66 USHER : THE INFLUENCE OF NON-ELECTROLYTESVIII.-The InJtuence of Non-electroltjtes on theSolubility of Cadma Dioxide in Water.By FRANCIS LAWRY USHER.THE object of the investigation to be described in this paper wasto place upon record a larger number of accurate measurementsof the solubility of a gas in solutions of non-electrolytes than hashitherto been available. Although there is a considerable massof experimental data relating to the solubility of gases in saltsolutions, the non-electrolytes examined have been confined tosome sugars, chloral hydrate, carbamide, and a few slightly dis-sociated organic acids (Roth, Zeitsch. physikal Chem., 1897, 24,114; Braun, ibid., 1900, 33, 721; Knopp, ibid., 1904, 48, 97;Christoff, ibid., 1905, 53, 321; Hufner, ibid., 1906, 57, 611;Steiner, Wied.Annalen, 1894, 52, 275). The experiments recordedhere were all carried out at 20°, and, except in the case of sucrose,a t only one concentration, namely, semi-normal. The gas usedwas carbon dioxide, chosen because its comparatively largeabsorption-coefficient permitted greater accuracy in the cleter-minations, whilst at the same time its slight deviation from Henry’slaw is not sufficient to preclude its use as an indifferent gas. Thesubstances studied were sucrose, dextrose, mannitol, glycine, pyro-gallol, quinol, catechol, resorcinol, carbamide, thiocarbamide,urethane, acetamide, antipyrine, acetic acid, and n-propyl alcohol.EX PE R I M E NTALExcept in the case of the two liquid substances examined, namely,acetic acid and n-propyl alcohol, the solutions were of exactly semi-normal strength, and were prepared in the absorption vessel itsell.The apparatus was designed with this end in view, and a descriptionof it will now be given. The carbon dioxide was prepared in anordinary Kipp’s apparatus, from marble which had been boiledfor some time in distilled water in order to remove adhering andoccluded air, and concentrated hydrochloric acid.The gas wasfound to contain less than 0.03 per cent. of foreign gas. Theoutlet tube from the Kipp’s apparatus was connected through a tapwith a large U-tube filled with marble chips, which served toprevent any acid spray from being carried over with the carbondioxide. The gas was next led through a wash-bottle containingconcentrated sulphuric acid, and finally through a phosphoric oxidetube. Wherever practicable, connexions between glass portions ofthe apparatus were made by sealing the glass together with ON THE SOLUBILITY OF CARBON DIOXIDE IN WATER.67mouth-blowpipe; in fact, the number of taps and ruljber con-nexions was kept as small as possible, and the fatter, when it wasnecessary to use them, consisted of short pieces of pressure tubingof small bore, inside of which the glass ends were brought together.The tap of the Kipp’s apparatus was left permanently open, so thatthe internal gas pressure was always in excess of the atmospheric,and no air could leak into the apparatus. The measuring burette( B ) was of 100 C.C. capacity. Of this, 50 C.C.were containedin the narrow upper part, which was graduated in tenths of ac.c., and was calibrated by weighing out mercury, whilst theremaining 50 C.C. were contained in a bulb blown below thisgraduated portion. The burettt was connected by rubber tubingwith a mercury reservoir H , carrying a levelling tube L of thesame diameter as the graduated part of the burette, and wasprovided at the top with a three-way tap G, by means of which itcould be connected either with the carbon dioxide supply or withthe absorption vessel. The absorption vessel d was of about 220C.C. capacity, and was provided near the bottom with a tubulure T,about 14 mm. wide, carrying two small glass hooks, and whichcould be closed by a ground glass stopper carrying a second pairF 68 USHER : THE INFLUENCE OF NON-ELECTROLYTESof hooks, by means of which it could be held firmly in positionwith two elastic bands.The solid substance was introducedthrough this tubulure. On the opposite side, and nearer the top,was a short capillary tube carrying a three-way tap C , so arrangedthat either the absorption vessel could be connected with a vessel Sdelivering a known volume of gas-free water, or the latter couldbe connected, independently of the absorption. vessel, with thestore of gas-free water employed.A flexible copper capillary, 2 metres long and of 1.5 mm. bore,was used to connect the absorption vessel with the burette, andwas cemented into the glass tubing with marine glue, the junctionsbeing subsequently enclosed in plaster-of-Paris blocks, to preventthem from becoming loose when the absorption vessel was shaken.A thermostat was employed for all the experiments, which werecarried out at 20° & 0'02O.When the room-temperature rose above20°, a cooling coil was introduced into the thermostat. The methodof carrying out a determination is as follows:In the first place, a semi-normal solution of the substance to beexamined was prepared, and its specific gravity a t 20° was deter-mined, and from this was calculated the weight of substance whichwould give it semi-normal solution when dissolved in 117 C.C. ofwater, this being the amount of water used in every case. Theexact quantity of the substance was then weighed into the driedabsorption vessel through the tubulure, after which the stopperwas inserted and fastened in position.The absorption vessel wasnow connected through the three-way tap with a Topler pumpand completely exhausted. After closing the tap, it was fillgd withpure dry carbon dioxide by alternately filling the burette fromthe supply and allowing the gas in the burette to pass into theabsorption vessel, by suitably turning the three-way tap G. Theabsorption vessel was now placed in the thermostat and left thereuntil the temperature of the contained gas was constant at 20°,and it was arranged that when it was full of gas a t 20° under theatmospheric pressure, the level of the mercury should be near thetop of the burette. The burette reading was then observed, andthe room-temperature and barometric height noted.Pure gas-freewater had now to be introduced. Ordinary distilled water wasused, and was previously boiled out in a vacuum in a round-bottomed flask provided with a rubber stopper carrying two glasstubes, one short, the other passing to the bottom of the flask. Assoon as all air had been completely removed, the flask was closed,cooled to a little below 20°, and the longer tube was then connectedwith the branch P of the three-way tap C, and so with the vessel S,which was filled with mercury, and of which the volume betweeON THE SOLUBILITY OF CARBON DIOXIDE IN WATER. 69two marks on the capillary stem a t a and d was accurately known :this was 117.0 C.C. a t 20°. By suitable manipulation all air wasremoved from the capillary tubing, and the tap C was now turnedand the vessel X filled with gas-free water to the lower mark ur.C was next turned in the other direction, and by raising thereservoir K and lowering the reservoir H , the exact quantity ofwater contained between a and ur was driven into the absorptionvessel.C was now closed, and the absorption vessel disconnected a tP and Q and placed in the thermostat. It was then shakenvigorously until all the solid was dissolved, and the resultingsolution saturated with carbon dioxide at 20° under the atmosphericpressure. When the burette reading was constant, the barometerheight and room-temperature were again noted, and the deter-mination was now finished. Care was taken that no gas or liquidpassed back from the absorption vessel to the burette, consequentlythe gas in the burette was always dry, whilst that in the absorptionvessel was only dry a t the commencement of the experiment. Forthe purposes of calculation the volume, and hence the densities ofthe solid substances, had to be known.The values of some of thesewere taken from papers by Schrodter (Ber., 1879, 12, 1611; 1880,13, 1070), and some were redetermined.The absorption-coefficients for the liquids, acetic acid andm-propyl alcohol, were determined by Ostwald’s method in anabsorption vessel containing 246.3 C.C. Since the volume of gasabsorbed in the case of these liquids was much greater than thecapacity of the burette, the latter had to be refilled several times.Although the error of reading was repeated as often as the burettewas filled, the volume of gas dealt with was proportionately larger,and the probable error in the final result therefore remained thesame as for the other solutions, for which a single filling of theburette sufficed.Two determinations were carried out with every solutionexamined, and four in the case of water.The maximum differencebetween the results of two such experiments was 1 in 250, whilstmost agreed to within 1 or 2 in 1000.Calculation of Results.The absorption-coefficients (a) given represent the volume ofcarbon dioxide, reduced to Oo and 760 mm., which is dissolved by1 C.C. of liquid at 20° when the partial pressure of the carbondioxide is 760 mm.Calculation of absorption-coefficient for solutions of solids :Let = volume of absorption vessel up to beginning of copper(i)xcapillary70 USHER : THE INFLUENCE OF NON-ELECTROLYTESLet y = volume from beginning of copper capillary to markon burette.66 0 'I.,, b, and 6, = initial and final burette readings.7 9 A =97 a =y , 23 =Y Y P =9 9 tThe corrected- -volume of solution.volume of solid substance.barometric height, corrected to Oo.vapour pressure of solution at 20'.room-temperature.initial volume of gas in the apparatus will beand the corrected final volume will be:2 73hence the volume dissolved by A C.C. of the solution- P 273 (b,-b,) + -{-(-a) 273 P - p - ( z - A)}c.c.,760 273+t 2'33 76U 760andP-p' m tA a =I f the room-temperature is itself 20°, the expression simplifies to :(ii) When water is used instead of a solution of a solid, if Ti1 isthe volume of water taken :273293W a = =orif the room-temperature is 20°.(iii) I n the case of the solutions of liquids examined, theabsorption-coefficient, measured in the Ostwsld vessel, is equal to :2 73I' - p a 273 + tAwhere b and 6' are the burette readings before and after introON THE S0,LUBILITY OF CARBON DIOXIDE IN WBTEI'L.7 1ducing the gas into the absorption vessel, and T Y is the volumeof solution run out from the absorption vessel.Nature and Magnitude of Errors.I n all the experiments, an accuracy of 1 in 1000 was aimed at,and the values of a given may be taken as correct t o 1 in 500.Since the measured volume difference between the first and lastreadings was always about 100 c.c., an error of 0.1 C.C.involves anerror of 1 in 1000 in the value of a. By far the most importantsources of error were (1) inexact levelling when reading the volumeof gas in the burette, and (2) variations in the temperature of thethermosta,t.There was no difficulty in reading the burette with a maximumerror of 0.05 c.c., but the exact adjustment of the levelling tubewas not so easy. Ia the initial reading, when the volume of gasin the apparatus is about 226 c.c., an error of 1 mm.*in thelevelling involves an error of about 0.33 C.C. In the final reading,when the volume of gas is only half as great, a similar error inlevelling involves an error of 0.17 C.C.The actual uncertainty oflevelling was probably about a quarter of a millimetre, and ifthe errors in the two readings were additive, this involves an errorof 1.2 in 1000 in the value of a. Thus the probable error intro-duced by this inaccuracy is 0.6 in 1000.Thus,if the temperature of the thermostat is not exactly 20°, not onlyis the volume of gas soluble in the liquid changed by expansionor contraction, but the absorption-coefficient is also directlyinfluenced. Supposing that at the initial reading the true tem-perature in the thermostat is (20--8)O, and a t the final reading itis (20 + O)O, it can be calculated that the difference between the trueand the apparent volume of gas absorbed is:Variations in temperature affect the resulks in two ways.where /3 is the difference between the volume of gas soluble in theliquid at 20° and the volume soluble at (20 + 8)O.This expression,after neglecting 8 and 82 in the denominator, and evaluation of fifrom the temperature-coefficient of solubility of carbon dioxide inwater, is equal to about 48 C.C. Now the maximum variation (28)in the temperature of the thermostat was 0*04O, and the maximumerror from this cause is therefore 0-08 c.c., that is, about 0.8 in1000.Compared with the above, the other errors are unimportant.The barometer was read with an accuracy of 0.1 mm., and theheight of the mercury column was corrected for expansion fro72 USHER : TEE INFLUENCE OF NON-ELECTROLYTESOo to the room-temperature.The time interval between the initialand final readings was usually about haIf an hour, and no sensibleerror was introduced through employing the mean of the twobarometer readings, since these never differed by more than 0.2mm., and could not as a rule be detected. Similarly, the room-temperature was always sufficiently constant during an experimentto preclude the introduction of any appreciable error throughemploying the mean temperature for the calculations. There isno doubt that the gas in the absorption vessel was completelysaturated with water vapour a t the second reading, on account ofthe vigorous shaking which always took place, or that the gas inthe burette was dry, since the burette tap was always closed whilethe absorption vessel was being shaken, and gas was never alIowedto pass from the latter into the burette.As already stated, twodeterminations were made in every case, in which the averagedifference is 1.7 in 1000, and consequently the probable accuracy ofthe results is about 1 in 1000.R e s d t s.A bsorption-coefficient in Wat er.-Four determinations weremade, and the values found were (i) 0.8775; (ii) 0.8766; (iii)0.8755; (iv) 0.8766; mean, 0-877.This value is in good agreement with that given by Bohr (Wied.Annalen, 1899, 68, 503), namely,Table I gives the absorption-coefficients (a') in the solutions ofsucrose examined, and the specific gravities of the solutions at 20°.=0*878O.TABLE I.Concentration. a'. Mean a'. 8p. gr................ 0.846 1'01518 N/8 {%E...............0.~15 N/4 { 8::::;N/2 -K;:t3" ............... 0.756N ............... { 8'E 0'6491.031251'06372i.12a09I n table I1 are given the absorption-coefficients in semi-normalsolutions of the other substances examined, and, for purposes ofreference, the specific gravities of the solutions a t 20° and thespecific gravities of the solid substances a t the same temperatureON THE SOLUBILITY OF CARBOX DIOXIDE 1N WATER. 73Substance.Dextrose ...............Msnnitol.. .............Glycine ...............Pyrogallol ............Quinol.. ........ - ......Resorcinol ............Catechol ..............Urethane ............Carbamide ............Thiocarbamide ......Antipyrine .........Acetamide ............Acetic acid ............n-Propyl alcohol ...TABLE 11.Mean a'.0.7920.7820.8430.8530'8870*9010.8680.8690'8640-8590.8590.8790.8680.869SP. €7.:N/2-solution.1 *03281 -030311 '01 4 131 '01 71 81'009461'009581'01071'00371.007151-009171.013391 *00051 -00260'9939Sp.gr.solid.1-56 *1-46 *1-61 *1-451 '331 -271-340 *991.33 *1.42 *1.19 *1.56- -* Redetermined.Biscussiom of Results.Since any theoretical deductions from the results must dependon the way in which the latter are expressed, it is first of alldesirable to consider briefly the methods which are usuallyemployed for this purpose. In order t o compare together a numberof different substances with respect to their influence on thesolubility of a third substance, it is, of course, only permissibleto employ solutions of the same molecular concentration. Herethe usual difficulty arises with regard to the calculation ofmolecular concentration (compare Abegg, Zeitsch.physikal. Chem .,1894, 15, 248). The semi-normal solutions used in this investi-gation all contained half a gram-molecule in a litre of the solution,and this concentration is in many cases considerably different fromthat of a solution containing the same weight of substance in1000 grams of water. Except in the case of sucrose, no experimentswith several different concentrations were carried out, but thefigures for this substance certainly suggest that the volume-norma1is more convenient for our present purpose than the weighbnormalmethod of calculation, and also that solutions of semi-norma74 USHER : THE INFLUENCE OF NON-ELECTROLYTESstrength are still sufficiently dilute to permit inferences whichmay be applied without serious error to very dilute solutions.I n table I11 are given values for the inolecular depression ofsolubility calculated according to the volume-normal and weight-normal methods for solutions of sucrose. a and at denote theabsorption-coefficients of carbon dioxide in pure water and thesolution respectively, N is the number of gram-molecules of sucrosein 1 litre of sohtion, and N’ the number in 1000 grams of water.TABLE 111.N. a - a’/N.0’125 0-2480 *25 0.2480.5 0’2421 ‘0 0 *228a - u’fN).0-2410-2340.2160-180It will be seen from these figures that the molecular depression,calculated by the volume-normal method, is constant for solutionsup to N / 4 , and is very little different for those of N/2-concentra-tion; and since sucrose produces a much greater effect than anyof the other substances examined, the deviations in the case ofthese must be still smaller. The purely empirical formula,a - af/iVe, suggested by Jahn for expressing the relation betweendepression of solubility and concentration in the case of electro-lytes, is obviously unsuitable for sucrose, and Roth (Zeitsch.phgsikal. Chew&., 1897, 24, 114) has shown that the figures forglycerol are better represented by the linear formula.* It istherefore probably safe to assume that for moderately dilutesolutions of non-electrolytes the effect on the solubility of carbondioxide, or of any indifferent slightly soluble third substance, isdirectly proportional to the amount of non-electrolyte present. Itnow remains to consider certain attempts a t generalisation 3n thelight of the experimental data recorded in this paper.One of the most recent of such attempts is that of Philip (Trans.,1907, 91, 711), who suggests that substances which have notendency to combine with the solvent are without influence on thesolubility of an indifferent gas, whilst those which do influence thesolubility do so because they remove a portion of the solvent byforming compounds with the latter.By “solubility” is heremeant the amount of gas dissolved by unit mass of the pure solvent;.For example, the absorption-coefficient of hydrogen in an aqueoussolution of chloral hydrate is smaller than in pure water (Knopp,Zeitsch.physikal. Chem., 1904, 48, 97), but the amount dissolved* It should be mentioned that Geffcken (Zeitsch. physikal. Chwn., 1904, 49, 257)has pointed out sources of error in the experiments of Gordon, Roth, and Rraun,which may invalidate the apparent support they afford to Jahn’s empirical formulaON THE SOLUBILITY OF CARBON DIOXIDE IN WATER, 75by 1000 grams of water is the same in each case. This statementis not true, however, for solutions of sucrose; and in order tobring this substance into line, it is assumed that the statementwould be true if it were not that a certain fraction of the wateris withdrawn by the sucrose, and the average number of watermolecules attached to one molecule of sucrose is calculated in away consistent with the theory.This method of calcuTationinvolves three assumptions : it presupposes (1) that hydrates areformed, (2) that the gas is insoluble in the dissolved substance, and(3) that it is insoluble in the hydrate.That such assumptions its these have little foundation in factmay be inferred from the following table, which shows the volumeof carbon dioxide dissolved by 1000 grams of pure water for thedifferent semi-normal solutions examined, calculated on the assump-tion that the water alone is responsible for the absorption of gasobserved.TABLE IV.Carbon dioxidedissolved by1000 grams ofSolution.water, in C.C.Water ............... 878Sucrose.. ............. 797Dextrose ............ 841Mannitol ........... 833Glycino.. ............. 864Py rogal lo1 ......... 894Quinol ............... 928Resorcinol ......... 946Carbon dioxidedissolved by1000 grams ofSolution. water, in C.C.Ca techol ............ $08Urethane ............ 907Carbamide ......... 884Thiocarbarnide ... 885Antipyrine ......... 935Acetamide ......... 906Acetic acid ......... 893n-Propyl alcohol.. . 902It is noticeable, in the first place, that in eleven out of the fifteensolutions examined a larger quantity of carbon dioxide is dissolvedthan can be accounted for if the water only is responsible for theabsorption.This fact alone suffices to show that if we wish toexpress the solvent properties of these solutions in terms of theproperties of their components, any conclusions depending on theassumption that the dissolved substance has no solvent power areworthless.It is instructive also to compare, from the point of view of thehydrate theory, the behaviour of some of these substances withthat of the solutions used by Jones and Getman (Amer. Chem. J.,1904, 32, 308) in their cryoscopic investigations. I f the " averagemolecular hydration " is calculated in the way described by Philipfrom the figures given above for sucrose, dextrose, and mannitol;we arrive at the conclusion that sucrose is hydrated to the extentof 3.6 molecules of water, dextrose 4.6 molecules, and mannitol 5.2molecules-values which are completely a t variance with thosededuced by Jones and Getman from their data for the molecula76 USHER : THE lNFLUENCE OF NON-ELECTROLYTESdepression of the freezing point.These authors conclude thatsucrose in semi-normal solution a t Oo forms complexes containingabout 5 molecules of water, whereas dextrose is hydrated only toa small extent, and the tendency of mannitol to form hydrates isinsignificant. It is not intended here to dispute the existence ofhydrates in solution, but it may be permissible to raise the questionwhether much is to be gained by referring abnormal depression ofsolubility to this cause, when by so doing it becomes necessary tomake other and more improbable assumptions.Of a different character is the generalisation deduced, on thermo-dynamic principles, by Jahn (compare Roth, Zeitsch.physikal.Chem., 1897, 24, 115). According to this, the molecular con-centration of a gas remains the same when it is dissolved tosaturation in a dilute solution of an indiff erentf non-volatile sub-stance as when it saturates the pure solvent under the sameconditions of temperature and pressure. I n other words, if C,denotes the ratio of the number of gas-molecules to the sum of themolecules of gas and solvent, and C2 is the ratio of gas-molecules tothe sum of those of gas, solvent, and third substance, the theoryrequires that C,jC2 shall be equal to unity, provided that thefollowing conditions are fulfilled: (i) The gas must exert nochemical action on the solvent or the solution; (ii) it must havethe same molecular weight in the liquid as in the gas phase;(iii) the solution must be dilute.I n a limited number of cases, results have been obtained whichare described as being in good agreement with the theory.Thus,Roth (Zoc. cit.), using nitrous oxide, found C,/C2=1.009 forcarbamide a t about semi-normal concentration, 1.013 for glycerol,and 1-009 for oxalic acid. Braun (Zoc. cit.) found 1.037 forcarbamide and 1.023 for propionic acid in the case of nitrogen;and, in the case of hydrogen, 1.015 for propionic acid. Knopp(Zoc. cit .) found €or chloral hydrate, when hydrogen was used,0.993, and when nitrous oxide was used, the ratio was 1.010 forchloral hydrate and 1.037 for propionic acid.Here we have values of C,/C, deviating by anything between0-7 and 3.7 per cent.from the theoretical value; but in view ofthe fact that the actual change of solubility effected lies betweenthe limits 1.3 and 4.5 per cent., it becomes obvious that theapparently good agreement has no significance whatever. It shouldbe mentioned that Knopp's experiments were all carried out at20°, whilst those of Roth and Braun were carried out a t 5O, loo,15O, 20°, and 25O. The figures quoted above refer to the meanof the results obtained a t these five temperatures; the actual valueof C,/C, was found to diminish with rising temperatureON THE SOLUBILITY OF CARBON DIOXIDE IN WATER. 77In table V the values of C,jC2 for the substance3 employed inthe present investigation are tabulated, and it is interesting t ocompare the percentage deviations from the theory, given in thethird column, with the percentage change in solubility of the gasbrought about by the non-electrolytes used, given in the fourthcolumn.TABLE V.Solution.N-Sucrose ...............N/8-Sucrose ............N/4-Sucrose ............N/B-Sucrose ............N/2- Uextrose ............N/B-M~rtnnitol.. ..........N/%Glycine ............N/2- Py rogallol .........N/2-Thiocarbamide ., .N/e-Carbarnide ........N/Z-Urethane ............N12-Antipyrine .........N/ 2 . Catec hol ............M/B-Quinol. ..............N12- Acetamide .........A7/2-Acetic acid ....Nj2-n-Propyl alcohol.N/i-Resorcinol .........Cl/G1.1011'0121 -0241 -0481 -0561.0651.0260.99211 *0021 -0030.97890.94940.97620-95540.93960.97950-99360.9817Percentage deviation Percentage changefrom Cl/C, = 1.of solnbility.10-1 26 .O1.2 3.52.4 7-14.8 13.85-6 9.76 -5 10.82-6 3.90.8 2-70.2 2.00.3 1 -52.1 1.05-1 2.02.4 1 '04.5 1.06.1 2.72 *1 0 . 20.6 1.01 -8 1-0It will be seen from this table that in no less than seven instancesthe deviation from the theory actually exceeds the magnitude ofthe effect which is being studied; and, in general, the extent ofthe discrepancy increases with the amount of the effect produced.There can, however, be no doubt that the formula deduced byJahn is inapplicable to the experimental data hitherto available ;in other words, that the conditions for which the formula is validare not fulfilled in the experiments. Indeed, the influence of mostnon-electrolytes is s,o small that it may reasonably be doubtedwhether a rigid confirmation of the theory is possible with ourpresent methods of determining gas-solubility ; and in any casethere is the possibility that the dissolved substance, althoughchemically indifferent, may itself be capable of dissolving the gasemployed.An attempt has been made by Roth (Zeitsch.physikal. Chem.,'5903, 43, 539) to find some common factor which will bring thedeviations from Jahn's thermodynamic formula into line withdeviations from van't Hoff's freezing-point law, and it was shownthat a parallelism did exist in the cases of glycerol and sucrose,both of which give too large a depression of the freezing point,and too great a value for C,/C',. Thiocarbamide and glycine,however, give too small depressions, whereas the values of C,/C78 POPE AND HOWARD: THE CONDENSATION OFfor these substances are respectively 1.002 and 1.026 in semi-normalsolution a t 20°, and would, of course, be greater still at Oo.It seems, therefore, that the effect on the solubility of a gas *produced by non-electrolytes is not capable of explanation byreference either to the formation of hydrates in solution or todeviations from the theory of osmotic pressure-conclusions whichhave already been expressed by Levin (Zeitsch. physikal. Chem.,1906, 55, 513); but that since these effects are, as Geffcken (Zoc.cit.) has shown, practically independent of the solubility of thegas and almost entirely determined by t.he nature of the solventor solution, it is only possible a t present to refer them to mutualinteraction among the molecules. It only remains to be mentionedthat the lasbnamed author has already called attention to aparallelism which exists between depression or elevation of solu-bility and such properties as compressibility and surface tension,and it is possible that the whole problem might be more success-fully attacked from this point of view.This research was carried out a t the suggestion ofProf. Rothmund, t o whom I am greatly indebted for much kindlyadvice snd criticism, as well as for placing at my disposal thenecessary apparatus and material.INSTITUTE OF PHYSICAL CHEMISTRY,GERMAN UNIVERBITY,PRAGUE