This latter order was the inverse of the order 45 of magnitude of maximum osmose. Comparison of the rates of passage of chlorides into the water compartment showed the cation order in dilute solution to be Al > Ca > K. Exactly the opposite order K > Ca > Al was found for concentrated solutions. The order for magnitude of osmose with dilute solutions, bore no apparent relation to the cation diffusion order, however, the order of magnitude of osmose with concentrated solutions was the inverse of the cation diffusion order. 4. With medium porous membranes (i. e., with pore diam- eters of about 0.93 microns) the anion order for rate of diffu- sion of potassium salts from dilute solutions was SC>4 > Cl > /// FeC 6 N 6 and with concentrated solutions was Cl > SO 4 > e. This was again the inverse of the order of maximum osmose for concentrated solutions. The cation order for rate of diffusion was Ca > K > Al in dilute solutions, and Ca > Al > K in concentrated solutions. 5. For very permeable membranes (i. e., with pore diameters of about 1.6 microns), comparison of the amounts of potas- sium salts diffused during the period of maximum osmose, /// showed the anion order to be Cl > SO 4 > FeC 6 N 6 for both dilute and concentrated solutions. This order is the inverse of the order of maximum osmose. For chlorides, the cation order with the most porous membrane was K > Ca > Al in both dilute and concentrated solutions. This order was the same as the maximum osmose order in very dilute solutions, but was the converse of the order of maximum osmose in con- centrated solutions. 6. If the pore diameter of an unstretched membrane, i. e., one not subjected to the high pressures such as were used in measuring pore diameters, is as we believe, slightly less than the diameter as measured with the stretched membrane, these membranes then have pore diameters of just about the same magnitudes as the copper ferrocyanide membranes in- vestigated by Bar tell 67 which were on the border line between osmotic effect and no osmotic effect. 67 Bartell: Jour. Phys. Chem., 16, 318 (1912). 46 7. Tinker 18 made a microscopic study of copper ferro- cyanide gels and found: "The structure seems to be that of a somewhat irregular network having a mesh of the order 0.5 to 1.0 microns. In this respect, it seems to be similar to most of the other ordinary gels, such as gelatine, silicic acid, etc., which have been examined with great thoroughness by Butschli, Hardy, van Bemmelen, von Weimarn, Quincke, Zsigmbndy, Pauli and others. There seems to be a general agreement at the present time that most gels consist of a lattice work sys- tem in which a solid or semi-solid phase encloses a more liquid phase." The results of this investigation tend to corroborate the above view. The lattice work of the collodion membrane may be formed in such a manner that the net work forms a mesh of the order 0.5 to 1.0 microns. If the liquid phase is replaced by water when the lattice work is of this order the lattice work remains unaltered for months even though con- stantly subjected to nitration tests. 8. The work described in this section has clearly brought out the fact that the pore diameter of an osmotic membrane is a highly important factor in determining the exact nature of osmose. Furthermore, it seems probable that the phe- nomena of anomalous osmose and the attending salt diffusion is governed largely by the precise pore diameter of the osmotic membrane. V. EFFECT OF STIRRING SOLUTIONS DURING OSMOSIS The experimental work described in this section of the thesis was undertaken to test further the hypothesis of the author that passage of solute through the membrane was largely respon- sible for the N shaped curves observed with collodion mem- branes. If this hypothesis is correct, we shall expect that even with the least permeable collodion membrane (pore diameter = 0.7 micron) the N shaped curve will be initially apparent in the case of potassium salts when the solutions are stirred. In 47 addition, we should expect a much greater positive osmose in the shaken cell with potassium salts giving the correct elec- trical orientation for abnormally positive osmose, for the rea- son that in the shaken cell, diffused solute should be rapidly removed from the face of the membrane on the water side, thereby maintaining a greater potential and concentration gradient across the membrane. In the cases of CaCl 2 and Aids, salts giving an electrical orientation favorable to neg- ative osmose, we shall expect a greater negative effect in the shaken than in the stationary cell. Diffusion of solute in the shaken cell may be so augmented by the effect of the con- tinuous stirring that an N shaped osmose curve may also be observable with A1C1 3 , which hitherto has shown no such tendency in our experiments with membranes of different pore size. It has been shown by Kahlenberg 21 that the osmose rate is materially changed by stirring the solution in the cell. Cohen and Commelin 68 also report osmose experiments in which the solutions were stirred, but they could not duplicate their results. A decided change in osmose values has often been observed in this laboratory when the stationary osmose cells have been accidentally disturbed. The contents of the osmose cell of the T-type was found to be stirred to uniform composition very quickly by placing a metal ball in each compartment, then rocking the entire cell slowly back and forth. This method of stirring possessed none of the mechanical difficulties such as are encountered when externally driven rotary stirrers are used. Changes in con- centration in each compartment of the cell were noted by measuring the conductivity of the solutions. Construction and Assembly of Rocking- Cell The construction of the cell, shown in Figure 13, was the same as used in our previous work with the exception of the introduction of two circular platinum electrodes, connections 68 Cohen and Commelin: Zeit. phys. Chem., 64, 1 (1908). 48 to which were made through glass inseals in each compartment. The inseal glass tubing was cut off at an appropriate length so as to serve as a mercury cup through which electrical contact with the outside of the cell could be established. The cell was assembled in the usual manner. The plati- nizing of the electrodes and the washing was done in the usual way and before the membrane was put in place. The as- sembled cell was rinsed with several changes of water and then placed under an hydrostatic pressure to clean the membrane capillaries. The conductivity constant for each chamber of the cell was determined before each pair of osmotic experiments Fig. 13 Osmose Cell, Rocking-Type with Electrodes (M/50 KC1 solution being used for this purpose). It was found that the specific conductance of the M/50 KC1 at first changed gradually at constant temperature. This was traced to adsorption of KC1 from the solution by the membrane. It was found necessary to allow the membrane to come to equi- librium with the KC1 solution after which this solution was replaced with fresh KC1 solution before the cell constant could be determined. This manipulation was found to yield reproduceable results. After the cell constant had been de- termined, the cell was rinsed with water several times, and then the hydrostatic pressure applied; finally the cell was again rinsed with conductivity water. 49 The Thermostat and Rocking Machine The thermostat was a large air bath, approximately 2 x 4 x 6 feet high, electrically heated, and cooled by a large metal cold water coil. Rapid circulation of air was maintained by an electric fan. The temperature was 25 0=t 0.02 C. During operation the current in the heater was thrown off and on by a sensitive toluene regulator approximately every five seconds. Within the thermostat was an electric motor which through reducing gears drove the rocking platform, which carried the osmotic cell, at a slow and even rate. Uprights were attached to the platform, and carried between them a cross piece with narrow slots cut at the proper distance to support the osmo- meter tubes. Two coiled wire springs of equal tension were at- tached from the base of the apparatus to each side of the up- right. This avoided any jerking motion of the rocking platform during operation. The front and one side of the thermostat were of glass. The front was in two sections, the upper and lower halves of which could be raised or lowered as occasion required. Method of Setting up Cell for Experiment Previous to setting up the cell, the solutions and conduc- tivity water to be used were allowed to come to the tempera- ture of the thermostat. The long osmometer tubes were then half filled with solution or water, depending on the chambers with which they were to be used. Each compartment of the osmose cell was then rinsed twice with the solution it was to contain, filled with solution, air bubbles removed and a monel metal ball lowered into each compartment. The osmo- meter tubes were next inserted and adjusted to the same level. The apparatus was then fastened onto the rocking platform and the stoppers of the osmometer tubes waxed into place. At the close of an experiment, the rocking machine was stopped, the cells taken out and the solutions and metal balls removed by taking out the stoppers in the two ends of the cell. The osmometer tubes were then removed and the cell washed in the usual manner. 50 X 3 W 3 ^ *O O CO tQ O O ^O 1^ rH t^ f^ r" O 1>- TH rH C^ CO If d 11 TH.CO oo co co a bo 5=1 i i d CO "^^ ^^ CO i~~i t^ CNl rH CO 1 03 1 1 CO t^ CO 00 Oi CO 00 00 rH 00 CNJ rH 1 w d '? rH a ^ 00 CO rH CO Q o TH (M *s d a . ,_, (M O 00 CO O5 8 5 rH^ CXI o H 10 d O CO CVJ 00 (M Cf Oi TH O Oi rH T (M rH TH r- a o rH OO >O CO I>- O5 O r 1 d 1 & s LO CO iO 1>- C^J 00 O CO | TH ti !> rH t>. !> t- 1> TH rH Ol CO C* rH bJO _ ,a . CO 1 & rH {H C rH rH rH rH CM C tions.witl 1 1 Osmose t^ CO 00 CO CO L( CO TH TH CO LO C C\q r _l JY^ s d D w: +H ^ TH CSI 00 C ^.a o "S CJ s n iO (M CO 00 CO CO C5 00 rH O C^ rH I tH O ; s d a> t/5 6 ^ TH TH O C<1 TH s ^.3 *-p 'o 02 a Osmose O iO CO CO l>- GO rH Ci I> TH CO Jp CD CO 5 d II TH 00 TH 00 00 CNl rH rH rH 'g .1 S a C/3 LO O O CO TH CX| TH LC rH rH o ^Q O o t/3 ' < JJ a, ^ g 1 1 1 Q k >f k>* k> ""^ r O ^ .51 Osmose Results in the Rocking 1 Cell Readings were taken every half hour for the first two hours and then every two hours thereafter, until maximum osmose values were reached. The results at the end of the two hour period, twelve hour period and maximum osmose are given in Tables XI and XII. Comparisons of the effect of stirring and not stirring on the osmose of A1C1 3 and K 3 FeC 6 N 6 are shown in Figures 14 and 15, respectively. The following generalizations concerning the effects of stirring on osmose were disclosed by the foregoing data. 1 . The shape of the ' 'initial osmose concentration curves' ' (2 hour readings) and the ' 'maximum osmose concentration curves" for the potassium salts, were of the N type throughout and were more strongly ac- centuated than when the solu- tions were not stirred. 2. With dilute solutions stirring increased both the initial and maximum osmotic effects in the case of KC1, CaCl 2 , and K 2 SO 4 , but de- creased the effect with AlCls and the maximum effect with K 3 FeC 6 N 6 . 3. With concentrated solutions of electrolytes, greater than 0.1 M, stirring had but little influence on either the initial rate or maxi- mum osmose. 4. Stirring produced a tendency toward the N shaped curve even in the case of AlCla Fig. 14 which tendency hitherto had Osmose Rate of A1CU comparing Effect not been observed, of Stirring vs. not Stirring Solutions 5. With SUCrose SOlu- 52 tions, stirring increased the initial osmose rate, but decreased the maximum osmose. 6. The above results are in harmony with the hypothesis that passage of solute through the membrane is largely re- sponsible for the appearance of the N shaped curve in osmosis through collodion membranes. Passage of Salt through the Membrane during* Osmose in the Rocking- Cell For the purpose of measurement of the concentration of solute in each compartment during the osmotic process, the specific conductance of the re- spective solutions was deter- mined. The specific conduc- tance of the various electro- lytes was calculated at all known dilutions from molar and equivalent conductivity data already published by Kohlrausch 69 and Jones. 70 It was originally planned to determine the hydrogen ion concentration of each com- partment as osmose de- veloped. It was believed that this might throw some light on the problem. This would be of particular interest in cases of highly hydrolyzable salts, such as A1C1 3 , where the Osmose Rate of KsFe (CN) 6 comparing HC1 Undoubtedly passes into Effect of stirring vs. not Stirring the water compartment more rapidly than A1(OH) 3 . These data could not be obtained in the time available for this re- search, but a correction can be applied to the conductivity 8//-Y //// ~' MOLAR COHCENTRAT/ON 69 Kohlrausch and Holborn: "Leitvermogen der Elektrolyte" (1898). 70 Jones: Carnegie Institute of Washington, Publ. 170 (1912). 53 measurements hereinafter given, from data which may be pre- sented at some future time. The ratio of salt concentrations on the two sides of the membrane was calculated from conductivity measurements for the two hour period and also for the maximum osmose period. Several experiments continued beyond the maximum TABLE XIII Molar Concentrations of Solutions Two Hours after starting Osmose Experiment Rocking Cell Salt solution Initial molar cone, of salt Molar cone, of salt. Compart- ment Ci Molar cone, of salt. Compart- ment Ca Ratio Qi/Ci K 3 FeC 6 N 6 0.001 0.000981 0.0000212 46.3 0.004 0.0*03795 0.000218 17.4 0.01 0.00919 0.000862 10.65 0.1 0.0907 0.01036 8.76 0.5 0.422 0.0597 7.29 K2S0 4 0.001 0.000956 0.0000405 23.6 0.004 0.003806 0.000248 15.3 0.01 0.00907 0.000947 9.58 0.1 0.0879 0.01201 7.31 0.5 0.4151 0.0760 5.47 KC1 0.001 0.000977 0.0000768 12.7 0.004 0.00347 0.000613 5.65 0.01 0.00872 0.00152 5.75 0.1 0.0784 0.0183 4.77 0.5 0.423 0.1195 3.54 CaCl 2 0.001 0.000971 0.0000684 14.0 0.004 0.00373 0.000443 8.42 0.01 0.00899 0.00124 7.25 0.1 0.0899 0.0124 7.25 0.5 0.456 0.0661 6.89 A1C1 3 0.002 0.00178 0.000198 9.0 0.008 0.00674 0.000671 10.03 0.02 0.01601 0.00165 9.71 0.2 0.1628 0.0219 7.43 osmose period showed that eventually identical concentra- tion in both chambers was reached and at this point the osmose value had diminished to zero. The "least per- meable" membranes were used throughout this part of the work. 54 Summary The foregoing data warrant the following conclusions: 1. Stirring increased the rate of passage of all electrolytes through the membrane at all concentrations. The relative increase in diffusion rate due to stirring was greatest in dilute solutions. TABLE XIV Molar Concentrations of Solutions when Maximum was registered Rocking Cell Salt solution Initial molar cone, of salt Molar cone, of salt. Compart- ment Ci Molar cone, of salt. Compart- ment C 2 Ratio Q/C, K 3 FeC 6 N 6 0.001 0.000761 0.000183 4.16 0.004 0. '002378 0.001384 1.72 0.01 0.00595 0.003876 1.53 0.1 0.0635 0.0438 1.45 0.5 0.2973 0.2128 1.40 K 2 SO 4 0.001 0.00746 0.000276 2.70 0.004 0.002526 0.001662 1.52 0.01 0.005895 0.00424 1.39 0.1 0.0676 0.03868 1.75 0.5 0.2966 0.2190 1.35 KC1 0.001 0.000640 0.000471 1.356 0.004 0.002315 0.001755 1.320 0.01 0.00612 0.00410 1.493 0.1 0.0546 0.0426 1.281 0.5 0.2765 0.2338 1.182 CaCl 2 0.001 0.000792 0.000237 3.34 0.004 0.003024 0.001237 2.45 0.01 . 0.00618 0.00420 1.47 0.1 0.0656 0.0432 1.52 0.5 0.3182 0.2300 1.38 A1C1 3 0.002 0.00121 0.000714 1.70 0.008 0.00445 0.003017 1.47 0.02 0.0088 0.00658 1.65 0.2 0.1154 0.0803 1.44 2. The rate of passage of solutes into the water compart- ment was greater for concentrated solutions than dilute, but in no case proportional to concentration, except in two cases with CaCl 2 . 3. In the region of 0.01 M concentration, the percent 55 increase in mols of salt passing through the membrane in two hours, due to stirring, was as follows: KC1 140, K 2 SO 4 13, K 3 FeC 6 N 6 63, CaCl 2 72, A1C1 3 95. 4. The ratio of concentrations of solute in the two com- partments decreased in magnitude as osmose progressed. In TABLE XV Millimols of Salt passing through Least Permeable Membrane into Water Compartment during Two Hour Osmose Period Salt Initial molar cone, in salt compartment Stationary cell A Rocking cell B B A K 3 FeC 6 N 6 K2S0 4 KC1 CaCl 2 AlCla 0.001 0.004 0.01 0.1 0.5 0.001 0.004 0.01 0.1 0.5 0.001 0.004 0.01 0.1 0.5 0.001 0.004 0.01 0.1 0.5 0.002 0.008 0.02 0.2 0.00036 0.00371 0.01465 0.1760 0.9840 0.00069 0.00421 0.0163 0.2040 1.291 0.00130 0.0104 0.0258 0.3110 2.032 0.00118 0.00755 0.02105 0.2105 1.125 0.00336 0.01141 0.02801 0.3722 0.0015 0.0090 0.164 2.46 1.62 1.07 0.0034 0.0144 0.1803 1.23 1.13 1.15 0.0020 0.0107 0.217 5.20 2.41 1.33 0.00275 0.01225 0.199 . 2.76 1.73 1.05 0.00587 0.0234 0.282 1.94 1.21 1.32 concentrated solutions this ratio approached 1 : 1 (approx- imately 1: 1.4) at the maximum osmose period. For more dilute solutions, this ratio remained much greater. 5. With potassium salts, the relation of anion valence to rate of passage of electrolyte through the membrane showed 56 the following order in each case: Cr>SO 4 ' r >FeG 6 N 6 ' / . This order was the same at all concentrations and was the inverse of the order of magnitude of maximum osmose. With chlor- ides, the cation order was K>Ca>Al at all concentrations. This cation order was the inverse of the order of magnitude of maximum osmose in concentrated solutions. The above orders were the same as those obtained when the solutions were not stirred. 6. Stirring does accentuate the N shaped osmose curve. VI. GENERAL SUMMARY. 1. The theory of anomalous osmose as advanced by Bartell and co-workers has been examined with respect to collodion membranes and found to apply. 2. Certain anomalous effects involving time as a factor in their appearance (the N shaped curves in the advance stage of osmose), not directly covered by the above theory, have been shown to be due to the passage of solute through the membrane into the water compartment. This materially alters the potential and ' concentration gradients across the membrane and gives the observed abnormal osmotic effects. AN INITIAL FINE OF 25 CENTS WILL BE ASSESSED FOR FAILURE TO RETURN THIS BOOK ON THE DATE DUE. THE PENALTY WILL INCREASE TO 5O CENTS ON THE FOURTH DAY AND TO $1.OO ON THE SEVENTH DAY OVERDUE. * A " 4 19* RECD UD 1 AN 1 18 1938 nrT6 '64-8 PM Uvl u IV!*V 9,1 '#40 JAhj 1*2 194 *\K iey5*?% '.(_r> i A .. i ^^ jftN ! (61 WJ^^ W" LD 21-95m-7,'37 53959! UNIVERSITY OF CALIFORNIA UBRARY