[ 463 ] XT .
J he Density and Coefficient of Expansion of Ice .
By J. H. Vincent , D.Sc .
, B.A. , St , John 's College , Cambridge , Communicated by Professor J. J. Thomson , F.R.S. Received January 22 , \#151 ; Read February 6 , 1902 .
There are perhaps no subjects in the domain of experimental physics which call more urgently for attention , than investigations into the properties of water in its various states of aggregation .
And of the various points which still need study , the latent heat of fusion is without doubt the most pressing .
The method which promises to yield a reliable result for this determination , requires a knowledge of the density of ice at 0 ' C. The Bunsen Ice Calorimeter has , in the hands of Dieterici and other Continental physicists , recently become an instrument of precision , but the results which this apparatus is capable in itself of yielding , are unavailable to Science owing to the lack of an accurate knowledge of the density and latent heat of ice .
But as Griffiths remarks , " There can he but little doubt that the mass of mercury expelled from a Bunsen Calorimeter by the subtraction of a definite thermal unit , is a quantity that can be and doubtless will be determined with accuracy .
" ( Griffiths , ' Phil. Trans. , ' A , vol. 186 , 1895 , p. 265 .
) It was with the object of contributing something to the solution of this problem that the investigation to be detailed subsequently was undertaken .
Previous Methods and Residts .
The first paper of importance , as regards scientific accuracy , on these subjects was published by Brunner in 1845 ( ' Pogg .
Ann. , ' vol. 140 , p. 113 , 1845 ) , hut before treating of his paper , we may glance at the state of knowledge on the subject when he attacked it .
He was led to take up the research by the fact that Petziioldt ( Petzholdt , ' Beitrage sir Geognosie von Tyrol , ' 1843 ) had announced that ice expanded when its temperature was lowered .
Petzholdt obtained this result experimentally , and proceeded to found thereon a new theory of glacier action which had the effect of bringing his paper into prominent notice .
The idea that ice contracted on warming was an old one , and had been originally mooted by Musschenbroek , ( a 310 .
) 13-5-1902 a hundred years previously .
( Musschenbroek , 'Essai de Physique , ' Leyden , 1739 .
) Mairan also supported it by experiments published ten years later .
( M air an , 'Dissertations sir la Glace , ' Paris , 1749 .
) But Heinrich in 1807 had obtained a positive coefficient .
( Heinrich , 'Gilbert 's Annalen , ' vol. 26 , p. 228 , 1807 .
) His result , obtained by the direct determination of the change in length of a bar of ice , yields the value '000024 as the linear coefficient for a degree centigrade .
This observer also found the density of ice to be '905 .
Thus the subject stood when Brunner commenced his experiments .
Brunner started experimenting in the direction of preparing air-free ice from boiled distilled water , but failed to obtain it free from air bubbles .
Even when he covered the surface of the water with turpentine immediately after boiling , the product had still to be rejected owing to its being full of small cracks ; so that he was led to use selected pieces of river ice .
The method consisted in weighing the ice in air , and in either turpentine or petroleum , which latter liquid had the advantages : 1 .
Of its smaller density ; 2 .
Its freedom from solvent action .
He determined the density of the liquid by weighing a piece of glass in it immediately before and after weighing the ice , and he subsequently weighed the same piece of glass in water at different temperatures .
He satisfied himself by direct experiment that by determining the temperature of the oil he also obtained the temperature of the ice suspended in it .
The whole of the operations were conducted in a laboratory , the temperature of which never rose above freezing point .
After making due allowance for the buoyancy of air , the result for the specific gravity of ice at 0 ' C. , referred to water at 0 ' C. , was '9180 , or *9179 as the density in grammes per cub. centim. The linear coefficient of expansion was '0000375 , which Brunner remarks was greater than that previously found for any other solid .
The paper of Petzholdt also set Struve to work about the same time ( Struve , ' Pogg .
Ann. , ' vol. 66 , p. 298 ) .
He obtained the value '0000531 for the linear coefficient per 0 ' C. , using long bars of artificial ice in his experiments .
Marchand ( s Journ. f. prakt .
Chemie , ' vol. 35 , p. 254 ) , using a dilato-meter of glass containing mercury and the ice to be experimented on , obtained '0000350 for the linear coefficient , but did not state the kind of ice used .
The dilatometric method was also employed in 1852 by Plucker and Geissler ( 'Pogg .
Ann. , ' vol. 86 , p. 265 , 1852 ) , who determined the density and dilatation of artificial ice .
They used a dilatometer of a remarkably elegant design ( see fig. 1 ) .
The cylinder , M , of thin glass is open at the bottom , and has a capillary tube , c } inside it .
This tube is sealed into the cylinder , M , and also into the outer cylinder , N , at one end .
At the other end of the outer cylinder was another capillary tube , and in the preliminary part Fig. 1 .
OF CUBICAL EXPANSION OF ICE .
of* the experiment , this tube reached as far as o only .
The coefficient of expansion of the glass was first determined by preliminary experiments with mercury .
To partially fill the inner cylinder with water , a bulb provided with a small opening at the top was sealed on at c. This bulb was filled with distilled water , and after this had been boiled for some time the upper orifice ot the bulb was sealed .
During this operation the opening at o had been closed , but , after cooling , both the upper sealed point in the bulb and o were simultaneously opened .
Water flowed into M , expressing an equal volume of mercury .
The tube at o was again closed , and the water in the fine capillary at c was displaced by slightly warming the apparatus .
This caused the mercury in which the central capillary dipped to rise , and the bulb was then removed by sealing off at c.Finally the capillary tube d o was sealed on at o , and the position of the end of the column of mercury marked on the tube after the whole apparatus had been reduced to 0 ' 0 .
On freezing , the water in the inner glass vessel expanded , and breaking the inner cylinder , relieved itself from constraint .
The mean result for the increment in volume of unit of volume of water at 0 ' C. , on changing to ice at 0 ' , was '09195 , which is equivalent to '91567 for the density at 0 ' , while '0001585 was obtained for the coefficient of cubical expansion .
The next observations with which it is necessary to deal are those of Dufouk ( ' Comptes Rendus , ' vol. 54 , p. 1080 ) .
Having previously experimented by finding the density of a mixture of alcohol and water in which ice floated in neutral equilibrium , he published in 1862 an account of experiments in which a mixture of chloroform and petroleum was used in preference to the former liquid , which dissolves ice .
By taking the mean of 16 experiments , he obtained '9178 , with a probable error of '0005 , as the specific gravity referred to water at 0 ' C. He employed the value '000158 for the coefficient of expansion for reducing his results .
The ice used was prepared from water boiled in vacuo , and although free from air bubbles was " opaline " in appearance .
Bunsen 's celebrated paper on Calorimetry appeared in 1870 Fig. 2 .
( ' Pogg .
Ann. , ' vol. 141 , p. 1 , 1870 ) .
Amongst other researches included in this memoir is a determination of the density of ice at 0 ' C. , by a dilatometric method which , according to the illustrious ' author , completely eliminated the errors which had rendered previous estimations uncertain .
Bunsen 's dilatometer is shown in fig. 2 .
It consisted of a thick-walled U-tube of hard glass drawn out at a , and this was filled with mercury up to the level , b , which was boiled for some time .
Boiled water was sucked into the apparatus , and rested on the mercury at b. This water was then boiled in the tube for half an hour , the end a being , by means of a rubber tube c , led under the surface of water , which was also kept boiling .
The dilatometer was allowed to vol. cxcviii.\#151 ; A. 3 o cool , while the vessel into which c dipped was kept boiling ; the side of the tube a b then became completely filled with air-free water , and the point was then sealed off .
By weighing before and after filling with water , the mass of water taken was obtained .
The other limb of the U-tube was now filled with mercury , and the water was frozen by subjecting it to cold in such a way that the water froze from above downwards .
The ice thus formed was absolutely clear .
The cork and capillary tube shown in the figure were then inserted , and the whole apparatus surrounded with dry snow .
The mass of the vessel B and its contained mercury was noted .
On melting and again reducing to 0 ' , re-weighing the vessel B provides the other datum requisite to compute the density of ice at 0 ' C. Bunsen 's mean value was " 91674 .
No experiments on these subjects seem to have been published again until quite recent times , when Nichols brought out his paper on the density of ice in 1899 ( 'Physic .
Beview , ' vol. 8 , January , 1899 ) .
Leaving the theoretical portion of this memoir out of consideration for the present , we find that Nichols determined the density of artificial and natural ice by several methods .
Method 1 .
Specific Gravity Bottle.\#151 ; The apparatus consisted of a specific gravity bottle fitted with a tube ( see fig. 3 ) , round which a cylinder of ice was formed .
The unfrozen water was shaken out , and the whole again subjected to cold ; by weighing in a laboratory , whose temperature was below freezing point , the mass of ice taken was found .
The bottle was now filled up with cold mercury , the stopper inserted , and the whole left in an ice bath overnight with the stopper dipping in mercury .
Finally , the stopper was dipped into a weighed quantity of mercury , the ice permitted to melt , and the whole apparatus again reduced to 0 ' C. The loss of weight of the mercury into which the stopper dipped , gave the means of computing the density of the ice mantle at 0 ' C. free from any error due to deformation of the flask on filling with mercury .
The result for the density from a single experiment was '91619 .
Method 2 .
Brunner 's Method.\#151 ; Nichols employed refined petroleum , and weighed several varieties in it , again working in a laboratory below freezing point .
The results were reduced to 0 ' C. , by employing the value '00015 for the coefficient of cubical expansion .
The results obtained were :\#151 ; _ i r rp ' t j Kind of ice .
Artificial ... ... ... Natural ... ... ... ... ( icicles ) Method 3 .
Determination of the Volume of the Ice by \#151 ; Nichols next attacked the question by the employment of an absolutely original method .
An Density at 0 ' C. Kind of ice .
Density at 0 ' C. .
-91603 Natural .91792 ( new pond ice ) .91795 Natural .91632 ( pond ice , 1 year old ) Fig. 3 .
OF CUBICAL EXPANSION OF ICE .
iron box of special construction , having a capacity of about 2 litres , and rectangular in shape , was nearly filled by means of a regular block of new pond ice , and the rest of its interior was filled up with mercury .
From the weighings of this mercury and the ice , and a knowledge of the volume of the box , the density of the ice at 0 ' C. was computed .
The ' ice contained a small quantity of air , the amount of which was separately determined and allowed for .
The final value for the density came out at .91760 .
Nichols next turned his attention to the determination of the linear coefficient of expansion of ice .
No work had been done on the dilatation of ice since 1852 .
The method employed ( ' Physic .
Review , ' vol. 8 , p. 184 , 1899 ) was similar to that used by Struve in 1845 .
A bar of commercial artificial ice , which had been manufactured some months previously , was used , and Nichols again had the felicity of working in a laboratory which was never warmer than \#151 ; 3 ' C. during the work .
The readings were obtained by measuring , by means of a dividing engine , the distance between the centres of two tiny drops of mercury resting in depressions in the ice about 40 centims. apart .
The range of temperature was from \#151 ; 8 ' to \#151 ; 12 ' C. Four sets of readings were taken , with the mean result *0000540 for the linear coefficient .
Nichols 's Theory .
In order to explain the remarkable discrepancies between the values obtained by previous observers , Nichols put forward the theory that in reality there are two kinds of ice which have been under experiment ; the density of artificial ice being about '916 , and that of natural ice more than one part in a thousand greater .
This immediately throws the subject into a more tangible form , but the serious consequences of such a dual character for ice demand most careful consideration .
It seems to me that if there are really two kinds of ice , differing in density so largely , these varieties would also have different latent heats , and , what is perhaps more important still , different melting points .
Nichols 's theory is , however , supported entirely by his own work , and also by most of the results of previous observers .
In this connection it must be pointed out that , according to one of Nichols 's experiments , natural ice assumes a density approaching to that of artificial ice if the natural ice has been kept some time .
The value obtained for the density by Dufour for artificial ice is larger than that of other observers using the same variety .
But the method of neutral equilibrium is far inferior in exactitude to the methods employed by Plucker and Geissler , Bunsen , and Nichols .
Neither must the results of Bunsen be accepted as being of extraordinary reliability in spite of his assurance that he had eliminated all error .
The U-tube dilatometer suffers from the disability that any small difference in the method of holding it may cause considerable change in its voluminal contents .
It must also be remembered that the ice in Bunsen 's experiment was probably under 3 o 2 considerable pressure , since each new layer as it was formed became the vehicle ol transference of heat upwards from the underlying water .
Ice is one of the most contractible of solids by fall of temperature , and thus when the whole of the water was frozen , it must have been considerably denser than it would be at 0 ' C. The sides of the somewhat narrow tube would tend to prevent the ice assuming its proper density as the temperature rose to 0 ' C. It should be noted that if the mean temperature of Bunsen 's ice column was 4 ' or 5 ' below zero , this would suffice for the somewhat high value which he obtained .
There is another and more serious objection to Bunsen 's method .
Any attempt to get ice exactly at 0 ' C. by surrounding it with an ice jacket may result either in the resulting temperature being lower than 0 ' 0 .
, through the observer not allowing a sufficiently long time for the equalisation of temperature or , on the other hand , may result in some of the ice melting .
In either case the value obtained for the density will be too high .
None of these objections apply to Pluoker and Geissler 's work .
Barnes ( ' Physic .
Review , ' July , 1901 ) has recently determined the density of natural ice by weighing selected specimens in water .
His residts give the same density for old and new ice .
The mean value obtained ( expressed in grammes per cub. centim. ) was *91649 .
Synopsis of Previous Work .
In order to facilitate reference , the results of previous workers have been set out in Table I. , which gives the methods adopted , the variety of ice used , and the results obtained by different observers .
In Table II .
the results for the two kinds of ice are separately set forth .
The work of Marchand is omitted altogether from this table , as he did not state what kind of ice he used .
The value of the density obtained for old pond ice by Nichols is also not included .
The mean result for the density of natural ice at freezing point is *9176 gramme per cub. centim. , while that of artificial ice is *9165 gramme per cub. centim. If , however , we neglect Dufour 's value , we obtain the result *9162 gramme per cub. centim. for artificial ice .
Only one estimation of the dilatation of natural ice is available .
It is *0001125 for the cubical coefficient of dilatation for 1 ' C. , while three results are available for artificial ice .
The mean value is *000160 for the cubical coefficient of dilatation for 1 ' C. Only one direct determination of the cubical expansion of artificial ice is to hand .
This was obtained by Plucker and Geissler , and is *0001585 for the cubical coefficient of dilatation for 1 ' C. In both Tables I. and II .
the cubical coefficient only has been tabulated .
In those cases in which the linear coefficient was actually determined , the cubical coefficient OF CUBICAL EXPANSION OF ICE .
Table I. Date .
Observer .
Method .
Kind of ice .
Density .
Coefficient of cubical expansion .
( grammes per 1845 Brunner .
Weighing in liquid Natural ( ub .
centim. ) .9179 .0001125 1845 Struve . .
Direct measurement of linear co- Artificial .0001593 efficient 1845 Marchand .
Dilatometric ?
.0001050 1852 PLUCKERand Dilatometric Artificial .91567 .0001585 Geissler 1862 Dufour . .
Neutral equilibrium in liquid . .
Artificial .9177 1870 Bunsen : .
Dilatometric Artificial .91674 r Dilatometric Artificial .91619 | f Artificial .91603 !
Natural .91795 \#151 ; VT 1 ( icicles ) 1899 Nichols !
Weighing in liquid 1 | Natural .91792 \#151 ; ( new pond ice ) Natural .91632 \#151 ; ( old pond ice ) Volume by displacement ... .
Natural .91760 \#151 ; ( new pond ice ) 1899 Nichols Direct measurement of linear co Artificial \#151 ; *0001620 efficient 1901 Barnes . .
Weighing in water Natural .91649 \#151 ; 1 Table II .
Natural .
Artificial .
Observer .
Density .
Co. of Cub. Exp. 1 Density .
Co of Cub. Exp. Brunner Struve Plucker and Geissler . . .
Dufour Bunsen Nichols ' Barnes .9179 .91795 .91792 .91760 .91649 .0001125 __ .91567 .9177 .91674 .91619 .91603 .0001593 .0001585 .0001620 Mean .9176 .0001125 .9165 or .9162 neglecting Dufour .000160 i has been tabulated as three times the linear , but the legitimacy of this procedure is open to grave doubt in the case of a body like ice which is endowed with hexagonal symmetry of structure .
Principle of the Method employed .
Since the question of the density of ice was still , in spite of all the labour that had been spent upon it , in a far from satisfactory state , and since a direct determination of the Cubical Coefficient of Expansion had not been attempted since 1852 , I was desirous of employing a method which should yield both results .
In order that the work should have any value , it was necessary to employ some device other than any which had been used previously .
The method of weighing ice in mercury was one which naturally suggested itself .
For the purpose of a sinker two metals are available , tungsten and platinum .
Tungsten is difficult to work , but is readily procurable in any amount ; if mercury attacks tungsten this could be avoided by protecting it by an iron shell .
Although this direct method was not employed , I believe that the use of a platinum or tungsten sinker for weighing ice in mercury would be well worth the attention of future workers .
The necessity of a sinker can be done away with if the buoyancy of the ice is obtained by determining the tension of a wire which moors it to the bottom of the vessel .
The tension of the wire may be found by passing it through a small hole in the bottom of the mercury-containing vessel , which latter must be closed at the top so that the mercury will not pour out of the hole through which the wire comes .
This method obviates the use of a balance , for a scale pan may be hung on the wire , and equilibrium obtained by suitably adjusting the suspended weights .
Joly has used this principle in the construction of a balance ( Joly , ' Phil. Mag. , ' September , 1888 ) , and my apparatus differs from his , in that I introduce the material whose density is to be determined and use it as the float .
The advantages to be derived from the use of mercury are several .
It can easily be obtained quite pure , is without solvent action upon either water or ice , and its coefficient of expansion is of the same order of magnitude as that of ice .
Further , this coefficient is known with greater accuracy .
The use of a liquid whose coefficient of expansion is near that of ice is helpful in determining the density at 0 ' C. , and also in the determination of the coefficient of cubical expansion .
In order to determine the density at 0 ' C. it would be sufficient under ideal conditions of temperature to find the buoyancy of an inverted vessel in mercury , to introduce air-free water into this vessel , to determine the buoyancy of the vessel and also that of the water in its liquid and solid state .
The experiments were performed by equilibrating the ice at temperatures below zero , and to find the density at these temperatures we must allow for the contraction of the mercury and of the vessel containing the ice .
The equilibration of the water was always performed at zero .
OF CUBICAL EXPANSION OF ICE .
Description of Apparatus .
The vessel which constituted the reservoir for the liquid used in the hydrostatic balance is shown in section in fig. 4 .
It consisted of a funnel-shaped vessel , A , of Fig. 4 .
( f size .
) cast iron .
This was turned up in a lathe , and care was taken to have the inside surface quite free from " blow-holes .
" Into this vessel at B , a tube of steel , N , was driven .
The lower end of this tube carried a tightly-fitting screwed piece C , the bottom of which was closed , except for a central hole about '4 millim. in diameter .
This hole was for the wire to pass through downwards to support the scale pan , and upwards to tether the vessel which held the ice or water .
The vessel shown in fig. 4 may be called ( to save circumlocution ) the funnel .
The top of the funnel could be closed by an accurately fitting steel circular plate , shown in section as I ) .
In order to fasten this plate down securely on to the top of the funnel , a three-armed piece of iron , shown in fig. 5 a , was used .
This was provided with three screws , E , the lower ends of which bore directly on the circular plate I ) , immediately over the p annular plane-bearing surface , F. A central screw , G , with G a milled head , served to hold the plate D in position in some manipulations which did not require that the surfaces at F should fit very closely .
I he three-armed piece was provided with three stout pegs , one of which is shown as H in fig. which is a section of one of the portions of the tri-radiate clamp of which fig. 5a is a plan .
These pegs could traverse round the channel , I , fig. 4 , and their pressure on Figs. 5\lt ; i\#151 ; 8 .
5 a 5 b 67 8 ( | size .
) ( i size .
) ( f size .
) ( 5 times size .
) ( f- size .
) the upper roof of this channel , provided reactions for the pressure of the four screws E , E , E , G. In order to place the tri-radiate clamp in position , three small gaps were cut at angular intervals of 120 ' in the outside of the top rim of the funnel , and three similar gaps were provided in the plate D. Dll .
J. II .
VINCENT ON THE DENSITY AND COEFFICIENT The funnel was provided with a screwed collar , J , which was permanently shrunk on to the cylindrical portion of its outer surface .
This collar served to support a removable ring of iron , K. The walls of this ring ( which we may call the mercury collar ) were higher than the top surface of D , when this latter was in position .
The mercury collar served two purposes ; it provided a means of sealing the whole of the top of the funnel by flooding with mercury when the closing plate D was in position , the mercury it contained surrounded the bulb of the thermometer which was used to find the temperature of the contents of the funnel .
When the funnel was not in position for the actual determinations of buoyancy , it could be held in a vice by the flat surfaces cut in the thick metal at B. The vessel which served to contain the water or ice , while it and its contents floated in the mercury in the funnel , is shown in fig. 6 .
It was somewhat of an umbrella shape , and was cut out of a solid block of mild steel in the lathe .
It was perfectly smooth , and provided no lurking places for air .
The sides , L , were made of decreasing thickness downwards as also was the central stem , M , which was pierced at its lower end with a hole which served to attach the wire by which the scale pan was supported .
Steel wires of two diameters were used in the experiments\#151 ; one , about '17 mm. in diameter , was used in the preliminary investigation of the dilatation of the umbrella , the other , about '2 mm. in diameter , was used in the actual experiments when the umbrella held ice or water .
The wire passed through the mercury in the funnel , down the centre of the tube , N , through the small hole in .
C , and had a specially constructed clamp attached to the end outside C. This clamp was made so that when held up close to the hole o by the buoyancy of the umbrella and its contents , it could be enveloped by the screwed closed tube , P ( fig. 4 ) .
The clamp is shown in fig. 7 .
The peculiarity in the construction of this clamp was that it was pierced by the holes Q and B , through which the wire passed as well as being held by the jaws .
This arrangement made the chance of the wire slipping very small .
The course of the wire is indicated by a broken line in fig. 7 .
The lower portion of the clamp was also drilled with a screwed hole , S. This hole served to receive a hook by which the scale pan used to hold the weights in the equilibrations was supported , and was screwed inside so as to enable the clamp to be attached to the piece C ( fig. 4 ) always at a definite distance by means of a second screwed piece which could be attached to C. The object of this arrangement was to ensure always that the wire was the same length .
In fig. 8 an iron cylindrical vessel is shown in section .
This reservoir could be put on the tube N ( fig. 4 ) , which was made slightly conical at its lower end in order to fit into the hole T. A bent wire , U , was fixed into the side of the reservoir , and served to keep the stopper , P , submerged when the reservoir was full of mercury .
The use of this portion of the apparatus will be referred to in describing the process of filling the funnel with pure , dry , air-free mercury .
OF CUBICAL EXPANSION OF ICE .
Fig. 9 .
( i size .
) The method of holding the funnel during the determinations , and the arrangements for surrounding it with ice or freezing mixture , are shown in fig. 9 .
The tube , C N , of the funnel , A , passed through an india-rubber bung , V , which closed the lower orifice of a large glass jar , W. The funnel was supported by a framework of iron , X , which consisted of two rings joined by three bent wires .
The lower ring-rested in the concavity of the glass jar and bore the |g weight of the funnel .
The jar in turn was supported by a larger rubber bung , Y , which fitted into a hole in the bottom of the lower wooden box , Z. Three stout brass pieces , a , only one of which is shown in the drawing , prevented the jar from tipping sideways .
The upper wooden box , b , rested on the lower box , Z , while the whole was surrounded on five sides by the non-conducting cases , c. These cases wTere removable boxes of wood loosely filled with cotton wool .
The thermometer , d , passed through a piece of ebonite , e , which rested on the top of the non-conducting case , through a hole in the case , down a wide brass tube , f in the upper wooden box , and rested with its bulb in the mercury in the collar , K. The removable thin metal vessel , g , was supported by three lugs , h , which were bent over so as to engage the rim of the glass jar .
The upper and lower boxes were fitted with the holes i , j , which could be closed by bungs .
The whole of the apparatus shown in fig. 9 rested on three levelling screws , which in turn were supported by a strong table having a hole in its centre through which the wire from the funnel passed .
The lower surface of the lower box was protected from the access of heat by filling in the space between it and the table with loosely packed cloth .
The description of any other parts of the apparatus which may he necessary will be more conveniently given when dealing with the conduct of the experiments .
Determination of the Buoyancy of the Umbrella.\#151 ; In order to find the buoyancy of the ice and water , it was necessary to determine that of the umbrella at different temperatures .
If we know the load on the scale pan necessary to obtain equilibrium when the umbrella only is tending to float in the mercury , then the weights added when the umbrella and its contents are equilibrated , gives us the buoyancy of the contents .
The funnel was taken and a thin steel wire ( '17 millim. in diameter ) passed upwards through the hole o ( fig. 4 ) until it could be threaded through the hole in the stem of the umbrella , when it was fixed by twisting the end round itself .
This VOL. cxcvm.\#151 ; A. operation was done in such a manner as to ensure that the length of wire thus used in the fastening was the same in each experiment .
The wire was then pulled tight and the little clamp attached with its jaws about 1 centim. distant from o by the method previously mentioned .
The whole length of wire used was thus always the same .
The stopper , P ( fig. 4 ) , was then screwed on C and mercury poured into the funnel .
The mercury used throughout these experiments was first cleaned by the ordinary chemical methods , then boiled in air , and distilled twice in a vacuum .
After being used in one experiment it was filtered , boiled and distilled twice again before being used in a fresh determination .
The plate D was jdaced on the funnel and the whole was turned on one side to enable any air imprisoned under the umbrella to escape .
This was repeated until no more mercury could be poured into the funnel , when the latter was heaped up with mercury and the plate D was slid on to the top of the funnel and firmly screwed into position by means of the tri-radiate clamp .
It was found necessary to carefully grind the bearing surfaces of the closing plate and of the funnel together before each experiment in order to obtain a perfect fit .
After thus filling with mercury , the funnel was removed from the vice and placed in an inverted position on a tripod .
The stopper was then unscrewed from the tube , when the little clamp could be seen supported on the top of a straight piece of wire projecting a few millimetres through the hole at the end of the tube .
The mercury reservoir ( fig. 8 ) was slipped on to the conical tube and the mercury in the funnel was then boiled by applying the flame of a large Bunsen burner to the closing plate .
The mercury which came out of the small hole in this process , partially filled the reservoir , which on removing the flame was filled with mercury boiled in another vessel .
The whole was left to assume the ordinary temperature , and then the stopper was inverted and plunged beneath the mercury in the reservoir , where it was prevented from rising to the surface by the wire U. The boiling was again performed , and then the mercury in the tube and reservoir was also boiled .
Operations of alternate heating and cooling were continued until I felt satisfied that no air or water remained in the funnel .
The stopper was then taken from the wire in the reservoir , and manipulating it so as never to permit the fingers to come beneath its orifice , it was screwed on to the piece C , and the funnel was thus closed .
The reservoir was then removed , and the funnel was again put in the vice ; this time with the mercury collar round it ready to be screwed on .
The tri-radiate clamp and the closing plate were removed and the bearing surfaces were covered with a thin film of vaseline before pouring an excess of recently boiled mercury into the funnel and sliding the closing plate again into position .
The plate was then fastened down securely and the mercury collar screwed on .
The funnel was then installed in the apparatus shown in fig. 9 , when the collar was OF CUBICAL EXPANSION OF ICE .
filled with mercury and the top of the funnel thus completely sealed .
The stopper was then unscrewed and a small vessel of hot recently boiled mercury was placed so that the hole o dipped beneath its surface .
This caused the mercury in the tube to expand and to displace any air from about the orifice of the tube as the exuded mercury flowed through it .
The funnel was now levelled by adjusting the screws on which the bottom box rested , the level being placed with its legs on the top of the closing plate .
The wire mooring the umbrella down then coincided with the axis of the tube .
In order to find the buoyancy of the umbrella at 0 ' C. , the vessel g was not used , but the jar and the boxes were filled with table ice , and the cases c placed round the apparatus .
Cloths were packed under the lower box , and the whole was left overnight .
When the equilibration was to be performed , the small mercury vessel into which C dipped was removed , the scale pan attached , and the weights adjusted so that equilibrium was attained when the wire projected 4 millims. from the hole .
The reading having been taken , the pan was removed and the hole was again closed with hot mercury .
In order to get readings below7 0 ' C. , the ice was all removed , g was put in position filled with ice and salt , and the boxes wTere filled vdth the same mixture .
Then the thermometer was inserted so as to have its bulb in the mercury collar .
The space round the funnel was clear of the freezing mixture , and the funnel thus changed in temperature so slowly that the thermometer readings could he relied on as giving the temperature of the funnel and its contents .
All readings were taken with the thermometer slowly rising .
The thermometer used in these experiments was made by Hicks , and had the portion which projected above the cotton wool case graduated from 1 ' C. to 10 ' C. in tenths of a degree .
I tested its accuracy at 0 ' C. , and could find no error .
The temperature rose ( after the apparatus had been left a day or so ) about a degree in three hours , and readings of the buoyancy could be obtained at intervals .
The results for the weights necessary to be added to the pan , which itself weighed about 60 grammes , are set out in Table III .
, and shown also in fig. 10 .
It v7ill he seen that the last weighing taken after five others agrees closely with the first , showing that no air gained access to the umbrella in the process of equilibration .
A new set of platinized weights by Oertling was used in these experiments .
They were tested after the experiments , and were found to be consistent with themselves .
Their absolute mass is not involved in the determination .
Fig. 10 .
sX 1 ' ' X \\ " \j \ \gt ; -10 -5 TfAtP .
c. Table III.\#151 ; Buoyancy of Umbrella .
Order of Weighing .
Circumstances .
Temperature .
Weights in grammes .
1st After 15 hours in ice 0 ' 5-890 2nd After 21 hours in 1st freezing mixture -8 ' 6-071 3rd After 33 hours in 1st freezing mixture - 4''48 5 .
995 4th After 45 hours in 1st freezing mixture - 0 ' .
1 5-895 5 th After 45 hours in 2nd freezing mixture - 9'*42 6-120 6 th After 72 hours in 2nd freezing mixture -0''l 5-910 Determination of the Density of Ice at different Temperatures .
The wire and clamp used in equilibrating the water and ice were not the same as those used with the empty umbrella .
The values for the buoyancy of the umbrella as read from the unbroken straight line on fig. 10 are thus subject to a correction of *012 gramme , which must be subtracted from the values thus found .
The results for the buoyancy of the umbrella are thus taken from the broken line in this figure .
Let W = the number of grammes necessary to equilibrate the water at 0'C .
; i.e. , the load in the pan less the corrected buoyancy of the umbrella , I == the number of grammes to equilibrate the ice at \#151 ; C. , i_t = the density of ice at \#151 ; t'C .
, wQ = the density of water at 0 ' C. , h0 \#151 ; the density of mercury at 0 ' C. , h_t \#151 ; the density of mercury at \#151 ; t ' C. Then equating the two values obtained from the above for M , the mass in grammes of the material taken , we have -/ *0-M = W and M .
h_ , - M = I ; W " 1- , Let K = and - q ; W hQ \#151 ; w0 therefore W .
I. K \#151 ; wo h-t \#151 ; %-t then qh-l K + q ' which was the formula used in computing the results .
Since the number K only OF CUBICAL EXPANSION OF ICE .
depends on the ratio of the weights , no correction for the effect of the buoyancy of the air is necessary .
The density of water at 0 ' C. was taken as '999884 , and that of mercury at the same temperature as '135956 , while the density of mercury at lower temperatures was found from the formula of Chappuis ( ' Proces-verbaux des Seances du Comite International , ' 1891 , p. 3 7 ) .
The results needed were read off from a plotted curve .
The funnel having been filled with pure dry air-free mercury in the manner already described , the closing plate was removed and air-free water introduced under the umbrella .
This was accomplished by means of the apparatus shown in fig. 11 .
This consisted of a glass bulb , k , which had a capillary tube , l , sealed into it above , and which terminated in a stout tube , m , below .
The whole could he supported by fixing this tube in a clamp .
A flexible rubber tube , communicated with the bidb and with a somewhat larger reservoir not shown in the figure .
Mercury was poured into this reservoir , and it was raised until the mercury poured out of the orifice , o , in the capillary tube , which was plunged into a beaker of water which had been kept boiling for half an hour .
The reservoir was then lowered , and the boiled water rushed into the bulb , k. On again raising the reservoir , this water was expelled , and boiled water was then introduced in its place .
This was repeated seven or eight times , when the mercury reservoir was raised and the beaker removed .
The water spurted out of in a rapid stream , so that no air could pass back into the bulb .
The point o was then dipped into the mercury of the funnel , and as the water Was ejected from the capillary tube it floated up and occupied the upper portion of the inside of the umbrella .
This was kept in position during filling by a set of three stout iron wires fixed in a wooden board , which was fixed on the rim of the funnel so that the lower ends of the wires pressed on the flat top of the umbrella .
This was necessary because the umbrella was only stable when it contained more than a certain quantity of water .
The point o was so made that although the water was projected upwards , the point offered no projection for the rim of the umbrella to catch upon , so that when sufficient water had been introduced , the tube could be removed .
The water used in these experiments was prepared from ordinary distilled water by re-distilling in a new block-tin still , the earlier products of the second distillation being rejected .
After filling , the funnel and its contents were allowed to cool ; the funnel was then closed with the closing plate as already described , and was placed in the apparatus , fig. 9 , the boxes having been previously washed out to get rid of the salt from the freezing mixture .
The mercury to seal the top was poured into the collar , and the Fig. 11 .
( I size .
) Table IV .
Equilibrations .
j .
1 | I. II .
III .
IV .
V. VI .
VII .
VIII .
IX .
X. XI .
Water at Ice at Temp. W by subtracting Buoyancy of I by subtracting numbers in ^nnirnn v 1 K i-t _ K +qi Experi- 5 * 886 from umbrella at I K + q. h-t meant .
0 ' C. - fC .
= -fC .
numbers in column II .
- r C. , from fig. 10 .
LUIU .
II1I1 V X from those in column III .
- w ' q = *07938 .
from fig. 11 .
( a ) 646*225 ( b ) 709*350 -1*95 640*354 5 * 930 703*420 1*09849 1*17787 13*6004 *916603 1 ( e ) 646*255 Mean of ( c ) 709*675 - 10*02 \#151 ; 6*112 703*563 1*09871 1*17809 13*6204 *917780 and ( = 643*240 ( d ) 709 * 300 -1*01 \#151 ; 5*909 703*391 1*09844 1*17782 13*5981 *916487 2 ( a)860*055 ( b ) 945*235 -2*90 854*169 5*952 939*283 1*09965 1*17903 13*6028 *915863 ( c ) 945 * 100 - *37 \#151 ; 5*894 939*206 1*09956 1*17894 13*5965 *915509 3 ( a)822*495 ( b ) 903*450 -8*37 816*609 6*074 897*376 1*09890 1*17828 13*6163 *917355 ( \lt ; 5 ) 903*080 -1*02 \#151 ; 5*909 897*171 1*09865 1*17803 13*5981 *916324 ( a ) 792*690 ( b ) 871*060 -2*60 786*804 5*945 865*115 1*09953 1*17891 13*6020 *915903 4 ( c ) 870*760 -6*59 \#151 ; 6*034 864*726 1*09904 1*17842 13*6119 *916950 ( d ) 870*580 -3*15 5*957 864*623 1*09890 1*17828 13*6034 *916487 478 DR. J. H. VINCENT ON THE DENSITY AND COEFFICIENT OF CUBICAL EXPANSION OF ICE .
funnel was then levelled .
The jar and boxes were filled with ice , and the equilibration of the water was performed after waiting about eighteen hours .
The ice m the boxes was then exchanged for freezing mixture , the glass jar was emptied of its ice , and the metal vessel g was filled with freezing mixture and put into the jar .
After leaving for two days , the ice was equilibrated , and this could be done at different temperatures , as the ice very gradually rose in temperature .
The ice could then be examined , or a new equilibration of the water could be obtained , by allowing the ice to melt and bringing the whole again to 0 ' C. The results of the different equilibrations and the computation of the density of ice for different temperatures below 0 ' C. are set out in Table IV .
In this table the letters prefixed to the equilibrations in columns II .
and III , indicate the order in which these numbers were obtained .
In finding the numbers in column IX .
, the values of q is only needed to four significant figures ; but to compute the density of ice at different temperatures ( column XI .
) this same quantity q is -0793829 .
The whole of the equilibrations performed with the final form of apparatus are given in the table .
An improvement of the filler was introduced subsequently and anomalous results were obtained for the only experiment carried out .
This was traced to the fact that the wire had become damaged during the filling , and the results for this experiment were rejected .
In the case of the first experiment the water was equilibrated at the beginning and end , and the mean value was used for computation .
The ice in this experiment was not examined .
In Experiments 2 and 3 , the ice was taken out after the last equilibration and was found to be free from milkiness and air bubbles , but it had fine circular cracks running round it concentric with the central stem of the umbrella .
The fourth experiment was conducted differently from the others , and the values obtained proved unmistakably that the same specimen of water may assume different densities on freezing .
After the value b had been obtained , the ice was permitted to melt either partially or completely .
A fresh freezing mixture was put in the apparatus , and two subsequent equilibrations of the new specimen of ice were performed .
The second specimen of ice had a considerably greater density than the first although it was made from identically the same water .
The results given in column IX .
are plotted in fig. 12 .
The points are marked with numbers indicating the experiment .
The unbroken straight lines drawn through the points give us , by extra-polation , four values for the density of ice at 0 ' C. , wdiile a broken line drawn parallel to the straight line for Experiment 2 through the point given by the first specimen of ice in Experiment 4 furnishes a fifth value .
The numbers thus obtained and the weights to be assigned to them in computing the mean are set out in Table V. , the weight assigned in each case being equal to the number of equilibrations of the ice .
Table V. Experiment .
Density of Ice at 0 ' C. Weight Assigned .
I .916335 3 2 *915460 2 3 .916180 2 l-915540 1 4 t-916060 2 Weighted mean . . .
-9160 We thus obtain *9160 gramme per cub. centim. as the density of the ice at 0 ' C. This result depends upon the assumption that the density of ice is a linear function of the temperature on a mercury-in-glass thermometer .
Systematic error in the Fig 12 .
.9175 .9170 .9160 thermometer , so long as the zero is correct , is eliminated .
The result likewise depends on the values assumed for the density of water and of mercury at zero , but is independent of the absolute mass of the weights employed .
The Coefficient of Cubical Expansion of Ice .
The errors of the thermometer are , however , involved in computing the coefficient of cubical expansion which also depends upon the particular law of dilatation of OF CUBICAL EXPANSION OF ICE .
mercury used , but these errors are probably such as will not affect the result to the accuracy with which it is given below .
The four values which can be found from the data available are set out in Table VI .
Table VL Experiment .
Coefficient of cubical expansion .
*000155 'l *000152 l *000153 j *000148 j Mean *000152 Comparison of Results .
The value '9160 for the density of ice at 0 ' C. , is lower by two parts in 10,000 than the mean of the results obtained by Plucker and Geissler , Bunsen , and Nichols .
It is 1 part in 10,000 less than the mean of Nichols 's values , but is 7 parts in 10,000 lower than Bunsen 's value .
The value *000152 for the coefficient of cubical expansion is 4 per cent , lower than that of Plucker and Geissler , the last published value for the directly determined cubical coefficient of artifical ice .
It is 5 per cent , lower than the mean value given in Table II .
Conclusion .
The results of this determination of the density and coefficient of cubical expansion of ice are , that Nichols 's value for the density is confirmed , and that Bunsen 's value is probably too high ; but as the same specimen of water can freeze into specimens of ice having different density , the use of the Bunsen ice calorimeter in absolute determinations must be limited to an accuracy of probably about 1 in 1,000 .
The coefficient of cubical expansion seems to be 4 or 5 per cent , less than the mean of previous determinations .
The expenses of this research have been in part defrayed by a Government grant from the Boyal Society , and in part by the Cavendish Laboratory .
I wish to thank Professor J. J. Thomson , F.B.S. , for his kind encouragement , and my thanks are also due to Mr. Griffiths , F.R.S. , through whom I was led to undertake the investigation .
VOL. CXCVHI.--A .

