T H E BOILING-POINTS OF AQCEOUS SOLUTIOSS* BY WILDER D. BANCROFT AND HERBERT L. DAVIS
I n the field of physical chemistry probably no generalization has been of greater aid or served to suggest more experimental investigations than the van’t Hoff treatment of the osmotic pressure of solutions, with its demonstrations of the relation between this and the other colligative properties of solutions. This was soon followed by the empirical statement of Raoult relating the vapor pressure of an “ideal” solution and its molecular concentration, and then by the hrrhenius Theory of Electrolytic Dissociation to account for the behavior of solutions of electrolytes. Most chemists have been content to accept these generalizations as adequate or to confine their work to a study of very dilute solutions in which the deviations from the theories are not marked. Of the properties of solutions, the one most suitable for determination is usually held to be the freezing-point, and some very splendid measurements have been made of this property. The boiling-point is the property next in favor but, principally from a fear of the unknown magnitude of superheating, relatively little work has been done on the boiling-points of aqueous solutions of electrolytes. The data that are available do not cover the very dilute solutions nor the concentrated ones, much less show the changes over the whole range.
Boiling-Points at One Atmosphere A stimulus to this investigation was the very disturbing work of Kahlenberg’ who showed that many common solutes in water do not behave a t all as the ionization theory would require. I n dilute solutions the agreement is fair but in concentrated solutions their behavior is most atrocious from the standpoint of a pious adherent to the Theory of Electrolytic Dissociation. Kahlenberg showed that, while electrical conductivities at oo and at 9 jogave values for the degree of dissociation of common salts in qualitative agreement with the theory,-the values calculated from the freezing-points and boilingpoints of the same solutions were markedly different, indicating in some cases that the solute salt was more than completely dissociated. On the basis of these, as yet, uncontradicted and unexplained results Kahlenberg continues to be one of the most able opponents of the dissociation theory. This theory has been so valuable in other directions that we feel justified in assuming that possibly other factors have entered to give rise to these apparently anomalous results. Such factors as the displacement of the water equilibrium or abnormal heats of dilution may well account for these phe* This work is preliminary to the programme now being carried out a t Cornell University under a grant from the Heckscher Foundation for the Advancement of Research established by August Heckscher at Cornell University. I J. Phys. Chem., 5, 339 (1901).
592
WILDER D. BAXCROFT ATD HERBERT L. DAVIS
nomena. At least a restatement of the problem is not amiss since it may finally lead to a n explanation of these abnormalities which have been too long ignored. For many years Professor Bancroft has been proclaiming that the most profitable way of attacking this problem is not by a study of the dilute solution or "slightly polluted water" but by a study of the concentrated solutions. Pragmatically this seemed true since the attack on dilute solutions showed so little progress. An idea of the situation may be gained if one considers a large number of curves of various kinds and of various eccentricities, all tangent to the axis at' the origin. A study of the infinitely small portion of each curve near the origin will tell nothing about the true nature of the curve:. Only as one considers the extensions of the curves does the specific nature of each become apparent. If we plot properties of solutions against concentrations, only in the concentrated solutions will the specific character of these properties become apparent. It may be admitted readily that the study of concentrated solutions has its own set of difficulties; but when as much time and thought shall have been spent on these as have been expended on the properties of dilute solutions, more light will be enjoyed than we now have. While we have been most interested in boiling-points, the abnormalities of concentrated solutions are shown also by the freezing-points. These data from a number of authors are given in the tables of Landolt, Bornstein and Roth. In solutions containing from 7.09 to 24.6 grams of potassium chloride in I O O grams of water, the apparent molecular weight of the salt varies between 42.7 and 43.0 so that the curve of molecular weight against concentration is practically a straight line parallel to the concentration axis. For solutions of potassium bromide containing from 0.3 t'o 45.6 grams per I O O grams of water the values are between 62.5 and 67.7 and for solutions containing more than 8.1 grams the values are between 67.0 and 67.;. The apparent molecular weight of potassium iodide varies from 8 8 . j to 91.8in solutions containing from 1.0t o z 1.4 grams per I O O grams of water. I t should be noted that the apparent molecular weight and therefore the apparent degree of dissociation of these three salts is practically constant a t about 80 percent over a considerable range of concentration. The dissociation theory and the dilution law require that, if the values of the apparent molecular weight or of the degree of dissociation be plotted against concentration of the salt, the apparent molecular weight should give a curve extrapolating to half the value of the formula weight in the infinitely dilute solution and approaching the value of the formula weight in the concentrated solutions, The degree of dissociation extrapolates to I O O percent in the dilute solut,ion and approaches zero in the concentrated solutions. Potassium nitrate and silver nitrate appear to be in qualitative agreement with the theory, although the degree of dissociation of these salts can not be made to fit the simple dilution law.
THE BOILISG-POINTS O F AQUEOUS SOLVTIONS
593
In the study of these relations, the device was used of plotting AI/F (the ratio of the apparent molecular weight, 11,to the formula weight, F) against the concentration. This relation is simple and is comparable for all the salts ionizing into two ions. Its connection with a, the degree of dissociation, has been shown by Lewis1 to be:
In the dilute solutions the apparent molecular weight ought to extrapolate to half the formula weight if the theory be followd. Before leaving the freezing-point data, reference may be made to the experiments of Fluge12 and of Roth3 who have measured the freezing-points of some dilute solutions of common salts. T h i l e in every case extrapolation points to complete dissociation of the salts in the infinitely dilute solutions, neither freezing-points nor conductivities can be made to agree with the dilution law. In the more concentrated of their solutions the same appearance of constancy of dissociation with increased concent,ration is evident, which continues to be marked as really concent’rated solutions are reached. The data on boiling-points reveal a similar state of affairs with the abnormalities greater in most cases. This paper will deal mostly with solutions of potassium chloride, potassium bromide, potassium iodide and potassium nitrate, as common salts showing typical abnormalities. Kahlenberg’s results for potassium chloride show that in the most dilute solution ( 2 . 1 2 2 g in I O O g of water) the apparent molecular weight is 37. j . The next solution ( j . 1 2 4 g) gave 38.1 and then the apparent molecular weight decreased to 33.0 for the solution containing 31.84 g in I O O g of water. For this salt complete dissociation is indicated when the apparent molecular weight is 37.3 while the observed value, 33.0, represents 1 2 j% dissociation. Since this is a manifest absurdity, some explanation is called for. The behavior of solutions of potassium bromide and potassium iodide is similar. In t’he most dilute solutions of potassium bromide ?\I= 66.0, then 66.1, and then decreases gradually to 55.7 for a solution containing j 1 . 2 g KBr in IOO g of water. The value of 11 for potassium iodide decreases from 04 t’o 78.4. In all these cases and in others, the apparent molecular weight decreases t o a value which indicates more than complete dissociation instead of increasing toward a value indicating no dissociation as the concentration is increased. I n regard to the boiling-points of their solutions silver nitrate and potassium nitrate are again in qualitative agreement with the theory. After plotting Kahlenberg’s data, the question arose as to whether these halide solutions would extrapolate to complete dissociation as a t the freezingpoint, or to some other value, as appeared possible. The early experiments were designed to confirm those of Kahlenberg and t o answer this question. l
“A System of Physical Chemistry,” 2, 139. Z. physik Chem., 79, j77 (1912). 2. physik. Chem., 79. j99 (1912).
594
WILDER D. BAKCROFT BKD HERBERT
L.
DAVIS
The first apparatus employed was one built by N r . Ross Babbitt for use in the course in Physical Chemistry. This apparatus has proven quite useful for certain purposes such as following the sucrose inversion by observing the rise in boiling-point, or the determination of the molecular weight of nonelectrolytes. The apparatus consisted of a Dewar flask of about z j o ce capacity fitted with a rubber stopper with five holes. Through these holes passed: the Beckmann thermometer; the reflux condenser tube; a small stoppered tube for introduction of solute ; and two mercury-filled glass tubes sealed to the platinum heating spiral and supporting, just above the spiral, a Pyrex tube with vigreux tips to break the bubbles as they rise and pump solution around the bulb of the t'herniometer and out over the tube. Several runs showed the molecular weights of the potassium halides to be considerably higher in this apparatus than those found by Kahlenberg, and to vary widely and irregularly. Some electrolysis of the iodide solution was observed and the temperature readings were quite largely influenced by changes in the depth of solution and position of the thermometer. I t was found, however, that if a solution containing about two grams of salt in one hundred grams of witer were used as the standard, and elevations of the boiling-point' measured from it rather than from pure water, the apparent molecular weight. came out nearly the same as found by Iiahlenberg. This phenomenon could be best explained on the assumption that there was marked superheating of the water which decreased as salt was added. This question of superheating is the one chat requires more attention than any other in the determination of boiling-points of aqueous solutions. Superheating seems to arise from two principal causes. The first of these is an inherent reluctance of the liquid in contact with the heating surface t o form bubbles of vapor. A common remedy for this is the introduction of large roughened surfaces. A second cause is the fact that bubbles formed 3 . j cm. under the surface of water will be about 0.1' hotter than if formed at the surface. The principal problem in the design of boiling-point apparatus is the removal of this superheating, bringing the solvent vapor into equilibrium with the solution under no hydrostatic head. This problem becomes even more important when one considers the work of Gerlach,' who showed that the boiling-point of a solution is affected greatly by the material of the vessel used and the character of the heating surface. Furthermore, the effect olc various salts on the superheating in a given vessel was quite erratic and unpredictable, certain salts increasing it, certain ones having a constant effect, while in others the superheating decreased with increasing concentration of the salt. If salts do thus change the superheating of a solution in an unknown specific way, it is unsafe to assume this a constant, error throughout a set of determinations and the only apparatus that may be used for such experiments is one in which there is no superheating of the pure water, or at least one in which this superheating is completely removed.
* Z. anal. Chem., 26, 413
(1887).
T H E BOILIXG-POISTS O F AQTEOCS S O L T T I O S S
595
l'he Kashburn-Read' modification of the Cottrell apparatus har found much favor and was believed to accomplish this. Some preliminary runs on potassium bromide in this apparatus gave a fairly constant value of 11 a t 6 j to 66, thus agreeing with Kahlenberg and others. Then a test for thc superheating was devised. If this apparatus really removes superheating, the thermometer ought to read the same whether pure water and steam arc being pumped over the bulb, or the bulb is in steam only. Closing the pump tube of the Cottrell apparatus by a stopper caused the thermometer to read
ill
U
FIG.I
from 0.03' to 0.04' lower than when the pump was throwing water and steam over the bulb. This means that the superheating is not completely removed in the Cottrell apparatus and that, in the familiar form, it cannot be used for accurate results in dilute salt solutions. Probably the error introduced is much less in the organic solvents t o which its use has hitherto been confined. h similar test with the Babbitt apparatus showed a difference of about 0.4', to explain the results previously obtained. A modification of the Cottrell apparatus was designed which showed a superheating of about 0.008~,
*
J. Am. Chem. Soc., 41, 729 (1919).
596
WILDER D. BASCROFT A S D HERBERT L. DATIS
which was an improvement: but would still be too large, since the observed rise in boiling-point for the dilute solution might well be half that value. It was believed that the superheating in the Cottrell apparatus was mostly due to the hydrostatic head under which the bubbles were formed and entered the pump tube. If this large superheating were in part relieved by causing the bubbles to pass freely through a mass of the solution before entering the pump, the rest of the superheating might well be removed in that tube. To this end the pump tube was shortened and the body of the apparatus constricted in such a way that the vapor bubbles after passing upward freely through about half the liquid are deflected into the pump tube. The new apparatus is shoryn in Fig. I and, in addition to the above, will be found to differ from the Cottrell forni in some respects. The new position of the condenser A is more in accord with good theory. I n the older form the condenser opened beloiv the large stopper B and outside the sheath C, the main function of the sheath being to protect the thermometer from the cold refluxing water. If the elimination of air around the thermometer in the older form were incomplete, the reading would be too low and would rise as the air was expelled. In this form air is! expelled as soon as boiling starts. Practical considerations of construction and operation may make the lower position of the Condenser advisable without great sacrifice of accuracy. The present apparatus is made of Pyres, except the condenser, the withdrawal tube F, and its stopper, vhich are of ordinary glass. Ground joints for these parts are made very tight by the unequal expansions of the two kinds of glass on heating to the boiling-point. Solute is introduced as pellets down the condenser or as a powder through the condenser opening, boiling being interrupted and the condenser removed. In general the same procedures as given by Rashburn should be followed. The principal advantage of this apparatus lies in the fact that superheating of the pure water is really removed, since the thermometer reads the same whether the pump be operating qr not.
Experimental Work on the New Apparatus Jablcyznski and Kon' reported the construction and operation of a n ebullioscope in which the thermometer dips into the liquid, superheating being minimized by rapid stirring to produce a whirl which entraps vapor bubbles and brings them into contact with the solution. They admit a superheating of 0.01' and probably have more, but explicitly assume this superheating t o be constant. As a test solution they employ boric acid and were able to show that its apparent molecular weight varies irregularly between 61.4 and 61.86 as the concentration increases from 1.87 to 14.5 grams of boric acid in 73 grams of water. This agrees with the findings of Kahlenberg over a much wider concentration range. I n his experiments the molecular weight changed irregularly between 61.2 and 65.2 while the concentration increased fr3m 3.161 to 36.407 J. Chem. Yoc., 123, 2953 (1923).
THE BOILISG-POISTS OF AQUEOUS SOLUTIOSS
597
grams of boric acid in I O O grams of water. The change in molecular weight in both cases is so irregular that it is probably within the experimental error. I n Kahlenberg's results the most dilute solution gives 63.2 and the most concentrated 62.9 and there is no sign of a definite trend with concentration. This property of the constancy of molecular weight is shown by very few solutes in water and may possibly be accounted for on the assumption that these solutes are practically inert toward the solvent water. It certainly cannot be explained by the statement that since boric acid is so weak a n acid we would scarcely expect it to be able to show any change on dilution. Nobody postulates any dissociation or change of dissociation with cane sugar and yet its apparent molecular weight changes most abnormally as evidenced by these figures from Kahlenberg. Boiling-Points of Sugar Solutions Grams sugar in I O O g vater 20.7j 29.51 Rise in boiling-point 0.30' o 42' Apparent molecular weight 3 60 366
36.1; 289.4 o.jjo 7.10" 3 42
212
The formula weight of cane sugar is 342 and yet the apparent molecular weight decreases regularly from this value to 2 12. The volatility of boric acid in steam is well known and it was suggested that this might be affecting the boiling points of its solutions. Since these were used as test solutions, an answer to this question was sought. Koningh' found that no appreciable loss of boric acid took place on boiling aqueous solutions, until they were reduced nearly to dryness when a rapid volatilization of boric acid takes place. Skirrow2 reports the distillation of ten cubic centimeter samples from two liter solutions of boric acid of increasing concentration. The concentration of boric acid in the distillate rises rapidly at first and then flattens to a practically constant value when a solution containing 2 0 0 grams of boric acid in a liter gives a distillate containing 0.62 grams of boric acid in a liter. This indicates that the partial pressure of the boric acid is negligible. Boric acid was used in the new apparatus and gave results which are in agreement with the other investigators. Boiling-Points of Boric Acid Solutions Grams acid in I O O g water 2.41 4.82 7.23 9.64 14.44 19.28 ,617' ,817' 1.269' 1.609" Riseinboiling-point o.zojo .417' Apparent molecular weight 60.6 60.1 60.8 61.3 j9.3 62.3 Solutions of potassium nitrate present a n unexpected phenomenon. Data for these solutions are given in Table I. The formula bveight is 101.1.
J. Am. Chem.
Soc., 19, 38j (1897).
* Z. physik. Chem , 37, 84 (1901).
WILDER D. BASCROFT AKD HERBERT L. DAVIS
598
TABLE I Boiling-Points of Potassium Kitrate Solutions Grams KSOS in roo g Kater
Rise in
B. P.
0.626 I ,260 I ,910
o.oj9" 0.113 0.166
2 .j4I
0.220
3,175 4.434 6.350 9.827
0.272
0.383 0.538 0.807
JI calc.
55.2 57.9 59.8 60.0 60.6 60.2 61.4 63.3
Grams K S O S in IOO g water
Rise in B. P.
10.0
0.813'
Ij.0
I . 182
20.0
1.544
2j.o
I . 892 2.232 2 567 2.884 3 ' 493
30.0 35.0 40.0 jo.0
'
31
calc.
64.0 65.9 67.4 68.7 70.0
70.8 72.1
74.4
These data are found to lie on a curve similar in form to the dilution law curve, except that between 0.2 j and 0.7 j molar the values of the degree of dissociation are decidedly higher than would fit a regular curve. This hump in the values of a is accompanied by the observation that for these concentrations in the apparatus the bubbles formed are much smaller and much more numerous, so as to raise perceptibly the level of the boiling solution. This results in the thermometer being heated abnormally by the body of the solution and the rise in boiling-point observed being abnormally high. Above or below these concentrations the appearance of the boiling potassium nitrate solutions is not markedly different frorn that of pure boiling water. Apparently the salt a t these concentrations is having a depolymerization effect on the water. This tendency of the potassium nitrate solutions to give little change in the apparent degree of dissociation for this intermediate concentration range was found also by Kahlenberg from conductivities a t 9jo. A similar phenomenon appears in the boiling-points of the potassium halide solutions. I n these, as well as in the potassium nitrate solutions, the values for the dilute solutions have been found to extrapolate to complete dissociation in the infinitely dilute solution. Further work is now being done on these solutions and will probably be communicated later as a complete system. Consideration of these facts suggested the possibility that a t higher pressures and temperatures such solutions as here mentioned might prove to be more or less abnormal than a t one atmosphere and 100'. Some preliminary runs were made and are reported in the following sections. Description of Pressure Apparatus The water or solution to be studied was put in a Pyrex test-tube d 29 nini in diameter and I j o mm deep, with a capacity of about 100. cc, the test-tube being about three-fourths filled. The test-tube was held in a steel bomb B closed a t the top with a heavy-threaded cap which was tightened by a spanner wrench. Making this joint tight proved very troublesome but was finally
THE BOILISG-POIPiTb OF AQUEOUS b O L U T I O S >
599
achieved by machining two ridges on the cap and one on the body of the bomb so that a thin, well-annealed copper gasket was pressed into a “v” and thus closed the joint, Fig. 2 . The bomb was heated in a bath of Crisco, furnished with a motor stirrer and heated partly by a direct gas flame and partly by an electrical resistance Jvound around the bath just inside the heavy pipe-covering insulation.
FIG. 2
Dipping into the solution tested x a s a monel metal thermometer well, C, brazed into the bomb cap. This well was supplied with an extension sleeve bo keep the bath liquid from the thermometer and the well was filled with sufficient mercury to immerse the 76 mm immersion Nurnberg thermometer to the mark. This thermometer had a scale from r40° to 230’ in fifths of a degree and was calibrated by two Bureau of Standards’ thermometers having a coininon point at 200’ and extending, one above, and the other below that point. Dipping into the solut’iontested was also the monel metal lower end 01 the reflux condenser tub?, fitted with holes just inside the bomb for the escape of steam, and provided with a beveled lo7Yer. end to deflect the condensed water froni the thermoriirt~erwell and to causr it to reenter the solution near
600
WILDER D. BASCROliT AND HERBERT L. DAVIS
the side wall of the test tube. In the test-tube glass beads were used to facilitate boiling until it became apparent that they were markedly attacked by the water at the high temperatures; therefore they were replaced by strips of platinum. The monel metal lower end was attached 9cm above the top of the bomb, by a ground-joint connection, to the quarter-inch brass pipe that made up the rest of the open system. The first vertical length of pipe was surrounded through part of its length by a brass condenser D with the cold water inlet a t the lower end to ensure quick condensation. At the top of this first section of pipe was a capacity chamber E made up of iron pipe of capacity about 300 cc. Here air pressure was applied from an air tank through a Parr needle valve F. The brass pipe was then carried down to the gage. To the height of the piston on the absolute gage this pipe was filled with oil which first actuated a very sensitive Bourdon pressure gage G made by the American Schaeffer and Budenberg Corp. This gage served as first approximation and as an index of pressure changes but was not relied on for the final readings. Beyond this gage, the oil actuated the absolute pressure-gage H built around a piston and cylinder very kindly furnished by Prof. F. G. Keyes and practically of the same design as the gages built and used by him and his associates.' -1valve between the gage and the Bourdon permitted the use of the Bourdon alone to read pressures and another valve between the gage and the oil injector I permitted the direct contact through the air column between the boiling liquid and the absolute pressure-gage. The motion of the piston was magnified by means of a long pointer J moving over a scale. By observation of the motion of the pointer, an indication was obtained as to whether the weights used on the gage were sufficient to balance the pressures developed. Also variations in pressure caused by the boiling of the liquid were transmitted to the gage and were plainly perceptible by the oscillations of the pointer even though the capacity chamber had been included to minimize this effect. Final readings were always taken under these conditions after they had been shown to be constant for some time. The piston, yoke, pans, and supports which constitute the fixed load on the piston were weighed and an additional weight prepared so that the total minimum load on the piston was 10,000grams and this sum was added each time to the removable weights employed. The weights were all calibrated against standards. The constant of the gage as determined by Prof. Keyes was 0.49912 mm per gram load. This meant a total load of about 26,000 grams on the piston when the vapor pressures were about 1 7 atmospheres. h small motor actuating an oscillator, kept the piston from sticking during a run. The pressures were all converted to mm of mercury and reported as such. The whole apparatus was mounted on a pipe stand raising the gage I I 7 cm from the floor. 1
J. Math. Phys., Mass. Inst. of Tech.. 1, 196 (1922);J. Am. Chem. SOC., 41, 589 (1919).
THE BOILISG-POITTS O F AQUEOUS SOLCTIOTS
601
For tightness it was necessary to solder all joints of the apparatus except the brass ground-joint connections. These were coated with a shellac mixture and screwed tight. The bath itself had to be brazed since the temperatures employed are above the melting point' of solder. In the earlier part of the experiments, lead washers were used under the monel metal thermometer well and condenser reflux pipe where they screwed into the bomb cover. The hot water soon attacked and rendered these ineffect,ive and brazing had to be resorted to there too. In spite of the fact t,hat. inside the pipe, t,he steam going to the condenser, was up to roo' hotter than the melting point of solder, the soldered parts of the ground-joint connection, joining the monel metal end and the brass condenser pipe, held fairly satisfactorily. The margin of safety was too small however, and on those occasions when the bath temperature was permit'ted to exceed 2 2 j" or 2 3 0 ° , this solder gave way, necessitating complete tearing d o m and resoldering. Cp to this point, apparently, the condensing water running down the pipe was sufficiently cool to keep the solder from melting. It would probably have been better design to have replaced the two ground joints by one such joint above the condenser, and to have brazed the joint bebween the monel metal end and the brass condenser tube. Such a single ground-joint would permit the necessary adjustments of the bomb.' Operation of Pressure Apparatus I:or purposes of reproducibility, constancy of superheating, etc., the bath temperatures were ten degrees hotter than the bomb temperatures it TVRS desired to measure. The latter temperatures were set arbitrarily at 18c', 190°, 2 0 0 ° , and 2 1 0 ' on the Surnberg thermometer, which gave from the calibrations, 178.9', 188.9'; 198.9'and 2 0 8 . 3 ' respectively as the corrected temperatures. As heating of the bath was started, a slight' excess of air pressure x i s applied above that t'hought to be necessary for the desired boiling-point. Then as the bomb temperature approached the desired reading, t,he air pressure was slightly reduced so that the temperature did not too greatly exceed the temperature wished. Khen a series of readings of temperature and pressure oyer possibly half an hour indicated equilibrium had been reached, these data rrere taken as valid. This procedure gave quite satisfactory results with water and probably viould with any pure solvent. But the case of solutions offered another difficulty. I n order to avoid local superheating, the bath had been designed to produce fairly uniform heating on the surface of the bomb and rather rapid stirring of the bath made this possible. But since the surface of the liquid in the tube was under slightly less pressure than the main body of the liquid, it was the portion that boiled more readily. As boiling continued and condensation started, the upper portion of the solution became increasingly diluted and the rapor pressure increased, or the temperature fell if the pressure remained constant. The authors wish to express their deep appreciation of the very able machine work done
o n this apparatus by l l r . Harry Bush, the mechanician of Baker Laboratory at Cornell.
602
WILDER D. BAXCROFT A S D HERBERT L. DAVIS
Early readings on salt solutions showed this; but the reason for it was not understood until later. Then the device was adopted of stirring the solution by a sudden release in pressure causing a vigorous ebullition with consequent stirring. The pressure was then quickly increased up to about the value expected to be in equilibrium a t the desired temperature. The temperature rose readily and a few trials established the proper pressures. The appearance of equilibrium would be maintained for possibly ten minutes or more and then the pressure would increase, or more probably the temperature would fall slightly. Readings thus obtained, were taken as representing the vapor pressures of the solutions studied and the data are here represented. Before starting the actual runs, it was necessary to discover if the bath temperature affected the readings of the thermometer in its well in the bomb. special run with water in the tube and with the system above the condenser open to the atmosphere through the safety pet-cock, showed that a thermometer in the well maintained a constant reading of 100' although the bath temperature went a? high as I~O'. S o higher reading than this was shown by the bomb thermometer although the bath was kept a t the elevated temperatures for long periods. The conclusion seemed justified that the boiling of the water in the bomb produced sufficient cooling of the thernionirter well to balance the heat carried to it by met'allic conduction from the bath. I t therefore seemed reasonable to assume that temperatures of the bomb, measured when the bath was ten degrees hotter than the bomb, would not be affected by this difference, or, if slightly affected, the error would be R constant one. Naturally the first runs of the apparatus were made on water. Table I1 shows the results of two such runs compared with the vapor pressures of water a t the same temperatures as given by the Regnault-Smithsonian Tables. T.4BLE
Temp. I I1 Average Tables
li8.90 jj21
i549
-,. 133.5
7357
mni
11
188,9'
'98.93
9321 9319 9335 9218
11412
11445
11428 11123
208.3' 14007 14065 14,036 13848
While the agreenicnt n.ith the valucs of Regnault leaves much to be desired, the agreement of the runs themselves is satisfactory. h possible csplanation for the failure of the results to agree with those of Regnault inay be found in the calibration of the ?;urnberg thermometer. The zoo3 point was the only point that could be compared directly with the standard thermometers, the other points being interpolation calibrations. A t the 198.9' point, the agreement with the values of Regnault is satisfactory. After water came thc study of some solutions in water. -As indicitpd above, boric acid has been found to be a good testing solution and was, thrrefore, employed in this appaiztus.
T H E B O I L I S ( ; - P O I S T S O F AQLYE0L-s YOLUTIOSS
603
Before proceeding with the data, it might be well to give the method used for the calculations of the molecular \?eights a t these elevated temperatures. -1s indicated by Iiendall' the data of Frazer and Im-elace? can be expressed rather well by the equation,
.kt the same time, Iiendall adds that this is "the dilute solution equation" constituting the limiting case of the more exact equation,
nhrre p. is the vapor pressure of the pure solvent, p1 that of the solution, and n and 9 are respectively the number of rnols of solute and of solvent in the solution. Since this is the form we have recently shown to be valid,gwe have used it to calculate the apparent molecular weight of the different solutes. Expressing this equation in ternis of solute in I O O grains of water, it becomes n = 12.;8 log
e. Of course, n Pl
=
>I
where g is the number of
grams of solute in I O O grams of water and 31is the molecular weight of the solute. Solutions w r c used containing I O grams and 2 0 grams of boric acid per I 0 0 grams of water, heating where necessary to cawe solution. The solutions were placed in the test-tub? in the bomb and run exactly as the water had been run. The pressures and calculations are shown in Table 111.
TABLE I11 Roiling Points of Boric Acid Solutions Boric Acid (IO granis per I O O granis of water.) Temp. I j8.9' 188.9' 198.9' _ J-apor press. n.atei , 3 3-3 , _ 9335 11428 J-apor press. Solution 7326 9043 11090 llolecular weight 64 08 56.j o 60.00 dverage 60.16 Boric .kid Vapor Press. Solution lIolecular rveight
(20
grains per TI24 64 2 4
-4verage
IOO
grains of water.) 8;j 6 10757 56.29
59.j;
208.3'
14036 13620 j9.87
13213
j9.62
59.g 2
The formula weight of boric acid is 61.84. and it is rvidcnt that the average value.: her? found differ from this value by about three percent. This discrepancy may bc accounted for on the very reasonable assumption that 3 . ; cc of water were continuously out of solution as yapor and refluxing condensed water. -__~___
~
-
.J. .4m. ('hem 8 S ~ ~43, r.. r 3 t p 1921 .J. .im. ('hem. doc.. 36, 2439 , 1 g r 4 1 . .J. Phys. ('hem., 32, I [ r g - b i . 8 .
''
604
WILDER D. BASCROFT A S D HERBERT L DAVIS
The runs on the potassium chloride solutions proved less satisfactory as the design of the apparatus showed increasing signs of inadequacy. Leaks developed and it became more difficult to recognize and maintain the boilingpoint. The data in Table IT' show the average values for some runs on solutions containing 20 grams of potassium chloride in I O O grams of water.
TABLE IT Boiling-Points of Potassium Chloride Solution Temp. Vapor press. (H20) V.P. (solution) Nolecular weight calculated
175.9' 188.9' 7535 ~ U I I 9335 6995 8642 48.3j 46.7 I
198.9' 11,428 10643 j o ,63
208.3' 14,036
12~964 45.35
If we may ignore the third value, these data indicate an increase in dissociation with rise in temperature. This is contrary to the findings of Xoyes' in his high-temperature conductivity measurements in more dilute solutions than this one. On the other hand, Kahlenberg's results at 100' by the boilingpoint method give an apparent molecular weight of 3 j . 1 2 . It is probable that an improved apparat'us designed possibly somewhat similarly to that used by Soyes will be capable of giving much more reliable and consistent results in the study of this interesting phenomenon. Xore recently there appeared an interesting paper by AIonrad and Badger2 who determined the boiling-points of electrolytic caustic solutions under reduced pressures. Their apparatus was of monel metal and included a Cottrell pump and a platinum resistance thermometer. Certain details would have to be altered to permit high-pressure work but the general design seems as well suited for the problem here outlined as it proved for the investigation they made. Summary I. X study of the data on the boiling-points of aqueous solutions reveals striking divergences from the commonly accepted theories. 2. The Washburn-Read modification of t,he Cottrell boiling-point apparatus is shown to be ineffective in removing the superheating of water. The test for this is described. 3 . A new modification of the Cottrell apparatus is described and is shown capable of removing the superheating of pure water, thus making possible the determination of the boiling-points of dilute aqueous solut'ions. 4. Dilute solutions of potassium nitrate and of the potassiuni halides are found to be completely dissociated in the infinitely dilute solutions. j . A preliminary study of the boiling-points of aqueous solutions under from ten to twenty atmospheres pressure is described. 6. Results with the pressure apparatus indicate that potassium chloride is as abnormal at the higher temperatures as it is known to be a t oo or a t 100'. Cornell 1-niversity
Pub. Carnegie Inst., S o . 63 I 1907:. Ind. Eng. Chern., 21, 40 (1929).