QD 
 
 672 
 
EXCHANGE 
 
 
SEP 2 
 
 ' 
 
 A Contribution to the Investigation o< the 
 Temperature Coefficient of Osmotic Pressure; a 
 Redetermination of the Osmotic Pressures of Cane 
 Sugar Solutions a 
 
 DISSERTATION 
 
 SUBMITTED TO THE 
 
 Board of University Studies of Johns 
 Hopkins University 
 
 IN CONFORMITY WITH A REQUIREMENT FOR THE DEGREE OF 
 
 DOCTOR OF PHILOSOPHY 
 
 BY 
 
 WILLIAM MANSFIELD CLARK 
 
 B A L T I M () R 1- 
 
 1910 
 
A Contribution to the Investigation of the 
 Temperature Coefficient of Osmotic Pressure; a 
 Redetermination of the Osmotic Pressures of Cane 
 Sugar Solutions at 20 & & & & & & & 
 
 DISSERTATION 
 
 SUBMITTED TO THE 
 
 Board of University Studies of Johns 
 Hopkins University 
 
 IN CONFORMITY WITH A REQUIREMENT FOR THE DEGREE OF 
 
 DOCTOR OF PHILOSOPHY 
 
 BY 
 
 WILLIAM MANSFIELD CLARK 
 
 BALTIMORE 
 1910 
 

 CONTENTS. 
 
 Page. 
 
 Acknowledgment 3 
 
 Preface 4 
 
 Purification and Analysis of the Cane Sugar. . 5 
 
 Improvements in Apparatus 9 
 
 The Measurements: 
 
 Preliminary Discussion i I 
 
 Tables 17 
 
 Discussions of Results 1 19 
 
 Theoretical Considerations 27 
 
 Summary 30 
 
 Biography 31 
 
ACKNOWLEDGMENT. 
 
 This investigation was carried out under the supervision of 
 Professor Morse, to whom the author wishes to express his 
 gratitude for instruction in lecture room and laboratory. The 
 measurements of osmotic pressures were made in collaboration 
 with Dr. Zeis, Dr. Holland, Mr. Carpenter and Mr. Meyers, 
 and special credit is due Dr. Zeis, who was in charge of the 
 measurements from Oct. i to March i. 
 
 I desire to thank Dr. Turner and Dr. Pfund for helpful 
 suggestions, and to express my obligation to President Remsen, 
 Professor Jones, Professor Acree, Professor Bliss and Dr. 
 Pfund for instruction. 
 
 239378 
 
PREFACE. 
 
 The investigation here recorded was a study of the osmotic 
 pressures of cane sugar solutions at 20. It is one of a series 
 of investigations having for their object the establishment of 
 the temperature coefficient of the osmotic pressures of cane 
 sugar solutions, ranging in concentration from one-tenth 
 weight 1 normal, to normal at intervals of 5 between o and 
 25-30. 
 
 Those phases of the subject which are dealt with in this 
 paper are : 
 
 1. Purification and analysis of the cane sugar. 
 
 2. Improvements in apparatus. 
 
 3. The measurements at 20 and their discussion. 
 
 1 Morse, Frazer and Dunbar, Am. Chem. J., 38, 2, '07. 
 
PURIFICATION AND ANALYSIS OF THE CANE 
 
 SUGAR. 
 
 The methods in use at this laboratory for measuring osmotic 
 pressures having reached a high state of accuracy it became 
 desirable to make the measurements with a single stock of 
 sugar of known purity. Previously, while the methods were 
 in a less satisfactory state, the sugar used was the purest 
 obtainable commercial rock candy, the uniformity of which was 
 only assured within certain limits. 
 
 The method of purification employed was essentially that 
 described by Cohen and Commelin 1 . The original material 
 was a rock candy crystalized by a local firm. This was ground 
 in a porcelain mortar, and dissolved in previously boiled, dis- 
 tilled water, warmed in a porcelain dish over a water bath for 
 not longer than twenty minutes, and at a temperature never 
 allowed to exceed 60. When approximate saturation was 
 obtained in this manner, the solution was quickly filtered with 
 suction, and precipitated with 98% ethyl alcohol, which had 
 previously been distilled from lime through a column of beads. 
 Samples of this alcohol were tested from time to time and 
 found free of lime. The precipitated sugar was filtered with 
 suction, and, after draining three to six hours, was washed 
 with 85% ethyl alcohol. To make the washing the more 
 thorough, the sugar was each time removed from the funnel, 
 and stirred vigorously with the alcohol. Thirty-two kilo lots 
 of the sugar at this stage were very carefully mixed, and then 
 in small lots again subjected to the same process of purification. 
 In addition to the washing with ethyl, hot methyl alcohol was 
 used for the final washing. The separate precipitates were 
 then thoroughly mixed with each other, dried at 60, and bot- 
 tled after another thorough mixing. Sixteen kilograms of this 
 twice precipitated sugar constitutes the supply which has been 
 drawn upon for all measurements since March 26th, 1909. 
 Previously, since January ist, a smaller sample of three times 
 precipitated sugar had been used. A large portion of the stock 
 has been kept over calcium chlorid to insure desiccation and 
 the removal of alcohol before use. 
 
 In the following analyses "Sample A" designates the original 
 rock candy, B the once precipitated, C the twice precipitated, 
 and D a small preliminary batch which was carried through 
 three precipitations. 
 
 IZt. Phys. Chem., 64, 1. 
 
Combustion With Electric Furnace. 
 SAMPLE A. SAMPLE B. 
 
 6.432 ...42.156 
 
 (6.506 42.335) 1 
 
 6.495 42.081 
 
 6.477... ...42.090 
 
 6.488 
 6.482 
 
 .006 
 
 42.109 
 42.083 
 
 + .026 
 
 6.430 42.116 
 
 6.451 42.059 
 
 6.465.... 42.151 
 
 6.445 
 6.482 
 
 .037 
 
 42.108 Average 
 42.083 Theoretical 
 
 + .025 Difference 
 
 SAMPLE C. 
 
 6.466 42.151 
 
 6.420 42.081 
 
 6.471... ...42.116 
 
 6.452 
 6.482 
 
 .030 
 
 42.116 
 42.083 
 
 + .033 
 
 SAMPLE D. 
 
 6.484 42.047 
 
 6.410 42.104 
 
 6.453 42.009 
 
 6.487 42.031 
 
 6.485.... ...42.101 
 
 6.464 
 6.482 
 
 .018 
 
 42.058 
 42.083 
 
 Average 
 Theoretical 
 
 025 Difference 
 
 1 Not averaged, although no error in the determination was detected. 
 
 Ash. 
 
 A large known weight of each sample was incinerated in a 
 platinum crucible protected from the free flame by a porcelain 
 dish. All samples showed an ash content of from 21 to 50 
 parts per million. Each residue gave a test for iron, although 
 the sugar came into contact with iron at no time during the 
 process of purification. In order to make sure that no contam- 
 ination by the lime used in concentrating the alcohol had oc- 
 curred, thirty-five grams of Sample C were charred, and partly 
 incinerated to reduce its volume, and a spectroscopic examina- 
 tion for Ca was made with the electric arc and a Roland grat- 
 ing, and again with the Bunsen flame and a prism spectroscope. 
 Duplicate examinations with different incinerated samples were 
 made in each case. In no case was a trace of a calcium line 
 discovered by visual or photographic exploration in the visual 
 region of the spectra. Dr. Pfund, of the Department of 
 Physics, who made the examination with the Roland grating, 
 then photographed for the brilliant H and K lines of the ultra 
 violet, and found them only in about the same intensity as in the 
 photograph with the comparate blank carbons. 
 
Reducing Sugars. 
 
 Analyses for reducing sugars were made in accordance with 
 the directions described in* Methods of Analysis Adopted by 
 the Association of Official Agricultural Chemists* "for mate- 
 rials containing i% or less of invert sugar and a high percent- 
 age of sucrose." Since the precipitate in each case contained 
 considerably less than 50 milligrams of copper, the lowest 
 figure for copper in Hertzf eld's table, and since there is no 
 statement either of the method of extrapolation or of the 
 standard sugar used in preparing this table, the method was 
 abandoned for that of Soldaini as modified by Ost. 2 Prelim- 
 inary experiments with aliquot portions of known glucose 
 solution gave, 
 
 i. 0.0089 g taken 2. 0.0089 g taken 
 
 0.0088 g found 0.0089 g found 
 
 Using 10 g. portions of each sample the following data 
 were obtained. Ost's table for conversion of Cu to the equiv- 
 alent of invert sugar formed the basis of the calculations : ;i 
 
 Grams Cu. Grams invert %of invert sugar Sample 
 
 found. sugar calculated. in sample. 
 
 0.0277 0.0081 
 
 00280 0.0082 
 
 0.0293 0.0085 
 
 0.0283... ....0.0083 
 
 0.0083 0.083?; 
 
 0.0100 0.0017 
 
 0.0137 0.0019 
 
 0.0142 0.0018 
 
 0.0018 0.018^ B 
 
 0.0085 0.0017 less than 
 
 0.0108 0.0018 
 
 0.0106 0.0017 
 
 0.0062 0.0017 less than 
 
 less than 0.0017 less than 0.017^ C 
 
 0.0065 0.0017 less than . , .,.^ 
 
 less than 0.0017 less then 0.017^ D 1 
 
 x Only one analysis of D made because of low supply. 
 
 According to these analyses, a normal solution of Sample C 
 would be about ^ with respect to invert sugar. If we 
 assume that one molecule of invert sugar takes the place of one 
 
 1 Bulletin 4G, U. S. Dept. Agriculture. 
 
 2 Wiley; Practice of Agricultural Chemistry, Vol. III. 
 
8 
 
 molecule of sucrose, and that the second molecule of the two, 
 which together approximate the weight of one molecule of 
 sucrose, is the only one which would produce an error in the 
 osmotic pressure, then we may say that a normal solution of 
 Sample C would have a true normality of approximately I-f-gjfr. 
 A further striking qualitative difference between the original and 
 purified samples should be noted. The original, at the moment 
 of incipient boiling with the Soldaini reagent, formed a heavy 
 green cloud, while with the purified sugars the reagent kept its 
 fresh color and the precipitate was only noticed when filtered. 
 
 The United States Bureau of Standards has been given 
 samples of A, B, C, and D, upon which it is conducting tests. 
 A formal report of these has not yet been received. It will 
 include determinations of the invert sugar, and of the rotations, 
 in both of which the Bureau is making every effort to attain 
 the greatest accuracy at present possible. A preliminary re- 
 port 1 on the tests for invert sugar has been received in which 
 it is stated that: 
 
 Sample A contains between 0.09% & 0.08% 
 B " " 0.02 & 0.01% 
 
 C " " 0.02 & 0.01% 
 
 D less than 0.01% 
 
 The relative difference between B and C is about 0.003% 
 and between C and D is 0.005%. A comparison of these deter- 
 minations with those made in this laboratory will show that the 
 agreement is fairly close in the third decimal place. 
 
 I think it permissible to quote from the informal report, 
 that Sample D as regards invert sugar compares very favorably 
 with the best prepared at the United States Bureau of 
 Standards. 
 
 Freezing Point Lowering s. 
 
 Dr. Turner, of this laboratory, who made some very careful 
 determinations of the freezing point lowerings of normal solu- 
 tions of each sample, has kindly allowed me to publish his 
 final averages. The details of his work will be published later. 
 The following molecular lowerings for the normal solutions 
 employed are accurate +o.ooi, with their absolute, but not 
 relative values subject to a possible slight modification when 
 the thermometer is still more carefully calibrated. 
 Sample A 2. 037 
 
 " B -2.0 4 I 
 
 " C -2.0 4 I 
 
 D -2. 048 
 
 1 By R. F. Jackson. 
 
A slight but distinct gain in the molecular lowerings is seen 
 to have resulted. That this occurred in spite of the distinct 
 lowering in the content of invert sugar is strange. That it 
 could be due to alcohol in the sugar is improbable, because 
 some very careful tests failed to reveal any indications that 
 alcohol was present, and, furthermore, because desiccation in 
 vacua over calcium chlorid gave ample opportunity and time 
 for the removal of alcohol before the freezing point determina- 
 tions were made. One consideration to which great import- 
 ance is not attached is this. Hot methyl alcohol was used for 
 the final washing of each of the purified samples. According 
 to its comparatively greater solubility in methyl alcohol, rani- 
 nose, a normal constituent of cane sugar, should therefore have 
 been rapidly removed. The assumption that raffinose of higher 
 molecular weight than saccharose was rapidly removed by the 
 methyl alcohol is in harmony with the observed increase in 
 molecular lowerings. 
 
 A summary of the analyses of Sample C, the stock used in 
 the measurements of osmotic pressures, is now given. 
 
 % Carbon 42.166 
 
 '% Hydrog-en 6.452 
 
 f* Invert Sugar 01-0.02 
 
 %Ash about 0.003 
 
 Iron trace 
 
 Calcium none 
 
 Alcohol none 
 
 Normal molecular lowering- of freezing point 2. 041 
 
 IMPROVEMENTS IN APPARATUS. 
 
 The apparatus and methods used at this laboratory in deten- 
 mining osmotic pressures, have been described in previous pub- 
 lications. 1 A few minor improvements are to be noted. 
 
 Those who have followed the work will recall that the prin- 
 ciple of temperature regulation employed is to pump the water 
 rapidly over coils of pipe kept by running water a little below 
 the temperature desired in the bath, and then over an electric 
 stove, which is controlled by a thermostat. By regulation of 
 the thermostat any desired temperature may be maintained 
 between that to which the cooling pipes can lower the water 
 and that to which the electric stoves can warm it. The air 
 space is kept at constant temperature in accordance with this 
 same principle, a fan taking the place of the pump, and an 
 electric light serving as stove. Because of the different heat 
 
 1 Morse & Frazer, Am. Chem. Jour., 28, 1 ('02). 
 
 Morse & Frazer, Am. Chem. Jour., 36. 1 ('05). 
 
 Morse & Lovelace, ibid, 40, 4 ('08). 
 
 Morse & Mears, ibid. 40, 3 ('08). 
 
 Morse & Holland, ibid, 41, 2 ('09). 
 
10 
 
 capacities of air and water and the different volumes of each 
 in the bath, it was found advantageous to separate into two 
 sections the cooling pipes which had previously been in series 
 for the cooling of both air and water spaces. Each may now 
 be separately controlled. Through the winter months the 
 cooling pipes in the air space were not used because the tem- 
 perature of the outside room was sufficiently low to produce 
 the proper cooling. A more even temperature environment 
 for the bath was furnished by the installation of a fan in the 
 room outside the bath. Because of' the relatively low heat 
 capacity of air, the air space in the bath is the more difficult 
 space to keep at constant temperature, but it can be done with 
 a fair degree of success if the heating and cooling sources are 
 nicely adjusted so that the air will not pass over either an 
 excessively cool or an excessively warm surface. 
 
 For the maintenance of constant temperature we are of course 
 dependent in large measure upon the sensitiveness of the mer- 
 cury and glass thermostats. The least sparking of these will cause 
 fouling of the mercury, and early in the year considerable 
 trouble was definitely traced to this cause. The means which 
 had been used to prevent this was the spanning of. the spark 
 gap by the proper resistance, either that of a 75 volt 16 candle 
 power electric light filament, or that of graphite painted over 
 a porcelain tube. To eliminate the sparking altogether, tin foil 
 condensers are now used to span the spark gaps. Although these 
 tin foil condensers alone give every evidence that they com- 
 pletely absorb the induced currents, which have caused spark- 
 ing in the thermostats at the "break," we have "made assur- 
 ance doubly sure" by leaving in parallel with each condenser as 
 a "by pass" the resistance formerly used alone. Several ther- 
 mostats which have been used in various baths about the 
 laboratory, have worked perfectly since their spark gaps have 
 been "spanned" with these condensers. 
 
 It has been noted that when the platinum point and the sur- 
 face of the mercury are very clean the mercury will wet the 
 platinum. The result is that, when the thermostat cools, the 
 mercury still clings to the platinum point, and the breaking of 
 the circuit will therefore occur at a slightly lower temperature 
 than the "make." This is easily remedied by rubbing the 
 .finger over the platinum tip until it is found that the platinum 
 will touch and leave the mercury meniscus sharply. 
 
 After these two sources of trouble were remedied the water 
 space kept at 20 for a period of two months at a stretch to 
 within about + o.oo5 except at three observations, when the 
 deviation was -fo.oi, due to stoppage of the water in the 
 cooling pipes. The air space during this period though far 
 
II 
 
 more difficult to keep constant, did so with only an occasional 
 observed deviation of + o.o5, except, of course, when the 
 bath was opened to change the cells. 
 
 In order to insure constant temperature of the nitrogen 
 volume of the manometers, it seemed desirable to eliminate as 
 far as possible the heating effects of the electric light used in 
 the readings. Accordingly, a device was made by which the 
 infra red waves of the light were cut off. A 4% solution of 
 nickel sulfate was used at the suggestion of Dr. Pfund, of 
 the Department of Physics, and this was enclosed between 
 glass plates, forming a screen about one inch thick. An eight 
 candle power electric light bulb, frosted to provide a soft, dif- 
 fused light, was placed behind this, and the intensity of the 
 illumination increased by a reflector made of a silvered watch 
 glass. Tests made with a thermocouple showed that the ra- 
 diant energy emitted through the nickel sulfate screen was but 
 i% of that given by the unscreened light. The green color 
 obtained by this method of illumination is also thought to 
 enhance the detail of both mercury meniscus and silvered 
 meter bar. One of these devices has been used in the deter- 
 mination of the normal volumes of the manometers. 
 
 THE MEASUREMENTS PRELIMINARY DIS- 
 CUSSION. 
 
 Since this investigation had for its object the establishment 
 of one link in the chain of data whereby the temperature co- 
 efficient of osmotic pressures may be established, relative 
 rather than absolute pressures were sought. With this in view 
 the technique and conditions employed in measurements at 
 other temperatures were followed as closely as possible. 
 
 In this connection attention sfiould be called to two well 
 recognized errors, which have but little significance in the 
 present investigation, but which render the data somewhat 
 inaccurate for the comparison of high with low pressures. In 
 our manometer data the volume correction in calibration units 
 for the double meniscus is taken as two-thirds the radius of 
 the bore. This is based on the assumption that the meniscus 
 in these small -bores, of about half a millimeter diameter, is 
 approximately hemispherical. Since it is known that the 
 meniscus leaves the glass even in capillaries of this size at an 
 angle much greater than zero, and since actual observation of 
 the form of the meniscus shows it to be considerably less than 
 hemispherical, the 2/3r correction is doubtless too large. 
 With low pressures the error will be insignificant, but when 
 the nitrogen is compressed to one twenty-seventh its volume 
 
12 
 
 under standard conditions by the pressure of a normal solution, 
 a small error in this small observed volume will produce a 
 considerable error in the calculated pressure. It would, there- 
 fore, be unjust to claim for the present series of investigations 
 great accuracy in the comparison of high with low pressures 
 for the purpose of substantiating a law similar to Boyle's. 
 But, since the normal volumes of the manometers are of 
 about the same order of magnitude, and, since they are com- 
 pressed to about equal degrees for the same concentration of 
 sugar at the various temperatures employed, the error is of 
 little significance in the purpose of the present series of meas- 
 urements. 
 
 The second error, to which attention was called in 1907,* is 
 one which, if it exists, makes the data inaccurate for the com- 
 parison of high with low pressures. But this again would not 
 alter the value of the data for use in determining the tempera- 
 ture coefficient except perhaps by an indirect effect to which 
 attention will be called later. The error lies in the possibility 
 that the membrane formers are not isosmotic. If, for ex- 
 ample, the concentration of the potassium ferrocyanid on the 
 interior is less than osmotically equivalent to the copper sulfate 
 on the exterior of the membrane, a constant positive correction 
 should be applied in order to obtain the true osmotic pressure 
 of the cane sugar. Obviously this constant correction would 
 be of greater significance to the low than to the high pressures. 
 But an example will show that this error would have to be 
 large to affect the sought for temperature coefficient. Let us 
 take a measurement of a O.5N solution that gave an observed 
 osmotic pressure of 11.890 atmospheres at o, and one of the 
 same concentration that gave a pressure of 12.750 atmospheres 
 at 20. The "theoretical gas pressure" of such a solution is 
 11.133 atmospheres at o, and 11.950 at 20. The ratio of 
 osmotic to gas pressure for the former was found to be 1.068 
 and for the latter 1.067. Assume now that the potassium fer- 
 rocyanid and copper sulfate, used to protect the membrane, 
 differ in osmotic pressure by as much as 0.03 atmposphere. 
 Unless the potassium ferrocyanid itself affects the osmotic 
 pressure of the cane sugar, we are justified in adding this 
 0.03 atmosphere to obtain the pressure due to the sugar alone. 
 We then obtain as the corrected pressures 11.920 at zero, and 
 12.780 at 20, giving for the ratio at zero 1.071 and for the 
 ratio at 20, 1.070. Each ratio is considerably changed by 
 this correction, but each by the same amount, so far as the 
 third decimal place is concerned, and hence the temperature 
 
 1 Am. Chem. Jour., May, 1907. 
 
13 
 
 coefficient is seen to have remained unaltered. In this calcula- 
 tion the assumption has been made that the temperature co- 
 efficient of dissociation for the two membrane formers are prac- 
 tically equal in the range of temperature used. 
 
 Sources of error which would render the measurements in- 
 accurate for the determination of the temperature coefficients 
 have not been overlooked. A change of temperature in the 
 air space of o.i will not affect the ratio seriously, and can 
 always 'be allowed for in the calculation ; but a similar fluctua- 
 tion in temperature in the water space, that is in the neighbor- 
 hood of the cell itself, cannot be allowed for in the calculation, 
 because it involves the expansion or contraction of cell, solu- 
 tion, brass cone and collar, and the mercury in the bulbs of 
 the -manometer. It has, therefore, been our practice to plug 
 the tops of the cans, into which the cells are set, with cotton, 
 so that any change of temperature in the air space could only 
 permeate slowly to the cell and the bulbs of the manometers. 
 Of the two hundred and eighty observations which are inclu- 
 ded in the following tables, fourteen were made when the 
 temperature of the water bath varied from 20 more than 
 o. 01, and in but two was the variation greater than o. 03. 
 The variations in each of these two cases was o. 3. The 
 maximum probable errors in weighing the solutions, of incli- 
 nation of the manometers to the meter bar, and of the barom- 
 etric readings, could only affect the ratio in the fourth decimal 
 place. A recent calibration of the meter bar used with the 
 cathetometer, shows a periodic error of 0.02 mm. for alternate 
 millimeter marks. This in conjunction with the unavoidable 
 errors in reading, may have affected the ratios of the higher 
 concentrations to a very slight extent. We have sought to 
 reduce the errors of reading to a minimum, both by check 
 readings by two or more observers, and by keeping ignorant of 
 the figures found on the previous observation. The agreement 
 has been very satisfactory. In checking the concentration of 
 the solutions before and after the measurements, by means 
 of the saccharimeter, the agreement of different observers, and 
 even of the same observer, has not been so satisfactory. After 
 an extended series of comparative readings we are of the con- 
 clusion that for a solution of average clearness, darkened as 
 it always is by the potassium ferrocyanid, o.i point on the 
 scale of our Schimidt and Haensch saccharimeter is the limit 
 of our accuracy. When the difference in rotation between the 
 solution as made up and the solution at the close of an experi- 
 ment, both compared at the same time and temperature, was 
 distinctly as great as o.i, we made it a practice to discard the 
 measurement. All differences less than this are recorded as 
 
14 
 
 "no loss in rotation." o.i point error in the saccharimeter 
 reading corresponds to o.ooi in the ratio for a normal solution, 
 0.002 in the ratio for a o.5n, and as much as 0.008 in the ratio 
 for a o.i normal solution; and only within these limits are the 
 ratios certain to be unaffected by change in the concentration 
 of the solutions. 
 
 We have, therefore, taken pains to prevent the dilution or 
 concentration of the solution which inevitably accompanies 
 four blunders in good technique. In the first place, the cells 
 before being set up are filled and kept in thymol water. A 
 thorough rinsing of the interior with the sugar solution is 
 therefore necessary, else the solution will become diluted by 
 the water which clings to the interior of the cell. Secondly, 
 the solution which has been finally poured into the cell will, 
 in exerting its osmotic pressure, cause water to permeate the 
 membrane with resulting dilution of the solution, unless the 
 manometer is quickly fastened in place and an initial "mechan- 
 ical pressure" brought to bear by forcing the cone down into 
 the cell so that the osmotic pressure of the solution is balanced 
 by the gas pressure in the manometer. This one point caused 
 considerable trouble in former years, when the means of secur- 
 ing the manometer in place was not developed. With the pres- 
 ent arrangements, and "team" work, but few seconds elapse 
 between the time when the solution is poured into the cell and 
 the moment when, with the manometer firmly secured, the 
 proper initial pressure is brought to bear by closing the exit 
 tube and forcing the cone down by screwing up the brass collar. 
 Rapid manipulation is also essential when the cell is taken 
 down. In the third place, this initial pressure must be nicely 
 adjusted to balance the probable osmotic pressure to be devel- 
 oped, since, if too small or too great an initial pressure is given, 
 the solutions will be diluted or concentrated by a volume of 
 water corresponding to the volume through which the mercury 
 in the manometer must 'be forced in order to establish the 
 equilibrium. The small bore of the manometer allows a little 
 range in this, but in extreme cases, of which we have had some 
 examples, concentration of the solution has been traced to the 
 production of too great an initial pressure. Such cases have 
 confirmed the statement, 1 "that the 'mechanical' pressure upon 
 the solution must never exceed the maximum osmotic pres- 
 sure." Concentration from this source may even escape de- 
 tection because of the operation of one or more of the causes 
 which subsequently produce dilution, and is consequently to be 
 feared more than dilution. The fourth matter to which atten- 
 
 1 Am. Chem. Jour., Sept., 1908, p. 276. 
 
15 
 
 tion should be paid, is the prevention of fluctuations in tem- 
 perature just after the cell is set up. The cell in its bottle of 
 o.in CuSO 4 is immediately placed in a small bath at the proper 
 temperature, from which it may be conveniently removed from 
 time to time for the adjustment of the initial pressure. But 
 the manometer projects into the air above, which is, of course, 
 at the temperature of the laboratory. The opening of a win- 
 dow, for instance, will alter the temperature of the nitrogen 
 in the manometer, and it becomes troublesome to tell its pres- 
 sure. We have been favored in working at 20 by the ease 
 with which it is possible to keep the laboratory at nearly 20 
 during this preliminary manipulation. 
 
 An uncertainty occurs in the measurement of low pressures 
 bv reason of the fact that the membrane is often slow in allow- 
 ing the passage of sufficient water to compensate fluctuations 
 in barometric pressure. The barometric pressure is, of course, 
 subtracted from the sum of the pressures of nitrogen, capillary 
 depression and liquids in the manometer to obtain the osmotic 
 pressure. If the barometric pressure suddenly drops, and the 
 membrane is slow in allowing the escape of sufficient water to 
 compensate for this drop, the observed pressure of the solution 
 is that of the osmotic pressure plus that of the former high 
 barometer, and an error is made in subtracting the new low 
 barometric pressure. This uncertainty is often evident when 
 measuring the lower concentrations, but since the barometric 
 pressure is small in relation to the osmotic pressure of high 
 concentrations, barometric fluctuations have no noticeable 
 effect upon these. 
 
 A very uncertain source of error lies in possible variations 
 in the bore of the manometers, and the consequent variations 
 in the capillary depression when the meniscus is at different 
 points. This error also enters into the determination 'of the 
 normal volume of a manometer. The capillary depression of 
 each manometer is determined experimentally, but only for one 
 point. From an inspection of the calibration curves of the 
 manometers the bore is seen to vary in each case, but it is 
 doubtful whether the variations are of sufficient magnitude to 
 alter the capillary depression more than a few millimeters. 
 This is a matter of some importance to determine experimen- 
 tally, and it will be made the subject of an investigation when 
 accuracy in other directions becomes sufficient to warrant it. 
 In this connection attention might be called to a little point 
 where care has been used in the present investigation. The 
 lower part of the manometer is filled with solution around to 
 the first bulb where it meets the mercury. If care is not taken 
 to have the mercury in this bulb at the widest part when the 
 
i6 
 
 final pressure is developed, and the surface is actually near the 
 upper constructed part, an appreciable capillary depression will 
 obtain which is not allowed for in the calculations. 
 
 In accordance with the usage in the investigations at other 
 temperatures, the capillary depression of each manometer has 
 been reckoned as 0.02 atmosphere. As experimentally deter- 
 mined, they actually range from 0.015 to 0.023 atmosphere. 
 Furthermore, the pressure of the solution measured from the 
 surface of the liquid outside the membrane to its height at the 
 point where it meets the mercury in the manometer, has in each 
 case been taken as o.oi atmosphere, altho it varies slightly with 
 each manometer and with the density of the solution. By 
 rounding off the values for both capillary depression and pres- 
 sure of solution, no great error has been introduced, but the 
 assumption should be noted. 
 
 It is to be regretted that more measurements which fulfill 
 the requirements to be mentioned were not obtained. That 
 they were not is believed to be due, in the main, to two causes. 
 The cells available for the measurements had been used at o, 
 5 and 10, and it has been observed, almost without exception, 
 that cells used at a low temperature are of little avail for meas- 
 urements at higher temperatures until after a long period of 
 treatment. A reason for this will be mentioned in another 
 connection. The cells used at 20 were, therefore, not in prime 
 condition until January, and indeed there is some evidence 
 that, in spite of the thymol, the slight growths of Penicillium 
 which have been so troublesome in the past, may have infected 
 parts of the cell wall. However that may be, by January 
 manometer 6 had developed a flaw, and 5 was broken. As 
 subsequent investigation has shown, we were left with only 
 one reliable manometer, number 15, and possibly 21. Since 
 the others were at the time supposed to be thoroughly reliable, 
 they were often used to the exclusion of 15, and not until 15 
 was used continuously and 9 came upon the scene late in 
 March, were reliable measurements obtained. 
 
 The criterions which the measurements have been required to 
 fulfill in all the series, as well as in this, before they were ad- 
 judged acceptable were: i. Constancy of temperature in air 
 and water spaces. 2. Constancy of osmotic pressure for a 
 period of at least twenty-four hours. 3. No change in the 
 concentration of the solution that could be detected with the 
 saccharimeter. 4. No fault which could be detected in the 
 technique or in the apparatus. 
 
 Of the 163 experiments made at 20, 127 had to be rejected 
 as not fulfilling one or more of the above requirements. The 
 remaining 36 are recorded in the following tables. 
 
888888888888888 
 
 -IA9Q 
 
 
 J -Q 
 
 d '0 
 
 9SBJ9AY 
 
 ggggggSSSgg 
 
 90U9J9jgiQ 
 
 O O O 
 ONGNOO 
 
 'd 'O 
 
 oofO 
 
 gou'B^sisg^ 
 
 OOO'OO^oi^r-lOOOOOO 
 
 O 
 iO 
 
 'IPO 
 
 
 jo - 
 
i8 
 
 'HOISTS 
 
 SDiiajajfia 
 
 r4 ,_( ^ ,-t r 
 
 888888S8S 
 
 o o 
 
 O O 
 
 o o 
 
 O O 
 
 r-i> i o^o^ i '-<<y>ONO 
 
 xOfO'O'O^OOOCCOOO 
 OOOOOOOO^t 
 
 3DU12}SIS92I 
 
 'IPO 
 
 Oit lr-1'O'O'Hr-I 
 
 rovovor-io^^ 
 
 o o 
 
 oo o^ t^ 01 ve 10 n o 
 r< ci 
 
 o to ^ vc i 
 
 odcddrH 
 
19 
 TABLE C. 
 
 
 I 
 
 
 
 .M. 
 
 , 
 
 
 
 d 
 
 
 fl 
 
 .2 
 
 I 
 
 y 
 
 
 u 
 
 
 
 * S 
 
 1 
 
 .2 
 
 "rt 
 
 
 rt 
 
 H 
 
 
 B 
 
 5 u 
 
 . *o 
 
 a 
 
 
 
 a 
 
 
 
 i 
 
 ' 1 
 
 
 C B 
 
 1 
 
 
 
 
 13 g 
 
 
 o 
 
 OT O 
 
 O 'O 
 
 .S o 
 
 fl 
 
 flj 
 
 jq 
 
 
 _4 
 
 C 
 
 .^H U 
 
 ^ G 
 
 X V-* 
 
 rt 
 
 
 jo 
 
 6 B 
 
 0^ 
 
 
 
 "cd ^ 
 
 rt 
 
 d rt 
 
 V 
 
 
 
 5 
 
 o 
 
 5 
 
 K 
 
 & 
 
 * 
 
 5 
 
 & 
 
 0.1 
 
 139 
 
 t 
 
 13 
 
 1.050 
 
 1.084 
 
 .005 
 
 .003 
 
 72 
 
 0.2 
 
 148 
 
 R 
 
 20 
 
 1.055 
 
 
 .001 
 
 .001 
 
 24 
 
 0.3 
 
 151 
 
 I 
 
 20 
 
 1.055 
 
 
 .002 
 
 .001 
 
 47 
 
 0.4 
 
 91 
 
 Y 
 
 13 
 
 1.042 
 
 1.068 
 
 .003 
 
 .001 
 
 45 
 
 0.4 
 
 93 
 
 X 
 
 13 
 
 1.046 
 
 1.071 
 
 .001 
 
 .001 
 
 25 
 
 0.5 
 
 85 
 
 P 
 
 11 
 
 1.049 
 
 
 .001 
 
 .001 
 
 49 
 
 5 
 
 86 
 
 O 
 
 13 
 
 1.045 
 
 1.071 
 
 .001 
 
 .001 
 
 49 
 
 0.8 
 
 105 
 
 z 
 
 24 
 
 1.080 
 
 1.101 
 
 .002 
 
 .001 
 
 78 
 
 0.9 
 
 71 
 
 H 
 
 22 
 
 1.072 
 
 
 .002 
 
 .001 
 
 27 
 
 0.9 
 
 154 
 
 F 
 
 24 
 
 1.096 
 
 1.115 
 
 .002 
 
 .001 
 
 72 
 
 1.0 
 
 118 
 
 A 5 (63) 
 
 24 
 
 1.106 
 
 1.124 
 
 .003 
 
 .001 
 
 51 
 
 THE MEASUREMENTS AND THEIR DISCUSSION. 
 
 Table A is a summary of the measurements which give 
 ratios agreeing with those obtained recently at o, 5, and 
 loV 
 
 In table B are to be found ratios lying between those obtained 
 in the earlier work and those of table A. 
 
 In table C are some determinations which fulfill all the re- 
 quirements of good measurements, except that a curious dis- 
 crepancy in the volumes of the manometers occur, which at 
 present cannot be explained. 
 
 The columns in the tables are, in order, the concentration of 
 the solution in terms of weight normal, the number of the ex- 
 periment, the date 2 when the experiment was started, the cell 
 used, the resistance of the membrance in ohms at the close of 
 the electrolytic renewal, the number of the manometer, its vol- 
 ume in calibration units at standard conditions of temperature 
 and pressure, the average observed osmotic pressure in atmos- 
 pheres, the theoretical gas pressure in atmospheres, the average 
 difference between the observed osmotic pressure and the theo- 
 retical gas pressure, the average ratio of osmotic to gas pres- 
 
 1 Dissertations of Zies & Gill, 15)09, and unpublished work at 10 done 
 
 in 1909. 
 
 2 In the academic year 1909-1910. 
 
20 
 
 sure, the maximum deviation in the ratio from the mean of all 
 the observations, the mean deviation of the same,- the time 
 period in hours during which the ratio remained constant. 
 
 In tables A and B only those measurements are included 
 which were made with manometers whose volumes at standard 
 conditions of temperature and pressure were determined by 
 the side tube method described in The American Chemical 
 Journal for October, 1908. It will be seen that the manometers 
 in table C are an entirely different set. The volumes of these 
 latter were determined by comparison with a standard mano- 
 meter. It was hoped, that, by the use of this standard mano- 
 meter, the normal volumes of the experimental manometers 
 could be calculated by employing a greater range of pressure. 
 Accordingly, several manometers were compared with the 
 standard. In the case of manometer 21 the volume as calcu- 
 lated by comparison with the standard under low pressures 
 agreed very well with the volume as calculated by the side tube 
 method. At pressures of from three to seven atmospheres 
 however, a minimum in the calculated volumes were observed 
 in all manometers compared with the standard, and at higher 
 pressures the calculated volume steadily increased with increas- 
 ing pressure. 
 
 This increase is probably due to the error in meniscus cor- 
 rection previously mentioned ; but, since neither this, nor pos- 
 sible causes of the minimum, have as yet been studied quanti- 
 tatively, the proper corrections cannot be applied. Obviously, 
 the only course to pursue at present is to bring all the mano- 
 meters to a comparable basis by determining their normal vol- 
 umes by one method, preferably the side tube method, which 
 is an absolute one. Unfortunately, a considerable number of 
 measurements were made with manometers n, 13, 20, 22 and 
 24 before there was an opportunity to determine their normal 
 volumes by the side tube method. When 13 and 24 were so 
 redetermined, their volumes were apparently altered. It is 
 impossible to say at present to what the discrepancy is due, and 
 consequently the ratios of table C cannot be placed with either 
 those of table A or B. A discussion of which values should 
 be taken in these measurements would be unprofitable in view 
 of the total uncertainty attached to the manometer data. Fur- 
 thermore, in certain cases the results are vitiated by the prob- 
 ability that the wrong volume of the manometer was chosen 
 as the basis for the calculation of the initial "mechanical" pres- 
 sure, and consequently, a concentration of the solution may 
 have occurred, as previously shown to be possible, with a sub- 
 sequent compensating dilution, and the saccharimeter would 
 have indicated no change in concentration. 
 
21 
 
 It is to be noted that the high ratios agree very well with 
 those found at o, 5, and 10 as the following table will show: 
 
 Weight Normal Concentration. 
 
 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0-8 0.9 1.0 
 
 1.061 1.059 1.061 1.069 1.076 1.084 1.094 1.104 1.115 
 
 5 1.082 1.083 1.059 1.061 1.067 1.074 1.084 1.093 1.101 1.115 
 
 10 1.08-3 1.061 1.061 1.066 1.072 1.083 1.092 1.102 1.114 
 
 20 1.083 1-061 1.062 1-059 1.067 1.073 1.083 1.091 1-103 . ... 
 
 mean 1.082 1.062 1.060 1.061 1.067 1.074 1.084 1.093 1.103 1.115 
 
 The ratios in table B, on the other hand, when compared 
 with the old values found at 2O , 1 and with the mean values of 
 the old series V to VIII, 2 are found to be above these older 
 values in each case, and to lie between the lower ratios of the 
 older series and the higher ratios of the more recent investiga- 
 tions. If it is permissible to omit experiment 29 simply be- 
 cause the 1.037 ratio is not in agreement with that of two other 
 measurements of the same: concentration, the relations of the 
 ratios in table B to the mean of the ratios of series V to VIII 
 and to the mean of the higher ratios given in the above table 
 are clearly shown in the plotted curves. 
 
 LOW RATIOS OF FORMER WORK COMPARED WITH LOW RATIOS 
 FOUND AT 20 THIS YEAR. 
 
 Weight Normal Concentration. 
 Series. 0.1 0.2 0.3 0.4 0-5 0-6 0-7 0.8 0-9 1.0 
 
 VIII 20 mean 1.055 1.051 1.038 1.042 1.045 1.060 1.066 1.077 1.084 1.093 
 of series. 
 
 V-VIII 10-25 1.054 1 047 1.039 1.040 1-045 1.056 1.060 1-070 1-081 1.068 
 
 Mean ratios in 1.057 1.050-56 1.064 1.081 1.089 1.100 
 
 table B. 
 
 Perhaps the low values of table B are too few to say defi- 
 nitely that they constitute a series in themselves, but the pro- 
 portionality between them is striking. 
 
 The sharp division between the ratios of table A and those 
 of table B calls for discussion. The search for possible ex- 
 planations will be found to be met with difficulties which are 
 insuperable in the present state of our knowledge, but a dis- 
 cussion may be of profit. 
 
 At first thought it might be suspected that the discrepan- 
 cies were due to the use of different manometers. Manometer 
 21 occurs only in table B and 9 only in table A. 9 is not sus- 
 pected ; and 21 is under only slight suspicion because a recent 
 recalibration indicates a slight change in its bore, due perhaps 
 to molecular changes in the glass. The curve of the recalibra- 
 tion has been used in the calculations. 
 
 Examining the dates of the measurements with 5, 6, and 15, 
 the manometers which occur in both tables, we find that the 
 
22 
 
 one low value with 5 occurred after four high values were ob- 
 tained with it, while four low values with 15 were obtained be- 
 fore the six high values. On the other hand, the one low 
 
 I fK> 
 
 ( 100 
 
 0.3 Oij 0.)' Oi, 0.7 
 
 0.1 1.0 
 
 HIGH RATIOS OBTAINED IN RECENT WORK COMPARED WITH 
 HIGH RATIOS AT 20. 
 
 value obtained with manometer 6 has its date sandwiched be- 
 tween those of the two high values. If, in subsequent work, it 
 should be proved that the manometers need more frequent 
 
 1 Morse & Holland, Am. Chera. Jour., 41, 4 (1900). 
 
 2 Ibid, page 275. 
 
23 
 
 standardization than has been considered necessary, consid- 
 erations such as the above may be of value. 
 
 Turning now to the cells, it is seen that D and O alone 
 occur in both tables. It would be unfair, however, to attrib- 
 ute great significance to this, for an examination of a consider- 
 able number of measurements, which were rejected only be- 
 cause of a slight gain or loss in rotation, shows that other cells 
 included in table A have given low ratios in spite of gains in the 
 rotation of the solutions. 
 
 We might find in the conditions or known errors of the ex- 
 periments some constant factor which will explain this sharp 
 division in the ratios. Let us inquire if any such factor can 
 be discovered. 
 
 It has been suggested that the low pressures might be due 
 to an accumulation of copper salts in the cell wall during the 
 electrolytic renewal of the membrane. Although it is difficult 
 to see how this should affect certain measurements always to 
 about the same degree, and not other measurements conducted 
 under apparently the same conditions, the suggestion was taken 
 as the basis of some experimental work. A cell, after subjec- 
 tion to the customary electrolytic process for the renewal of 
 the membrane, was soaked for ten days in frequent changes of 
 water. The total water was then evaporated, the residue of 
 accumulated thymol ignited and evaporated with sulfuric acid. 
 The residue of 6 mg. was considered as that of CuSO 4 soaked 
 from the cell wall. Had this diffused from the cell wall, and, 
 by very slow diffusion contaminated only that portion of the 
 solution adjacent to the cell wall, it might have had a slight 
 effect upon / the observed osmotic pressure. It is impossible, 
 however, to say what its distribution would have been. In 
 some cases salts may diffuse out more rapidly than in others, 
 while it is perhaps also possible, that the membrane, which in- 
 advertently accumulates on the cell's exterior, may serve in 
 some cases to retain inclosed salts quite completely. 
 
 Another cell, after the customary rinsing, was soaked for two 
 hours after electrolysis, the time usually allowed between elec- 
 trolysis and the setting up of a cell for a measurement. The 
 residue obtained in this case was considerably below that which 
 can be estimated by ordinary gravimetric means. 
 
 A curious fact, with which this idea that the cell wall accu- 
 mulates salt has been connected is observed when a cell is used 
 more or less continuously without giving it the "rest" and the 
 soaking out, which has become a regular part of the procedure. 
 When not given a sufficient "rest" with soaking, cells have de- 
 veloped constant but low pressures. A notable case of this was 
 observed some two years ago, and a possible explanation is, 
 
2 4 
 
 that in rushing the cells into frequent measurements the re- 
 peated electrolytic renewal of the membrane plugged the cell 
 wall with salts, for the soaking out of which sufficient time was 
 not allowed. In the present investigation each cell has been 
 soaked about a week between measurements made with it. 
 
 Another attempt was made to gain an experimental basis for 
 the supposed influence of salts accumulating in, and diffusing 
 out, from the cell wall. It has been found best to renew the 
 membranes in o.in CuSO 4 , but the concentration of CuSO 4 
 used as membrane former in a measurement is o.oin. It 
 would seem, however, that if the membranes were renewed in 
 a solution of the same low concentration as that into which 
 they are placed during a measurement, there would be less 
 chance of carrying over to this dilute solution, copper sulfate 
 which might have accumulated while the cell was in the more 
 concentrated solution ordinarily used. This was tried, but 
 the uncertainty with any cell is so great that no conclusions can 
 be drawn. 
 
 Because of the difficulty of drawing conclusions from the 
 data obtained in any of these methods of attack, the question 
 was investigated directly by setting up cells with very dilute 
 solutions and open manometers capable of showing very slight 
 differences in pressure, the plan being to see whether any dif- 
 ference of pressure could be observed between cases when 
 every chance was given for the cells to lose the accumulated 
 salt, and cases when the cells were supposed to have had an 
 opportunity to accumulate salts within their walls. The ex- 
 periments though, at first sight, both confirming and contra- 
 dicting the hypothesis they were designed to test, were found 
 to be so involved by reason of some unexpected phenomena 
 that the evidence desired can only be untangled after further 
 investigations. 
 
 It is, of course, possible that a minute leak in the membrane 
 could just counterbalance the tendency of the solution to de- 
 velop a maximum pressure. It would seem, however, that 
 leaks, even if they allowed constant pressures to be observed, 
 would cause greater divergence in the results obtained in dif- 
 ferent instances. Attention is called to the agreement between 
 experiments 27 and 102 and between experiments 62 and 70. 
 Furthermore the fact that a very careful study of the original 
 and final solutions with the saccharimeter in experiments 128 
 and 102 showed no change in concentration after the period of 
 89 hours in one case, and 16/j. 1 hours in the other, argues 
 against leakage. 
 
 1 The maintenance of constant pressures for periods of this length 
 would seem to indicate a truly semipermeable membrane. 
 
25 
 
 A calculation will perhaps bring this point out more clearly. 
 If the solution should leak through the membrane and not be 
 replaced by water, we would observe a steady drop of the mer- 
 cury in the manometer. In the case of two o.Qn measurements, 
 both made with manometer 15, no such steady drop was ob- 
 served, and the actual difference between the observed volume 
 which indicated the low ratio and that which indicated the high 
 ratio was only 0.3 calibration unit, or about 0.06 cubic milli- 
 meter. It is, therefore, necessary to assume that the observed 
 pressure was kept constant by the intake of sufficient water to 
 replace the volume of solution which had leaked out. Dilu- 
 tion of the solution must therefore have occurred. Now, let 
 us say that 10 milligrams of sugar escaped, and the solution 
 containing this was replaced by an equal volume of water. The 
 capacity of the cell is about 20 c.c. and, had twenty c.c. of the 
 o.Qn solution been diluted by the loss of 10 milligrams of sugar, 
 its rotation would have been diminished by almost 0.2, an 
 easily detected loss. But let us assume that even this escaped 
 detection, and, furthermore, that while the cell was quiet, the 
 dilution only occurred in the fourth of the total 20 c.c. which 
 lay nearest the membrane. This 5 c.c. would then have had 
 its sugar content reduced from the original 1.5273 g. to 
 1.5173 g. 1 The observed osmotic pressure of the first concen- 
 tration is 23.721 atmospheres, and of the second would be about 
 23.566 atmospheres, a difference of 0.155 atmosphere. The 
 actual difference in pressure between experiments 155 and 70 
 was 0.306 atmosphere, or about twice the amount which could 
 be accounted for when every advantage is given to the above 
 argument. In this calculation an assumption has been made 
 which is perhaps false to the actual phenomenon. It was as- 
 sumed that the observed low pressure was a function of the 
 dilution of the solution. A pseudo equilibrium might have 
 obtained, the solution leaking out with extreme slowness, and 
 the water entering at the same slow compensating rate. The 
 dilution in this case might have been inappreciable at the end of 
 even a considerable period. The data of the intake of water 
 through the membranes are insufficient to base any calculations 
 upon, and therefore it can only be said that a remarkably exact 
 and constant adjustment of intake to .leak must have occurred 
 in each case to produce a pseudo equilibrium of the constant 
 magnitude observed. 
 
 Again, if the original solutions were wrongly made up, the 
 error would be detected in the rotation. 
 
 Let us now consider another point. In calling attention to 
 
 1 The error in assuming the weight concentration to be a volume con- 
 centration is negligible in this calculation. 
 
26 
 
 the necessity of very accurate temperature regulation, Morse 
 and Holland stated in their article, The Regulation of Tempera- 
 ture in the Measurement of Osmotic Pressure / that "the ideal 
 would be a regulation so exact that not even the form of the 
 mercury meniscus in the manometer could be sensibly affected 
 by variations in the volume of the inclosed solution," and " . . . . 
 even the variation in the form of the meniscus becomes a matter 
 of importance whenever it is attempted to measure the osmotic 
 pressure of concentrated solutions." If, for example, with 
 manometer 15 and the pressure of a 0.911 solution, the observed 
 volume plus the double meniscus correction is 22.57, and, 
 owing to slight variations in temperature, the mercury has 
 risen and then fallen leaving the meniscus flat, then one half 
 the double meniscus correction has been added erroneously. 
 One-half the double meniscus correction for manometer 15 is 
 0.08 calibration unit. Subtracting this from the 22.57 we g et 
 for the true volume of the nitrogen 22.49 calibration units. 
 In the particular observation taken as the basis for this calcula- 
 tion the 22.57 volume was found to give a ratio of 1.090. Had 
 22.49 been the true volume it would have given in the calcula- 
 tion a ratio of 1.094, while the maximum obtained for a o.9n 
 solution is 1.103. Obviously, errors in meniscus correction 
 produce the greatest effect in the calculated pressures of high 
 concentrations, and, since the discrepancy which might be at- 
 tributed to a complete flattening of the meniscus effects the ratio 
 for a o.9n solution but four points in the third decimal place, 
 as shown, it would have proportionally less influence upon the 
 ratios for lower concentrations. Further, more than a slight 
 flattening of the meniscus would not have escaped notice in the 
 reading. Therefore it is to be concluded that this source of 
 error, when taken alone, is insufficient to explain the differences 
 between the ratios of table A and those of table B, although it 
 is of considerable import, as the 0.004 point possible error in 
 the o.9n ratio will show. 
 
 It is, of course, possible that a "heaping" of all these errors 
 in one direction could cause a low pressure to be observed in 
 one case, while the maximum was allowed to develop in another 
 case. This is highly improbable in view of the fact that only 
 one distinctly anomalous value has been observed. There is, 
 therefore, every evidence that some constant difference in the 
 conditions, as yet not fully known, has caused the discrepancy. 
 Certain possible sources have long been recognized ; but it has 
 been the policy in the researches in this laboratory to attack 
 those problems which seemed of greater moment, and to leave 
 
 1 Am. Chem. Jour., Feb., 1909. 
 
27 
 
 other smaller sources of error till the larger ones have been 
 removed. At present our time and thought is concentrated 
 upon the manometers, and manometers of a type which will, for 
 the most part, eliminate the errors of meniscus correction are 
 uriiier construction. With the development of these and other 
 improvements in the apparatus the time is approaching when 
 the accuracy of the measurements will exceed even that which 
 now obtains, and the absolute, rather than the relative, osmotic 
 pressures, may be the subject of future investigations. It may 
 therefore be well to call attention to certain conditions which 
 far more important matters have forced into the background 
 to await more thorough investigation, but which are of live in- 
 terest, not only because they are recognized as having a possi- 
 ble influence upon the true osmotic pressure, but because they 
 may possibly have had something to do with the discrepancies 
 observed in the present investigation. Because of the scanti- 
 ness of quantitative evidence for or against these considera- 
 tions, they will be discussed under the heading 
 
 THEORETICAL CONSIDERATIONS. 
 
 The membrane formers, that is the potassium ferrocyanid, 
 which is put into the sugar solution and the copper sulfate into 
 which the cell is placed during a measurement, were calculated 
 to be isosmotic from considerations of their freezing point 
 lowerings and conductivities. 1 The concentration of osmotically 
 active particles in solutions of each of these salts at the dilu- 
 tion used is still in doubt, but we have some evidence that the 
 solutions in use as membrane formers are not isosmotic. 2 In 
 addition it is known, that there is a considerable difference of 
 potential, when a potassium ferrocyanid solution of the con- 
 centration used is placed in the cell, and the cell dipped into a 
 copper sulfate solution of the concentration used in a meas- 
 urement. When measured with platinum electrodes, by a po- 
 tentiometer, and at constant temperature, the two solutions 
 separated by the membrane of cell Z, shortly after Z was taken 
 down from a measurement, showed a difference of potential of 
 0.452 volts, declining to 0.449 volts in five minutes. With cell 
 P, two hours after it had been subjected to the electrolytic 
 process for the renewal of the membrane, the difference of 
 potential was 0.752 volts, which declined steadily to 0.519 volts 
 at the end of three hours. In each case the electrode in the 
 copper solution was positive to that in the ferrocyanid solu- 
 tion, as might be expected if part at least of the difference in 
 
 1 Am. Chern. Jour., 34, 31; 34, -311. 
 
 2 Attention was called to the significance of this on page 
 
28 
 
 potential is due to the establishment of a liquid element. 
 The curious behavior of a cell's electrical resistance during the 
 round of its treatment is also to be noted. When treated elec- 
 trolytically for the renewal of the membrane, the resistance 
 gradually rises in the course of an hour to its customary max- 
 mum, which for some cells is as high as 500,000 ohms, and has 
 been known to reach 1,000,000 ohms. When removed from 
 the electrolizing bath, and allowed to soak, the cell is found to 
 show a declining resistance, which, in the course of a few 
 hours, may become one-fifth its former value; and, if the re- 
 sistance is taken immediately after the cell has been taken down 
 from a measurement, it is found to vary considerably in dif- 
 ferent cases, and in certain instances to be close to the initial 
 resistance, and in other cases to have declined greatly. * 
 
 From the above considerations it may be inferred that we 
 may possibly have to deal, not only with a difference in the 
 osmotic concentrations of the two membrane formers, but also 
 with the possibility that adjustments towards electrolytic equi- 
 librium may take place in varying degrees. It is to be hoped 
 that further study of this problem may throw some light on the 
 discrepancies observed in the present investigation. 
 
 A further theoretical consideration is now presented, not be- 
 cause any great importance is to be attached to a certain paral- 
 lelism between the theory and certain facts which have become 
 prominent in the work, but because it is hoped that the idea 
 may be of interest. 
 
 It has been observed, that to produce a membrane capable of 
 allowing the development of a maximum osmotic pressure, 
 there are required weeks and often months, during which the 
 copper ferrocyanid membrane is repeatedly packed by the 
 electrolytic process using as high a potential 1 as is consistent 
 with the safety of the membrane. Weak spots in the mem- 
 brane are then burst by subjecting them to high osmotic pres- 
 sure, with the membrane formers present, and the rents are 
 mended by electrolysis. Furthermore, it has been found nec- 
 essary to renew the membrane between each measurement. 
 Membranes which have not been so repacked just previous to 
 a measurement have been found not to allow the development 
 of the highest attained pressures, but permit a steady decline in 
 pressure. At times a new cell will give a constant pressure 
 below the maximum, but never, until after continued treatment, 
 a maximum. There is also some evidence, obtained this year 
 by Dr. Holland and Mr. Meyers, and confirmed by past expe- 
 rience, that the efficiency of a membrane is improved when it 
 
 1. 110 volts has been found best. 
 
2 9 
 
 is packed at a higher temperature than that at which the meas- 
 urement with it is made, the supposition being that the coeffi- 
 cients of expansion of membrane and cell wall ace such, that, 
 when the temperature is lowered, the membrane is packed 
 closer. There is, therefore, some basis for believing that the 
 development of a maximum osmotic pressure is intimately 
 connected with a close texture of the membrane. 
 
 Let us now assume that the membrane is semipermeable by 
 reason of a net-work of exceedingly fine capillaries. Moore 1 
 on the basis of this assumption, deduces the following relation 
 connecting the surface tension of solutions with their osmotic 
 pressure : 
 
 (i.) 2HrT = nr 2 P 
 
 whence P = ^ 
 
 "Where T is the difference between the surface tension of 
 the solution and that of the pure solvent, P the osmotic pres- 
 sure 2 in this solution, supposed to be due to the action of this 
 difference in surface-tension, and r the radius of a capillary 
 opening, placing the solution and the solvent in communica- 
 tion." 
 
 According to the equation, P = 2 r , P is a function of 
 r and must increase as r diminishes ; that is, as the membrane 
 is packed closer and closer and the capillaries become smaller 
 and smaller. Moore then deduces that when the radius of the 
 capillaries become so small that the surface tension acts not 
 only at the perimeter but over the' whole cross section of a 
 capillary, a condition which will be realized when the diam- 
 eter of the capillary approaches the diameter of molecules, 
 then equation (i) becomes: 
 
 n r 2 1 = n r 2 P 
 
 or t=P 
 
 where t is the difference for maximum value of molecuar at- 
 traction for solution and solvent. Therefore, with capillaries 
 of molecular dimensions, the osmotic pressure would cease to 
 be a function of r and a maximum would obtain. 
 
 It may be well to call attention to the fact that equation ( i ) 
 is the condition of a pseudo equilibrium. According to well- 
 known thermodynamical reasoning P, "the osmotic pressure" 
 as a function of r could in no sense of the word be taken as 
 
 1 Phil. Mag., 38, 279 (1894), Fifth Series. 
 
 2 Referring to Moore's article, it will be seen that less danger of con- 
 
 fusion would have been incurred had P been termed the hydro- 
 static pressure. 
 
30 
 
 the true osmotic pressure of a solution in equilibrium with the 
 solvent. 
 
 Attention is called to the deduction simply to show the in- 
 teresting parallelism between what it suggests to the mind, and 
 the inference which is aroused by a consideration of the meth- 
 ods found necessary in this laboratory to produce perfect 
 membranes. 
 
 In conclusion it may be said, that no one of the considera- 
 tions, experimental or theoretical, which have been advanced in 
 this paper can be claimed in the present state of our knowl- 
 edge to be adequate for the explanation of the discrepancy be- 
 tween the two sets of ratios ; and, with one -exception, that the 
 constancy of the discrepancy is such as to seem to preclude 
 the argument that errors were "heaped." Yet all considera- 
 tions alike seem to militate against the low and in defense of 
 the high values. The high ratios are therefore provisionally 
 judged to be the more reliable. 
 
 SUMMARY. 
 
 A large supply of cane sugar of known high purity has been 
 prepared, in order that the measurements in progress during 
 the past year, as well as those to be made for some time to come, 
 may have the advantage that they were conducted with a single 
 stock of sugar of guaranteed purity. 
 
 Minor improvements have been made in the apparatus 
 used at 20. 
 
 The measurements made of the osmotic pressures of cane 
 sugar solutions at 20 have furnished, on the one hand, data 
 which confirm the conclusion deduced from measurements at 
 other temperatures, namely, that the temperature coefficient 
 of the osmotic pressure of cane sugar solutions ranging in con- 
 centration from one tenth weight normal to normal follows 
 closely the temperature coefficient of gases within the range of 
 temperature o-2O. 
 
 And, on the other hand, there have accumulated in this in- 
 vestigation a few data which, if taken without due considera- 
 tion, might be claimed to stand at slight variance with the 
 former conclusion. But these, it is believed, are of more scien- 
 tific value when used in calling attention to certain conditions 
 and certain problems, which, it is hoped, further investigation 
 will clarify to the advantage of whatever conclusions the evi- 
 dence yet to be accumulated may point. 
 
BIOGRAPHICAL. 
 
 The author was born at Tivoli-on-the-Hudson, New York, 
 August i /th, 1884. He prepared for college at the Hotchkiss 
 School, Lakeville, Connecticut. He received his A. B. at Wil- 
 liams College with the class of 1907, and the year after grad- 
 uation was Assistant in Chemistry. The degree of Master of 
 Arts was conferred upon him in 1908 by Williams College for 
 studies in Chemistry and Physics, and the following year he 
 entered upon his work for the degree of Doctor of Philosophy 
 at the Johns Hopkins University, with Chemistry as major 
 subject and Physical Chemistry and Physics as subordinate 
 studies. He has held a University Fellowship during the aca- 
 demic year 1909-1910. 
 
UNIVEESITY OF CALIFORNIA LIBRARY 
 BERKELEY 
 
 THIS BOOK IS DUE ON THE LAST DATE 
 STAMPED BELOW 
 
 Books not returned on time are subject to a fine of 
 50c per volume after the third day overdue, increasing 
 to $1.00 per volume after the sixth day. Books not in 
 demand may be renewed if application is made before 
 expiration of loan period. 
 
 MAY 9 1917 
 JAN 12 1948 
 
 
 50m-7,'l 
 
239378 
 
 Ck.