Membranes will be, similarly, semipermeable to solvent or solute, when only one of these is soluble in the membrane, or is capable of forming an unstable compound with it. For instance, salts, holding water of crystallization which is readily lost and recovered, may easily be conceived of as assuming the rôle of semipermeable membranes, allowing the passage of water say from a wet atmosphere to a dry one, or from pure water to a solution; and Tammann[31] has realized such membranes by the use of zeolites—silicates, which hold water of crystallization but are insoluble in water. Kahlenberg[32] has recently used rubber membranes, that are permeable for solvents like benzene, pyridine, etc., which are soluble in rubber, but not permeable for water, which is insoluble in rubber. [Illustration: FIG. 5.] «Osmosis.»—The recognition of this rôle of the semipermeable membrane leads to the second question raised, namely as to the mechanism of the process by means of which osmotic pressures are measured directly with the aid of such membranes (p. 11). [p023] The answer hinges on the question of the mechanism of the diffusion of the solvent into the cell, a diffusion which is called its osmosis.[33] If we consider the pure solvent, say water, on one side of a semipermeable membrane, and a solution (‹e.g.› of sugar in water) on the other side, it is obvious that the ‹solvent› itself has a ‹higher concentration› on the side where it is pure, than on the side of the solution, where it is diluted—distended by the solute in it. The solvent is soluble in the membrane, and its solubility will be proportional[34] to its own (the solvent's) concentration; it will, consequently, be more soluble in the membrane on the side of the pure solvent than on the side of the solution. If we bring such a membrane first into contact with the pure solvent (Fig. 5), the membrane will take up the solvent (from the side ‹A›) until it is ‹saturated› with it. Let the solubility, which represents the concentration of the solvent in the membrane at this stage, be called ‹c›. The membrane may then be considered to be taking up in unit time just as many molecules from the solvent as it gives up to it (dynamic equilibrium), exactly as, when water is in equilibrium with water vapor, we consider the water to be vaporizing just as fast as vapor is condensing to form water. Now, if a solution of sugar is placed on the other side of the membrane, the solvent will pass out of the membrane into the solution just as fast as it passes back into the pure solvent. At first the concentration of the solvent on the surface ‹B› of the membrane is just as great (‹c›) as on the surface ‹A›; but ‹the membrane will here receive the solvent more slowly from the solution, which is less concentrated as to the solvent›; and consequently the membrane ‹will lose water to the solution›. The solubility (‹c′›) of the solvent at this surface ‹B› of the membrane, corresponding to the smaller concentration of