]\gt ; Dr. P. Phillips .
[ Apr. 18 , \mdash ; Five discharges rendered the lines , and ; with four discharges the line 27803 is alone visible and very feeble .
Carbon.\mdash ; Pure Ceylon graphite was taken for the electrodes .
With five discharges only one line , , appears .
Vanadium.\mdash ; With one spark discharge the following lines appear distinctly : \mdash ; 2893.5 3190.8 3102.4 2371.2 * This is a distinct line at 3280 , and a distinct pair , , constitute with it a group .
The Viscosity of Qarbon Dioxide .
By P. PHILLIPS , D.Sc .
, B.A. ( Communicated by Prof. F. T. Trouton , F.R.S. Received April Read May 23 , 1912 .
) The viscosity of carbon dioxide was determined under considerable variation of temperature and pressure by Warburg and Babo*in 1882 , but their results are rather incomplete .
They determined a few viscosity-pressure isothermals at temperatures above the critical temperature , one such isothermal for the liquid at C. , and a few values of the viscosity of the liquid under the ) ressure of its saturated vapour at different temperatures .
Recently E. N. da C. Andrade , working at University College , began to determine the viscosity of carbon dioxide at different temperatures under the critical pressure by a continuous flow method .
This involved the use of large quantities of , which he obtained in the usual trade cylinders , and which he kept at the critical pressure by immersing the cylinder in a bath at the critical temperature .
The cylinder communicated with the capmary tube , which was surrounded by water at any desired temperature , and through which the was slowly allowed to escape into weighed potash bulbs .
The mass of absorbed in a certain time gave the quantity of flowing through the capillary tube per second , and a manometer gave * Warburg and Babo , ' Annalen der Physik und Chemie , ' 1882 , vol. 17 , p. 390 .
1912 .
] The Viscosity of Carbon Dioxide .
the difference of pressure at the two ends of the capillary .
The weakest point in the method was the use of from the ordinary cylinders , for it contains a considerable amount of moisture , which has a effect on the viscosity , especially near the critical point .
Andrade obtained a number of results*but they were very irregular , and he had to leave the work unfinished at a point where he had decided that the apparatus would need to be almost entirely remodelled .
His results are in as good agreement with the corresponding results in this paper as their irregularity will allow .
These are the only experiments on the viscosity of at high pressures of which I am aware . .
O. Rankine 's device*for , the viscosity of small quantities of a gas suggested a convenient form of apparatus and led the author to determine some viscosity-pressure isothermals both above and below the cntical temperature .
Appxratus.\mdash ; The essential part of the apparatus\mdash ; an adaptation for use at high pressures of Rankine 's apparatus\mdash ; is shown in fig. 1 .
The weight of a pellet of mercury contained in a glass drives the liquid or gas through fine capillary tube , and the time which the pellet takes to fall between two marks is determined .
The two tubes and are connected together at each end by passages bored in mild steel blocks and , and are fixed into the blocks by means of stuffing boxes , S. These stuffing boxes gave some considerable trouble at first by leaking .
washers were not used at first because indiarubber absorbs carbon dioxide and allows it to diffuse through .
Other packings which were tried had to be screwed up so tightly , and became so hard , that they broke the tubes before they were air-tight .
In the end indiarubber packing was tried and was found to be admirable , the absorption of carbon dioxide proving to be a great advantage , for the indiarubber swelled up and made a ectly t joint .
The diffusion of the carbon dioxide through it was so slow that no leak was detected for several weeks .
Into the upper steel block another steel block was screwed , and this contained two pin valves and , and had attached to it a pressure gauge reading to 150 atmospheres .
The valve served to separate the pressure gauge from the lower part of the apparatus , while .
separates this apparatus from the arrangement for evacuating and .
The screwed connection to the evacuating and filling apparatus stands out from the plane of the paper , and is indicated by the dotted circle at A. The mercury pellet was introduced at by unscrewing the small nut , and it ran into the passage * A. O. Rankine , ' Roy .
Soc. Proc 1910 , , vol. 83 , p. 266 .
VOL Dr. P. Phillips .
[ Apr. 18 , remaining there while the apparatus was being filled .
The ivory scale fixed on to the tube by clips , served to measure the height of fall of the FIG. 1 .
mercury pellet and the length of the pellet itself .
After the apparatus was filled it was placed in a large water-tight copper-lined box cm .
1912 .
] The Viscosity of Dioxide .
inside ) , with the pressure gauge and the two valve spindles , and projecting , and with the connection A communicating with the exterior by means of rubber tubing .
The box was supported on two pivots so that it could be rotated about an axis perpendicular to and about in the centre of fig. 1 .
A plate glass window in one side of the box allowed a complete view of the tube , and of the pellet and scale .
The box was filled with water , which was heated to the desired temperature by means of a hot water geyser , and the final adjustment of the temperature was made by introducing small quantities of cold or hot water and with the main bulk of the water .
The mixing was very effectually done by turning the box on its pivot .
The temperature of the large mass of water remained constant to within one-tenth of a degree for quite a considerable time\mdash ; quite long enough to time the fall of the pellet several times\mdash ; and , therefore , the temperature adjustment was very simple and accurate .
The carbon dioxide was prepared by the action of hydrochloric acid on pieces of marble , and was washed by being passed through water , and dried by being passed through calcium chloride and phosphorus pentoxide drying tubes .
It was temporarily stored over water in a aspirator of 9litres capacity .
In order that no air might be introduced into the gas , all the water was boiled before used .
On absorbing a volume of the gas in strong caustic soda solution the residue was less than a thousandth part of the original volume , and therefore the was very free from air .
The gas was introduced into the apparatus in fig. 1 by means of the arrangement in fig. 2 .
A and are two steel cylinders , A of 1800 .
capacity , and of 2000 .
They were mounted on a board in the relative position shown in the figure , and were connected by a steel tube with a pin valve The cylinder was filled with mercury up to the gauge tube , which was fixed into a stuffing box on the top of the cylinder .
Above , and fixed to it by another stuffing box , was a connection to a compression pump which would compress to 200 atmospheres , if desired .
Above the cylinder A were a gauge tube and two steel valve blocks with valves and .
The lower valve block was connected by a steel tube to the connection A in fig. 1 , and the valve served to shut off the cylinder A from the apparatus in fig. 1 .
The valve served to cut off the aspirator and the evacuating pump from the cylinder A. In order to fill the apparatus the valve and the tap of the aspirator were closed , the valves , and were opened , and the whole apparatus , with the connecting tubes and drying tubes , was evacuated to about 1 mm. pressure .
The aspirator tap was then turned on a very little so as to slowly fill the apparatus with through the drying tubes .
In order Dr. P. Phillips .
[ Apr. 18 , to get rid of the last traces of air from the apparatus the evacuating and filling with was repeated .
The valve was then closed , was opened , and the compression pump was started .
This forced the mercury from into , and so compressed the into the essential part of the apparatus .
When the mercury appeared in the gauge tube , , the valve was closed the compression pump was stopped , and the pressure in released .
Some FIG. 2 .
of the mercury in A then ran back into until the pressure of the small quantity of gas in A became less than atmospheric pressure , and then the valve was opened , allowing the to enter , and allowing the mercury to fill the cylinder .
The valve was then closed and the compression repeated .
The with and compressing it into the essential part of the apparatus was repeated until the desired pressure was reached .
The valve was then closed and the apparatus in fig. 1 was disconnected and placed in its box ready for the actual determination of the time of fall of the pellet .
The capillary tube was calibrated by a slight variation of 1912 .
] The Viscosity of Dioxide .
the method of W. J. Fisher , the value of for the equivalent uniform tube .
The length of the tube cm .
The volume between the two marks on the wide tube was found by filling the tube with mercury and running out the mercury between these two marks and weighing it .
In the earlier part of the experiment the marks were cm .
apart and the volume between them was found to be .
Later , in order to increase the time of fall , the distance was increased to 20 cm .
and the volume between these was found to be Errors and Corrections.\mdash ; The chief correction which had to be made was for the reduction in the pressure of the pellet due to capillarity .
When the mercury is moving , the front surface has a smaller radius of curvature than the rear surface , and there is , therefore , a back pressure due to this difference of curvature .
The ideal thing would have been to take a number of observations with different lengths of pellet at each temperature and pressure and thus evaluate the correction for each reading in the same way as Rankine did .
This would have been very troublesome with the apparatus adapted for high pressures , and , since a large number of values of the viscosity had to be found , would have meant prolonging the experiment enormously .
Consequently the correction was found under a few conditions , and the mean value was used throughout the calculation .
The length of the pellet was varied by running parts of it above the valve ( fig. 1 ) and then closing the valve and repeating the observation of the time of fall .
The correction decreases with rise of temperature and with increase of density of the , as is shown by the following four values which were chosen because they represent the widest changes in temperature and density:\mdash ; A few other determinations of the correction were made at odd times , and it was never found to be below and never higher than 0.36 .
The largest error introduced by taking the mean value was therefore cm .
or about per cent. 'Phys .
Rev Feb. , 1909 .
Dr. P. Phillips .
[ Apr. 18 , A small correction is needed for the rounded ends of the mercury pellet .
Assuming that the two ends were hemispherical while the length was being measured , the volume of the two hemispherical ends was , whereas if the ends had been cylindrical the volume would have been .
There is , therefore , a reduction in the effective length of , i.e. of cm .
cm .
The total reduction of pressure due to these two causes was , therefore , cm .
of Hg dynes per cm.2 .
There will also be an error due to the momentum of the fluid in the capillary tube .
This will probably be very small , as the two ends of the capillary tube are similar , and would be zero if we could assume stream-line motion at both ends .
The author has , however , been unable to evaluate it either experimentally or by calculation .
of -Bjsults.\mdash ; Pankine has shown in the same paper that the calculation of the coefficient of viscosity for a gas is just the same as for a liquid , i.e. where coefficient of viscosity ; radius of capillary tube ; ence of pressure on the two sides of the mercury pellet ; length of capillary tube ; volume swept out by the pellet ; time of fall .
When the pressure is low , the obeys the gas laws , and therefore the formula is exact .
When the pressure is high , is such a very small fraction of the total pressure that the formula is true whatever the law relating the volume and the pressure .
In this experiment dynes/ cm.2 , where 136density of mercury ; density of the length of pellet ; .
when of fall was cm .
, Therefore when height of fall cm .
or , , , , As a sample of the calculation , take the values obtained at 83 atmospheres and C. 1912 .
] The Viscosity of Carbon Dioxide . .
Length of pellet .
Height of fall .
The density of , calculated from Amagat , Method of Observation.\mdash ; After the apparatus had been filled , placed in its box , and surrounded by water at the desired temperature , it was placed in the horizontal position , the pressure was read , and the valve was closed .
It was then turned into the vertical position , so that the mercury pellet ran into the space in fig. 1 .
The box was then rotated gently in the clockwise direction , looking at the apparatus as in fig. 1 , until it was inverted .
This caused the mercury pellet to run into the glass tube G. The times of fall in the inverted and upright positions of the box were then determined four times , and the mean of the times calculated .
In no case did the times differ between themselves by more than one-fifth of a second .
The box was then placed in the horizontal position , and the length of the pellet observed by means of the ivory scale .
This could be done accurately to a quarter of a millimetre , the scale being divided into half millimetres .
The temperature was observed during the timing .
The box was then turned into the upright position again , and the pellet run into the space , and kept there while the temperature of the water was altered , or while the pressure of the was reduced , by allowing a little to escape through the valve ( fig. 1 ) .
The observations were then repeated , and so on for all the variations of temperature and pressure .
The Results .
Viscosity-pressure isothermals were determined at , and C. , and the values are given in the five tables below .
Dr. P. Phillips .
[ Apr. 18 , Temperature C. Temperature C. 1912 .
] The Viscosity of Carbon Dioxide .
Temperature C. Temperature C. The observations in the above five tables are plotted in figs. 3 and 4 , the former giving the viscosity-pressure isothermals , and the latter the kinematic viscosity-pressure isothermals .
At the lower pressures the values are only plotted for , and , in order to avoid confusion .
For the same reason , in fig. 4 , the and the isothermals are left out entirely .
The minimum values of the kinematic viscosity at , and are nearly equal , the value being .
The critical value of the kinematic viscosity will , therefore , also be 0 .
, and since the critical density is , the critical viscosity will be Putting this value into fig. 3 , it would seem that a law of rectilinear diameters holds here , as is shown by the dotted line .
Ths uncertainty as to the exact position of the points and would quite account for the slight divergence that exists , but as there are only two pairs of points from which to construct the diameter , and , as one of these pairs is a little uncertain , Dr. P. Phillips .
[ Apr. 18 , oo Pres 1912 .
] The Viscosity of Carbon Dioxide .
very little emphasis must be placed on this point .
The form of the viscosity-pressure isothermals is very similar to that of the densitypressure isothermals , but the former cross one another , whereas the latter do not .
They must necessarily cross , since the viscosity of a gas increases with rise of temperature , whereas the viscosity of a liquid decreases .
It is noticeable that the curves cross before the gas is liquefied , thus showing that it is beginning to act like a liquid as regards viscosity before condensation takes place .
Whether this would happen further from the critical temperature is , of course , doubtful ; the crossing might then happen in the passage from the liquid to the gas .
At the only two temperatures at which liquefaction takes place in this experiment , the kinematic viscosities of the liquid and gas seem to be equal at the point of condensation .
This means that in fig. 4 there is no discontinuity of position in the curves .
There is , apparently , a discontinuity of curvature .
At and atmospheres the kinematic viscosity of the liquid is , while that of the gas is FIG. 6 .
Dr. P. Phillips .
[ Apr. 18 , At and 71 atmospheres the kinematic viscosity of the liquid is , and that of the gas .
Here again there are only two pairs of such values , and therefore too much weight must not be given to this point .
If the viscosity-pressure isothermals , as their form seems to suggest , are represented by a van der Waals or Clausius equation , then the parts of a curve representing the gaseous and liquid states must be connected by a Square of Density FIG. 6 .
continuous curve as indicated by the dotted line in fig. 3 .
The portions AB , , will represent the viscosity of the superheated liquid , and the parts , the viscosity of the supercooled vapour .
These would be represented in fig. 4 by loops of some sort , such as those indicated by the dotted loops , and these would shrink to zero at the critical point .
In figs. 5 and 6 the density and the square of the density of the are plotted respectively against the viscosity .
It will be noticed from these that in the greater part of the range of the experiment the viscosity seems to 1912 .
] The Viscosity of Dioxide .
depend almost entirely on the density , for the points corresponding to different temperatures are all confused together about the same line .
The change in temperature is not great , whereas the change in density is large , so that any effect due to slightly different velocity of molecules is probably quite masked by the effect due to the great alteration of their distance apart in the critical region .
At very low densities the scale is too small to show the fact that the viscosity is independent of the density , while at the highest densities it will be noticed that the different temperatures are beginning to get separated .
Fig. 6 shows that between densities of about and the viscosity is a linear function of the square of the density .
This would mean that in this region of very rapid alteration of density with temperature or pressure , the change in viscosity is due almost entirely to the change in attraction between two adjacent layers of the fluid , for this would be proportional to the square of the density .
In other words , the viscosity in this region depends almost entirely on the term in the van der Waals equation .
At higher densities it is probable that the alteration in kinetic energy of the molecule begins to play a larger part , whereas at low densities the attraction between the molecules becomes quite negligible .
In conclusion , I wish to thank Prof. Trouton for his kindly interest throughout the experiment , which was carried out in University College Physical Laboratory , and to thank Assistant Prof. Porter for some useful suggestions .
I also wish to acknowledge the Government Grant , which defrayed most of the cost of the experiment .

